Professional Documents
Culture Documents
The Geography of Competition Firms, Prices, and Localization (PDFDrive)
The Geography of Competition Firms, Prices, and Localization (PDFDrive)
John R. Miron
The Geography of
Competition
123
John R. Miron
Department of Social Sciences
University of Toronto Scarborough
1265 Military Trail
Toronto ON M1C 1A4
Canada
john.miron@utoronto.ca
This book focuses on three questions. First, what are the impacts of location deci-
sions of firms—by affecting prices in various markets—on the locations of their
customers, suppliers, and competitors in a market economy? Second, how, when,
and why does this result in the clustering of firms in geographic space? Third, when
and how is society—overall or by sector within it—made better or worse off as a
result? At its heart, this book is in an area of scholarship that I label competitive
location theory. Broadly, this is the study of how competition among firms leads to
localization: geographic patterns of concentration among firms in markets in equi-
librium. Competitive location theory takes the form of a set of interesting problems
or stories. In this book, each story is constituted as a model, and placed within a
set of models for the purposes of comparison. Methods of analysis in this book take
the form of model construction and interpretation. What is meant by these terms?
Dorfman (1960, p. 579) suggests three aspects to model building: (1) inventing
symbols for the components and writing down the relationships connecting them;
(2) creative hypothesizing wherein behavioral and technological assumptions are
introduced; (3) quantification and statistical estimation. Location theory, as presented
in this book, emphasizes primarily the first two aspects. In this book, I reinter-
pret 11 sets of basic models in competitive location theory focusing on these three
questions. From this reinterpretation, I conclude that (1) competitive location the-
ory offers diverse, rich, and profound ideas about the nature of a regional economy
and that (2) the conceptualization of geography is central to economic analysis.
Answers to the three questions above are of great interest to students and schol-
ars in a variety of disciplines: e.g., City Studies, Civil Engineering, Development
Studies, Economics, Geography, Housing Studies, Management, Public Finance,
Public Policy, Real Estate, Regional Science, Regional Studies, Transportation, and
Urban Planning. Much of competitive location theory is drawn from Economics. It
is mathematical and logically rigorous. As such, students in other disciplines who
would benefit from its insights and could contribute to its debates do not readily
grasp it. A geographer myself, I have written this book in part to make this area of
scholarship more accessible to students from outside Economics. My goal here is
to enable the kind of intellectual breakthroughs that are made possible by a broader
discussion of ideas. What is it that makes study in this area problematic for students
outside Economics? These students quickly discover that the discipline is perhaps
vii
viii Preface
unique among the social sciences. More than any other discipline, Economics starts
from a core body of theory. As this book is focused on microeconomic applications,
I take this to include neoclassical theory of consumer demand, theory of the firm,
and welfare economics. Students begin to learn this theory from their first course in
the subject. As I remember, it was a rush; students are challenged to use the theory
from their first course and often feel empowered after just a few courses. Just as their
professors, they are soon able to assess and critique the work of others. Contrast this
with other social sciences such as Geography, Anthropology, Sociology, or Political
Science where the absence of a common integrated core means students often have
to learn multiple perspectives over a longer period of time before they can begin to
critique work in a meaningful way. This has the further implication that observation
and measurement play a more important role in the early stages of study in social
science disciplines other than Economics.1 Arguments are made there partly on the
basis of evidence (whatever the intellectual lens used to see this evidence) and partly
on the basis of theory. The same is true in Economics, but here there is an emphasis
from the outset on the idea that evidence and core theory must be jointly consistent.
Economic reasoning has a distinctive logic. As a child, I remember a particu-
lar set of economic stories my father told me. They were fascinating accounts of
the nature of money and banking, risk, insurance, and investment. In each case, as
appropriate to a storyteller, the tale was cast in the simplest of terms to enable the
listener to see how and why something worked. My father’s purpose was to get me
to better understand the working of business. Later, as an undergraduate student
studying Economics, I came across similar stories told in lectures. Of course, at
the University level, these stories were more sophisticated and disciplined, heav-
ily graphical and mathematical, and even more intriguing. Strangely, though, my
father always was uncomfortable with what might be thought to be straightforward
extensions of economic reasoning. When I came home from university for a holi-
day, he would have assembled newspaper articles or quotes from radio or television
interviews in which an economist had made a prediction about the future based on
economic reasoning. He would pounce on cases where he thought the prediction
was incorrect or improbable. He could not understand that the economist was not
actually predicting the future per se. What the economist was doing was to show
economic analysis could be relevant to a matter of popular concern or public policy.
The caveat, always implicit in the economist’s argument, is that a certain economic
outcome will follow based on a certain set of assumptions. But, my father scoffed,
if the assumptions are not tenable, what is the value of the prediction? It took me a
while to realize his own economic stories also were based on assumptions and that
the conclusions to his stories were actually predictions: as flawed as those of the
economists he scorned.
1 See, for example, the classic statement by Sauer (1924) on the survey method in geographic
research. I am not arguing here that the teaching of economics ignores empirical evidence; it does
not. What I am arguing is the relative importance of core theory early on in a student’s study of
economics.
Preface ix
I will return to that strand of thought in a moment. First, let me make a con-
nected argument. My minor subject at university was Mathematics. On the one
hand, my studies in Mathematics helped a lot with my studies in Economics. Areas
of Mathematics provide tools economists use to advance their thinking. However,
it quickly becomes apparent to students of Mathematics at the University level that
the subject is much more than that. Mathematics enables a kind of reasoning that
is pure; it is tied only to assumptions (axioms) from which a set of deductions can
then be derived. This is pure reasoning; there is no requirement these deductions be
applicable to anything in the real world. At the same time, even a mathematician
recognizes that it is the application that makes pure reasoning relevant to a broader
audience. They too have to worry about a reader’s suspension of disbelief and the
criticism that they are otherwise being self-indulgent.
In any science, mathematical tools—and the stories we build with them—must
be appropriate to the real world in which they are to be applied. It is never enough
simply to lay out some assumptions and then draw conclusions from them only in
the abstract. Adding to the confusion, some economists do not lay out all of their
assumptions at the outset of their model: perhaps because it might make the story too
dry or formal. Instead, they reveal assumptions—sometimes explicitly, sometimes
not—at appropriate points in the course of telling the story. This can be madden-
ing to readers from outside Economics: my sense is that they like to know what
is being assumed at the outset of an argument. However, this raises a fundamental
problem in economic analysis. Is it ever possible to outline fully the assumptions
that underlie a real-world problem in economics? Take one simple economic prob-
lem. A market consists of N identical individuals. Some are endowed only with one
unit of commodity A. The rest of them, otherwise identical, each have only one unit
of commodity B. Assume the proportion endowed with commodity A is pa and that
the two commodities are each indispensable. These N individuals somehow gather
in a market to exchange commodity A for commodity B. However it is determined,
what will be the market equilibrium rate of exchange between A and B? Economists
here usually assume in such cases an auctioneer and motivated individuals who
have identical preferences characterized by a well-behaved utility function. Stated
as such, the problem is simple, even elegant. Notice, however, I say nothing here
about how a legal system, social institutions, and other mechanisms make the mar-
ket work. The reluctance of economists to define even the notion of a market adds to
the discomfort. Other things have to be assumed in order for there to be a predictable
market rate of exchange. Therein is the source of a key shortcoming in any attempt to
build economic models. Compared to Mathematics where the axioms constitute all
that can be assumed, economic models assume more than the author typically states
at the outset. Strangely, it can be viewed also as a source of richness. After all, what
makes economics and the other social sciences so intriguing (especially to those of
us who are relativists2 ) is the nuance: the ambiguity about underlying motivations
8I should be careful here. Not all economists concur with all aspects of the core theory.
Nonetheless, a widespread acceptance of the relevance of core theory within the economics dis-
cipline is sometimes thought by readers from other social science disciplines as verging on the
ideological.
9 See Roth (1989)
10 See Friedman (1953, pp. 3–43).
11 It also instances the question, raised above, of how close the analog has to be to generate
dependence or causality.
12 By rational, I mean simply that an economic actor makes choices consistently: see Becker
(1962). A consumer is deemed to be rational if, in choosing between alternatives, he or she exhibits
preferences among these alternatives that are (1) well-ordered and (2) stable over time. Here well-
ordered means that if the consumer prefers alternative A to B and also prefers B to C, then that
consumer will prefer A to C. The phrase “stable over time” is to suggest that if the consumer pre-
ferred A to B yesterday, then other things being equal, will still prefer A to B today. However,
economists typically impose additional constraints on rationality, including diminishing marginal
utility, limitations on separability, and notions of expected utility. For an interesting discussion of
the nature and paradoxes of rational choice, see Sugden (1991).
Preface xiii
13 A Case Study approach makes use of in-depth analysis involving one or a few examples. Here
typically—in a process sometimes called process tracing—proponents construct an explanation for
the case study and then show that the explanation is consistent with all information available about
the process.
14 The Deductive approach is based on a theory (definitions, classifications, and logic) and makes
use potentially of data for a large number of examples (alternatively, observations or cases) thought
to illustrate the phenomenon. Here proponents typically identify a theory to be evaluated (the
so-called alternative hypothesis) and a null hypothesis (its negation), a statistical population (the set
of cases generated by the same process), a sufficiently large sample of cases from that population,
and then a statistical test using the sample data to see if the null hypothesis can be rejected in favor
of the alternative hypothesis.
15 Rodwin (1945) nicely compares the two approaches with respect to industrial location analysis.
16 The exercise is similar in a Case Study.
xiv Preface
of fiction. We are more like writers of “how to” books than books of fiction.17
This book is no different; my objective is to use a deductive approach to tell
compelling stories: i.e., stories structured to pique the curiosity of readers while
leaving no escape from their sometimes startling conclusions.
• Reductionism. Reductionism is at the core of this book. A reductionist like me
believes the best way to improve our understanding is to focus on an aspect of
the process under study and simplify the problem enough to permit us to rep-
resent it as a model (often mathematical). Inherently, this involves a suspension
of disbelief (is the model informative even if it is unrealistic): i.e., the sense we
have assumed away something in fact important.18 In using a model, we hope to
better our understanding of the process under study: i.e., our ability to describe,
simulate, explain, predict, or control it. In an idealized (mathematical) setting, a
model is built from a set of assumptions (axioms); logical deduction then leads to
a set of outcomes (or hypotheses). The outcomes themselves sometimes initially
surprise: as in “I did not know these assumptions would necessarily give this out-
come!” The derivation of such outcomes can be boring, tedious and mechanical
but is always necessary; altogether too often in my own humbling experience,
errors arise when we take shortcuts in the derivations. At the same time, reduc-
tionists have an obligation to consider the broader implications of their thinking,
not just the immediate conclusions. To avoid the label self-indulgent, reduction-
ists need to consider just how generalizable are their findings.19 To a greater
extent than in the other social sciences, reductionism is at the heart of economic
reasoning. Economic reasoning consists of applying notions from core theory to
the sub-discipline. To do this, questions in the sub-discipline must be recast in,
or reduced to, a form consistent with core theory. There are several goals here.
One is to confirm that core theory is applicable to each sub-discipline: otherwise
the theory is not core. A second is to garner insights about the process. A third is
to create testable hypotheses that can then be refuted on the basis of data analy-
sis. If the data lead to rejection of a hypothesis consistent with core theory, this
may raise important questions about the core. In an explanatory approach, the
model can generally be subjected to an independent test. For instance, the model
to be statistically estimated allows for the possibility that either X affects Y or
it does not. Economists are fond of such models; (human) geographers—more
17 Worse than that for the wary reader, scholarly writers overtly proselytize; they seek to convert
readers from disbelievers into believers. The Case Study approach seeks to convince the reader
through immersion in detail applicable to that one case; the Deductive approach seeks that sus-
pension of the reader’s disbelief through an appeal to generalization (an escape from detail). Truth
be told, the drive for compellingness typically means scholarly writing often has the subtlety of a
sledgehammer.
18 I include here the translation problem raised by Dennis (1982a, 1982b, 2002), wherein
economists do not prove by deductive exercises what they claim to prove about their economic
subject matter.
19 This well-known point is made in Edgeworth (1888, p. 348, Footnote 6) and in Solow (1956,
p. 65).
Preface xv
model,21 something readers from Geography and other social science disciplines
can find mystifying, impenetrable, inappropriate, unrealistic, or too determinis-
tic. Economists favor a parsimonious writing style—at its best, breathtakingly
insightful—that other kinds of readers may see as either narrowly focused ways
of thinking about behavior or simply as too dense or terse. This book attempts
to overcome aspects of that by casting models in terms readers from other disci-
plines can more easily understand. In general, the objective here is to make some
of the tools of competitive location theory better understood in broad swaths of
the social sciences to which it should be of interest.22
• The origins of economic models. Where do models come from? Put differently,
when a location theorist first identifies a problem (i.e., a question to be addressed),
how does he or she come to represent it in the form of a model? Sometimes
the model is spawned directly from a body of theory; sometimes it is an exten-
sion to an already-existing model. And, sometimes, as Murphy and Panchanadam
(1997, p. 342) state, it is an analog (as in the functionalist approach), wherein
we abstract features of our problem, look for similar problems that have been
solved elsewhere, and then apply those solution methods to our own problem.
Part of the unease that other scholars have with neoclassical economics is the way
in which that area of study has borrowed models from Newtonian physics. Put
differently, these critics see microeconomic explanations as functionalist rather
than as explanatory or instrumentalist. In this book, I hope to convince skepti-
cal readers that the benefit to be derived from use of these models outweighs the
limitations.
• Refutability. Another source of bewilderment to some scholars is the emphasis
placed by economists on the refutability of scientific propositions.23 A model
is generally deemed inappropriate in Economics if it does not permit the possi-
bility of being refuted. In areas of the social sciences where the culture is one
of inclusiveness of ideas rather than exclusiveness, this emphasis on refutabil-
ity may seem unwarranted. However, I think that every scholar would agree that
it is important to cull arguments or ideas that can be shown to be incorrect or
inappropriate; in my view, refutability is valuable in that exercise.
21 Here, I use model to mean a simplified representation of the real world whose purpose is to
describe, simulate, explain, predict, or control a process under study. Although models can take
various forms from physical to chart to mathematical, I use the term in this book to mean only
something that can be expressed mathematically. In this situation, a model includes values that are
exogenously given (typically parameters and exogenous or lagged variables) and equations (both
behavioral and identities) that link these to the values of endogenous variables.
22 In large part, this is due to the way economists practicing competitive location theory look at
the world in general and location in particular; their ideas and methods, however insightful, do not
come easily to students in other disciplines. Of course, every discipline has its own sensibilities,
and the starting assumptions that characterize one discipline may well be the antithesis of another
discipline’s creed. Indeed, one might ask, why else have disciplines?
23 See, for example, Ellsberg (1954, p. 529).
Preface xvii
Let me conclude this preface with a few statements on the background assumed
and the pedagogical approach of this book. In this book, I assume the reader has
completed first-year university calculus. Even though I do not explain calculus con-
cepts and derivations here, readers without that background should still be able to
follow the main ideas presented in each chapter. The book uses other mathematical
methods as well: notably vector analysis, fixed point analysis, variational inequali-
ties, and mathematical programming. In such cases, the text does not assume readers
are knowledgeable: I therefore take time to explain important concepts and proce-
dures where needed. In keeping with the emphasis on mathematical basics, this book
uses elementary representations of geography. In every chapter, the importance of
distance is emphasized, typically for its effect on transaction cost. This generally
includes a shipping cost or commuting cost. Usually here, I assume unit transporta-
tion cost is strictly proportional to distance. The book also assumes familiarity with
introductory economics that includes basic elements in the theory of the consumer,
theory of the firm, and welfare economics. This is the kind of material students
acquire in their first two undergraduate semester courses in microeconomics. I make
no attempt to derive microeconomic fundamentals here, but I try to provide enough
details about concepts for readers to follow the main arguments in case their back-
ground in microeconomics is rusty. The book is written to accommodate different
styles of learning: mathematical, graphical, and verbal. The body of each chapter
is a verbal exposition of a model and its outcomes. Mathematical renditions of the
model are presented in tables at the end of each chapter. Graphical renditions are
similarly presented as a set of figures.
Finally, let me end on a personal note regarding conclusions. In each chapter
that follows, I present one or more models: 44 in total. As befitting the idea that
each model is a story, the chapter is in part an exposition on structure and in part
deductions or derivations that necessarily follow from this. In this, I strive for par-
simony. Some authors (and editors) like to end each chapter with a restatement of
conclusions. I do not. The chapters that follow end with a section titled “Final com-
ments” but nothing titled “Conclusions.” In part, my objection is philosophical. To
a relativist like me, a conclusion is just the point in your analysis that you reach
when you decide to stop thinking any further. In part, my objection is in terms of
parsimony. If a conclusion I reach midway through a chapter, say on p. 6, required
understanding the derivation and qualification on the pages that preceded it, how
do I then write a conclusions section at the end of the chapter without having to
repeat derivations and qualifications. To me, a model is like a well-told story. It is
something to be appreciated for the craft in telling it, not something with which to
bludgeon the reader at the end. I apologize now to the readers who get to the end of
each succeeding chapter and wonder why the conclusions are not restated.
To all the scholars who, through imagination, creativity, and questioning, make it
possible for others, including me, to better understand and think more critically
about this subject.
To students in my courses in location theory over the years, whose surprise, won-
derment, and enthusiasm reminds me constantly that competitive location theory is
a special part of social science thought.
To my graduate students over the years who, like me, frittered away their youth
lost in the puzzles that are competitive location theory only to find those ideas,
methods, and perspectives coming back to their aid wherever their careers have
since taken them.
To my university for fostering an environment where scholars can reach for their
highest goals.
xix
Contents
xxi
xxii Contents
1.1 Introduction
In 1776 Adam Smith proposed—in his famous book, An Inquiry into the Nature and
Causes of the Wealth of Nations—that division of labor is limited by the extent of
the market. Put simply, his idea was that the greater the output (scale) of a firm the
better able it is to take advantage of the added productivity from having specialized
labor.1 In my view, this important notion has five corollaries of particular signifi-
cance for this book. First, it implies ferocious competition2 among firms, each only
too well aware that improved efficiency and profit—even survivorship itself—hinge
on increased size and market share. Second, it implies that the firm has a keen inter-
est in anything—say an improvement in transportation infrastructure—that enables
it to grow its market. Third, it raises the question of the economic design of a firm;
what activities of a firm are central to profitability and why, where, and how does it
undertake these activities? Fourth, the division of labor implies that something—an
inefficiency I label congestion within or outside the firm—limits just how large a
firm can become. Fifth, and in my view most significant, is the implied simultaneity
of price and location; just as firms select locations where prices allow them to max-
imize their profits so also do the locations of firms come to affect those very prices
and profits.
In a market economy, how do firms get organized at the sites where they produce,
distribute, or sell their commodities or services? Put differently, what are the loca-
tional strategies used by firms to compete? In a sense, almost any strategy by which
a firm survives and prospers in competition has a locational aspect. Among those
commonly used are the following: siting to increase quantity sold, to increase price
obtained, to increase market share, to reduce cost of commodities in production, to
better access business (producer) services, to better access sites where inputs are
more productive, to better access information about new technologies and market
innovations, or to reduce risks. All these strategies share the idea that the firm is
using the organization of markets to help it cope and prosper. Not all of these strate-
gies are equally important to every firm and every industry.3 Taking into account
competition among firms as well as the impact of firm siting on the siting of their
suppliers and customers, how when and why does this locational behavior lead to
localization (clustering) of firms in geographic space, the growth of some places
(e.g., some cities or districts), and the decline of others? How do these locational
strategies themselves shape the operation of local markets? More broadly, if we
assume a local economy made up of workers, capitalists, and land owners whose
incomes are wages, interest and profit, and land rents, respectively, how is the distri-
bution of such incomes affected by the spatial configuration of economic activity.4
Equally important, how do changes in price locally change the translation of these
incomes into levels of economic well-being? The wrenching local changes (e.g.,
economic, social, cultural, political, and environmental) that locational behavior
appears to inflict on particular places (e.g., cities or districts) are the stuff of public
policy and practice. Explaining locational behavior is the stuff of the social sciences
in general and Economics in particular.
Economists typically argue that localization is driven by the efficiencies (more
correctly, the enhanced profit) made possible through agglomeration of production
5A consequence arising from an economic landscape whereby the firm finds it relatively more
costly or less profitable to marginally increase the quantity of a commodity that it supplies.
Congestion may arise, for example, because of (1) limitations in the firm’s ability to manage a
larger output, (2) limitations at the factory or in the supply chain (supplier, warehouse, mode trans-
fer station, or transportation network), or (3) a deterioration in the firm’s profit from the response
of competitors. Usually, congestion is measured over the short run: i.e., before the firm has the
opportunity to adjust its investment in land, plant, and equipment.
6 An attribute of a production function whereby the firm can produce more efficiently at a higher
level of output. In the usual conceptualization, indivisibilities in production technology are thought
to make possible a lower unit production cost possible when the firm achieves a particular scale
of output. In some cases, a firm is thought to experience economies of scale through all relevant
levels of output. In other cases, the firm is thought to have a most efficient scale of output (lowest
possible unit production cost) above which the firm begins to experience diseconomies of scale.
Economies and diseconomies of scale are generally measured over the long run: i.e., giving the
firm sufficient time to adjust its investment in land, plant, and equipment.
7 Dividing the US into 9 geographic regions, Kim (1995) concludes that, after a decline between
1860 and 1890, localization rose. Localization stopped increasing during the interwar period and
then fell substantially after the 1930s. US regions were less specialized by 1987 (the final year
examined) than they were in 1860. Kim argues that long-term trends in localization are more
consistent with production scale economies and Heckscher–Ohlin models than with explanations
based on external economies. See also Belleflamme, Picard, and Thisse (2000), Bennett, Graham,
and Bratton (1999), Cook, Pandit, Beaverstock, Taylor, and Pain (2007), Giarratani, Gruver, and
Jackson (2007), Head, Ries, and Swenson (1995), Nachum and Keeble (2003), Perry and Hui
(1998), Pinch, Henry, Jenkins, and Tallman (2003), and Rugman and Verbeke (2004).
8 Studies of this type include Atwood (1928), Lawrence (1934), and Hammond (1942) on cotton
and textiles, Birkett (1930), Carlson and Gow (1936), and Alexandersson (1961) on iron and steel
production, Breedlove (1932) on the ice industry, Landon (1935) on chewing gum, Brand (1937)
on fertilizer production, and Barnes (1958) on milk production.
4 1 The Craft of the Story Teller
given. When regional planners today strategize with respect to trade liberalization
and globalization, they look to support localizations (clusters) of kindred businesses,
often with a focus on the kind of jobs that require creative skills and pay well. When
such scholars or policy analysts first become interested in phenomena like the four
above, they might well start by collecting information. Take the case of the celery
farmers. We might want to collect information that helps us answer questions like
the following. Who are these farmers? What is their history? When, why, where, and
how did they get into the business? How important has celery production been to
them over the years? Why this locale? Why did particular farmers lead in the adop-
tion of celery production? Why did others follow? What kinds of social, business,
and professional ties link these farmers? Such information is helpful in developing
or assessing explanations (ideas or theories) about the phenomenon. This kind of
research, sometimes labeled Economic Sociology, is the stuff of what geographers
and others have also sometimes called behavioral geography, locational analysis, or
simply economic geography, among other labels. Here, a range of methods—from
exploratory data analysis to process tracing—help identify explanations that appear
consistent with the evidence collected. Along the way, of course, there always is
the risk of arriving at an explanation that is ad hoc (idiosyncratic): i.e., consistent
only with the evidence from the case study at hand—celery farming today at that
locale—and not generalizable (e.g., to this locale at another time, to other locales,
or to other kinds of agricultural production).
Let me now critique such approaches. A central theme of this book is that we
are looking at the same process in all these cases: be it celery farming, redecorat-
ing suppliers, factory location, or regional redevelopment. In each case, we look at
the localization of businesses in geographic space. To me, it is only natural to ask
whether and how these processes might in fact be the same. In so doing, we begin
to elaborate a theory of location general enough to explain a range of phenomena.9
What is the similarity that binds the four phenomena outlined here? In my view,
none of the descriptions above mentioned an important process in common: namely
how the location of a firm or farm affects prices locally, including the prices of com-
modities sold, the wages paid to labor, the market rents paid by firms for their sites,
or prices of other inputs. Even if we assume that each firm individually is in a com-
petitive market—that is, the firm is a price-taker—as more firms join a geographic
cluster, they do collectively affect prices locally. At the core of this book is the idea
that the common process underlying all of these is the way in which localization
shape prices which in turn impact on localization itself.
The notion that prices in an economy are interdependent is widely attributed to
Walras.10 The idea here is simple. Assume a demand curve for a consumer good
in which the quantity demanded in the region depends upon its price, the prices of
9 In an early paper advocating quantitative geography, Huntingdon (1927, p. 289) makes a similar
argument. In the language of Curry (1967, p. 265), I am a modeler: my focus is a pre-determined
set of relationships existing in a portion of reality, not the portion of reality as a whole.
10 Marie-Ésprit Léon Walras (born 1834), a French (Swiss) economist, first published his Élements
d’économie politique pure in 1874.
1.2 Geographic and Other Perspectives on Localization 5
other goods (complements and substitutes), the income of the consumer, and tastes.
Assume a supply curve for the same good in which quantity supplied depends on
its price, the prices of factor inputs (e.g., wages, land rents, interest rates) and other
inputs used to manufacture the good, and technology. Now, also envisage a simi-
lar demand and supply curve for every other good in the economy. Assume there
are N goods in total (inclusive of factors) and that we have competitive equilibrium
in every market in the economy. In that situation, there must be a set of N prices,
one for each good, such that the excess demand (the amount by which quantity
demanded at that price exceeds the quantity supplied) is simultaneously zero for all
goods (markets) in the economy.11 Prices here are linked in two important respects:
(1) goods are substitutes or complements for one another and (2) goods must com-
pete for inputs in the same factor markets. Such an economy is then said to be in
Walrasian equilibrium.
Let me express this differently. In the context of a geography, the Walrasian
assumption is that, in competitive equilibrium, all firms are price takers and the
good sells for the same price everywhere. Location theorists say that such a good
is ubiquitous . However, once we take into account shipping cost and commuting
cost, it is possible that a good is non-ubiquitous: that is, priced differently between
locations. As such, geography poses three problems for Walrasian equilibrium.
• How do we handle the idea that in competitive equilibrium price may vary from
one location to the next? One possibility here is to assume that the economy can
be portioned into geographic regions: say 2 regions. Then, instead of solving for
one price for each of the N goods, we solve for 2 prices: one for each region.
There may be some amount of the good being shipped from one region to the
other. Nonetheless, a gap in equilibrium price between the two regions implies
that it is no longer worthwhile for a trader to purchase any more in the region
where price is lower for resale in the other region. If there are more than 2 regions,
we could imagine repeating this process for every pair of regions. Of course,
we would also have to take into account situations where there might be only 1
geographic region (spanning the economy) for some commodities, 2 for others,
and still more, say k, for other commodities. In this case, the problem posed by
geography is that it means we generally need to find not just N prices but some
multiple of N prices.
• How do we take into account the idea that geography might create a local
monopoly for each firm? Inherent in Walrasian equilibrium is the idea that the
market for each good is competitive in the sense that each buyer and each seller
is a price taker. However, given shipping cost and commuting cost, any one firm
finds that it can affect the price paid by nearby customers when its competitors
are located sufficiently far away. Rather than the traditional models of perfect
competition on which Walrasian equilibrium is normally predicated, geography
implies the role of imperfect competition across space.
• How does geography affect the way in which factors (inputs) such as land, labor,
and capital get priced? Wage rates, net of the cost of commuting, can easily differ
across the landscape. Differences in shipping and commuting costs by location
also may have the implication that land rents differ with location. A geographic
setting makes Walrasian equilibrium more problematic as a concept.
My perspective here is Walrasian in origin. I want to know (1) how the location
decision of a firm changes prices across markets, (2) how these changes then influ-
ence the locational behavior of other firms, and (3) who is made better off or worse
off by these changes.12 However, I seek to do this in ways that recognize explicitly
the roles played by geography.
This book is about the development of a theory of location focused on prices,
applicable to any competitive firm, and that therefore might be empirically tested
as part of a Deductive approach (see Preface). Specifically, it is about the conse-
quences of competition on the locations of firms generally in a geographic setting in
a market economy. From an economic perspective, a decision to locate or relocate
is inherently a decision to invest. In so doing, the firm is thought to weigh the cost
of an investment (the initial outlay) against return (the present value of the stream of
future benefits net of recurring costs) taking into account a planning horizon, riski-
ness and risk aversion, opportunity cost of capital, and scrap value. Firms sense that
some places might be more profitable or less risky for their factory, warehouse, or
store than are others. To them, relocation is also an investment; its cost is the expense
of relocation, and the future annual benefit is the net increase in profit or reduction
in risk associated with the new location. Where such an investment is attractive, we
can expect firms to compete for best locations. The use of geographic location as
part of a firm’s competitive strategy, for example, has been extensively evidenced
and described elsewhere.13
This book is part of an area of scholarship popularly known as location theory.
In fact, location theory broadly consists of two overlapping subareas: optimal loca-
tion theory and competitive location theory. Optimal location theory deals with how
firms, industries, or societies might be positioned in geographic space: e.g., to be
most efficient, most profitable or to make society best off.14 Optimal location the-
ory is not the focus of this book. Instead, this book is concerned with competitive
location theory, which looks at how competition among firms leads to geographic
12 Walras (1954 [originally 1900], p. 47) was unequivocal, indeed stirring, on the importance of
mathematics to the study of Economics.
13 See, for example, Ghosh and McLafferty (1987) and Netz and Taylor (2002).
14 See, for example, Bigman and ReVelle (1978, 1979), De Felice (1972), de Smith (1981), Eiselt
and Sandblom (2004), Hamacher and Nickel (1998), Hurter and Martinich (1989), Mirchandani
and Francis (1990), Teitz (1968a), and Thomas (2002).
1.2 Geographic and Other Perspectives on Localization 7
scale.24 To this, he added the role played by skilled labor nearby. What advantages
arise because workers in a skilled trade are localized? Marshall saw two mech-
anisms at work here: (1) children learn the trade from their parents; (2) workers
implement innovations that they hear about from friends and neighbors. Marshall
also saw advantages to the firm. A local market25 for skill means that employ-
ers have a choice of workers with the skill they require nearby, and workers find
plentiful employers.26 Nowadays, we tend to label all such advantages localization
economies.27 Included under localization economies as well are the advantages to
customers from being able to efficiently do comparison shopping when suppliers are
localized.28 Marshall also discussed what we now call urbanization economies.29
For some reason, having a diversity of firms in other businesses nearby makes your
firm more efficient.30 Marshall mentioned (1) the development of subsidiary trades
locally that improve efficiency in the supply of materials and services, (2) the insur-
ance principle31 in the local labor and consumer markets, (3) joint demand by firms
24 Marshall (1907, pp. 278–279) saw three kinds of economies of scale: economy of skill (division
of labor enabled by a larger scale of production), economy of machinery (indivisibilities enabled
by a larger scale of production), and economy of materials (less wastage in a larger operation).
25 A local market is a set of agents (suppliers and demanders) engaged in the sale and purchase of a
good wherein market price—not a single price set in a global market—varies from one local market
to the next because of some impediment (characterized by a unit shipping cost between markets).
Where a mechanism links participants so that market price varies systematically from local market
to local market (i.e., local markets are not in autarky), local markets can be termed submarkets. A
domestic market or home market each instance a local market. The counterargument to Marshall’s
labor pooling is the notion of labor poaching: see Combes and Duranton (2006).
26 Strangely, Marshall did not observe the flip side to this argument. If a firm has an advantage
in technology or process, would it not want to keep knowledge of the advantage away from its
competitors and their employees?
27 Localization economies are reductions in unit production cost that arise when several firms in the
same industry locate in close proximity. For some reason, having other firms in the same business
in close proximity allows your firm to be more efficient. Glaeser, Kallal, Scheinkman, and Shleifer
(1992) refer to these as Marshall-Arrow-Romer externalities to take into account the contributions
of Arrow (1962) and Romer (1986).
28 Keir (1919, p. 48) concurs with Marshall overall but cites some disadvantages of localization:
distance from markets for materials and outputs, the localization of suffering in hard times, the
strength of labor unions, and the creation of a labor class.
29 Urbanization economies are reductions in unit production cost made possible when firms in
different industries locate in close proximity. Glaeser, Kallal, Scheinkman, and Shleifer (1992)
refers to these as Jacobs’ economies after Jacobs (1969, 1984).
30 Javorcik (2004)—looking at foreign direct investment in Lithuania in the form of joint
ventures—found urbanization economies in the form of productivity spillovers from the foreign
affiliates to local suppliers.
31 With insurance, a consumer or firm incurs a small upfront cost (the premium) now to protect
themselves against the possibility of a substantial, though unlikely, loss during some future period.
What enables a market in insurance is the prospect of a profit by the insurer; that total payouts to
the insured plus other costs of business do not exceed the revenue earned from premiums paid by
consumers willingly insured. This is the insurance principle. In general, it requires that the risk of
loss for each customer be small and that the occurrences of loss be statistically independent.
10 1 The Craft of the Story Teller
for common services (e.g., transportation, telegraph, and printing press), and (4) the
benefit from having a variety of jobs (e.g., for men and women)32 to better address
the complementary work needs of households. At the same time, Marshall noted the
offsetting tendency of land rents to rise the more industry localizes at one place.
At the same time, he was careful to avoid oversimplification. For one, Marshall
(1907, pp. 273–274) saw the tradeoffs here between inputs and outputs: a lessening
of shipping cost or tariffs tends to increase the range and market area, of a given
factory but on the other hand also increases people’s readiness to migrate from one
place to another and thereby bring production closer to places of consumption. For
another, he saw the importance of technology and its transmission across geographic
space but also pointed out (p. 270) that technology is easily spread where favored by
the character, cultures, and ideals of the people, and by religious, social, economic,
and political institutions. Finally, he explicitly recognized the importance of great
political events and strong personalities historically on the development of regional
economies over time.
Over the last century, economists have added to Marshall’s ideas. Keir (1919,
p. 48) discussed the branching of new plants from old, the subdivision of produc-
tion, the acquisition of skills by labor locally, and the emergence of new factories
that make use of production waste. Stigler (1951, p. 192) builds on the ideas that
(1) localization allows individual firms to achieve the gains of specialization and (2)
auxiliary and complementary industries that must be nearby to operate efficiently.
In recent years, we have come to see the connections between the economics of
agglomeration and networks. Feldman and Florida (1994, p. 210) adds that innova-
tion is increasingly dependent on geographic networks of firms that relay knowledge
and expertise, channel new scientific discoveries and applications, and give access
to expertise in product marketing. Networks of economic actors—perhaps not even
in close physical proximity—can create benefits that might otherwise be available
only to actors clustered at the same place.33
Much has happened in economic thought since the time of Marshall. Marchionatti
(2004, p. 443) summarizes early critiques of Marshall: loose definitions and generic
concepts; a failure to use mathematics; no general economic equilibrium analy-
sis; and no use of systematic statistical analysis. There are a variety of views as
to how economic thought evolved in the twentieth century in light of these criti-
cisms. Marchionatti (2004, p. 454) focuses on the increased use of mathematics.34
32 Reflectinghis Victorian sensibilities, Marshall also includes here jobs for children.
33 SeeJohansson and Quigley (2004).
34 Others have argued that economists borrowed much from Physics in this process. See Mirowski
(1984) and Turk (2006). In recent years, there has been a growing critique of the mathematical
approach to economic thought. See, for example, Velupillai (2007).
1.4 The Development of Economics Since Marshall 11
Mindful of the significance of key works such as Arrow and Debreu (1954), Bowles
and Gintis (2000, p. 1411) argues that major streams of economic thought in the
twentieth century—behavior based on self-interest, exogenous preferences, and
complete and costless contracting—built more closely on a Walrasian approach than
a Marshallian approach. Baumol (2000, pp. 2–3) suggests two major advances in
the discipline since Marshall: (1) new empirical tools and the insights they give and
(2) the widespread use of theory and econometric analysis in application.35 Stiglitz
(2000, p. 1441) focuses on the importance of innovations in how we think about
information in modern economic thought.36 Other scholars today critique Marshall
for not paying enough attention to geography in the dissemination of knowledge.37
Finally, economic thought has been significantly shaped since Marshall by the idea
that it is valuable to break down production by firms (and possibly consumption by
consumers too) into processes and distinct activities that can be analyzed separately
and therefore packaged in different ways (i.e., differences in vertical integration) at
the level of the firm or geographic location.38 At the same time, Bowles and Gintis
(2000, p. 1411) argue that Marshall’s ideas continue to find application in new areas
within Economics: e.g., endogenous growth theory, behavioral and experimental
economics, and evolutionary game theory.39
From the perspective of location theory, also missing from Marshall’s analysis is
an emphasis on the implication of distance—and the associated shipping cost—for
the notion of competitive behavior in general and perfect competition in particular.40
Also missing from Marshall’s account was any anticipation as to how economic
conditions would change over the course of the twentieth century: e.g., establish-
ment of a national power grid, better transportation systems, improved functioning
of the markets for capital as well as for raw materials, technological change and
the reallocation of labor,41 and the second industrial revolution that characterized
the twentieth century.42 These changes were seen to be conducive to the emergence
of great clusters of factories that came to form manufacturing belts: e.g., the US
Midwest.43 Keir (1921, p. 83) argued that firms became increasingly better able to
overcome restrictions in materials, labor, power or fuel, or capital that otherwise
would have limited them to specific locations. At the same time, Keir concludes
that market and transportation considerations each continued to be important in the
placement of factories.
Building on Marshall and consistent with the directions of economic thought
and changes over the twentieth century listed above, the subject matter of com-
petitive location theory today is generally thought to include the following:
market formation and areas; trade and spatial equilibrium; location of a firm
(Weber-Launhardt problem); spatial pricing (Cournot-Enke-Samuelson problem);
transportation pricing (Dupuit problem); locational competition (Hotelling prob-
lem); equilibrium for an industry (Lösch problem); locational interdependence
(Koopmans-Beckmann problem); agricultural location (Thünen problem); urban
spatial equilibrium (Alonso problem), and the existence and growth of cities.44 The
breadth and variety of subjects here is indicative of the diverse ways in which prices
and localization have been linked. Because the purpose of this book is to explore
these linkages in a way that helps readers think more comprehensively and cre-
atively about prices and localization, I have strived to cover a variety of the most
promising of these topics.
As a field of study, competitive location theory has a long history. It dates back
at least to early work by Thünen first published in German in 1826.45 Early con-
tributors include Cournot in 1838 Dupuit46 beginning about 1842, Launhardt47 in
1885, Marshall starting in 1890, Weber48 in 1909, Predohl49 in 1928, Palander50 in
1935, and Lösch51 in 1939.52 Among the early American scholars writing in the area
are Fetter53 in 1924, Hotelling54 in 1929, Hoover55 in 1937, Hitchcock56 in 1941,
Koopmans57 in 1947, Samuelson58 in 1952, and Alonso59 in 1964.60 Since then,
economists have written extensively on competitive location theory.61 Geographers
and others have also done important work in location theory.62
51 August Lösch (born 1906), a German economist, worked on equilibrium in a spatial economy.
His main contribution was Die Räumliche Ordnung der Wirtschaft , published in 1939. An English
translation of this book, The Economics of Location, was first published in 1954.
52 See Krzyzanowski (1927) and Blaug (1979) on the German origins of location theory.
53 Frank Fetter (born 1863), an American economist, authored “The economic law of market areas”
published in the Quarterly Journal of Economics in 1924.
54 Harold Hotelling (born 1895), an American economist and statistician, published a seminal arti-
cle on locational competition entitled “Stability in Competition” in the Economic Journal in 1929.
See also Samuelson (1960).
55 Edgar Malone Hoover (born 1907), an American economist, published Location Theory and the
Shoe Leather Industries in 1937.
56 Frank Lauren Hitchcock (born 1875), an American mathematician, published “The distribution
of a product from several sources to numerous localities” in Journal of Mathematics and Physics
20 (1941): pp. 224–230.
57 Tjalling Charles Koopmans (born 1910), a Dutch-born American economist and Nobel Laureate
in 1975, published “Optimum utilization of the transportation system” in D.H. Leavens (ed.).
The Econometric Society Meeting (Washington, D.C., September 6{18, 1947; Proceedings of the
International Statistical Conferences, Volume V, 1948, pp. 136–146.
58 Paul Anthony Samuelson (born 1915), an American economist and Nobel Laureate (1970), pub-
lished “Spatial price equilibrium and linear programming” in the American Economic Review in
1952.
59 William Alonso (born 1933), an American regional scientist, published Location and Land Use:
Toward a General Theory of Land Rent in 1964.
60 For a historical review, see Krzyzanowski (1927) and Ponsard (1983).
61 Over the last six decades or so, important markers among writings by economists for me include
Arnott (1986), Berliant and Dunz (1995, 2004), Berliant and ten Raa (1988, 1992), Beckmann
(1968, 1999), Beckmann and Thisse (1987), Borts and Stein (1964), Bressler and King (1970),
Capozza and Van Order (1978), Combes, Mayer, and Thisse (2008), Dunn (1954), Duranton and
Puga (2001), Eaton and Lipsey (1977, 1997), Economides and Siow (1988), Fujita, Krugman, and
Venables (1999), Gabszewicz and Thisse (1986a), Greenhut (1970, 1995), Greenhut and Norman
(1992, 1995), Greenhut and Ohta (1975), Greenhut, Norman, and Hung (1987), Harris (1973),
Harris and Nadji (1987), Henderson (1988), Herbert and Stevens (1960), Huriot and Thisse (2000),
Isard (1954), Johansson, Karlsson, and Stough (2001), Koopmans and Beckmann (1957), Lee
and Averous (1973), Markusen (1986, 2002), Mathur (1979, 1982, 1989), Meardon (2002), Mills
and McDonald (1992), Moses (1958), Nickel and Puerto (2005), North (1955), Peneder (2001),
Puu (2003), Rossi-Hansberg (2005), Sakashita (1968), Sasaki (1996), Siebert (1969), Smithies
(1941), Stahl and Varaiya (1978), Steininger (2001), Stern (1972a), Takayama and Judge (1971),
Takayama and Labys (1986), Weinschenck, Henrichsmeyer, and Aldinger (1969), Wheaton (1974),
and Wilson (1987). Seminal related work thought to be more in the area of economics, but of impor-
tance to location theory, include Dixit and Stiglitz (1977), Myrdal (1957), Ohlin (1933), Ricardo
(1821), and Scherer (1975).
62 Here, I include Alao (1974), Batty (1978), Boots (1980), Bunge (1966), Berry, Parr, Epstein,
Ghosh, and Smith (1988), Curry (1964, 1972, 1976a, 1976b, 1978, 1984a, 1984b, 1985a, 1985b,
1985c, 1986, 1989), Daly and Webber (1973), Chisholm (1970, 1979a), Dicken and Lloyd (1990),
14 1 The Craft of the Story Teller
Epping (1982), Erickson (1989), Fik (1988), Fik and Mulligan (1991), Fotheringham (1979),
Garrison (1959a, 1959b), Garrison and Marble (1957), Ghosh and McLafferty (1987), Golledge
(1967, 1970, 1996), Golledge and Amedeo (1968), Griffith (1986), Hartshorne (1927), Huff and
Jenks (1968), Jones (1984a, 1984b, 1988), Jones and Krummel (1987), Jones and O’Neill (1993,
1994), Kellerman (1989a, 1989b), Lentnek, Harwitz, and Narula (1988), Lentnek, MacPherson,
and Phillips (1992), Lo (1990, 1991a, 1991b, 1992), McCann (1993, 1995), McCann and Shefer
(2004), McCann and Sheppard (2003), Miller and Finco (1995) Miron (1975, 1976, 1978a,
1978b, 1982, 2002), Miron and Lo (1997), Miron and Skarke (1981), Mu (2004), Mulligan
(1981), Mulligan and Reeves (1983), Nijkamp and Paelinck (1973), O’Kelly and Miller (1989),
Papageorgiou (1973, 1976, 1978, 1979, 1980, 1990), Papageorgiou and Pines (1999), Parr (1993,
1995a, 1995b, 1997a, 1997b), Peet (1969), Penfold (2002), Pitts and Boardman (1998), Pred
(1964, 1969a, 1969b), Rushton (1971a), Scott (1988, 1998), Sheppard and Curry (1982), Solomon
and Pyrdol (1986), Stimson (1981), Teitz (1968a, 1968b), Thrall (1987), Wagner (1974), Webber
(1984), Wendell and McKelvey (1981), Wilson (1967), and Zhang (2007).
63 Johann Heinrich von Thünen (born 1783), a German economist, published the original version
of his Der Isolierte Staat in 1826. An English translation of this book, Isolated State: An English
Edition of Der Isolierte Staat, was first published in 1966. See Blaug (1992).
64 Antoine Augustin Cournot (born 1801), a French economist and mathematician, published the
original version of his Recherches sur les principes mathématiques de la théorie des richesses in
1838.
65 See Blaug (1992).
1.4 The Development of Economics Since Marshall 15
better able to compete in foreign markets.66 The second stream draws a distinc-
tion between price (the net amount received by the supplier selling the commodity)
and effective price (the amount paid by the purchaser). An effective price includes
price plus any transaction costs paid by the purchaser related to search and informa-
tion gathering, negotiation, and acquisition (including freight and transfer, storage
and inventory, agency and brokerage fees, credit, cost of insurance67 and other loss
risks,68 installation and removal, warranty and service, and taxes and tariffs), inclu-
sive of normal profit (the profit attributable to an unpriced factor of production
such as entrepreneurial skill or owner equity), with respect to the commodity.69
The supplier firm is also seen to modify effective price through its choice of pricing
strategies (e.g., f.o.b. price,70 basing point pricing, uniform pricing, and discrim-
inatory pricing).71 The net amount received by the supplier, as described above,
is net of any such cost alternatively borne by the supplier. For ease of exposition,
I refer to the transaction costs collectively as unit shipping cost throughout this
book.72 Unit shipping cost can make a firm less able to compete when it is badly
located. Put differently, to compete from a distance, there must be some advantage
to the firm at least sufficient to offset unit shipping cost. Ideas about the distinction
between price and effective price are applicable to commodities or services sold by
the firm as an output, or purchased by the firm as an input. Specifically, they apply
to the labor market where the firm obtains the workers needed to produce its output;
here, we can substitute wage for price and commuting for shipping. The kinds of
66 Davis (1998), Davis and Weinstein (1999, 2003), Head and Ries (2001), and Hanson and Xiang
(2004) discuss and present evidence on the home market effect.
67 For an early analysis of insurance in shipping, see Bernoulli (1954, pp. 29–30) originally
published in 1738.
68 For simplicity of exposition, I imagine here that there are competitive markets in which vendors
and purchasers purchase protection against some risks (i.e., insurable risks), but other risks (i.e.,
uninsurable risks) must be borne by the vendor or purchaser as appropriate.
69 Cournot in 1838 argued similarly for a broad view of transaction cost in this regard. See
Cournot (1960, p. 117). Launhardt—in mentioning costs for packing, freight, storage, taxes and
tariffs, interest losses, insurances, provisions for agents, advertising, and profits of the merchants—
employs a notion similar to a transaction cost. McCann and Sheppard (2003) also puts an emphasis
on the role of transaction costs in rethinking location theory. Caves (2007, p. 2) uses a transaction
cost approach in his analysis of multinational enterprise.
70 Abbreviation for “free on board.” Firm sets price at factory, warehouse, or store; customer pays
freight from that place–also known as mill pricing. According to Marshall (1907, p. 325), the label
“f.o.b.” arises from the practice of merchants to quote a price for their commodity on board a vessel
in port, each purchaser then incurring any shipping cost from there.
71 For further discussion of spatial pricing policies, see Anderson and Ginsberg (1999), Capozza
and Attaran (1976), Chisholm (1970, chap. 7), DeCanio (1984), Deutsch (1965), Furlong and
Slotsve (1983), Gilligan (1992), Greenhut and Ohta (1975), Haddock (1982), Hughes and Barbezat
(1996), Levy and Reitzes (1993), Mulligan (1982), Needham (1964), Ohta, Lin, and Naito (2005),
Soper, Norman, Greenhut, and Benson (1991), and Thisse and Vives (1992).
72 Among others, Louveaux, Thisse, and Beguin (1982) discuss the role of transportation costs in
location theory.
16 1 The Craft of the Story Teller
transaction costs involved in the labor market may be different from those in an
output market (e.g., the cost of delay in commuting may be more important than
the cost of delay in shipping), but they are costs nonetheless. The two streams are
connected; what differentiates the foreign market from the home market is gener-
ally a substantial unit shipping cost. This book explores location theory using the
perspectives of these two streams of thought.
What has motivated scholars to think about location theory? Two principal
motivations come to mind.
One group of scholars appears to have used geographic space as a way of eval-
uating the reasonableness of emerging economic theory.73 Thünen, Dupuit, and
Cournot fit into this category and so too do scholars working in the new economic
geography that began with Krugman (1995).74 Thünen, for example, wanted to
understand whether arguments about the efficiency of the then-new English system
of crop rotation were correct. He also wanted to develop a marginal productivity
theory of income distribution to see if it could explain the relatively low wage paid
to agricultural workers in Germany at his time. From the time of Walras, economists
have been interested in the notion of market price equilibrium across markets75 : for
example, if the price of bread were to rise, what would happen to the price of butter?
Location theory adds a spatial dimension; what is the effect of a rise in the price of
bread in city A on the price of bread (or, for that matter, the price of butter) in city B?
Location theory also adds mechanisms for this price interaction: e.g., if butter and
bread producers require production sites, they compete with each other for available
land. Location theory also contributes to economic thought about the nature and
organization of the firm. From the perspective of Coase (1937), production happens
inside a firm when nonmarket allocation is more efficient than allocation through a
market (price) mechanism. Usually, this happens when market allocation imposes
a substantial transaction cost. Location theory provides a useful venue for thinking
about the role of transaction costs in this regard.
The second motivation has been to better understand, from a economic
perspective, modern policy issues such as industrialization and postindustrial-
ization,76 transportation investment and finance,77 the growth and decline of
73 In seminal works in Economics, Coase (1937, pp. 402–403) and Solow (1956, p. 65) each exem-
plify their ideas about Economics using a locational example. Spengler (1974, p. 532) suggests
that the development of modern location theory was among the great achievements of Economics
after 1945.
74 See also Baldwin, Forslid, Martin, Ottaviano, and Robert-Nicoud (2003), Fan, Treyz, and Treyz
(2000), Fujita and Thisse (2009), Krugman (1998a, 1998b), Meardon (2002), and Neary (2001).
75 And the process (tâtonnement ) by which market equilibrium comes to exist.
76 Back to at least Clark (1887), there was concern whether industry would be competitive in the
same way as the market for agricultural commodities. In the early days of industrialization, there
was also concern over monopoly pricing by railroads: see Macdonell (1891).
77 An early source here is Cooley (1894).
1.4 The Development of Economics Since Marshall 17
78 To illustrate the range of thinking by economists in this area over the years, see Hart (1890),
Gras (1922), Haig (1926a, 1926b), Lampard (1955), Harris (1973), Quigley (1998), Anas, Arnott,
and Small (1998), and Aranya (2008).
79 An early example here is Giersch (1949).
80 A system of global production wherein firms in industrialized nations do product design, distri-
bution and marketing, and other aspects of good production and delivery but assembly to plants
(branch or contractor) in emerging nations. See Hanson (1996).
81 An early source here is Jones (1905).
82 I take to heart here the critique by Massey (1973) that sees as impossible the idea of an
autonomous location theory. In my view, it is inextricably linked to economic thought. At the
same time, it is useful to scholars and practitioners in a wide variety of disciplines.
83 See Chamberlin (1962).
84 See Leontief (1951).
85 In its initial application—which was to scheduling problems—programming was an appropriate
descriptor. Since then, the same label has been used to refer generally to optimization subject to
constraints.
86 John von Neumann, the mathematician, was important here. He wrote extensively on the subject
although nothing was published at the time. Dantzig (1991, p. 24) acknowledges his contributions.
18 1 The Craft of the Story Teller
Unfortunately, competitive location theory has too often been analyzed and pre-
sented in ways that will seem complex or difficult—even esoteric or impenetrable—
to many readers who might otherwise benefit from it. Why is that? What makes
competitive location theory so problematic? My sense is that part of the problem
is any discipline needs scholarship in terms of both appeal to a broader audience
(breadth) and appeal to specialists in the field (depth), but that competitive location
theory has focused on the latter.91 In part, the purpose of this book is to redress that.
Why has it been difficult to present location theory to a broader audience? I think
the problem here is the complexity that location theory brings to even the elementary
87 For a brief introduction to linear programming, see Spivey (1962). Early applications pertinent
to location theory include Garrison (1959b).
88 The Kuhn-Tucker conditions generalize the classical Lagrange method (optimization subject
to equality constraints) to allow for inequalities in the constraint set. See also John (1948). An
even-earlier unpublished analysis of the same problem is to be found in the MSc thesis (Minima
of functions of several variables with inequalities as side conditions) of W. Karush at University
of Chicago in 1939. In the 1950s, applications were mainly linear programs, and the main con-
straint was computing power. Later, the focus shifted to nonlinear programming and a variety of
computational techniques emerged: e.g., recursive programming, separable programming, gradient
methods, dynamic programming, integer programming, and quadratic programming. Bodington
and Baker (1990) review the history of application in management science. Day (1961) initiated
recursive programming.
89 The early history of mathematical programming is described in Hadley (1962, pp. 17–21, 1964,
pp. 14–16), Dantzig (1963, chap. 2), Lenstra, Rinnooy Kan, and Schrijver (1991), and Murphy and
Panchanadam (1997). Important work in game theory includes von Neumann and Morgenstern
(1947), Marschak (1950), Hurwicz (1953), Simon (1955, 1959), Koo (1959), Bishop (1963),
Harsanyi (1965, 1966), Loomes and Sugden (1982), and Sugden (1991).
90 On the relationship between location theory and mathematical programming, see Takayama and
Judge (1971), Lahr and Miller (2001, pp. xxv–xxviii), and Fernandez Lopez (2002).
91 In my view, the books on competitive location theory with broad appeal include Beckmann
(1968), Greenhut (1956, 1970), Isard (1956), Lefeber (1958), Richardson (1969), and Siebert
(1969). Anthologies in the area include Thisse, Button, and Nijkamp (1996).
1.5 Why Is Competitive Location Theory Problematic? 19
concept of a market.92 Marshall (1907, pp. 112, 324–325) illustrates the conundrum
here. Early on in his book, Marshall envisaged a market as a district—containing
many traders (buyers and sellers) all keenly on the alert and well acquainted with
one another’s affairs—wherein as a result the price of a commodity is effectively
everywhere the same. He then undermines this definition with three caveats: (1)
in practice, it is difficult for any trader to know exactly what price was paid in a
transaction among others; (2) the geographical limits of a market are seldom clearly
drawn; and (3) if the market is large, allowance must be made for the fact that each
purchaser might pay a different price on account of the cost of delivery. Later in
his book, Marshall muddies the water further in asserting that the central point of a
market is a mechanism (e.g., a public auction or published price list) whereby traders
are able to be in close communication with each other. What is not clear here is how
traders are to be “keenly on the alert and well acquainted with one another’s affairs”
when their knowledge of the market is limited to a published price list.
How do I propose to conceptualize a market in this book? In a market in equi-
librium where I (for the moment) ignore geography, there is neither an incentive for
either another firm (supplier) to add to the quantity supplied, nor for a new customer
(demander) to add to quantity demanded in that market. We can imagine a supply
schedule that shows the amount of the commodity offered at any particular price
and a demand schedule that shows the amount of the commodity demanded at any
particular price. Market equilibrium is then associated with the price where demand
equals supply. I take that to be the price at which the market clears.93 This is the stuff
of a first undergraduate course in Economics. Now, let me introduce geography into
this. To this point, I have left unstated the question of how a market is defined. As
a rule, economists are loath to define a concept like a market: and understandably
so. However, geographic distance is often thought to separate one market from the
next: whether it is because of shipping, commuting, search or communication costs,
or externality effects.
The notion of a market itself might be endogenous94 to the notion of spatial equi-
librium. This is an added complexity addressed in this book. At least three additional
complications arise. First, to the extent a market has a geographic limit, a spatial
equilibrium must incorporate the notion that there is no incentive for a potential
supplier or demander either to move into this market or to leave it. Second, to the
92 A market can be described as succinctly as a locus where buyers and sellers intersect. The
purchase or sale of a good typically involves search activity on the part of both buyers and sellers—
activity that is costly and time-consuming. The classical notion of a market is a combination of
institutions and mechanisms operating at a site at which (or portal through which) offerings of
a commodity are on view and sales are recorded. In this way, markets are seen to facilitate the
efficient exchange of the commodity.
93 For a further discussion of market-clearing versus equilibrium price, see Fehr, Kirchsteiger, and
Riedl (1993) and De Vroey (2007). The modeling of market clearing has been of interest in areas
ranging from environmental pricing to automated exchanges: see Flaam and Godal (2008).
94 In a model, an endogenous value is an outcome; a value predicted by the model based on other
(endogenous) values. Endogenous variables may have a stochastic component.
20 1 The Craft of the Story Teller
extent firms can vary both what they produce and how they produce it, spatial equi-
librium means there must not be any further incentive to change what is produced
or how. Third, the existence of transportation costs inherent in a spatial economy
makes necessary the assumption of something akin to increasing returns to scale;
after all, in the presence of transportation costs, would not the geographic size of
a market otherwise be very small?95 However, this need not (indeed, in this book,
typically does not) take the form of assuming increasing returns to scale.96
95 See Eaton and Lipsey (1977) and Fujita and Thisse (2009).
96 Instead, many models in this book use constant returns to scale. I follow the approach of Nerlove
and Sadka (1991, p. 100) here.
97 See, for example, McCann (1999). See the counter-argument in Alao (1974, p. 59) who decries
the efforts spent on criticizing assumptions as opposed to amplifying, refining, and evaluating the
conclusions or theorems so that we might better extend and improve our understanding.
98 Geography students in particular might benefit from the parallel argument in Ekelund and Hébert
(1999, p. 6).
1.6 Location Theory and Geography 21
historical and geographical context of its production. This argument itself sounds—
to people like me—strangely like a universal law. Nonetheless, it leads Barnes and
others to argue in favor of a kind of nuanced analysis (including triangulation99 )
wherein social, political, economic, and cultural aspects are interwoven. Advocates
point out that such an approach diminishes the problem of suspension of disbelief
inherent in the reductionist approach. From a reductionist’s perspective however, it
can be difficult to know whether any one hypothesis, or part of the overall story,
is more correct than some alternative. The fear here is we wind up with no way of
testing among competing hypotheses. Further, reductionists would argue that any
attempt to write about (i.e., to encapsulate) the world around us, regardless of the
degree of richness or nuance, of necessity requires a suspension of disbelief.
A third objection is that, to some geographers, theirs is a discipline of differ-
ence.100 They see different things happening locally around the globe and seek local
explanations for this diversity. In contrast, most economists see theirs as a discipline
of generalization. To them (and to me), competition leads other firms to adopt (copy)
the behavior of those who are successful, and this copying leads firms inexorably
to behave similarly, not differently.101 Further, the market tends to drive out firms
that do not conform to best practice. So important are these concepts of adaptation
and adoption that models in Economics typically start by assuming that economic
actors in a group are identical. To some outside Economics, the relentless assump-
tion of otherwise-similar economic agents may make the discipline seem colorless,
irrelevant, or evidence its status as the dismal science. To economists (and to me),
the strength of this perspective is the unrelenting emphasis it puts on the idea of
how economic actors behave under competition and the clarity this brings to the
conclusions that follow.
A fourth objection of some geographers is the undue emphasis put by location
theory on the role of unit shipping costs. In the 1960s and 1970s, for example, it
was commonly argued the marginal cost of transportation shrank because of heavy
investment (both private and public) in transportation infrastructure.102 At the time,
critics argued that the marginal cost of transportation would steadily drop in the
future and that locational models based on transportation costs were therefore of
little use. However, this argument misses several key points. First, unit shipping
cost—envisaged in this book as the cost of arbitraging between two geographic
markets—is broader in scope than just the cost of physically moving a commodity
from one place to another. Second, while the marginal cost to users may well have
declined, the overall cost of construction, operation, and repair of transportation
99 See, for example, England (1993), Tarrow (1995), and Downward and Mearman (2007).
100 See, for example, Sauer (1924, p. 17).
101 See Alchian (1950).
102 In an early study in this area, Chisholm (1961) concludes that reductions in shipping cost
began in the nineteenth century and that refrigerated freight also became of great importance in the
twentieth century. A similar argument is made in Peet (1969, see Fig. 3 on p. 293).
22 1 The Craft of the Story Teller
facilities has remained substantial.103 Further, the social and environmental costs of
transportation facilities are large. That transportation services may have been one
of the great mispriced commodities of the twentieth century104 does not take away
from the costliness of trips and shipments to society. In recent decades, as munic-
ipalities, regions, and nations consider how to implement road pricing and other
rationing schemes, the substantial marginal costs of transportation have become
more apparent. Third, we need to keep in mind how public discourse is shaped
by private interests. In Chapter 2, I present a model (2E) in which a monopolist,
facing the possibility of entry by a competitor into a remote market, engages in uni-
form pricing across local markets despite incurring shipping costs. In Chapter 12, I
present another model (12C) in which—given an inelastic demand for its product—a
monopolist firm finds uniform pricing to be most profitable within a given geo-
graphic market area despite incurring shipping costs. In such situations—to avoid
the connotations associated with entry deterrence or discriminatory pricing and give
the appearance of being a good corporate citizen—the firm has an incentive to say
that its uniform pricing is simply the result of a negligible marginal shipping cost.
1.7 My Approach
As a field of study, competitive location theory is littered with work that is incorrect.
I am referring here to errors of deduction, not the relativist’s problem where a new
explanation renders previous explanations obsolete and hence incorrect. Although
this is perhaps true of any field of study, I suspect competitive location theory may
be a special case. In part, I think the problem is that even the simplest model in com-
petitive location theory must, of necessity, be more complicated than other related
models in Economics for the reasons listed above. In part, the problem is that, in
their search for generalizations, scholars have tried to use complex mathematical
forms that sometimes have unexpected consequences. In part, the problem also is
that scholars have not always been able to cross-check their understanding of a
model against numerical simulations. As with all intrepid authors in the area, I hope
to have avoided such errors but, of course, readers may yet prove me wrong as well.
103 Supporting evidence can be found in Bukenya and Labys (2005) who examine convergence in
world commodity prices over much of the twentieth century for coffee, cotton, wheat, lead, cop-
per, and tin. In addition to reductions in unit shipping cost, the authors argue that convergence
has been aided by (1) improvements in communication and information technology, (2) centraliza-
tion of commodity markets (e.g., the London Metal Exchange), (3) the rise of coordinating central
bank activities in developing countries, (4) the globalization of enterprise, and (5) trade liberaliza-
tion policies at the national level. Overall, their empirical results do not support the convergence
hypothesis but rather a pattern of fluctuating divergences. They suggest the following might explain
the lack of convergence: (1) political unrest, wars, and climate change, (2) international business
cycle conditions, oil price shocks over the years, and (4) the persistence of substantial unit shipping
cost.
104 See the opening comment in Vickrey (1963, p. 452).
1.7 My Approach 23
The purpose of this book is to present and interpret applications in the form of
models that illustrate the linkages between localization and prices. I present mod-
els of general locational problems that are starting points for analysis. Much work
remains to be done on this topic. I leave it to readers—indeed encourage them—to
think further about how to extend or generalize from the models presented here. Let
me now explain how and why my approach in this book is different from past work
in location theory.
First, I rely on relatively simple notions related to the operation and clearing of
markets. A market is a locus of buyers and sellers that facilitates efficient exchange
of the commodity. Inherent in that conceptualization is a process (like auctioning)
wherein a price gets established such that demand equals supply (i.e., a price that
clears the market). I treat that process as instantaneous and ignore issues related to
the timing of consumption or production. The market process results in a single price
in common to all suppliers and all demanders. Of course, if the market were not to
clear, there would be potential vendors with product they would like to have sold
at that price (i.e., inventory) and/or potential consumers with unmet demand at that
price. Of course, it is possible to have markets that do not clear, or are slow to adjust,
but this would add unduly to the complexity of our models here. Unless otherwise
stated, I ignore uncertainty.105 I also ignore here the impacts of legislation that may
affect the operation of markets: e.g., anti-trust and environmental regulation.106
Second, I rely mainly on simple versions of the models developed by others.
On the spectrum of approaches from complex to simpler models, I think that this
book is most valuable for its intended audience if it focuses on simpler models.107 I
use simple demand functions where quantity demanded depends only on own price
(and possibly income); to keep the models simple, I generally ignore direct cross
price effects. As well, where others might use a generalized utility function108 or
one that presumes only homothetic preferences, I use a log-linear utility function.109
105 On the treatment of uncertainty in location theory, see Alperovich and Katz (1983), Asami and
Isard (1989), Cromley (1982), Dean and Carroll (1977), Hsu and Tan (1999), Mai (1984, 1987),
Mai, Yeh, and Suwanakul (1993), and Mathur (1983, 1985).
106 For an example of the incorporation of antitrust regulation into location theory, see Bobst and
Waananen (1968). Ulph and Valentini (1997) model the role of environmental policy.
107 See the parallel treatment of complexity based versus evidence-based models in Coelho and
McClure (2005, 2008).
108 Following neoclassical theory, I use a utility function here simply to mean an ordered (i.e.,
ordinal) scoring of choices (bundles of commodities whose consumption is desirable)—that reflect
consumer preferences—and that are transitive and evidence diminishing marginal utility: see
Becker (1962). As a ranking, a given utility function
is said to be unique up to a monotonic
transfor-
mation. For example, the utility functions f x, y = xb y1−b , where 1 < b < 0 and g x, y = axb yc ,
where b > 0, c > 0, and b + c < 1, calculated at consumption of x units of good 1 and y units of
good 2, generate the same rank ordering: i.e., g [x, y] is a monotonic transformation of f [x, y].
The two utility functions above exhibit diminishing marginal utility. I find Gorman (1976) useful
in thinking about the use of utility functions.
109 A log-linear utility function in two commodities takes the form ln [U] = a ln q
1 +
(1 − a) ln q2 or, equivalently U = qa1 q1−a 2 , where q1 is the quantity of good 1 and q2 is the
24 1 The Craft of the Story Teller
Something is lost here in terms of the generality of results. Economists are driven by
a desire to generalize their conclusions. They do not want to be limited to results that
are peculiar to the particular mathematical model employed. However, at the same
time, something also is gained by use of a specific mathematical form: tractability
and explicit solutions. With specific and simple mathematical forms, I can under-
take numerical simulations using standard spreadsheet software. While this does
not eliminate every error, it enables me to catch mistakes that might otherwise go
unnoticed. I also avoid extensive use of game theory, largely because these typically
rely on a specific conjectural variation110 and a rigidly structured decision-making
process. I cannot entirely avoid game theory since there is much in location theory
that is interesting to analyze in this way. However, for most of the readers I would
hope to reach with this book, game-theoretic approaches are too abstract, mechan-
ical, contrived or unwieldy. Similarly, I avoid the use of fixed-point (homotropy
based) analysis for similar reasons. Fixed-point analysis is useful in game theory;
it is also useful in studying Walrasian multimarket equilibrium. However, in this
book, I do present models for thinking about multimarket equilibrium that do not
necessitate fixed-point analysis.
Third, my focus is on interpretation rather than on proof. I leave it to others, for
example, to consider second-order conditions for optimization and the uniqueness of
solutions (typically, equilibria) to the models. These are important considerations,
but my sense is that they can quickly turn into insurmountable obstacles for the
intended audience. My focus instead is on understanding what the models can and
cannot tell us. I apologize here to more-advanced readers who correctly understand
the importance of such considerations and are surprised by their absence here.
quantity of good 2. This utility function has the desirable properties that marginal utility is positive
for each commodity and that there is diminishing marginal utility. At the same time, a log-linear
utility function has the special properties that each good is indispensable, that the proportion of the
consumer budget spent on each good is fixed (a for good 1, 1 – a for good 2), and that the cross
price elasticity of demand is zero (in other words, the demand for one good is independent of the
price of the other good).
110 A conjectural variation is a player’s perception (assumption) about the reaction of another
player to the first player’s choice of action. See Greenhut and Norman (1992).
1.8 What This Book Is About 25
not necessarily a price taker with respect to customers nearby even if other com-
petitors elsewhere mean that it is a price taker for customers further away. The
diverse links between geographical space and a monopoly that is at least local in
extent make it difficult to know when or how to apply basic economic ideas about
the consequences of competition. How, and to what extent, do the models in this
book allow us to do this?
• Geography. Location models represent geography in a variety of ways: e.g., as (1)
a bounded or unbounded line or ribbon, (2) a ring, (3) a bounded or unbounded
rectangular plane, or (4) a transportation network. In parts of the book, assuming
the world is stretched out along a line or ribbon allows me to represent geogra-
phy as one-dimensional (rather than the two-dimensional representation inherent
in the rectangular plane). Generally, this readily yields a solution that is easily
grasped. While it is true there may be particular geographies in the real world that
can be thought to be one-dimensional, my main purpose in presenting such mod-
els is to then contrast them with the same problem in two-dimensional space. In
the two-dimensional case, the solution is generally not so simple. Where I assume
geography takes the form of a plane or line that is boundless, my purpose is to put
aside complications that arise because of boundedness. Given an emphasis on the
efficient firm, I therefore focus on the shortest distance between two places (two
points on a map). In a spherical world, the calculation of shortest distance (great
circle route) is a challenging mathematical exercise. For better or for worse, loca-
tion theory has instead largely assumed that the world can be represented as a
rectangular plane (i.e., as flat) and that Euclidean distance measures the shortest
path. These models typically assume competitors, customers, and suppliers are
uniformly spread out across this geography and, economic activities are seen to
be either points on a map (punctiform) or to completely occupy space in simple
ways (tessellations). How important is the specific geography chosen to the out-
comes or conclusions of the model? What might have been the consequences of
using other geographical representations? Why represent geography in the ways
we do?
111 It
therefore excludes a large literature on aggregate regional economic growth that begins with
Borts and Stein (1964), Grether (1948), Nourse (1968), Richardson (1969), Siebert (1969), and
Vining (1946).
1.9 What This Book Is Not About 27
outcomes.112 Locational analysis overlaps with this book but only to the extent that
locational analysis makes use of competitive location theory.113 Behavioral geog-
raphy is another area of study concerned with the ways in which human behavior
might be studied in the context of a given physical and social setting.114 Competitive
location theory provides one set of tools that might be used in locational behavior.
Behavioral geography might be thought to include locational behavior
The book presents, extends, and reinterprets 11 sets of models that have been
developed in the literature: 44 models in all. I make no apologies for my selection
of the 11; my purpose here is to provide an introduction and indicate the scope
and potential of competitive location theory, not necessarily either to identify the 11
most important sets of models or exhaust the range of models.115 I have designed the
book so that these models are all largely static.116 I constructed this book using the
firm—consisting of one or more establishments—that produces a commodity or ser-
vice for profit as starting point. I did this for simplicity of presentation. I could have
used other starting points: e.g., the household, the consumer, or the local, regional,
or national government. I find the perspective afforded by the firm is especially
useful because it allows us to address a wide range of issues important to location
theory.
I do not look explicitly at issues that arise when firms are multinational enter-
prises.117 This includes issues related to corporate organization, foreign direct
investment, and international production. This book does consider a firm with more
than one factory but does not take into account the range of considerations—from
currency hedging to trade restrictions—thought to be central to a multinational
enterprise.118
112 Examples include Curry (1998), Ghosh and Rushton (1987), Haggett, Cliff, and Frey (1977),
Raina (1989), and Wolpert (1964, 1970). To illustrate the breadth of application of locational anal-
ysis across sociology, management science, ecology, and Geography, see Alba and Logan (1993),
Francis, McGinnis, and White (1983), Minta (1992), and Scott and Angel (1987), respectively.
113 See, for example, Harris and Hopkins (1972), Pred (1969a, 1969b), and Reggiani (1998).
114 Examples in this area include Aitken and Bjorklund (1988), Clark (1992), Couclelis and
Golledge (1983), Gale (1972), Gold (1980), Golledge (1970), Golledge and Stimson (1987), Gould
(1963), Hart (1980), Harvey (1966), Krumme (1969), Mark, Freksa, Hirtle, Lloyd, and Tversky
(1999), Pipkin (1977), Rushton (1971b), Scott (2000), Walmsley and Lewis (1984, 1993), and
Wolpert (1964). See also the collection of papers in Cox and Gollledge (1969, 1981) and Golledge
and Timmermans (1988).
115 Interestingly, 2 of the 12 sets of models included might not, I think, be regarded by the origi-
nating authors as centrally in the area of competitive location theory. Rather, they represent work
with fascinating implications for the central theme of this book.
116 For readers interested in dynamic models, Duranton and Puga (2001) present a model of local-
ization economies that emphasizes how product life cycle and process innovation interact with
localization economies.
117 See Buckley and Casson (1991), Caves (2007), Dunning (1981), Dunning and Lundan (2008),
and Markusen (1995).
118 Leading works in the field include Buckley and Casson (1991), Caves (2007), Dunning (1981),
and Markusen (1995, 2002). See also McCann and Mudambi (1984).
28 1 The Craft of the Story Teller
Undergraduate students often say that they become interested initially in compet-
itive location theory because they want to start a business and would like to know
where best to locate. This question—interesting on its own—is not centrally the
subject of this book.119 In this book, I present tools that might be used to look at
such a question, but the principal focus is on the question of how competition drives
firms into spatial equilibrium. Put differently, my interest is not to explain where
someone should locate their business; it is to say something about the locations of
firms once market equilibrium is established.
In this book, unit shipping costs are treated as exogenous.120 However, they could
also be viewed as endogenous. For both producers and consumers, an investment in
transportation facilities that brings about a reduction in the marginal cost of trans-
portation may be seen as mutually beneficial. This book ignores locational questions
that arise because of congestion on transportation networks. In reality, network con-
gestion can lead a firm to a location that it would not otherwise choose. In effect,
network congestion makes the location of firms interdependent. This can be mod-
eled, but the level of complexity is substantial. I therefore ignore it here and assume
unit transportation cost is constant (a given) to generate models that are readily
solved. I also ignore here the fact that commodities are typically shipped in a vessel
(or vehicle). If shipments occur only from Place A to Place B, the shipper is then
saddled with the cost of returning the empty vessel to Point A for its next shipment.
In such cases, a firm in the business of providing shipping services might well find
it profitable to cross-subsidize cross-hauling from Place B to Place A.
Finally, this book is focused on equilibrium, not on dynamics. I am not trying
to argue here that processes of change are not important; they are.121 However, in
trying to make this book accessible to as wide an audience as possible, a focus on
equilibrium models only is helpful.
(Greenhut–Manne Problem)
A monopolist sells a product at home and remote places. There is a cost to ship
the product to the remote place or, alternatively, to build a second factory there. In
Model 2A, a monopolist with one factory sells its product only at the home place.
In Model 2B (localization), the firm also supplies the remote place from the same
factory. In Model 2C (no localization), the firm builds a second factory at the remote
place and serves customers at each place from their own factory. Model 2D consid-
ers a monopolist choosing whether to build one factory or two. Where the number
of customers is sufficiently large—which in turn may require that the two places be
sufficiently close together—there will be at least one factory and it will be located
at the place with the larger demand. If the unit shipping cost is sufficiently low,
there will be localization (i.e., one factory serves both places), and soap will be
priced differently at the two places (partial freight absorption). Model 2E shows
how one might think about entry deterrence at the remote place. If unit shipping
cost is sufficiently high, and the smaller place has enough customers, the firm builds
a second factory there and prices will be the same at the two places. In this chapter,
localization and prices (one for each place) are joint outcomes of profit-maximizing
behavior.
I begin with the notion of a firm which I take to be a person or group of per-
sons engaged in the production of soap for the purpose of earning a profit. For the
moment, I don’t distinguish between a firm and the narrower concept of an estab-
lishment; that is, a branch of the firm that carries on business at a particular site.
I don’t address here how the firm is organized (e.g., in terms of research,
development, production, distribution, finance, and marketing) or about vertical
integration and outsourcing because what the firm does and how it does it may
1 The amount by which a firm’s revenue for a period exceeds its costs inclusive of a normal return
on any unpriced factors such as owner equity or management skill. Also known as excess profit.
2 A feature of a model wherein geography plays no role. Typically, shipping cost and/or commuting
costs are assumed zero or are otherwise ignored.
3 A market sufficiently small in area that we can ignore shipping costs on shipments of the good
within that market.
4 For this good, local producers and demanders transact in this and only this market. No one else
transacts product there: e.g., no one purchases there for the purpose of reselling elsewhere or resells
there a good purchased elsewhere.
2.2 Model 2A: Non-spatial Monopolist 31
depend on where it locates its chosen activities. Put another way, in the perspective
of Coase (1937), any firm can be thought to use a combination of market prices for
outputs and inputs together with some hierarchical (command) allocation and the
balancing of these, in the context of location, is at the heart of this book. However,
I ignore such balancing at this early stage of the book.
Assume the firm operates in a (paper) fiat money economy5 and plans to build a
factory at Place 1 for which K dollars of capital are needed.6 Assume this factory
will last forever with regular annual maintenance. Suppose the annual opportunity
cost of capital7 is a yield rate r. In this case, the opportunity cost of the firm’s
investment is the profit that normally could otherwise be earned on this capital:
rK. I use a period of 1 year here, but the analysis would be similar if I used say a
month, a week, or a day.8 Note that rK (the opportunity cost of capital) is a fixed
cost9 ; the firm incurs it regardless of the level of production. I assume here that
the other possible investments by the firm all have a comparable level of risk—an
assumption made to avoid the complexities of comparing investments with different
yields and risks. For simplicity, assume the firm has no other fixed costs. As well,
assume the firm has no choice as to the size of factory. The model does not permit
the firm to build either a smaller factory (presumably at less cost) or a larger factory
(presumably at higher cost). In this sense, I can imagine an indivisibility10 in scale
of factory.
In making its investment decision, I imagine the firm in two distinct scenarios.
In the capital-constrained scenario, the firm has a fixed amount of capital available,
say K, and chooses where or how to invest it. In the unconstrained scenario, the
firm can obtain as much capital as it wants at a given opportunity cost of r annually
per dollar invested. These scenarios are similar in that a profit-maximizing firm
5 Ritter (1995, p. 134–135) describes the emergence of paper (fiat) money economies in the twenti-
eth century and its relationship to seigniorage (the profit earned by a government on the difference
between the nominal value of a coin and the cost of minting it). I assume here that the paper money
supply is maintained by an authority (central bank) to ensure that the currency is a good store of
value (or, equivalently, that the level of inflation is low).
6 Economists might prefer to measure capital physically (e.g., number of machines) rather than in
dollars. Capital—measured in dollars—is a given physical amount of capital times the price per
unit of capital. In turn, the price of a unit of capital can be thought to be determined in a market for
plant and equipment where suppliers of capital commodities interact with demanders including our
firm. A problem with using dollars, as done here, is that a change the price of a capital asset will
have no direct effect on the amount of output that can be produced, say daily, from a given physical
amount of capital. However, such productivity concerns are not relevant in this model since, as I
comment below, the model assumes an unlimited wellspring of production once the factory is built.
7 The return on the best alternative investment opportunity available to the firm at a similar level of
risk.
8 A caveat is in order here. Later in the chapter, I argue that the market clears. Depending on the
nature of the production technology, that might be difficult to ensure if the period were very short,
say one second to the next.
9 A cost incurred by the firm for a period of operation that does not depend on the quantity of output
produced.
10 An attribute of a production process such that production cannot be replicated at a smaller scale
with the same efficiency.
32 2 The Firm at Home and Abroad
would seek to use capital efficiently. However, looking at the two scenarios gives us
complementary perspectives on the behavior of the firm.
Assume the marginal cost is a constant C dollars per unit of soap produced. There
is no congestion here. The marginal cost curve11 —the schedule of unit cost arrayed
by quantity produced—is a horizontal line: see DF in Fig. 2.1. This includes the cost
of manufacturing (e.g., commodity inputs, labor, and energy), the regular mainte-
nance required to offset the wear and tear of production, advertising and promotion,
and normal profit.12 The notion of a constant marginal cost is easiest to envision if I
assume, over the relevant range of output considered, the firm has a Leontief technol-
ogy13 (and therefore a linear expansion path),14 purchases its inputs in competitive
markets (so the purchase price of each unit of an input is the same), regardless
of the number of units purchased, and experiences no constraints on capacity,15
no congestion in its factory (that causes unit cost to rise as the scale of output is
increased), and no economies of scale. Put differently, having invested K to construct
the factory, and thereby incurring a fixed annual opportunity cost rK, the firm finds
itself with an unlimited wellspring of production16 at a constant marginal cost of
production17 (C).
Assume here the firm is a price taker18 in markets for its inputs. As far as the
firm is concerned, r and C are each simply a fixed expenditure required (per unit of
K and Q, respectively) that reduce profit. At the same time, r and C reflect the prices
of inputs. In purchasing them, the firm competes in markets for inputs along with
11 A schedule showing the marginal (additional) cost incurred by the firm (usually over the short
run wherein capital invested is held constant) as a function of the quantity to be supplied.
12 The profit attributable to an unpriced factor of production such as entrepreneurial skill or owner
equity.
13 A production technology characterized by the fact that each unit of output produced requires
exactly the same amount of an input regardless of the relative prices of inputs. In other words, there
is no substitutability among inputs in production. Put differently, a doubling of output requires
a doubling of each and every input used in production. It is named after Wassily Leontief, an
American Economist and Nobel Laureate (1973), whose main area of research was input–output
analysis.
14 As used here, a condition of a production function wherein, if as firm expenditure (output) is
increased by a fixed proportion holding prices of inputs constant, the efficient firm purchases the
same proportion more of each input.
15 The amount of output that can be produced by a given factory over a stated period of time. In a
simple case, we imagine that the unit variable cost of production is fixed per unit for any quantity
up to capacity (marginal cost of production is a horizontal line up to capacity). To allow for the
concept that no quantity greater than capacity is possible, we typically assume the marginal cost
curve then becomes vertical. In other words, no matter how much the firm might want to further
increase quantity, it cannot produce a quantity in excess of capacity.
16 In effect, rK is akin to a lump-sum licensing fee paid annually by the firm that gives it authority
to produce as much of the commodity as it wants during that period.
17 The increment to the firm’s total cost incurred by the last unit produced.
18 A condition under which a market participant (supplier or demander) is unable to affect the price
they receive or pay for a unit of the product by varying the quantity that they supply or demand. The
supplier (demander) sees the demand (supply) for its product as horizontal: i.e., infinitely elastic at
the given market price.
2.2 Model 2A: Non-spatial Monopolist 33
P N
M H
I
E
D F
0
J C Quantity B
Fig. 2.1 Model 2A: monopolist located and selling at Place 1 only.
Note: α = 15; β = 1; C = 3; K = 50,000; N1 = 200; r = 0.05. To maximize profit, firm sets
P1 = 9 and Q1 = 1200. Horizontal axis scaled from 0 to 4,000; vertical from 0 to 16
others demanding those inputs. To the extent it affects the price of r or the prices of
other inputs making up C, such competition may affect the location of our firm.
Suppose the firm plans to produce quantity Q1 annually with this factory. Since
I assume that markets clear throughout this book, I use Q1 to refer interchange-
ably to the quantity produced by the firm and the quantity demanded by customers.
Therefore, inventory is zero (ignored). The firm’s total cost19 and average cost,20
both inclusive of the opportunity cost of capital, are now given by (2.1.1) and (2.1.2).
See Table 2.1, wherein I summarize the equations, assumptions, notation, and ratio-
nale for localization in Model 2A. Average cost is the sum of marginal cost (assumed
constant above) and average fixed cost (which drops the greater the quantity of soap
over which to spread the total fixed cost, here only rK). This firm has a declining
average cost; the greater the output, the lower the average cost of units produced. See
the average cost curve and marginal cost curve: labeled GHI and DF, respectively,
in Fig. 2.1.
19 For a firm, the sum of variable and fixed costs of production inclusive of any unpriced resources
such as entrepreneurial talent.
20 Fora firm, total cost divided by the amount produced.
34 2 The Firm at Home and Abroad
Table 2.1 Model 2A: monopolist located and selling at local Place 1 only
Total cost
rK + CQ1 (2.1.1)
Average cost
rK/Q1 + C (2.1.2)
Notes: Rationale for localization (see Appendix A): Z1—Presence of a fixed cost; Givens (para-
meter or exogenous): a—Intercept of individual linear inverse demand curve: maximum price;
b—Negative of slope of individual linear inverse demand curve: marginal effect of quantity on
price received; C—Marginal unit production cost; K—Capital required to build factory; N1 —
Number of consumers at Place 1; r—Opportunity cost of capital. Outcomes (endogenous):
N1 ∗ —Minimum number of consumers required; P1 —Price of unit of soap at Place 1; q—Per
capita consumption of soap; Q1 —Quantity of soap supplied to Place 1; ε11 —Price elasticity at
Place 1 at market equilibrium.
2.2 Model 2A: Non-spatial Monopolist 35
21 An efficient firm (1) incurs the least possible cost in achieving its desired output and (2) seeks
the maximum revenue possible from that output. It adopts an organizational structure than enables
it to be efficient. It has a production function which shows, for each combination of inputs, the
maximum possible output that can be produced with those inputs. The firm also has a cost function
which shows, for each level of output (Q), the minimum possible cost of achieving that Q. Finally,
the firm knows the demand for its product and exploits that information along with the knowledge
of its cost and production functions to maximize its own profit.
22 See (2.1.3). A demand function is generally expressed as a schedule of quantity demanded (Q)
at various prices (P): i.e., Q = f [P]. An inverse demand function rearranges this as the price
consumers are willing to pay as a function of the quantity supplied to the market: that is, P =
f −1 [Q].
36 2 The Firm at Home and Abroad
profits to its owners. More generally, we can think of money as just another desirable
commodity used by the firm (1) to barter with suppliers or (2) for whatever other
purpose ownership of that commodity might be used.
What kinds of assumptions are implicit in Equation (2.1.3)?
• There is no multiyear setting here. The customer demands q units each year
unless something (i.e., P, α, or β) changes from 1 year to the next. This is con-
sistent with the idea that the individual starts the year with no inventory of soap
and consumes all q units by the end of the year. Implicitly, soap is therefore
perishable—perhaps because it has a “best before” date—and cannot itself serve
as a store of value.23
Person Demand
1 q = α/β − P/β
2 q = α/β − P/β
... ...
N q = α/β − P/β
All Nq = Nα/β − NP/β
• The demand curve tells us nothing about how frequently the customer shops for
soap during the year or how much is purchased at a time. A commodity may be
consumed on the spot at time of purchase (e.g., an ice cream cone). In other situ-
ations (e.g., a bar of soap), the commodity might be brought home and consumed
slowly over time. In the latter case, the customer maintains a stock (e.g., a partial
bar of soap), but I presume this disappears by the end of the year.
• Equation (2.1.3) is a particularly simple demand curve; it ignores income and the
prices of complements and substitutes. Where does a demand curve come from?
I find it helpful to think that (1) the customer is simultaneously participating in
markets for commodities, (2) these commodities are, to varying degrees, pair-
wise substitutable or complementary, and (3) the demand for any one product is
downward sloping in price in part because of diminishing marginal utility and in
part because of substitution effects.24 When I see a linear inverse demand curve,
I understand it to be an approximation to a process in which the customer is trad-
ing off the price of that commodity compared to the prices of other commodities
23 A commodity would be a store of value if it was nonperishable and could be readily resold in the
future: i.e., either converted back to cash at a low transaction cost or offer the possibility of capital
gains.
24 Economists in general moved away the notion of diminishing marginal utility of commodities
because it was seen to be based on a cardinal measurement of utility. Instead, the ordinal assump-
tion of a diminishing marginal rate of substitution between commodities was invoked. In this book,
I retain the use of diminishing marginal utility because students find it helpful. Where it becomes
problematic (e.g., in Chapter 9), I will discuss the matter further.
2.2 Model 2A: Non-spatial Monopolist 37
I assume here that the firm knows—and is able to exploit—the individual demand
curve.25 The firm is not a price taker in the market for soap (as is assumed in a per-
fectly competitive market); by varying the quantity supplied, the firm can affect
price. After all, that is what makes it a monopolist. The intercept may be inter-
preted as a maximum price26 ; the price paid rises to α as quantity supplied to the
customer annually approaches zero. This corresponds to OA in Fig. 2.1. The cus-
tomer views soap as an expendable commodity27 ; it would sooner go without soap
at all than pay a price in excess of α. In a corresponding sense and with apolo-
gies to economists who may take affront with my casual usage here, I think of the
marginal cost, C, as like a minimum price28 ; a firm presumably could not sustain
selling at a price below marginal cost indefinitely. Rewrite this as a demand curve
(q = α/β − P/β) once for each customer in a list as shown above. Summing
both sides of this set of individual demand equations, I get an aggregate demand
equation: N1 q − N1 (α/β) − N1 P/β. Aggregate demand is Q = N1 q. Because
demand depends on the number of customers at Place 1, this is best envisaged
strictly as a local demand29 , that is, assuming no demand (consumption) by non-
residents. After substitution and rearrangement, I get the aggregate linear inverse
demand curve (2.1.4): see curve AB in Fig. 2.1. Of course, since all customers are
identical, q = Q1 /N1 is simply consumption per customer. Here, the intercept (α)
and slope ( − β/N1 ) are the parameters of this aggregate demand curve. By defini-
tion, the larger the market (N1 ), the less quickly price drops for a given increase in
25 A skeptical reader here might well ask how a firm can know the demand curve of its cus-
tomer when in fact it can, at best, try varying the price it charges to see how much of the good
is demanded. This raises more generally the problem of whether economic agents ever have the
full information about the market that is presumed by economists in these models. For an impor-
tant paper on this topic, see Alchian (1950, pp. 220–221) on the famous adoption-versus-adaptation
argument.
26 In a linear inverse demand curve, maximum price is the Y-intercept: the price at and above which
the quantity demanded is zero.
27 An attribute of consumer demand for a product such that there is a price above which the con-
sumer demands none of it. The opposite of an expendable good is an indispensable good: i.e., a
good which must be consumed in some quantity, however small, even when its price is high.
28 The lowest price at which a profit-maximizing firm would participate in the market. Revenue
just covers the variable cost of production. Minimum price is not sustainable over the longer run
because fixed costs of production are not covered.
29 In a region, this is the demand for a product by consumers for local consumption and firms for
local production. Specifically, local demand does not include demand by arbitrageurs for resale in
another region. Marshall (1907, p. 112) invokes a similar notion when he refers to those who buy
for their own consumption, and not for the purposes of trade. Larch (2007) uses a similar approach
to local demand in international trade theory. In practice, there is no generally accepted standard in
Economics for measuring local demand. In empirical practice, local demand is often taken to mean
simply that the quantity demanded varies from one location to the next: see, for example, Megdal
(1984) and Justman (1994).
38 2 The Firm at Home and Abroad
30 The addition to the firm’s total revenue created by the last unit supplied by the firm to the market.
If the firm’s demand curve is horizontal, it is a price taker (i.e., in a perfectly competitive market)
and therefore its marginal revenue is simply the price. On the other hand, if the firm’s demand
curve is downward sloping, marginal revenue is the price of the last unit sold minus the revenue
2.2 Model 2A: Non-spatial Monopolist 39
much total revenue increases for the last unit produced at different levels of output:
MR = d(PQ)/dQ = P + Q(dP/dQ). See curve AC in Fig. 2.1. Because the aggre-
gate inverse demand curve (2.1.4) is linear in quantity in this model, application of
calculus shows us that marginal revenue is also linear in quantity and has the same
Y-intercept (α) and a slope twice as steep ( − 2β/N1 ) as the demand curve.31
As the firm increases quantity supplied, its semi-net revenue32 continues to
increase as long as price stays above marginal cost. In Fig. 2.1, semi-net revenue
is the rectangle DLKED.33 To maximize excess profit34 (semi-net revenue minus
fixed cost), the firm chooses the quantity where marginal revenue equals marginal
cost.35 As is well known to Economics students, the firm does not maximize profit
by setting price as high as possible (i.e., P = α); instead, the firm takes into account
how profit is also affected by the quantity sold. Note here also that excess profit will
be larger when C is smaller; to the firm, C is a drain on profit. In Fig. 2.1, total rev-
enue of the firm is the rectangle OLKJO, and excess profit is the rectangle MLKHM
at the profit-maximizing quantity Q.
From another perspective, excess profit arises because competitors for some rea-
son cannot or do not produce the same commodity. If enough competitors were to
produce the same commodity, the firm’s own demand curve would become horizon-
tal and excess profit would disappear as new firms enter the market. Why does a
firm have such an advantage? Often, the advantage can be tied to an input. It might
be attributable, for example, to a technology not available to other firms (because
of a patent for instance) or to a managerial skill unique to the owner of that firm.
In such cases, the excess profit can be thought to be a Ricardian Rent36 associated
with that advantageous input.
As a consequence of assuming the aggregate inverse demand curve is linear in
quantity, the profit-maximizing price (OL in Fig. 2.1) is halfway between the min-
imum price (C) and the maximum price (α): OD and OA, respectively, in Fig. 2.1.
Since C, α, and P can be measured in dollars, (2.1.6) has the desirable feature that
lost on all other units sold because the market price has now been reduced because of the marginal
unit of product supplied.
31 See (2.1.5). For ease of exposition throughout this book, I do not discuss the second-order
conditions for profit maximization.
32 For a firm, revenue minus variable cost. In this book, variable cost includes both production and
shipment. The firm’s net revenue (profit) is semi-net revenue minus fixed cost.
33 Throughout this book, I use the convention that variables (algebraic labels), equations, and
expressions are italicized while other figure labels (representing points, lines, and areas) are not.
All figures are labeled by their vertexes, and I indicate a polygon by repeating the first vertex again
at the end of the label.
34 Revenue of the firm in excess of all costs, including normal profit on unpriced inputs like the
firm’s capital and entrepreneurial skill.
35 See (2.1.6) and (2.1.7) and the corresponding excess profit (2.1.8). For those readers whose
microeconomics is rusty, to the left of Q in Fig. 2.1, the firm finds that increasing Q also increases
semi-net revenue. To the right of Q, semi-net revenue falls if we further increase Q.
36 An excess profit that arises because of an asset or market situation unique to a firm that prevents
competitors from entering the market and/or earning the same profit.
40 2 The Firm at Home and Abroad
there is no money illusion.37 Also, price is the same for every customer (since they
are all identical and the marginal cost curve is the same for every one of them); it
is unaffected therefore by size of market. In turn, the profit-maximizing quantity
is positive as long as maximum price (α) is higher than minimum price (C). This
implies soap is expendable when P1 exceeds α in that, if price is or above α, cus-
tomers at Place 1 would eschew soap. This itself is a consequence of the linearity of
demand: i.e., because the demand curve crosses the Y-axis. Note also that although
the number of customers at Place 1 (N1 ) does not affect the profit-maximizing price
(2.1.6), it does affect aggregate quantity demanded (2.1.7) and profit (2.1.8). Finally,
as is well known to students of Economics, the monopolist always chooses a point
(price and quantity) along the aggregate demand curve where the price elasticity of
demand is larger than one (along segment AN in Fig. 2.1): i.e., where demand is
price elastic.38
Here, in thinking about the behavior of the firm, I distinguish between the short
and long term. To survive in the short term, the firm must usually sell its prod-
uct at a price sufficient to cover the variable cost39 of production. In other words,
P1 is greater than or equal to C. However, to be profitable over the long term, the
firm must also recoup its fixed costs (that is, the opportunity cost of capital): i.e.,
(P1 − C)Q1 − rK ≥ 0. There is an implication here for size of market. If the min-
imum required level of excess profit in (2.1.8) is zero, then the minimum number of
customers (N1 ∗ ) needed to achieve this is given by (2.1.9). Put differently, satisfy-
ing (2.1.9) creates a kind of home market effect in the sense that the market is large
enough to enable production at Place 1.
Given a demand curve and a supply curve for a market, we know that as we hold
the supply curve constant and shift the demand curve up or down, the equilibrium
market outcomes (P1 and Q1 ) change in a way that has the effect of tracing out the
supply curve. Now, imagine a thought experiment in which we alter the intercept of
the demand curve (α) in Model 2A. As we do, we know from (2.1.6) that P1 will
be higher if α is made larger. We also know from (2.1.7) that Q1 will also be higher
if α is made larger. In Fig. 2.2, I show—as curve A3 DEFC—the locus of price and
quantity that result from different values for α. In this sense, we can think of curve
A3 DEFC as a quasi supply curve. To be clear, however, curve A3 DEFC is upward
sloped not because the firm’s unit costs are increasing with the scale of output (if
anything, they are decreasing); instead, this supply curve is upward sloped because
the monopolist captures the higher price now profitable because of the increase
in α.
More generally, in this model, there are six givens (α, β, C, K, N1 , and r) and
three outcomes (P1 , Q1 , and N1 ∗ ). How would the outcomes change here were the
37 Ina demand model, “no money illusion” means that price is relative to the units in which other
money variables are measured. Put differently, if money quantities were all to double, quantity
demanded would be unchanged.
38 See (2.1.15).
39 A cost incurred by the firm for a period of operation that varies with the quantity of output
produced.
2.2 Model 2A: Non-spatial Monopolist 41
A15 C
Price
A9 E
A3
0
B3 B9 B15 B21
Quantity
Fig. 2.2 Model 2A: tracing out a local supply curve by varying α.
Note: β = 1; C = 3; K = 50,000; N1 = 200 r = 0.05. Here, α is set to 3, 9, 15, or 21. Curve
A3 DEFC is the local supply curve: i.e., the locus of combinations of P1 and Q1 that maximize
profit as α is varied. Horizontal axis scaled from 0 to 4,500; vertical from 0 to 25
givens different? In Economics, such questions are the stuff of comparative stat-
ics.40 In this model, comparative statics can be derived from inspection of (2.1.6),
(2.1.7), and (2.1.9): see Table 2.2.
Outcome
P1 Q1 N1 ∗
Given [1] [2] [3]
C + − +
K 0 0 +
N1 0 + 0
r 0 0 +
α + + −
β 0 − +
wholesale price received by the factory, the markup used by the retail establishment,
and its inventory holding.41
Third, the firm earns its semi-net revenue by producing at unit cost C a commod-
ity that it then sells at a price P1 . Put differently, it earns a profit to the extent the
prices of its inputs are low relative to the price at which it sells soap. In general,
therefore, the firm—indeed any firm—can be thought to engage in a kind of arbi-
trage: buying in (input) markets where prices are low for resale in a (output) market
where price is high. What the firm does is transform inputs from low-priced markets
into a commodity sold in a higher priced market.
Fourth, for some reason, the ability to produce soap at a marginal cost of C
is restricted to just the firm or a small number of rival firms. This might be, for
instance, because of proprietary knowledge, a particular managerial skill, purchas-
ing power, or some economy of scale or division of labor possible in production.
Assume, instead, K= 0 and every customer at Place 1 could produce soap them-
selves at the same unit cost, C, as the firm. I label this home production. Faced
with paying a price P1 in excess of C to acquire soap why would not customers
simply produce soap themselves? In this case, the competition confronting the firm
is home production. Of course, in the real world, the firm may have any of the
advantages listed above that make it more efficient than home production. However,
this idea gives us an interesting way (though not the only way) to think about α.
Since α is the maximum price, presumably it might correspond to the unit total
cost (i.e., C plus unit capital cost) customers incur in home production. Suppose
the firm sets its price just below α. On the assumption customers have diminish-
ing marginal productivity in the activities in which they allocate their time, capital,
and other household resources, they will give up some home production in favor
of product purchased from the firm. If the firm were to further lower its price, the
customer might give up more home production (a production substitution effect) as
well as consume more of the commodity in total (a price substitution effect and an
income effect). Presumably, such changes are reflected in the shape (slope) of the
individual’s inverse demand curve.
Suppose the distance from Place 1 to Place 2 is x km. Let s be the (constant
by assumption) cost per unit of soap shipped 1 km; hereinafter, I call s the unit
shipping rate. In total, therefore, it costs sx to ship one unit from the factory at
Place 1 to a customer at Place 2. In Chapter 1, I introduced the notion of a unit
shipping cost. There, I defined it to be “transaction costs paid by the purchaser
related to search, negotiation, and acquisition (including freight and transfer, storage
and inventory, agency and brokerage fees, credit, cost of insurance and other loss
risks, installation and removal, warranty and service, and taxes and tariffs) with
respect to the commodity”. As noted in Chapter 1, there are two ways to think about
sx. One is that sx is exogenous: simply a resource cost for overcoming distance: e.g.,
fuel consumption, storage, and or costs related to information gathering. The other
is that sx is endogenous to a regional economy: it includes services provided by
shippers, brokers, service agents, retailer, and others, as well as the prices of these.
Assume here that the firm is a price taker in these input markets. In this model, as
far as the firm is concerned, sx is simply an exogenous aggregate fixed expenditure
required (per unit shipped) that reduces profit.42
Before proceeding, let us consider the components of unit shipping cost.
Specifically, when might these be proportional to distance as assumed here?
42 In assuming sx fixed, I ignore the possibility of congestion on transportation networks that might
cause sx to vary with the level of shipments. In this book, except where otherwise noted, I also
ignore the idea that unit shipping cost might somehow vary directly with price: see Azar (2008)
for a model where consumers perceive unit shipping cost to be relative to price.
2.3 Model 2B: Monopolist Selling at Two Places; Factory at Place 1 Only 45
be deployed and recycling/waste hauled away, such costs may increase with
distance shipped.
Warranty and service: This includes costs incurred by the vendor or purchaser
related to warranty coverage and/or service. To the extent that service agents
must be deployed and product must be shipped for repair or replacement,
such costs may increase with distance.
Taxes and tariffs: This includes sales and import taxes paid by the customer on
purchases of the commodity. These can be related to distance as, when, and
where tax or tariff is ad valorem based on a selling price that includes the
cost of shipping.
The dichotomization of geographic space that I use here usually is called puncti-
form. This is a commonly used abstraction of geographic space wherein economic
activity is portrayed as clustered at distinct geographic points (places). Put differ-
ently, economic activities themselves do not use land or otherwise occupy space.
In this abstraction, shipping between activities located at the same place incurs a
negligible (zero) cost; shipping between activities at two different places incurs a
nonzero cost that is invariant with respect to the number or volume of economic
activities there.
Throughout this chapter, I assume that all costs are borne by the firm; customers
at Places 1 and 2 incur no unit shipping costs. In this section, I assume that the
incremental unit shipping cost to a customer at Place 2 (over and above the cost
incurred in supplying a customer at Place 1) is strictly proportional to distance. I
assume here that the firm absorbs the unit shipping cost; it is free to set a delivered
price43 at each customer point. Put differently, the firm engages in discriminatory
pricing.44 I assume also that there are no impediments to shipment (e.g., quotas)
other than shipping cost. The total cost to the firm is now given by (2.3.1). See
Table 2.3. At the same time, assume that the firm can charge a distinct delivered
price at each place (here, P1 and P2 ) without fearing an arbitrageur might purchase
soap at the lower-price place for resale at the higher-price place. To the extent the
firm sets a higher delivered price at remote Place 2, it can recoup some, all, or
perhaps even more than all, of the cost of shipping incurred.45
Suppose each customer at Place 2 has the same individual linear inverse demand
curve as at Place 1. In other words, α is the same at the two places: so too
is β. Assume here as well that transaction costs are zero so that effective price
equals price. Place 2 differs only in the number of customers there (N2 ). Its
aggregate inverse demand curve is shown in (2.3.2). Here, once again, α is the
maximum price the firm could expect at Place 2; β/N2 is how much the price
at Place 2 falls as the firm increases the amount supplied annually by one unit.
43 Firm sets price for good delivered to customer; consumer does not pay a separate shipping
charge.
44 A pricing scheme used by a monopolist to enhance profit that results in different prices for
different markets or submarkets, it is sometimes called third-degree price discrimination.
45 Depending on local price elasticities, there is even the possibility that the firm could set a price
in the remote market that is lower than the price in the home market.
46 2 The Firm at Home and Abroad
Total cost
rK + CQ1 + (C + sx)Q2 (2.3.1)
Notes: Rationale for localization (see Appendix A): Z1—Presence of a fixed cost; Z8—Limitation
of shipping cost. Givens (parameter or exogenous): C—Marginal unit production cost; K—Capital
required to build factory; Ni —Number of consumers at Place i; r—Opportunity cost of capital;
s—Cost of shipping one unit of product one km; x—Geographic distance from Place 1 to Place 2;
a—Intercept of individual linear inverse demand curve: maximum price; b—Negative of slope of
individual linear inverse demand curve: marginal effect of quantity on price received. Outcomes
(endogenous): N2 ∗ —Minimum number of consumers required at Place 2; Pi —Price of unit of
soap at Place i; Qi —Quantity of soap supplied to Place i; X—Range of soap (kilometers); ε 12 —
Price elasticity at Place 2 at market equilibrium.
2.3 Model 2B: Monopolist Selling at Two Places; Factory at Place 1 Only 47
As at Place 1, the intercept here is assumed positive and the slope (here β/N2 ) is
negative.
Here I implicitly assume that the demands at Places 1 and 2 are separable; the
firm can set quantity and price separately at the two places. For some reason, the
firm does not have to worry whether a difference in price between the two places
might lead customers at the place with the higher price to purchase soap instead
where the price is lower. I return to this matter shortly.
Suppose the firm now chooses Q1 and Q2 to maximize excess profit. My assump-
tion that marginal cost is the same regardless of the quantity supplied to either Place
1 or Place 2 makes the problem separable: I can solve for Q1 without knowing
Q2 and vice versa. In this case, the solutions for Q1 and hence P1 remain as in
Table 2.1. With respect to Q2 , the firm then simply equates marginal revenue (2.3.3)
with marginal cost (C + sx) for customers at Place 2. The firm will find it profitable
to ship to customers at Place 2 as long as the maximum price at Place 2 (α) is larger
than the marginal cost of serving a customer there (C + sx) The firm will not find
it profitable to supply Place 2 if C + sx > α. If profitable, Place 2 will be part
of the market supplied by the firm with a profit-maximizing combination of price
and quantity there46 and a corresponding semi-net revenue (or, equivalently, added
excess profit) earned from customers there (2.3.6). This is in addition to the excess
profit (2.1.8) earned from customers at Place 1. Finally, note that the price elasticity
of demand is larger here47 compared to (2.1.15) because unit shipping cost pushes
the firm higher up the demand curve (i.e., up segment AN in Fig. 2.1) where demand
becomes more elastic.
Comparing profit-maximizing prices at the two places, (2.1.6) and (2.3.4), we
see the two prices differ by half the cost of shipping one unit of product from local
Place 1 to remote Place 2. This principle—termed the half-freight48 rule—arises as a
special case because I assume linear individual demand curves at the two places. If I
had used another demand function (e.g., q = aP−P ), the half-freight rule would not
hold. Nonetheless, the model illustrates an important principle. For a monopolist,
the difference in selling price between two places is not necessarily equal to the
difference in shipping cost involved.
Now, imagine a thought experiment in which Place 2 is pushed further away
from Place 1 and soap thereby becomes more costly to ship. Eventually, given the
linear demand curve at Place 2, there would come a distance X at which the most
profitable price (P2 ) would equal the maximum price (α), and hence annual demand
there would drop to zero.49 In the view of the firm, that distance would consti-
tute what is called the range50 of soap. Of course, not every commodity need have
a range.51 Some commodities might be indispensable52 : that is, customers need
them regardless of price. In that case, the linear demand curve used here would be
inappropriate.
As indicated in (2.3.6), Place 2 may add to the profit of the factory built at Place 1.
Since the firm maximizes profit, it will ignore the second place unless it adds to the
total excess profit that would otherwise be earned from Place 1 alone. From (2.1.8)
and (2.3.6), the ratio of the semi-net revenue generated by a customer at Place 1
to the semi-net revenue of a customer at Place 2 is (α − C)2 /(α − C − sx)2 : put
differently this gives the number of Place 2 customers equivalent to one Place 1
customer. Given sufficient customers at Place 2, a firm would have built the factory
at Place 1 even if Place 1 alone would not support a factory in the sense of (2.1.9). In
(2.3.8), the minimum number of customers at Place 2 needed to do this is calculated.
On the right-hand side of (2.3.8), the left-hand pair of {} braces enclose the amount
by which N1 falls short of the requirement for profitability. The right-hand pair of
{} braces enclose the Place 2 equivalent.
Important here too is the conclusion that the delivered price at Place 2 does not
depend on the number of customers (N2 ) there. If the number of customers is insuf-
ficient, the firm may choose not to build a factory at all. However, if it is profitable to
build a factory, the price charged at each place is insensitive to the relative numbers
of customers at the two places.
Are there circumstances in which there would be arbitrage? Would a trader have
the incentive to purchase soap at Place 1 and resell it at Place 2? Comparison of
(2.1.6) and (2.3.4) tells us the price difference the monopolist would choose is P2 −
P1 = 0.5sx. In this case, on the presumption that the unit cost of shipping for the
arbitrageur is also sx, there is no incentive to purchase at Place 1 for resale at Place 2.
I have already characterized Place 1 as the home market and Place 2 as the remote
market. Potentially, there is a home market effect here. If the population of Place 1
were large enough to enable (2.1.9), the firm would build a plant there. If C + sx is
smaller than α, the firm would also sell product at Place 2 that it had produced in its
plant in the home market. If the population at Place 1 is not large enough to enable
(2.1.9), the firm would nonetheless build a plant there to serve both the home and
remote markets as long as the population at Place 2 is large enough to enable (2.3.8).
49 See (2.3.7).
50 For an expendable good sold at an f.o.b. price, the distance at which shipping cost is sufficiently
high to cause demand to drop to zero.
51 The starting point for this book is the firm producing a good or service. The book therefore
ignores firms whose business is the construction of a network. De Fraja and Manenti (2003) study
the extension of local telephone calling areas—within which long distance charges do not apply—
as a strategic variable chosen to maximize the carrier’s profit.
52 An expendable good is such that there is a price above which the consumer demands none of it.
The opposite of an expendable good is an indispensable good: i.e., a good which must be consumed
in some quantity, however small, even when its price is high.
2.3 Model 2B: Monopolist Selling at Two Places; Factory at Place 1 Only 49
In this model, there are eight givens (α, β, C, K, N1 , N2 , r, and s) and eight out-
comes (N1∗ , N2∗ , P1 , P2 , Q1 , Q2 , X). How would the outcomes change here were the
givens to change? In this model, comparative statics are readily derived from inspec-
tion of (2.1.6), (2.1.7), (2.1.9), (2.3.4), (2.3.5), (2.3.7), and (2.3.8). See Table 2.4.
Outcome
P1 Q1 N1∗ P2 Q2 N2∗ X
Given [1] [2] [3] [4] [5] [6] [7]
C + − + + − + −
K 0 0 + 0 0 + 0
N1 0 + 0 0 0 0 0
N2 0 0 0 0 + 0 0
r 0 0 + 0 0 + 0
s 0 0 0 + − + −
α + + − + + − +
β 0 − + 0 − + 0
Notes: See also Table 2.3; +, Effect on outcome of change in given is positive; –,
Effect on outcome of change in given is negative; 0, Change in given has no effect on
outcome; ?, Effect on outcome of change in given is unknown.
N2 ∗ increases because the firm once again needs more customers there to
make enough net revenue to cover the opportunity cost of capital. Finally,
X becomes smaller as s is increased.
α If α is increased, the individual demand curve shifts upward. The firm now
finds that consumers at both customer points are willing to pay more for any
given quantity than before. Both price and quantity increase in each market.
Because customers individually purchase more than before, both N1 ∗ and
N2 ∗ become smaller. Finally, as α becomes larger, it becomes possible for
Place 2 to be further away from the firm, that is, X becomes larger.
β If β is increased, the individual inverse demand curve sweeps clockwise
about (0, α). At any price below α, customer demand is now lower than it
was before at both customer points. For the profit-maximizing firm, price
in each market is unaffected but quantity decreases. Because each customer
is purchasing less, N1 ∗ and N2 ∗ must become larger. X is unchanged.
In this model, how does the firm retail to a customer at Place 2? One strategy
is to sell to both customers at Places 1 and 2 from the same retail outlet on-site at
Place 1. A second strategy is to have a retailer at Place 2 sell soap there at an agreed-
upon markup. In Model 2A, the firm’s marginal cost of production, C, presumably
includes the cost of retailing per unit sold to a customer at local Place 1. In Model
2B, the unit shipping cost, sx, presumably includes any extra cost of retailing per
unit sold to a customer at remote Place 2.
From the firm’s perspective, Places 1 and 2 can be said to constitute its market;
they are the customers and places where the firm sells soap. To an economist how-
ever, a market is something else. In Chapter 1, I describe a market as a locus of
buyers and sellers that facilitates efficient exchange of soap. Inherent in that con-
ceptualization is a process (like auctioning) wherein a price gets established such
that demand equals supply (i.e., a price that clears the market). The central idea is
that a market process results in a single price in common to all suppliers and all
demanders.53
In practice, it is rare to find markets that operate exactly like this. Nonetheless, we
commonly think of consumer commodities (e.g., for housing, automobiles, clothing,
or bread) in a metropolitan area as each being provided in a market. How do such
notion of markets diverge from the economic model? Two popular critiques come
to mind. Let me now respond to each of these.
One critique is that consumer commodities are rarely auctioned in the way envis-
aged: i.e., in a process that clears a market. In the short run—from minute to
minute for example—there may indeed be a divergence between demand and sup-
ply. However, this does not in itself negate the conceptualization of markets for
consumer commodities. When supply exceeds demand in the short term, inventory
accumulates in the hands of the supplier. When demand exceeds supply, inventory
gets drawn down or demand goes unmet. Each of these possibilities—which can
arise from the vagaries of demand and supply from minute to minute—imposes
costs on the vendor: costs for maintaining inventory and opportunity costs of unmet
demand. Auctioning too has its costs. Firms use strategies—from auctioning to
price discounting to supplier and customer contracting—to maximize their prof-
its. Put differently, firms use these strategies in a way that assists efficient clearing
of markets even in the absence of a formal auction process.
A second critique—using the market for bread as an example—is that the price
paid for a loaf will differ from white bread to whole wheat, from one neighborhood
of the city to the next, and from one type of retail outlet (e.g., convenience store
or supermarket) to the next. Doesn’t this mean that the market for white bread is
different from the market for whole wheat, different from neighborhood to neigh-
borhood, or different from one type of retail outlet to the next? Perhaps. However,
another way to think about these is that they are submarkets54 within a larger mar-
ket for bread. Submarkets are used to refer to the markets for commodities that are
alike, substitutable for one another, and whose prices therefore tend to change sim-
ilarly; when the price of one commodity increases, consumers tend to substitute for
a less-expensive alternative, and thereby normally drive up the price of the substi-
tute. Prices can differ among submarkets. After all, consumers might prefer whole
wheat bread, or it might be more costly to produce. As such, there might be a price
premium for whole wheat bread. However, the idea behind submarkets is that such
premiums are thought to be relatively stable over time, and factors that cause the
price of bread generally to rise or fall need not necessarily affect the price premium
in a particular submarket.
In this chapter, we have only one supplier (the firm) and demanders at Places 1
and 2. Consider first the case where sx = 0. In this case, the economist’s require-
ments for a market are met because there is a common price (since P1 = P2 ).
Alternatively, suppose sx = 0. Production and allocation appears to conform to a
market process, but with one notable difference. Here, the firm sets different prices
at Places 1 and 2. Does this difference in price mean that the two places are neces-
sarily each in a market of their own? Note here though that the difference in price
between the two markets (in other words, the price premium at remote Place 2) is
a constant amount: in this case, half-freight. If C were to rise by 1.00 for example,
then P1 and P2 would each rise by 0.50. In this respect, Places 1 and 2 appear to
be like two submarkets. At the same time, a profit-maximizing solution does not
54 To me, a market for identical, or similar, commodities is said to be formed of submarkets when
prices in the submarkets differ but are linked in some respect. In a strong version of submarkets,
the price in one submarket is a fixed premium on the price in another submarket. In a weak version,
a rise in price in one submarket causes the price in the other submarket to change, but there is no
fixed premium.
52 2 The Firm at Home and Abroad
concur with Marshall’s idea of a common market price wherein remote customers
pay a special charge on account of delivery; here remote customers are paying only
a half freight charge.
Because this is a book about location theory, it is appropriate to begin with a model
where the firm varies its price from one customer point to the next in response
to differences in unit shipping cost. However, a skeptical reader might well point
out that some monopolists in reality have other pricing practices in addition to
discriminatory pricing. Four popular alternatives come to mind.
F.o.b. price: The monopolist sets the same price at the factory gate for all cus-
tomers. Customers at Place 2 pay the same price as customers at Place 1
but then incur a shipping cost; the effective price is higher at Place 2, com-
pared to Place 1, by the amount of the unit shipping cost. Put differently, the
customer absorbs the unit shipping cost.
Uniform pricing: The monopolist makes the price the same for a unit of the
commodity delivered at different customer points even though the marginal
costs of serving customers there may vary. The firm absorbs the unit shipping
cost. Here, the firm must set both a price and a maximum distance beyond
which the customer is too costly to serve.
Pickup or delivery pricing: The monopolist sets two prices—an f.o.b. price and
a delivered price—and allows the customer to choose between them.55 In this
case, the unit shipping cost is absorbed either by the firm or by the customer
depending on the option chosen by the customer. In the case of delivered
price, the firm must set both the delivered price and a maximum distance
beyond which the customer is too costly to serve.
Basing point pricing: The monopolist chooses some customer points (so-called
“basing points”) and sets a delivered price at each of them.56 Customers else-
where then pay the unit shipping cost to shipping home the commodity from
the basing point. The firm in general here absorbs part of the unit shipping
cost.
to the other options.57 Why then might a monopolist choose f.o.b. or some other
pricing scheme and forego the extra profit possible with discriminatory pricing?58
Usually, the problem with discriminatory pricing is that it is difficult to enforce.
Why, for example, wouldn’t customers at Place 2 purchase soap at Place 1 for the
lower price and either consume it there or bring it back home with them? One might
initially want to argue that such travel or shipment can be ignored because the firm,
due to its scale of operation will already have found the least costly way to ship
to Place 2 and that its price is higher at Place 2 by only one-half of that amount.
However, this argument disregards the possibility customers might have to travel
from Place 2 to Place 1 for another purpose and the effective price of purchasing
soap at Place 1 for them is only P1 . Another aspect disregarded here is the rela-
tionship between inventory acquired by the traveling consumer and the unit cost of
shipping. The customer who makes a trip to Place 1 to purchase soap can reduce the
per unit cost of the trip by choosing to stock up on this trip: in effect, the customer
trades off the opportunity cost of holding an inventory of soap at home against the
cost of traveling to Place 1 to purchase it. I do not consider such inventory holding
explicitly in this chapter.59
Total cost
2rK + C(Q1 + Q2 ) (2.5.1)
Notes: Rationale for localization (see Appendix A): Z1—Presence of fixed cost. Givens (para-
meter or exogenous): C—Marginal unit production cost; K—Capital required to build each fac-
tory; Ni —Number of consumers at Place i; r—Opportunity cost of capital; α—Intercept of
individual linear inverse demand curve: maximum price; β—Negative of slope of individual lin-
ear inverse demand curve: marginal effect of quantity on price received. Outcomes (endogenous):
Pi —Price of unit of soap at Place i; Qi —Quantity of soap supplied to Place i; ε 11 —Price elasticity
at Place 1 at market equilibrium; ε22 —Price elasticity at Place 2 at market equilibrium.
(2.3.5), respectively. The contribution to firm profit arising from sales at Place 2 is
shown in (2.5.4).
Comparing (2.5.2) with (2.3.4), the price at Place 2 will be lower than in the one-
factory solution in Table 2.3. The elimination of the unit shipping cost allows the
firm to lower its price at Place 2 and still further increase semi-net revenue since the
quantity sold at Place 2 will be greater.
Further, assume demand at each place is constant from 1 year to the next;
Q1 and Q2 do not change if P1 and P2 remain unchanged. Here, we can think
the firm makes an investment now of K to build the second factory, the return
for which is an improvement in semi-net revenue that arises partly because of
2.7 Model 2D: Choice of Sites and Localization 55
shipping cost savings and partly because the firm can increase its quantity sold at
Place 2. The firm would find it profitable to build the factory at Place 2 where the
gain in semi-net revenue from local production at Place 2 exceeds the opportunity
cost of the capital required for the second factory: i.e., where (2.5.4) is non-negative.
In part, this is driven by the number of customers at Place 2 (i.e., N2 ) since the larger
the market the larger the semi-net revenue. Equation (2.5.5) indicates that the larger
the opportunity cost of capital (rK) or the price sensitivity of demand (β), or the
smaller the shipping rate (sx) the larger Place 2 must be to make the second factory
profitable.
The comparative statics of Model 2C are the same as in Model 2B except for
two implications that arise because shipments are no longer necessary. The first
implication is that the notion of a range (X) is not relevant in Model 2C. The second
is that variations in the unit shipping rate, s, will have no effect on price, quantity,
or minimum size of market. See Table 2.6.
Table 2.6 Model 2C: comparative statics of an increase in exogenous variable.
Outcome
P1 Q1 N1∗ P2 Q2 N2∗
Given [1] [2] [3] [4] [5] [6]
C + − + + − +
K 0 0 + 0 0 +
N1 0 + 0 0 0 0
N2 0 0 0 0 + 0
r 0 0 + 0 0 +
s 0 0 0 0 0 −
α + + − + + −
β 0 − + 0 − +
Notes: See also Table 2.5; +, Effect on outcome of change in given is positive;
–, Effect on outcome of change in given is negative; 0, Change in given has no
effect on outcome; ?, Effect on outcome of change in given is unknown.
In Model 2A and Model 2C, we might assume that the firm has an outlet counter
at the front of its factory from which it sells product and that the marginal cost
of production, C, includes the unit cost of retailing soap and servicing a customer
there. In Model 2B, however, the firm needs to have a way of retailing to customer at
Place 2. In this respect, Model 2C gives the firm a less-costly option for selling soap
that is reflected in the absence of s in Model 2C.
Finally, where, if at all, should the firm put a factory or factories? For example,
rather than assuming that the first factory is at Place 1 and looking at whether to put
a second factory at Place 2, how does our problem change if the firm has the option
also of building one factory only and putting this factory at Place 2? In all, the firm
has four options in the static world envisioned here: (i) build no factories, (ii) build
56 2 The Firm at Home and Abroad
only one factory and put it at Place 1, (iii) build only one factory and put it at Place
2, or (iv) build one factory each at both Places 1 and 2. The excess profit that arises
under each option is specified in Table 2.7. Here, Vij is the semi-net revenue from
having the factory at Place i supply customers at Place j where feasible: i.e., where
maximum price (α) is higher than minimum price (C plus any shipping cost). Of
course, the do-nothing option i has a zero excess profit. Also shown in Table 2.7 is
the rate of return for each of options ii, iii, and iv. In these, I have assumed symmetric
shipping costs.61
Using Table 2.7, I can now deduce whether and where the firm will build its fac-
tories. However, since this model contains 8 parameters—α, β, N1 , N2 , C, K, r and
sx—it is easiest to envisage solutions if I make some simplifying assumptions.
Table 2.7 Model 2D: the four options
Notes: Rationale for localization (see Appendix A): Z1—Presence of fixed cost; Z8—Limitation
of shipping cost. Givens (parameter or exogenous): C—Marginal unit production cost; K—Capital
required to build factory; Ni —Number of consumers at Place i; r—Opportunity cost of capital;
s—Cost of shipping one unit of product one km; x—Geographic distance from Place 1 to Place 2;
α—Intercept of individual linear inverse demand curve: maximum price; β—Negative of slope of
individual linear inverse demand curve: marginal effect of quantity on price received. Outcomes
(endogenous): Pi —Price of unit of soap at Place i; Qi —Quantity of soap supplied to Place i; Vij —
Semi-net revenue from serving customers at Place j from factory at Place i; X—Range of soap
(kilometers).
61 A feature of shipping rates such that the cost of shipping a unit of product from Place i to Place
j is the same as the cost of shipping a unit from j to i.
2.8 Two Markets Identical 57
62 Infact, if sx = 0 , the firm is indifferent between having its factory at Place 1 and Place 2 even
if N1 = N2 .
58 2 The Firm at Home and Abroad
I G
D
C
F
Profit
0
H E Unit shipping cost
shown as the curve OB in Fig. 2.4. The firm will choose option iv if, at the stated
unit shipping cost, curve OB lies above DE. Where the stated unit shipping cost is
OF in Fig. 2.4, the firm finds that the rate of return (FH) on the factory at one place is
larger than the opportunity cost of capital (OD), as is the incremental rate of return
(FG) on the factory at the other place.
Why does the rate of return on option iv take the shape of curve OB? To me,
there are four points of interest about OB. First, when unit shipping cost is zero,
the incremental rate of return on option iv is zero. In graphical terms, OB passes
through the origin of Fig. 2.4. Why? When unit shipping cost is zero, there is no
excess profit to be made by building a second factory. Second, the firm will always
find either or both options ii and iii have a rate of return that, at every sx, is greater
than or equal to the rate of return on the incremental investment in option iv. This is
because the firm chooses first the more profitable of options ii and iii and therefore
in option iv is looking at investing in a second factory, this time at the less profitable
place. Third, whether the firm builds the second factory depends on the position of
curve OB in relation to the line DE. As unit shipping cost becomes large enough,
it is uneconomic to supply a remote market. The greater is sx, the more attractive it
becomes to build a factory at the second place. Fourth, the rate of return on option
iv converges with the rate of return on one plant: at sx sufficiently large, each plant
serves only its local market
2.9 Differing Markets 59
Model 2D: Rate of return, locational choice, and unit shipping cost
AB Rate of return schedule under options ii or iii
DE Opportunity cost of capital (r)
FG Rate of return under option iv at stated unit shipping cost
FH Rate of return under options ii or iii at stated unit shipping cost
OB Rate of return schedule under option iv
A OF Stated unit shipping cost
H
Rate of return
G
D E
0
F Unit shipping cost
Suppose instead that there are more customers at Place 2 than at Place 1. In that
situation, the profit and rate of return associated with option iii increases compared
to option ii. Figures 2.5 and 2.6 show profit and rates of return, respectively, for
case where α = 15, β = 1, C = 3, r = 0.05, K = 50,000, and N1 = 200
as before, but now N2 = 250. The larger Place 2 now makes all options more
profitable. When sx = 0, the profit of options ii and iii are the same. At larger
values of sx, the larger size of Place 2 means that the profit associated with option
iii falls off less quickly than does option ii. Compare curves AH, AB, and CD in
Fig. 2.5. Comparing Figs. 2.6 and 2.4, the changes in rates of return are broadly
similar to the changes in profit. The main difference here is that, when sx is large,
the rate of return in option iii remains above the rate of return for option iv. The rate
of return for options ii and iv eventually converge for sx sufficiently large, as these
both then become ways of serving Place 1 alone. The implication of Fig. 2.6 is that
when Place 2 is larger than Place 1, option iii is everywhere preferable to option ii,
but when sx is sufficiently large, the monopolist still builds a factory at the smaller
place if the market there is large enough. The rates of return graphed in Fig. 2.6 tell
us that, as mutually exclusive alternatives, option iii is always preferable to option
ii when N2 > N1 , and that while option iv may generate greater total excess profit
60 2 The Firm at Home and Abroad
I
C D
G
F
Profit
0
E Unit shipping cost
than iii, the opportunity cost of capital has to be sufficiently low to make option iv
attractive.63
Finally, because I have assumed that the two places have a common currency and
no restrictions on the shipment of commodities other than a unit shipping cost, I do
not have to consider here the strategy of tariff jumping, where a firm builds a second
factory to avoid tariffs or other restrictions on trade. This is a rich literature on its
own.64
Let me close this section with an observation about the use of shipping here. Our
firm will ship either (1) nothing at all or (2) from the home to the remote market only.
Since the remote market would then normally have a price higher than the home
market by half-freight, the shipment is said to flow up the price gradient. There will
never be a shipment in the opposite direction: i.e., a cross haul or shipment down
the price gradient. This is not surprising. To have both shipments from A to B and
63 Backhaus (2002) makes a similar point about the importance of opportunity costs in location
theory.
64 The topic of tariff jumping appears to have begun with Brander and Spencer (1987), Motta
(1992), and Neven and Slotis (1996). See also Belderbos (1997), Belderbos and Sleuwaegen
(1998), Horn and Persson (2001), Hwang and Mai (2002), Neary (2002), and Norback and Persson
(2004).
2.10 Comparative Statics in Model 2D 61
Model 2D: Rate of return, locational choice, and unit shipping cost
AB Rate of return schedule under option iii
A AC Rate of return schedule under option ii
DE Opportunity cost of capital (r)
FG Rate of return under option iv at stated unit shipping cost
FH Rate of return under option ii at stated unit shipping cost
FI Rate of return under option iii at stated unit shipping cost
OC Rate of return schedule under option iv
OF Stated unit shipping cost
H
Rate of return
D E
0
F Unit shipping cost
from B to A happening at the same time would presumably require plants at both
places; in such cases, why would the firm ship at all?
Is there a case to be made here for behavior by a firm that we might label market
preemption or entry deterrence?66 We have seen here that building just one factory
(by spreading fixed cost over two markets) is more profitable than building two
separate factories (each with that same fixed cost) unless the shipping rate is too
high. However, this raises the question of how the firm behaves given the risk a
competitor builds a factory at the second (smaller) market in the firm’s absence to
serve just that market. For the firm with a factory only at the larger Place 1, its profit
at risk is (2.3.6). In effect here, I am treating Place 2 as a contestable market.67
We could set this up as a sequential game played by the firm (the incumbent) and
a competitor (the entrant). I do not do that here. My purpose is simply to show that
the models developed in this chapter give an interesting insight into the conditions
under which the incumbent might want to preempt: i.e., build a factory of its own at
Place 2 to forestall the entrant even though that would be less profitable than option
ii absent competition.
In what follows, assume the following specific scenario. An entrant with the same
costs (i.e., same r, C, and K) and product plans a factory at the smaller Place 2
to sell soap there only. Prompting the entrant here is the gap (0.5sx) between the
delivered price of the incumbent (absent the entrant) and the price the entrant intends
to set. Despite the fact that the incumbent finds it profitable to ship to Place 2, the
entrant does not plan to ship soap to Place 1. The entrant is myopic.68 Its conjectural
variation is that it does not expect the incumbent to respond by changing its price
at Place 2 or by building a second factory there. That entrant sets a price (2.8.1) and
expects to earn an excess profit (2.8.2) similar to the one-market solutions (2.1.6)
and (2.1.8), respectively. See Table 2.8. Here, price and profit are invariant with
respect to sx because I have assumed the entrant does not ship product to Place 1.
There are two implications here for the incumbent. First, the entrant will find
Place 2 profitable only if demand there is sufficient. The necessary condition can be
inferred from (2.8.3): akin to the one-market solution (2.1.9). If Place 2 is smaller
than this, the incumbent need not worry about an entrant. If Place 2 is larger than
this, consider a second implication: the entrant’s profit-maximizing price (2.8.1)
at Place 2 would be lower than that of the incumbent (2.3.4). Absent any difference
between their products, the entrant would therefore capture all the market at Place 2.
However, is there a catch here? Since I have assumed that the entrant is like
the incumbent in the sense of facing the same costs, whenever it is profitable for
Notes: Rationale for localization (see Appendix A): Z1—Presence of fixed cost; Z8—Limitation
of shipping cost. Givens (parameter or exogenous): C—Marginal unit production cost; K—Capital
required to build factory; Ni —Number of consumers at Place i; r—Opportunity cost of capital;
s—Cost of shipping one unit of soap one kilometer; x—Geographic distance from Place 1 to
Place 2; α—Intercept of individual linear inverse demand curve: maximum price; β—Negative
of slope of individual linear inverse demand curve: marginal effect of quantity on price received.
Outcomes (endogenous): Pi —Price of unit of soap at Place i; Qi —Quantity of soap supplied to
Place i; Vij —Semi-net revenue from serving customers at Place j from factory at Place i.
68 Any behavior of a firm such that it does not foresee any reaction by its competitors to its choices:
e.g., with respect to price, range, or quality of commodities sold, or geographic location.
64 2 The Firm at Home and Abroad
69 See (2.7.7)
2.13 Final Comments 65
N
R
P
D Q B
U W C
T
H I
K X
Y
V
0
J S Unit shipping cost
Fig. 2.7 Model 2E: rate of return on option ii as a function of shipping rate and presence of
competitor at Place 2.
Note: α = 15, β = 1; C = 3; K = 50,000; N1 = 200 N2 = 180, r = 0.05, sx = 4. Rate of
return is 0.045 on option iv, 0.78 on option 3, 0.083 on option 2 without competitor, and 0.068 on
option 2 with competitor. Horizontal axis is scaled from 0 to 9; vertical from 0 to 0.16
same price at both locations. Put differently, despite shipping costs that are not high
enough to warrant option iv, preemption drives the firm to price as though they were.
In this chapter, the principal model has been 2D. I included Models 2A through
2C to help readers better understand aspects of Model 2D. I included Model 2E
to show how one thinks about entry deterrence at a remote place. In Table 2.9, I
summarize the assumptions that underlie Model 2A through 2E. Many assumptions
are common to all these models: see panel (a) of Table 2.9. The models differ in
that (i) a remote place (market) is assumed in Models 2B through 2E and with it the
possibility of price discrimination, (ii) even then, shipping cost can still be ignored
in Model 2C because the firm produces at both places, (iii) Models 2D and 2E give
the firm the choice of how many factories to build and where to build them, and (iv)
66 2 The Firm at Home and Abroad
2A 2B 2C 2D 2E
Assumptions [1] [2] [3] [4] [5]
Model 2E allows for competition from a new entrant at the remote place and the
uncertainty that creates. Given the assumptions of the model introduced here, where
the number of customers is sufficiently large—which in turn may require that the
two places be sufficiently close together—there will be at least one factory, and it
will be located at the place with the larger number of customers (the home market).
If the unit shipping cost is sufficiently low, there will be localization (i.e., only one
factory) and soap will be priced differently at the two places (the half-freight rule). If
unit shipping cost is sufficiently high, and the smaller place has enough customers,
the firm will decentralize production (i.e., build a second factory there) and prices
will be the same at the two places. Put differently, localization (of production) and
price differentiation are joint outcomes in Model 2D.
In Chapter 1, I argue that prices are important in shaping the location of firms. In
this chapter, those prices are r, the price of a unit of capital inherent in our measure
of K, various input prices subsumed in C and s, as well as consumer income (typ-
ically derived from a labor market) and the prices of other consumer commodities
that are presumably incorporated in α and β. These other prices are all determined
in markets outside the scope of this model where the firm competes against other
firms and demanders. These input prices in turn help determine which of the four
2.13 Final Comments 67
options (each a combination of locations and prices) is most profitable to the firm
and thereby the extent of localization. To this point, I think Walras would argue that
the analysis has been only partial in the sense that we have not looked explicitly
at the simultaneity among prices in these markets. In later chapters, there will be
opportunities to do that. Nonetheless, even here, the decision of the firm to choose
option ii or iii implies that the price of soap will differ between the two places. To
the extent that the firm does not choose option iv, this price effect has the potential
(not explored in this chapter) to create an incentive for customers to relocate to a
place where the firm has built a factory.
What about the regional economy here? There are two problems. First, what is
the region here? Second, what is the nature and extent of economic activity within
it? The models in this chapter paint a picture of only one part of regional economy:
one firm and its customers. We don’t know anything about overall regional product
or regional income (factor payments); however that region might be defined. The
models in this chapter say little about the distribution of income in society among
factors (labor, capital, and land). We know how much profit the firm will earn. We
know also that the firm incurs costs, but (aside from the opportunity cost of capital)
these are not related explicitly to factor payments such as the hiring of labor or the
rental of land. Therefore, the only recognizable income gains that arise accrue to
the firm owner in the form of increased profit. We also know that the price set by
the firm affects the well-being of customers; in this chapter, such changes have been
measured by consumer surplus. Suppose Place 1 is large enough by itself to make
production profitable for the firm, Place 2 is not, and the firm therefore builds a
factory at Place 1. Assume also initially that the unit shipping cost is prohibitive in
the sense that C + sx > α. The firm is limited to its home market. In the absence of
information about other firms and their shipments, we might think of the region here
simply as the firm and its customers at Place 1. However, we do know that the firm
benefits from this market (as indicated by its monopoly profit) as do consumers at
Place 1 (as indicated by their consumer surplus). Now suppose the unit shipping cost
were to drop to a level where it becomes profitable for the firm to serve the Place
2 as well. In this case, we might think that the region now includes Places 1 and
2. We still don’t know anything about overall regional product or regional income.
However, we do know that there has been an addition to the firm’s monopoly profit,
a new consumer surplus for customers at Place 2, and no reduction in consumer
surplus for customers at Place 1.
Finally, this chapter has put much emphasis on the role of the opportunity cost of
capital as a fixed cost. That the firm has to invest capital in one or two factories and
incurs an opportunity cost thereby implies that unit cost falls the greater the scale
of production. That is the one and only rationale for localization envisaged in this
chapter. There is not, for example, any explicit consideration of the efficiencies that
might arise from division of labor in accounting for localization here.
Chapter 3
Logistics and Programming
Getting the Commodity to Customers
(Hitchcock–Koopmans Problem)
A firm has several factories that differ in terms of capacity and unit cost of produc-
tion. The firm sells its product at customer places, each with its own fixed quantity
demanded. How much does it ship from each factory to each customer place? To
solve this problem, linear programming is introduced. In Model 3A, the firm incurs
production costs, but unit shipping cost is zero everywhere. Model 3B includes a
unit production cost and a unit shipping cost that varies from one customer place
to the next for each factory. In addition to allocating production to customers, lin-
ear programs like 3A and 3B generate shadow prices: one for each constraint. The
shadow price on the capacity constraint at factory i tells us how much money the
firm could save (i.e., additional profit it could earn) if only it had one more unit of
capacity there. This is a kind of opportunity cost. There was no congestion (i.e.,
restriction on production) in Chapter 2. The models in Chapter 3 help us to better
understand congestion and its impact on locational decisions (how much, if any,
to produce at each factory). In this chapter, localization and the shadow prices on
capacity (one for each factory) are joint outcomes of cost-minimizing behavior.
In Chapter 2, I considered an efficient firm that sells at two places and operates either
a factory at one place only or alternatively a factory at each place. At each factory,
the firm manufactures soap—from materials purchased elsewhere and processed on-
site—and packages it for sale to customers. The firm there was assumed—having
invested K dollars in land, plant, and equipment—to have an unlimited wellspring
of production: it can expand its output as needed to meet customer demand. In this
chapter, I envisage a firm that might start business with one factory and then, over
the years, add factories (establishments) at other places by purchase of assets or by
construction and/or reorganize production activity including the division of labor
across establishments. In so doing, I assume that the firm has maximized profit in
setting the price for its product at each customer place.
In this chapter, let me again take a single firm producing soap and operating in
a fiat money economy but model it now differently. First, let us assume that there
are potentially more than two places (customer points) served by our firm. Second,
let us now take into account production capacity. In this chapter, I assume each
factory—including land, plant (buildings, improvements, facilities, and fittings), and
equipment—has been optimized for a particular production technology and scale of
production (capacity). Implicit here is a notion of economies of scale; that is, the
firm constructs a factory to take advantage of the indivisibilities possible at that
scale of production. Once the factory is built, we might reasonably expect a firm
to experience congestion at a scale of production above capacity in the short term:
that is, to be unable to further increase output without substantially increasing its
marginal cost of production.
In this chapter, I present what is generally labeled the Transportation Problem
(also known as the Hitchcock–Koopmans Problem).1 Koopmans initially solved this
model during the Second World War. Hitchcock developed an early formulation of
the problem. In the 1950s, Dantzig recast this as the total cost, which allowed for it
to be solved using newly developed simplex algorithm.2 Nowadays, such problems
are readily solved using off-the-shelf optimization software.3
In this problem, imagine you manage logistics4 for this firm over the short run.
Assume you have I factories in total, each producing soap. Assume each factory
has a fixed capacity (rate of output), say Si units of soap weekly for factory i. As
manager, you can set any level of production at factory i for the week from zero
up to but not exceeding Si .5 In this chapter—as in Chapter 2—assume a punctiform
landscape wherein the firm’s customers are distributed across a given set of J places.
1 In this Linear Program, the firm seeks to minimize the sum of the costs of production at a set of
factories, subject to capacity constraints (one for each factory), and the shipments from these fac-
tories to meet the demand requirements of customers at each distinct places. See also Appa (1973,
p. 79) on the early history here.
It was originally solved using a Stepping Stone Algorithm. In general, linear programs are
solved using the Simplex Algorithm. Hitchcock (1941) laid out the problem and sketched its solu-
tion. Among others who also contributed in important ways to the Transportation Linear Program
and its solution were Boldyre, Dantzig, Ford and Fulkerson, Kantorovich, Robinson, and Tolstoi.
Publication of solutions came shortly after the Second World War. Koopmans (1949) gave an eco-
nomic interpretation of the problem and its solution that was further elaborated in Koopmans and
Reiter (1951). Orden (1952) was the first to use the conventional notation now used for the model.
An early application of the Hitchcock-Koopmans Problem is found in Henderson (1955). Modern
applications include Bullard and Engelbrecht-Wiggans (1988), Rautman, Reid, and Ryder (1993),
and Cunha and Mutarelli (2007).
2 The use of simplex here is confusing. Dantzig had originally explored the use of simplexes
to solve linear programs, but the algorithm he eventually used (the simplex algorithm) did not
incorporate simplexes. However, others used simplexes in the fixed point approach to estimating
economic equilibrium. See Kakutani (1941).
3 A well-known early application is found in Harris and Hopkins (1972) and Harris (1973). This
model forecasts investment, production, and employment at the county level and uses the shadow
price on capacity constraints for each industry sector to help predict future investment locally. See
also Harris (1980), Harris and Nadji (1987), and the critique by Fjeldsted and South (1979).
4 The management of production, inventory, and shipments so as to enable a firm to achieve its
objectives.
5 Green, Cromley, and Semple (1980) extend the Hitchcock–Koopmans model to allow for the
possibility that there might be a minimum level of capacity utilization at each factory. In other
3.1 The Hitchcock–Koopmans Problem 71
Suppose further customers at each place demand a fixed amount (say Dj units at
Place j) of soap for the coming week. Further assume, aggregated across all demand
places, total demand (j Dj ) for next week is less than total capacity (i Si ).
In the models in Chapter 2, we did not know what determined the level of
demand. I assumed a downward sloped demand curve: see curve AB in panel (a)
of Fig. 3.1. Here in Chapter 3, I assume a vertical demand curve (also labeled AB)
shown in panel (b) of Fig. 3.1. Unlike Chapter 2, demand is fixed; it does not depend
on price. The objective here simply is to find the least costly way of meeting the
given demands of these customers.6
Note here the implication that the firm is assumed to receive a given but unstated
price at the destination for its soap. Perhaps this is because the firm is a monopolist
as in Chapter 2, has calculated the delivered price using discriminatory pricing that
maximizes profit for that part of the market, and now seeks the least-costly way of
supplying customers there.7 Alternatively, the firm might be in a perfectly compet-
itive market at the destination; it cannot affect the price it receives but nonetheless
wants to minimize the cost of serving customers there.8 In passing, note the contrast
A
A
Price
Price
0 0
B Quantity B Quantity
(a) Demand curve (AB) in Chapter 2 (b) Demand curve (AB) in Chapter 3
words, they allow for a level of production at each factory that has both a maximum (the capacity
constraint) and a minimum (to ensure efficient operation).
6 For discussion of the transportation problem in continuous space which includes the transporta-
tion problem on a line, see Beckmann (1952) and McCann (1999).
7 In practice, it is not clear how this might be done. A monopolist sets price so that marginal revenue
equals marginal cost. However, we don’t know marginal cost here until we determine which factory
is to provide the shipments to a given customer.
8 The model is silent on the possibility of arbitrage; however, the assumption of a fixed price
is perhaps easier to imagine under the assumption that there is only a local demand at each
destination.
72 3 Logistics and Programming
9 Perhaps, land too becomes depleted (e.g., as a result of quarrying associated with production) or
contaminated.
3.1 The Hitchcock–Koopmans Problem 73
Marginal cost
Marginal cost
A B
A B
0 0
Output S Output
(a) Marginal cost curve (AB) in Chapter 2 (b) Marginal cost curve (ABC) in Chapter 3
• The firm might expect ordinarily to meet demand entirely from its new, more
efficient, factories but retains older factories in case it unexpectedly needs more
capacity or loses use of a newer factory temporarily;
• Unlike Chapter 2, each factory is designed to operate most efficiently at a par-
ticular scale of production—its design capacity—and is simply not capable of
producing any and all output needed.
As a result, the firm may have factories available that range from more efficient
(state of the art) to less efficient (obsolete). In the short run, say one week or one
month, the firm can do little to alter a particular factory. Largely, it can adjust only
the amounts of materials and labor used in production at the factory or not produce
there at all. These variations are reflected in Ci for each factory. At an obsolescent
or older factory, unit production cost might, therefore, typically be higher than at a
newer factory. In a longer run beyond the scope of this model, the firm may have
the opportunity to repair, upgrade, or remediate land, plant, and equipment at an
existing factory.
Assume further the firm incurs a unit shipping cost for each unit of soap shipped
from factory i to customers at j. Assume this is a constant unit cost, sij . By con-
stant, I mean once again that the cost for each unit shipped does not depend on how
much is shipped: e.g., no economies of the large haul.10 I assume here that (1) there
is a minimum efficient size of shipment, (2) the firm ships annually multiples of
this amount, and (3) the firm decides how many multiples (one or more) to ship at
any one time. Let qij be the amount of soap produced weekly at i for shipment to
10 Inan early empirical study of this kind, Chisholm (1959) finds no evidence of economies of
scale in road goods transport in his case study of off-farm milk collection in England.
74 3 Logistics and Programming
customers at j. The shipping cost for this particular shipment is sij qij and the total
shipping cost for all shipments by the firm over the week is i j sij qij .11 The total
cost of production and shipment weekly is i j (Ci + sij )qij . Your job as manager
is to find the least-cost allocation of factory production to customer demand under
the assumptions that demand at every customer place is satisfied and that capacity at
every factory is not exceeded. The problem can now be written as in Table 3.1: there,
I summarize the equations, assumptions, notation, and rationale for localization in
Model 3B.
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z5—Capacity constraints and congestion. Givens (parameter or exogenous): Ci —Unit
production (variable) cost for Factory i; Dj —Amount of soap demanded at Place j in period
of time; I—Number of factories (supply points); J—Number of places (customer points); Si —
Amount of soap that can be produced at Factory i in period of time. Outcomes (endogenous):
qij —Amount of soap shipped to Place j from Factory i in period of time; Z—Total variable cost
of producing and shipping soap.
Above, I introduced the concept of the short run vs. the long run and then stated
that the logistics manager is concerned mainly with the short run. The short run is
the focus of the remainder of this chapter. However, before I consider specific ver-
sions of the model, it is helpful to think how much different the analysis would be
in a longer run perspective. In the short run, assume the firm cannot alter anything:
buildings, improvements, facilities, equipment, and fittings, or a factory design built
around a particular technology and scale of production. Therefore, there may be
a substantial difference between unit production costs at any two factories. In the
longer run, the firm can make an investment: e.g., replace or add equipment, facil-
ities, and buildings, and/or redesign a factory to increase its capacity. This should
bring unit production cost (i.e., reduce Ci ) at the more costly factory down closer to
that of the more efficient factory, or reduce shipping cost by enabling more pro-
duction at a more efficient factory (i.e., increase Si ). In making that investment
decision, the firm presumably uses a criterion similar to that discussed in Chapter 2,
11 Note the asymmetric treatment of congestion here; the model takes into account factory capacity
but otherwise assumes no congestion over a transportation network.
3.2 An Illustrative Example 75
namely, “Is the reduction in production and shipping cost sufficiently large to offset
the opportunity cost of the investment.” However, even in the long run, differences
among factories arise, for example, because inputs are less expensive for a factory
at one place compared to another (i.e., one or more inputs are not ubiquitous). See,
for example, Chapter 6.
Because I look at this firm in the short run, differences in marginal cost curve
from factory to factory and differences in costs of shipping to a consumer are givens;
that is, cannot be affected by the firm. Under this interpretation, more efficient fac-
tories generate excess profits compared to the marginal factory12 employed. To the
extent such excess profits arise because of the uniqueness of an input (e.g., an
advantageous location), these are also Ricardian Rents. In allocating output from
factories to the demand at a particular customer place therefore involves consid-
erations of opportunity cost. In this chapter, I introduce the concept of a shadow
price—a mathematical term—to implement the idea of opportunity cost here.
As in Chapter 2, this chapter includes material that would normally be part
of optimal location theory. Later in this chapter, I show how optimization by a
firm produces geographic outcomes—the allocation of its factories to customer
demand—that appear to be as though the firm’s factories compete with one another.
Put differently, in trying to be efficient, the monopolist allocates production and
shipments in a way that replicates a competitive market.
To illustrate the problem with a specific example, suppose you have customers at
three distinct customer places (labeled A, B, and C) that you can serve from any
combination of four factories (labeled 1, 2, 3, and 4) laid out on a rectangular plane:
see the map in Fig. 3.3. The pertinent details on demand, capacity, production costs,
and shipping costs are shown in Table 3.3. I have assumed here that shipping cost is
proportional to distance measured as the crow flies. I have constructed this example
to help clarify the application of the Hitchcock–Koopmans model. I intentionally
placed the factory with the lowest unit production cost (factory 1) far away from the
customer place where much of the demand is concentrated (namely, Place C). After
all, you might rightly ask why the most efficient factory is so far away from Place
C; presumably the firm might have wanted to build the factory closer to C to save
on shipping costs. Where inputs are ubiquitous, the firm would locate close to C on
the line joining C, B, and A that could be drawn on the map in Fig. 3.3. Presumably,
the reason here might be that either (i) there is a non-ubiquitous input that is less
costly near the site of factory i or (ii) the spatial pattern of demand has shifted over
time since factory 1 was built so that now customer demand is typically further
away.
12 For a firm with multiple factories or an industry composed of multiple firms each with its own
factory, the marginal factory is the factory, which produces the last (most costly) unit to the market.
76 3 Logistics and Programming
3 C 4
0 50 100
km km km
I constructed this example with specific ideas in mind. First, I deliberately put the
two factories with the lowest unit production costs far away from customer places
B and C where demand is relatively large. I did this to exemplify the tradeoff for
the firm between production and shipping costs. Second, for ease of exposition, I
have set the factory capacities in this example problem so that the firm will always
need three factories to meet the aggregate demand (4,200 units of soap weekly).
Therefore, one factory will always be unused. Also, one of the three factories used
will be marginal: that is, the factory whose output would be reduced by a unit if
aggregate demand was reduced by one unit (alternatively, in general, the factory is
producing, but with slack; that is, producing below its capacity).13
Let us begin to analyze this problem by considering a special case in which unit
shipping cost (sij ) is everywhere zero. See Table 3.2. In that situation, our model
becomes non-spatial since it does not matter where our customers or our factories
happen to be. Sum the demand at each customer place to give an aggregate quantity
demanded, QD = j Dj . I can also derive an aggregate supply curve for the firm.
Without loss of generality, index factories from lowest unit production cost to high-
est so that C1 ≤ C2 ≤ . . . ≤ CI . Thus, for any aggregate quantity demanded up to
S1 , the marginal cost is a constant C1 . Beyond that and up to a quantity of S1 + S2 ,
the marginal cost is C2 . Beyond that and up to a quantity of S1 + S2 + S3 , the
marginal cost is C3 , and so on. This yields a kinked14, 15 supply curve, also known
13 I ignore here the knife-edge possibility that two or more factories have the same unit cost
(production plus shipping) and may therefore both qualify as marginal factory.
14 A condition wherein a function (e.g., demand schedule, supply schedule, excess supply, excess
demand, or Price Difference Curve) takes the form of a polyline or piecewise-linear spline function:
i.e., forms a continuous function with a discontinuous derivative.
15 Samuelson (1952, p. 286) presents a diagram in which local demand and local supply curves are
each kinked, but offers no explanation or interpretation of this.
3.3 Model 3A: Non-spatial Version of the Model 77
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z5—Capacity constraints and congestion; Z8—Limitation of shipping cost. Givens
(parameter or exogenous): Ci —Unit production (variable) cost for Factory i; Dj —Amount of
soap demanded at Place j in period of time; I—Number of factories (supply points); J—Number
of places (customer points); Si —Amount of soap that can be produced at Factory i in period of
time; sij —Cost of shipping each unit of commodity from Factory i to Place j. Outcomes (endoge-
nous): qij —Amount of soap shipped to Place j from Factory i in period of time; Z—Total variable
cost of producing and shipping soap.
as a supply step function. A kinked line is also known as a polyline.16 I then simply
move to the right along the horizontal axis until I reach quantity QD , then travel
vertically until I intersect the supply step function. Suppose this happens along the
horizontal segment of the step function corresponding to factory m. I then know
all the more efficient factories (factories 1, 2, . . . , m − 1) will be fully used along
with some or all of factory m. Factory m is the marginal factory.17 The amount of
production is now clear. I use all of the output available from the m − 1 most effi-
cient factories with the remainder constituting QD coming from the marginal factory
m. However, the pattern of shipments, which output to ship to which customers, is
indeterminate. Because shipping costs are zero, it does not matter which of the most
efficient m factories are used to supply a given customer.
So far, we have thought about the firm’s problem from the perspective of logis-
tics: i.e., quantities to be produced and shipped. However, an economist usually
finds it helpful to think about prices and/or opportunity costs. How are we to do
this however when the problem is set out ignoring price setting by the firm? The
answer lies in the idea that each choice of shipment by the firm has a measurable
opportunity cost. Compared to the marginal factory, factory 1 is more efficient by
the amount Cm − C1 . Let us now calculate how much money the firm could save
if only it had one more unit of capacity at any one factory. If the firm had one more
unit of capacity at the most efficient factory 1, that is, S1 + 1 instead of S1 , it could
16 A piecewise-linear spline function composed of two or more segments. Each segment (piece) is
itself a linear function. The notion of a spline is that adjacent pieces share an endpoint: i.e., forms
a continuous function with a discontinuous derivative.
m − 1 < Cm < Cm + 1 .
17 I assume here for simplicity of exposition that C
78 3 Logistics and Programming
save Cm − C1 . If the firm had one more unit of capacity at factory 2, it could save
Cm − C2 , and so on up to factory m where the savings would be zero (Cm − Cm )
by definition. For any factory from m + 1 to I, the savings would be zero, since each
such factory is not currently in use, and therefore added capacity would not help at
all. Let us refer to such savings as the shadow price18 of capacity at each factory.
These shadow prices implement the idea of an opportunity cost.
Note the similarities here among shadow price, excess profit, and Ricardian Rent.
If excess profit is arising because of an input unique to factory 1, then that excess
profit is also the Ricardian Rent. If the quantity Qd can be sold at unit price P:
EP Excess profit attributable to factory: e.g., EP1 = (P − C1 )S1 .
SV Aggregate shadow value attributable to factory: e.g., SV1 = (Cm − C1 )S1 .
RV Residual difference between EP and CV (profit earnable on the marginal
factory and standardized to the capacity of factory): e.g., EP1 − SV1 =
(P − Cm )S1 .
To illustrate the ideas so far, consider a non-spatial version of the example problem
presented in Table 3.3. Here I use all data from Table 3.3 except to assume sij is
zero for all flows. In Fig. 3.4, I show the outputs forthcoming from the four facto-
ries in ascending order of production cost: see the polyline ABCDEFG. I also show
the aggregate quantity demanded in the three customer places as a vertical line at
4,200 units: see PQ in Fig. 3.4. The intersection of this curve with the cumulative
step function comes at unit production cost (Cm ) equal to 22: see OM in Fig. 3.4.
Therefore, 22 is the shadow price on demand at any of the three places: put dif-
ferently, if we reduce demand at any place by 1 unit, the firm would save 22. As
well, 22 is the unit cost of production at the marginal factory. Therefore, factory
1—the most efficient factory—has a shadow price on capacity of 22−20 = 2:
see AM in Fig. 3.4. Factory 2 has a shadow price of 22−21 = 1 : see LM in
Unit Factory
Factory production Unit shipping cost sij capacity
i cost (Ci ) j=A j=B j=C weekly Si
[1] [2] [3] [4] [5] [6]
1 20 6 7 9 1,000
2 21 6 7 9 2,000
3 23 6 4 4 2,500
4 22 6 4 4 1,500
Weekly demand
j=A j=B j=C
Dj 200 1,000 3,000
Minimize production plus shipping costs
Z = (20 + 6)q1a + (20 + 7)q1b + (20 + 9)q1c + (21 + 6)q2a (3.3.1)
+ (21 + 7)q2b + (21 + 9)q2c + (23 + 6)q3a
+ (23 + 4)q3b + (23 + 4)q3c + (22 + 6)q4a + (22 + 4)q4b + (22 + 4)q4c
Subject to the following conditions
Shipments must satisfy demand at each customer place
q1a + q2a + q3a + q4a ≥ 200 (3.3.2)
q1b + q2b + q3b + q4b ≥ 1,000 (3.3.3)
q1c + q2c + q3c + q4c ≥ 3,000 (3.3.4)
Shipments from each factory must not exceed capacity
q1a + q1b + q1c ≤ 1,000 (3.3.5)
q2a + q2b + q2c ≤ 2,000 (3.3.6)
q3a + q3b + q3c ≤ 2,500 (3.3.7)
q4a + q4b + q4c ≤ 1,500 (3.3.8)
Each possible shipment must be either zero or positive
qij ≥ 0 for i = 1, 2, 3, or 4 and j = a, b, or c (3.3.9)
Fig. 3.4. Factory 4 (the marginal factory) and factory 3 (unused) each have a shadow
price of zero. In this non-spatial model, total cost of meeting customer demand
is 1,000(20) + 2,000(21) + 1,200(22) = 88,400. In Fig. 3.4, the box AMNBA
shows the cost savings from Factory 1 and the box CNEDC is the cost savings from
Factory 2.
I have just determined the shadow prices when Qd is 4,200 units. How would
these shadow prices differ if Qd were something else? The idea here is to see how
shadow prices change as I go from a relatively low level of aggregate demand to a
relatively high level. Again, Fig. 3.4 is helpful here. Outcomes are summarized in
Table 3.4
• If Qd is under 1,000 units, use only factory 1. Factory 1 is the marginal factory.
The shadow price for each factory is zero. The shadow price on demand anywhere
is 20.
80 3 Logistics and Programming
N E F
M
L D
C
A B
0
H I Q J Quantity K
• If Qd is between 1,000 and 3,000 units, use both Factories 1 and 2. The
marginal factory is factory 2. Capacity at factory 1 now has a shadow price of
1. The shadow price at each other factory is zero. The shadow price on demand
anywhere is now 21.
• If Qd is between 3,000 and 4,500 units, use Factories 1, 2, and 4. The marginal
factory is factory 4. Capacities at Factories 1 and 2 now have shadow price of
2 and 1, respectively. The shadow prices at other factories are each zero. The
shadow price on demand anywhere is now 22.
• If Qd is between 4,500 and 7,500 units, use Factories 1, 2, 4, and 3. The marginal
factory is factory 3. Capacities at Factories 1, 2, and 4 now have shadow price of
3, 2, and 1, respectively. The shadow price at the now-marginal factory 3 is zero.
The shadow price on demand anywhere is now 23.
What I have just done here is to illustrate how the solution to a problem changes
if I were to change a given (in this case, Qd ). This kind of exercise, based on shadow
3.5 Model 3B: Spatial Version of the Model 81
Table 3.4 Model 3B: shadow prices and capacity surplus as Qd is varied
A 20 2 22
B 20 21 22
C 20 21 22
Notes: Min. slack is excess capacity at factory when QD is at maximum of range for that col-
umn. Parameter values given in Table 3.3 with the following modifications: DA =(200/4,200)QD ;
DB =(1,000/4,200)QD ; DC =(3,000/4,200)QD . Calculations by author.
How different is the solution where shipping costs are not zero. This is another
possible exercise in comparative statics. Put simply, the answer is that the solution
can change a lot! With shipping costs, it now matters which factory is used to supply
a given customer. A factory with a low production cost may no longer be an efficient
source if shipping costs from it are relatively high.19
19 Transportation linear programs have been applied widely. Illustrative of the breadth of applica-
tion, Kuznar (1991) uses such a model to look at herd management among Aymara alpaca herders
in the south central Andes in response to improved transportation systems. The focus here is on
how herders choose the size and type of herd constrained by the land and labor resources available
to them. The study concludes that herders are optimizing herd value, that sheep are a poor option
for Andean herders, and that this explains the reluctance of herders to adopt sheep herding.
82 3 Logistics and Programming
Maximize
Z = Σi ai Xi (3.5.1)
Subject to
Σi bij Xi ≤ cj at each j =1, 2, . . ., J (3.5.2)
Xi ≥ 0 for each i = 1, 2, . . ., I (3.5.3)
Linear programming20 can be used to solve any problem (the so-called primal21 )
that can be written as shown in Table 3.5. There, Z is an objective to be maximized,
X1 , X2 , . . . , XI are each an instrument variable22 whose value is selected to maxi-
mize Z, there are J less-than-or-equal-to constraints on linear combinations of Xi ,
and each Xi is nonnegative. The remaining terms (ai , bij , and cj ) are constants. This
is said to be a Linear Program because both the objective function (3.3.1) and each
of the J constraints can be written as a linear function of the instrument variables.
Linear programs are solved using numerical algorithms.23
The solution to the primal gives the values of the instrument variables—
X1 , X2 , . . . , XI —that optimize the value of Z. In the case of the Hitchcock–
Koopmans problem, the instrument variables are the shipments from factories to
customer places where customers demand soap. Given those shipments, the firm can
then readily calculate the total cost of production and shipment across all factories.
There is an important caveat here. A linear program is said to be solved when we
find a combination of instrument variable values—X1 , X2 , . . . , XI —that optimize
the value of Z. However, such a combination may not be unique. Typically, there
are other possible combinations of instrument variable values that also produce the
same optimimal value of Z. When we solve a linear program, it is therefore more
24 Inthis chapter, sij is a unit shipping cost. In Chapter 2, s was a unit shipping rate.
25 In Linear Programming, the amount by which the left-hand-side of a less-than-or-equal-to
inequality is less than the right-hand-side.
26 This has the further implication that the number of variables in the solution to the primal that are
nonzero cannot exceed the number of constraints.
27 Among early work on duality, see Gale, Kuhn, and Tucker (1951), Charnes, Cooper, and
Henderson (1953), and Wagner (1958).
84 3 Logistics and Programming
Minimize
Z = Σ j cj vj (3.6.1)
Subject to
Σj bij vi ≤ ai at each i = 1, 2, . . ., I (3.6.2)
vj ≥ 0 for each j =1, 2, . . ., J (3.6.3)
Duality theorem has two remarkable conclusions. First, the optimized value of the
objective function in the dual, Z , is the same as the optimal value of the objective
function in the primal, Z. In the Hitchcock–Koopmans primal, Z is the negative sum
of production and shipping costs. So too therefore is Z . The total cost (Z) to be
minimized in the primal is the same as aggregating the opportunity cost of each unit
of demand served (j Dj vj ) net of the savings made possible by each unit of capacity
(i Si ri ). The primal is about quantities (shipments) that are efficient, while the dual
is about the shadow prices that underlie this. Put differently, the dual reinterprets
firm expenditure as a sum of opportunity costs. Second, the optimal values of the
instrument variables in the dual (v1 , v1 , . . . , vJ ) are the shadow prices of the primal.
The Duality theorem is important here for two reasons. The lesser of the two
reasons is that it sharply reduces the amount of computation. Instead of computing
up to J + 1 linear programs to calculate a primal solution and its shadow prices, I
need to compute only 2: the primal and the dual. From the perspective of this book,
the more important of the two reasons is that the dual gives us an insight into the
comparative statics of the allocation process that are not evident from the primal.
Let me now explain why in more detail.
To see the significance of this latter reason in our case, I need to write the dual
of the Hitchcock–Koopmans problem. As a first step, I must rewrite the Problem
as a primal. See Table 3.7. Two changes are necessary here. One, as is done in
(3.7.1) is to convert the objective from a minimization to a maximization by multi-
plying through by –1. Minimizing a positive amount is like maximizing its negative;
(3.7.1) is equivalent in structure to (3.5.1). The capacity constraints (3.2.3) in the
Hitchcock–Koopmans problem already look like the inequalities (3.5.2) in the pri-
mal. However, demand constraints (3.2.2), which take the form of greater-than-or
equal-to constraints do not. This is remedied by multiplying both sides of (3.2.2)
by –1; this automatically converts the inequalities to a less-than-or-equal-to form.28
With these two changes, the Hitchcock–Koopmans problem is now in the form of a
primal.
The dual to this Problem can then be constructed as shown in Table 3.8. Note
here two sets of shadow prices: one set of shadow prices (ri ) for the capacity con-
straints and another set (vj ) for the demand constraints. I have already shown ri is
the cost saving that would arise were the capacity of factory i one unit larger. What
Maximize
− Z = − Σi Σj (Ci + sij )qij (3.7.1)
Subject to
− Σi qij ≤ − Dj at each j (3.7.2)
Σj qij ≤ Si at each i (3.7.3)
qij ≥ 0 for each combination of iand j (3.7.4)
Example problem
Maximize negative of production plus shipping costs
Z = − (20 + 6)q1a − (20 + 7)q1b − (20 + 9)q1c − (21 + 6)q2a − (21 + 7)q2b (3.7.5)
− (21 + 9)q2c − (23 + 6)q3a − (23 + 4)q3b − (23 + 4)q3c − (22 + 6)q4a
− (22 + 4)q4b − (22 + 4)q4c
about vj ? If the demand constraint at Place j were one unit less binding, the demand
there would be only Dj −1. In that case, the firm would be able to save production
and shipping costs. That cost saving is vj . It is not really a cost saving; the firm pre-
sumably makes a profit by selling that unit of soap and therefore would not want to
forego it simply to avoid a cost. Rather, I think of vj as the marginal cost of the last
(most costly) unit produced for customers at that place.
The second thing the dual tells us is something about the relationship between
the two kinds of shadow prices. Key here is the role of (3.8.2). To begin think-
ing, suppose I have more than sufficient capacity at every factory to satisfy all
demands that might be placed on it. In that special case, ri would be zero for
every factory by the Complementary Slackness Theorem. As shown, (3.8.2) states
ri ≥ vj − (Ci + sij ). Given the objective function in (3.8.1), I want to make each ri
as small as possible which means, given the nonnegativity of shadow prices, ri =
max [0, vj − (Ci + sij )] for the customer place with the largest vj − (Ci + sij ).
Intuitively, this should not be surprising; if there are no capacity constraints, the
opportunity cost of supplying the last unit demanded at Place j (vj ) is the cost of
acquiring soap from the lowest cost factory (Ci + sij ) and this should imply ri = 0.
86 3 Logistics and Programming
Minimize
Z = Σi Si ri − Σj Dj vj (3.8.1)
Subject to
ri ≥ vj − (Ci + sij ) at each i and j (3.8.2)
ri ≥ 0 at each i (3.8.3)
vj ≥ 0 for each j (3.8.4)
Example problem
Minimize
Z = 1,000r1 + 2,000r2 + 2,500r3 + 1,500r4 (3.8.5)
− 2,00va − 1,000vb − 3,000vc
Subject to
r1 − va ≥ − (20 + 6) (3.8.6)
r1 − vb ≥ − (20 + 7) (3.8.7)
r1 − vc ≥ − (20 + 9) (3.8.8)
r2 − va ≥ − (21 + 6) (3.8.9)
r2 − vb ≥ − (21 + 7) (3.8.10)
r2 − vc ≥ − (21 + 9) (3.8.11)
r3 − va ≥ − (23 + 6) (3.8.12)
r3 − vb ≥ − (23 + 4) (3.8.13)
r3 − vc ≥ − (23 + 4) (3.8.14)
r4 − va ≥ − (22 + 6) (3.8.15)
r4 − vb ≥ − (22 + 4) (3.8.16)
r4 − vc ≥ − (22 + 4) (3.8.17)
ri ≥ 0 for i = 1, 2, 3, or 4 (3.8.18)
vj ≥ 0 for j = a, b, or c (3.8.19)
Notes: See also Table 3.2. Here, ri is shadow price on the capacity
constraint at factory i. vj is shadow price on the demand at market j.
Alternatively, suppose there are binding supply constraints. In this case, even though
factory i may be the lowest cost supplier to Place j, the entire output of factory i may
be more efficiently committed to other customer places, and the opportunity cost of
supplying Place j may be larger than Ci + sij . Thus, the factory’s opportunity cost
of capacity (ri ) is the largest difference vj − (Ci + sij ) among the customer places.
This is a remarkable result.29 In Chapter 2, I looked at the outcomes in com-
petitive markets. Assume here that each factory was an independent, competitive
supplier and that shipping costs were either zero or sufficiently low for all suppliers
to compete for all customers. Then, based on the models presented in Chapters 2
and 4, I might expect market prices at each pair of geographic demand places to
differ by no more than a fraction of the shipping cost between the two places.
The Hitchcock–Koopmans dual tells us that something akin happens—substituting
29 Boventer (1961) came up with a similar interpretation by looking at the numerical procedure
used to solve the primal.
3.6 The Example: A Spatial Version 87
• On the one hand, this should not be surprising. After all, the monopolist is sim-
ply trying to be efficient here and hence mirrors the operation of a competitive
market.
• On the other hand, this might appear at first glance to be surprising. In Chapter 2
a monopolist sets delivered prices at two customer places, where the local
demand curves are linear in price, to differ by one-half the unit shipping cost.
In the Hitchcock–Koopmans dual, where the firm uses the same factory to serve
customers at two different places, the firm sets the two opportunity costs to differ
by the full amount of the unit shipping cost. Remember here though that oppor-
tunity cost and price are not the same thing and that the Hitchcock–Koopmans
model takes quantity demanded as given and is not designed to solve for the
profit-maximizing price charged at each demand place.
I return now to the example problem. The solution shipments and shadow prices are
shown in panel (a) of Table 3.9. In the least cost solution, the firm supplies customers
at A from factory 1, customers at B from factory 4, and customers at C from factories
3 and 4. Factory 3 is marginal and factory 2 is unused. The total cost of supplying all
customers is 200(20 + 6) + 2,500(23 + 4) + 1,000(22 + 4) + 500(22 + 4) =
111,700. That this is greater than the total cost (88,400) in the non-spatial version
above reflects the shipping costs incurred here. These shipping costs are sufficient
to make the firm give up use of factory 2, with its relatively low unit production
cost, in favor of factories 3 and 4 which are closer to the customer Place C where
much demand is concentrated. Shadow price on capacity is 1 at factory 4 and zero
elsewhere. Shadow price on demand is 26 at A and 27 at either B or C.
In panel (b) of Table 3.9, I present the example problem cast in terms of the dual
inequality (3.8.2). Column [2] shows the 4 shadow prices on capacity: one row for
each factory. In any row, Columns [3], [4], and [5] show the value of vj − (Ci + sij )
for each of the three sets of customers. For none of the customer places does an
extra unit of capacity at factory 1 help reduce costs: hence r1 = 0. The same is true
in the next two rows for factories 2 and 3. However, in the case of factory 4, an extra
unit of capacity would reduce the cost of serving customers at B by 1. The same is
true for customers at C. Therefore, r4 = 1. Even though this is a model to solve
logistics, the firm allocates production as though it were responding to customers
willing to bid the most: i.e., the places with the highest opportunity cost.
In the example problem, I consider the case where Qd = 4,200. Suppose I
experiment by trying a different Qd while holding the demands at the three places
in the same relative shares (200: 1,000: 3,000). Here, I use Qd = 4,200v where
v ranges from 0 to 1. Put another way, I maintain the relative spatial pattern of
88 3 Logistics and Programming
vj − (Ci + sij )
consumers (concentrated mainly at Place C) but set the aggregate number of con-
sumers either higher or lower. I do not look here at a region growing over the longer
run because that would require us to think about how the firm foresees growth and
invests accordingly in new plant and equipment. Instead, what I am doing here is
a kind of comparative statics analysis; we treat the populations of consumers at
Places A, B, and C at the stated Qd as given and solve for the least cost shipments.
See Table 3.10.
If aggregate demand is set low, say at Qd = 420, the least cost solution is to
supply the 300 units demanded at A from factory 1 and the 100 units demanded at
B as well as the 20 at C both from factory 4. Since there is slack in every factory,
r1 = r2 = r3 = r4 = 0. The shadow prices on demand are 26 for each of
the three customer places because factory 1 supplies customers at A at that cost as
does factory 4 for B and C. As I increase Qd , the shadow prices on capacity stay at
zero up until capacity is reached at one of the factories. First to run out is factory 4,
when Qd = 1,575. At that point, DB = 375 and DC = 1,125, which exhausts the
capacity at factory 4. Note that DA = 75 only, so there is much spare capacity at
factory 1. Factory 2 and 3 are unused here.
3.6
Table 3.10 Model 3B: shadow prices and capacity surplus as Qd is varied
v =0 0 < v <0.2 0.20 < v <0.33 0.33 < v <0.40 0.40 < v <0.67 0.67 < v <1
Min. shadow Min. shadow Min. shadow Min. shadow Min. shadow Min. shadow
i slack price (ri ) slack price (ri ) slack price (ri ) slack price (ri ) slack price (ri ) slack price (ri )
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13]
The Example: A Spatial Version
Note: Min. slack is excess capacity at factory when v is at maximum of range for that column. Parameter values given in Table 3.3. Calculations by author.
89
90 3 Logistics and Programming
When Qd rises above 1,575, B and C have to be supplied from the next best
alternative. There are two options here that are equally efficient.
The firm keeps expanding use of factories 1 and 3 to meet the growth in Qd until
it runs out of capacity at both factories. The shadow price on capacity on factory 4
is now 1 (because it costs us now 1 more to supply the marginal unit at B or C);
the shadow price on capacity at each other factory is still zero. The shadow price
on demand at A remains at 26; for both B and C, shadow price on demand at B
and C is now 27 because the marginal customer in each case is being supplied on
a more costly basis. As Qd is increased, the solution retains these characteristics
up to the level where the factories 1 and 3 run out of capacity. This happens when
Qd = 5,000. At that point, factory 1 supplies 238 units to A and the remainder of
its capacity to B. Factory 3 supplies 429 units of its capacity to B, and the remainder
of its capacity to C. Factory 2 is still unused here.
When Qd rises above 5,000 units, shipments begin from the least efficient factory
2. The shadow price on demand rises to 27 at A, 28 at B, and 30 at C because the
marginal customer in each case is now being supplied on a still more costly basis.
The shadow prices on capacity for the more efficient factories are now 4 at factory
4, 3 at factory 3, and 1 at factory 1.
This simple experiment illustrates the idea that shadow prices are tied fundamen-
tally to the notion of the marginal factory. As the increase in demand exhausts the
capacity of the marginal factory, the firm must switch to a new marginal factory and
then the shadow prices rise for factories already used to capacity.
Finally, I do one more experiment: this time varying the relative cost of shipping.
As in the previous experiment, I use the example problem from Table 3.3 (including
Qd = 4,200). This time, however, assume that shipping cost from any factory to
a customer place is a fixed fraction (v) of the shipment cost shown in Table 3.3.
For example, if I set v = 0.5, s1A would now be 3.00, s4C would be 2.00, and so
on. If I set v = 0, I get the non-spatial version of the model. If I set v = 1.0,
I get the solution that we have already discussed in Table 3.9. If I set < v < 1,
I get a problem in which shipping cost is everywhere proportionally less than in
Table 3.9. To understand these outcomes, look at Fig. 3.4 and keep in mind that
capacity constraints may mean the firm cannot always choose the least cost factory.
First, I consider what happens when v > 1. If shipping cost became sufficiently
high, that is, v much larger than 1, the firm would still find it least costly, capacity
permitting, to serve customers at A from factory 1; all factories are equidistant from
A, but factory 1 still has an advantageous unit production cost. This is like when
v = 1. Since DA = 200 and S1 = 1,000, there is sufficient capacity at factory
3.6 The Example: A Spatial Version 91
1. When v is large, the firm would want to serve customers at B and C—capacity
permitting—from factory 3 because it is closer than factories 1 or 2 and has a unit
production cost advantage over factory 4. Since DB + DC + 4,000, the capacity of
factory 3 is insufficient. Given v sufficiently high, the firm finds it less costly to meet
the excess demand at B and C from factory 4 than from either factory 1 or factory 2.
Hence, factory 2 will be unused and factory 4 will be the marginal factory. This is
like when v = 1. In other words, when v > 1, we get a similar solution—in terms
of the pattern of shipments—as when v = 1. This is akin to, but not the same as, the
situation of autarky30 to be outlined in Chapter 4. In Chapter 4, I assume shipping
costs large enough to prohibit any shipments. Here, we assume v is sufficiently large
to make unattractive all shipments except from the nearest supplier.
Panel (a) of Table 3.10 shows the minimum slack and shadow price on capac-
ity for each factory as v is varied from 0 to 1. Panel (b) of Table 3.10 shows the
shadow price on demand for consumers at each place. What happens to the least
cost solution as v is varied from 0.0 to 1.0?
30 A condition wherein regions do not engage in trade. Local supply in one geographic market is
not available to meet local demand in another geographic market for the same product. Autarky
arises when shipping costs are too high to permit arbitrage.
92 3 Logistics and Programming
and rises linearly from 0 to 0.67 at factory 4. The shadow prices on demand rise
linearly everywhere, from 22.20 to 23.00 at A, 22.40 to 23.33 at B, and 22.80
to 24.00 at C. At v = 0.33, the firm can do no better than to supply A from
factory 2 at a cost of 23.00, B from factory 2 at a cost of 23.33, and C from a
blend of factories 1, 2, and 4 at 23.00, 24.00, and 23.33, respectively. Factory 2
is marginal; factory 3 is still unused.
• At 0.33 < v ≤ 0.40, shipping costs are sufficiently high for the proximity of
factory 4 to Place B to offset its higher unit production cost. It is now less costly
to supply customers at B from factory 4 than it is from factory 2. However, this is
of no value to the firm since factory 4 is already fully used. Over this interval of
v, the shadow price on capacity remains at 1 at factory 1, zero at factories 2 and 3,
and rises from 0.67 to 1 at factory 4. The shadow price on demand rises linearly
everywhere: from 23.00 to 23.40 at A, 23.33 to 23.80 at B, and 24.00 to 24.60 at
C. At v = 0.40, the firm can do no better than to supply B entirely from factory
2 now at a cost of 23.80, A from factory 1 at a cost of 22.40, and C from a mix of
factories 1, 3, and 4 at unit costs of 23.60, 24.60, and 23.60, respectively. Again,
factory 2 is marginal; factory 3 remains unused.
• At 0.40 < v < 0.67. The least cost supplier for customers at C is factory 4
when v > 0.40. At v > 0.60. the second least cost supplier for C is factory 3.
Over this interval of v, the shadow prices on capacity are 1 at factory 1, zero at
factories 2 and 3, and 1 at factory 4. The shadow price on demand rises linearly
everywhere: from 23.40 to 25.00 at A, 23.80 to 25.67 at B, and 24.60 to 25.67 at
C. At v = 0.67, the firm can do no better than to ship 1,500 units each from
factories 4 and 3 to C at a unit cost of 24.67 and 25.67, respectively, 1,000 units
from factory 1 to B at a cost of 24.67, and 200 units to A from factory B at a cost
of 25.00. Factory 2 and now factory 3 are marginal; there is no unused factory.
• At 0.67 < v < 1.00, for customers at B, factory 4 becomes a lower cost alterna-
tive to factory 1. Over this interval of v, the shadow price on capacity falls linearly
from 1 to 0 at factory 1, remains at zero for factories 2 and 3, and remains at 1
for factory 4. The shadow price on demand rises linearly everywhere, from 25
to 26 at A, and 25.67 to 27.00 at B and C. At v = 1.00, the firm can do no
better than to ship 3,000 units to C—1,700 from factory 3 and 1,300 from factory
4—at a unit cost of 27.00 and 26.00, respectively, 1,000 units to B—800 units
from factory 1 and 200 units from factory 4—at a unit cost of 27.00 and 26.00,
respectively, and 200 units to A from factory 1 at a cost of 26.00. Factory 3 is
marginal; factory 2 is unused.
shipments get determined. By incorporating local capacity constraints into the anal-
ysis, the Hitchcock–Koopmans Model introduces a new dimension to thinking about
the geographic structuring of markets. Indeed, because it is possible that the cost-
minimizing firm may be able to forego production altogether at one or more of its
factories, the Model, like that presented in Chapter 2, makes it possible to begin to
think about where to locate as well as how much to produce and ship.
At the same time, the model reminds us of the importance of opportunity costs:
the shadow price of the capacity constraint at each factory. That shadow price
reminds us that we cannot, in general, look at one factory in isolation. Which cus-
tomers will be served by a factory depends in part on the other factories the firm
has available to serve these and other customer places. At best then, the existence of
shadow prices implies that the set of customers served by just one factory constitutes
a submarket. I return to this matter in Chapter 8.
Unfortunately, the Hitchcock–Koopmans Model is also of limited value in think-
ing about markets for three additional reasons. First is the fact that it considers just
the factories that belong to one firm (supplier) whereas a competitive market might
better be thought to be composed of numerous buyers and sellers. The second is that
the Hitchcock–Koopmans model, unlike the models to be presented in Chapter 4,
does not directly consider price. Without reference to the price of soap being sold,
it is difficult to know how one might delineate a market for soap. A different way
of posing the same question is to ask whether there is spatial price equilibrium
across places in this Model? I simply cannot answer such a question with this
Model. The third reason is that, because it ignores fixed costs including the invest-
ment in plant and equipment, the Hitchcock–Koopmans Model cannot be used to
look at the prospects for building a new factory; it looks only at the efficient use of
already-existing factories.
31 See, for example, Magnanti, Mirchandani, and Vachani (1995) and Mayer and Sinai (2003).
3.8 Final Comments 95
by their customers and therefore have nothing to say about the well-being of
consumers.
Finally, the models in this chapter are silent on local cost advantage. We do know
that if Factory i has a relatively low Ci it will be efficient to use that factory to supply
customers nearby. Place i might, for example, have a low Ci because it is larger and
takes advantage of an economy of scale or division of labor (e.g., when the firm
assembles its most efficient staff at the factory sites that will service the bulk of
demand). However, there are other possibilities too. Place i might, for example, have
access to inexpensive inputs for production. Until we know what it is that makes
some factories more efficient than others, it is difficult to infer how localization and
prices in general are linked.
Chapter 4
The Struggling Masses
Perfect Competition at Two Places
(Cournot–Samuelson–Enke Problem)
Two places, in isolation from the rest of the world, each meet the requirements of
a perfectly competitive market. In Model 4A, the unit shipping cost is prohibitive.
Each place is in autarky. Price locally reflects only local demand and local supply.
However, if the unit shipping cost is low, arbitrageurs purchase where price is low
for resale at the other place. In Model 4B, unit shipping cost is zero everywhere, and
there is a common equilibrium price at the two places. A change in any parameter of
local demand or local supply at either place can affect this price. In Model 4C, ship-
ping cost is neither prohibitive nor zero. Here, shipping occurs up the price gradient.
Because of the actions of arbitrageurs, the price difference between the two places
shrinks to the unit shipping cost. Corner solutions—in which either demand or sup-
ply drops to zero in one or the other of the two places—are solved and interpreted.
The models in this chapter are the competitive market equivalent of the models of a
monopolist in Chapter 2. However, as in Chapter 3, congestion in production means
that unit cost rises the more output the industry produces. Usually, we imagine that
competition causes excess profit to disappear. However, the notion of congestion
(an upward sloped supply curve) here in Chapter 4 means that some producers are
less efficient than others. More efficient suppliers earn a monopoly profit even in
competitive markets. In this chapter, localization of production, prices (one for each
place), and excess profit (for all but the marginal producer) are joint outcomes of a
competitive market.
In this chapter, I explore the localization of firms from the perspective of demand
curve, supply curve, and competitive market. In Chapter 2, I introduced local
demand; in this chapter, I now consider demand by arbitrageurs buying in a mar-
ket (or submarket) where the price is low for resale elsewhere where the price is
high; I treat this external demand as endogenously determined. This chapter comes
early in this book because it, as does Chapter 2, starts from material familiar to
anyone who has taken a first course in Economics.
When we look at a competitive market, our focus shifts from the firm to the
industry as a whole. With the monopolist in Chapter 2, I could examine directly
the potential return to capital invested in a factory at each location. In this chapter, I
rely instead on the notion of local supply.1 The local supply curve tells us how much
more local producers would make available in total if the local price were higher.
Put differently, the local supply curve is an ordering of local production from least
costly to more costly. See the right-hand side of Fig. 4.1.
A A
I H D
Price
Price
G F
G E
D
0 0
F C Quantity B E Quantity B
Chapter 2: Firm maximizes Chapter 4: Industry in
profit competitive equilibrium
AB Demand curve ABDemand curve
AC Marginal revenue CDSupply curve
DE Firm’s marginal cost curve CGFCProducer surplus
DIHGD Firm’s semi-net revenue GAFGConsumer surplus
ODGFO Variable cost incurred by firm OCFEOProducer cost
OF Quantity supplied by the firm OEQuantity supplied by the
OI Price received by firm industry
OIHFO Firm’s revenue OG Equilibrium price
This chapter distinguishes between local supply (the quantity offered by local
producers) and external supply (the quantity offered by traders who bring a com-
modity produced elsewhere into the local market). The chapter then uses the relative
size of external supply and external demand (the quantity demanded by traders for
resale elsewhere) to draw conclusions about the impacts of a change in market
conditions on the localization of production.
In Chapter 2, I looked at how a monopolist prices product and decides where
to produce (i.e., invest in factories). My treatment of market in that chapter was
asymmetric: there were many demanders (customers) but only one supplier (the
firm). The firm gets a lower price per unit the more the quantity it supplies to the
market. To maximize profit, the firm there chose the quantity Q where marginal
1 In a region, the supply of a product by firms from local production. Local supply does not include
supply offered by arbitrageurs importing from another region.
4.1 The Cournot–Samuelson–Enke Problem 99
revenue is equal to marginal cost, then sells it in the market at price P. See the
left-hand side of Fig. 4.1.
In this chapter, I consider a local industry consisting of a large number of firms
each supplying the same product. I say large here to invoke the notion of a com-
petitive market. Economists describe a competitive market as one with a sufficient
number of participants so that no one individually affects price: neither a deman-
der the price paid by varying the quantity they individually demand, nor suppliers
the price received by varying the quantity they individually supply. Each firm (and
consumer) individually is a price taker; equivalently the market exemplifies perfect
competition.2 Customers simply bypass suppliers with a higher price in favor of a
supplier at the market price. In the same way, suppliers ignore a customer willing to
pay only a lower price in favor of a customer willing to pay the market price.
I think of the local supply curve as the industry marginal cost curve3 there. Of
course, I can measure the industry marginal cost curve over either the short run or
the longer run. I assume in what follows that we are looking at the industry marginal
cost curve over a longer run in which capacity adjustments are possible. We want to
think about situations where unit shipping cost changes and affects the amount of a
commodity produced at a given place. Presumably new firms start up at some places
while other firms die off at others. However, we should be careful here. Over the
longer term, other things might be happening that also affect the local supply curve:
prices of inputs, technological change, and firm reorganization to take advantage of
economies of scale or Adam Smith’s idea that the division of labor is limited by the
extent of the market. Put simply, his idea was that the greater the output (scale) of
a firm the better able it is to take advantage of the added productivity from having
specialized labor. These factors can cause the supply curve to shift or twist. In the
relatively simple models used in this chapter, I don’t want to complicate the analysis
with such considerations.
In drawing a local supply curve, I am envisaging free entry of firms. There is
nothing to prevent new firms from entering the industry other than for considerations
of profit. If demand shifts so as to cause a rise in market price, new firms will keep
entering the industry until the price is driven down to the point where a potential
new firm finds that it is no longer profitable to get into that industry. In competitive
location theory, the concept of free entry is often associated with the work of Lösch
on imperfect competition in space. However, free entry is also central to perfect
competition in general and to the supply curve over the long run in particular.
In this chapter, I introduce models of spatial price equilibrium. In these mod-
els, assume all firms are competitive. There is no monopolist here pricing among
these markets to maximize profit. Instead, the emphasis is on the role of competition
2 An attribute of a market wherein each supplier and each demander is a price taker.
3A schedule showing the marginal cost to an industry (usually over the longer run where capi-
tal invested adjusts as needed) as a function of the quantity to be supplied. The market supply
curve is thought to be the same as the industry marginal cost curve. We can measure the industry
marginal cost curve over either the short run (no additional competitors or factories) or the long
run (additional competitors and/or factories possible).
100 4 The Struggling Masses
among firms and traders (arbitrageurs) who are each a price taker in markets for their
commodity. Location theory includes a substantial subfield with a long history that
deals with spatial price equilibrium.4 Early work in this area starts with Cournot5 in
1838 and includes Enke (1951) and Samuelson (1952).6 I first came across the price
difference curve solution method in Samuelson (1952) and hence refer to it here as
the Cournot–Samuelson–Enke Model: Model 4C later in this chapter. However, let
me first introduce two models as foils. Model 4B shows the solutions when unit
shipping rate is zero. Model 4A shows spatial price equilibrium in autarky; we can
think of autarky as a situation arising when unit shipping cost is prohibitive. Model
4C can be thought therefore to lie between Models 4A and 4B: unit shipping cost
larger than zero but not large enough to bring about autarky.
4 In the case of multiple places, a condition in which arbitrageurs have no further incentive to
purchase in a low-price market for resale in a high-price market.
5 See Cournot (1960, chap. 10).
6 Early writers in the field also mention an unpublished paper by William Baumol—dated 1952
and entitled Spatial Equilibrium With Supply Points Separated From Markets and Supplies
Predetermined—that might be similar to Samuelson (1952).
7 A supply function is generally expressed as a schedule of quantity supplied (Q) at various prices
(P): i.e., Q = g[P]. An inverse demand function rearranges this as the price needed by suppliers at
the margin in order to ensure that a given quantity is supplied to the market: i.e., P = g−1 [Q].
8 See (4.1.2) for Place 1.
4.2 Model 4A: Autarky 101
Notes: Rationale for localization (see Appendix A): Z3—Implicit unit price advantage at some
locales; Z8—Prohibitive shipping cost. Givens (parameter or exogenous): Ci —Intercept of inverse
supply curve at place i; Ni —Population of place i; α—Intercept of individual linear inverse demand
curve; β—Slope of individual inverse demand curve; δ i —Slope of inverse supply curve at place i.
Outcomes (endogenous): Pi —Price at Place i; Qi —Equilibrium quantity transacted at Place i.
G1 E1
C1
0
F1 B1 Quantity
a price given by (2.1.5) and a quantity given by (2.1.6). For ease of comparison,
assume now a horizontal supply curve here (e.g., δ1 = 0 in the case of Place 1).
There, autarky price in competitive equilibrium reduces to P1 = C1 , which is lower
than the price (2.1.5) charged by the monopolist. I have indicated this in Fig. 4.1
using a competitive price (OG) lower than the monopolist’s price, (0l). By impli-
cation, as shown in Fig. 4.1, the quantity transacted in market equilibrium (OE) is
greater than the quantity sold by the monopolist (OF) of Chapter 2. From (4.1.4),
I get Q1 = N1 (α − C1 )/β when δ 1 approaches zero, which is twice the quantity a
profit-maximizing monopolist would supply to the market (2.1.6). The competitive
market envisaged here is different from the monopolist of Chapter 2. The monop-
olist exploits the downward sloping demand curve to set marginal revenue equal to
marginal cost, and so supplies a smaller quantity and gets a higher price than would
a competitive firm. That the monopolist supplies exactly one-half of the amount
supplied in perfect competition is an outcome of the assumed linearity of demand;
nonetheless it illustrates the idea that the monopolist will supply less and get a higher
price than will firms that are perfectly competitive.
In what ways does the treatment of costs in this chapter differ from Chapter 2?
• In Chapter 2, I used C to denote the (constant) marginal cost. I also argued there
the firm would not normally supply product to the market, even in the short run,
4.2 Model 4A: Autarky 103
if the price of that product was below C. In the long run, price would have to be
higher to provide a return on invested capital as well. In this chapter, I interpret C1
as the price below which no quantity would be supplied to the market. Because I
think of the local supply curve as a long run marginal cost curve, C1 presumably
sums a marginal cost like C and a normal return on invested capital and any other
unpriced factors of production for the most efficient location.
• The industry marginal cost curve is not the same as a firm’s marginal cost curve.
What are key differences? For one, the firm’s marginal cost curve is usually
thought of in the short term only. Above, I argue that we look at the industry
marginal cost curve over the long run.9 A second difference is that, in Chapter 2,
the firm’s marginal cost does not include unit fixed costs. However, when I draw
the industry supply curve here, I must assume a price sufficient to keep firms in
the industry over the long term. Therefore, the industry supply curve, and hence
the industry marginal cost curve, must include unit fixed cost (at least over the
long term). A third difference is that, unlike Chapter 2, the link between profit
and costs is not clear. In Chapter 2, I assumed the firm was a price taker10 in the
markets for its inputs: the opportunity cost of capital r and the marginal cost of
production C were each simply a fixed expenditure required (per unit of K and
Q, respectively) that reduces profit. Here in Chapter 4, the supply curve tells us
about the quantity forthcoming from local suppliers at any given price, but this
does not say anything directly about input or unit costs like r and C. Nonetheless,
in interpreting the supply curve, it is helpful to assume that each supplier is an
efficient firm.
• In this chapter, I assume the supply curve (hence costs) may differ from one local
market to the next. In Chapter 2, I had assumed the firm faced the same costs of
production at each place.
• In this chapter, I assume the local supply curve is upward (positively) sloped. In
Chapter 2, I had assumed that the firm, by investing K, obtained an unlimited
wellspring of production at a constant marginal cost, C. The upward sloped sup-
ply curve is different from this. For some reason, it now costs more per unit to
induce a greater supply.
To me, the approach of this chapter raises three questions.
• Why is the local supply curve upward sloped? In this chapter, I assume unit cost
increases with level of output in the market. This is in contrast to the constant
9 Over the short term, the firm is not able to adjust its capital stock. In the short term, therefore,
the firm’s marginal cost rises as it attempts to increase the level of production, as the firm encoun-
ters congestion and capacity limitations. Of course, this is not what happens in Chapter 2 where
I assumed marginal cost is constant; the unlimited wellspring assumption allowed us to ignore
questions about the relationship between output and capital stock.
10 A condition under which a market participant (supplier or demander) is unable to affect the price
they receive or pay for a unit of the product by varying the quantity that they supply or demand. The
supplier (demander) sees the demand (supply) for its product as horizontal: i.e., infinitely elastic at
the given market price.
104 4 The Struggling Masses
marginal cost assumed in Chapter 2. In effect, I now allow for the possibility of
(1) congestion at the firm level that causes unit cost to rise eventually the more
output it produces, (2) competition among firms for the scarce resources they
need to produce more output, and (3) differences among firms that make some
more efficient than others. Why does this happen? In effect, the firms individu-
ally or the industry collectively experience congestion. That is, they are unable
to increase production as an industry without incurring a higher marginal cost
of production. In the standard introduction to competitive equilibrium in a first
course in Economics, it is typically argued that an industry is made up of iden-
tical firms with identical costs. As demand for the industry’s product increases,
the standard introduction envisages additional identical firms entering the indus-
try. In that case, the long run supply curve (industry marginal cost curve) is a
horizontal line. If every firm looked like the firm in Chapter 2 for example, then
local market price, P, would approach the marginal cost of production, C. In other
words, over the long term, the models in Chapter 2 ignore the congestion (within
the firm due to its structure and physical plant) and the competition for scarce
resources among firms that arises from producing more output from a given set
of firms using their existing factories and assume that all output needed can be
produced at the same unit cost.
• Where does a supply curve come from? In this chapter, I assume a local supply
curve that shows the amount forthcoming from local suppliers at each place con-
tingent on the market price there alone. This is an assumption made to simplify
the model. In fact, when an entrepreneur—like the firm in Chapter 2—is deciding
when and where to build a factory, it takes into account the price of the commod-
ity locally as well as the price that might be possible for a factory built elsewhere.
Put differently, in a world of two places, the supply forthcoming locally from any
place will depend on the prices at both places. In Chapter 10, I present a model,
which incorporates joint aspects of supply.
• Why do costs differ from market to market? Are costs that differ from market to
market inherently geographic (as in variations in labor or other factor costs) or
related to the efficiency or productivity of the firms that just happen to be found
in that market? In competitive equilibrium, as the number of consumers (N1 )
increases, autarky price increases. This is because more output will be demanded
at any price (below α) when the market is larger, but in autarky this can only
be obtained by moving to the right along the local supply curve and therefore
inducing more costly suppliers to enter the market.11 What makes the solution
differ here in part is that, in Chapter 2, the monopolist had a constant marginal
cost whereas here marginal cost increases with the level of output.12 In a compet-
itive market, excess profit disappears for the marginal supplier in equilibrium. In
11 This is different from the case of the monopolist in Chapter 2 where, in the case of a single
market, price was not affected by market size: see (2.1.5).
12 Here, I can approximate the condition of constant marginal costs by letting δ approach zero.
1
In that case as noted above, (3.2.3) implies P1 approaches C1 , and market size indeed no longer
affects price in a perfectly competitive market.
4.2 Model 4A: Autarky 105
other words, the supplier in the market with the highest unit cost—in this sense
the marginal producer—earns only normal profit. However, the upward slope of
the supply curve tells us some producers are more efficient and therefore do earn
excess profit even though the marginal producer does not. Whatever the particu-
lar advantage of the more efficient firm, that advantage gives rise to a Ricardian
rent.
In this model, I have assumed all consumers in the two markets have the same
individual demand curve for the commodity. Because of that assumption, differ-
ences in equilibria between the two markets under autarky do not arise because
consumers are different. However, differences can arise because of differences in
market sizes (N1 vs. N2 ) or the shapes of the supply curves (C1 vs. C2 or δ 1 vs. δ 2 ).
We can now imagine an autarky solution for Place 2. See the example in Fig. 4.3
where I have assumed Place 2 is smaller (demand curve AB2 ) and less efficient
(supply curve C2 D2 ) than Place 1 (reproduced from Fig. 4.2).
Suppose, for example, these two markets in autarky are similar in that C1 =
C2 and δ1 = δ2 , but that N2 > N1 . Here, at any given market price, the quantity
A D2
G2 E2
Price
C2
0
F2 B2 Quantity
demanded at Place 2 is greater than at Place 1. To coax that larger supply locally,
the autarky price at Place 2 would have to be larger than at Place 1 on the assumption
that δ1 = δ2 > 0. Put differently, the diseconomies of scale (i.e., the upward-sloping
supply curves) assumed in this model means that the price of the commodity will be
higher in a large market compared to a small market.
Now, instead, suppose the two markets are similar in that δ1 = δ2 and N1 = N2 ,
but that now C2 > C1 . In this case, the supply curve for Place 2 will be parallel to,
and lie above, the supply curve for Place 1. Therefore, P2 will be higher than P1 in
equilibrium under autarky.
Finally, suppose the two markets are similar in that C1 = C2 and N1 = N2 , but
that now δ2 > δ1 . In this case, the supply curve for Place 2, will be steeper than the
supply curve for Place 1 and the two will intersect at the Y-axis (since C1 = C2 )
Therefore, in equilibrium, P2 will once again be higher than P1 under autarky.
When prices differ between the two markets in autarky, the possibility of ship-
ments arises. Shipment of the commodity will occur, if at all, only from the market
with the lower price in autarky to the market with the higher price. There would
be no incentive to ship in the opposite direction. Note the importance here of our
assumption that all firms produce an identical product. If suppliers at Place 2 were
to produce a similar, but not identical, product then consumers at Place 1 might still
want to consume some of it even if P2 > P1 in autarky.
In principle, shipment is not the only way in which the two markets might be
affected by this difference in autarky prices. To the extent that firms are more effi-
cient than others because of an internal advantage (e.g., more skilled management),
they might be able to increase profits by shifting production into the market with
the higher autarky price. Similarly, consumers might be attracted to relocate into
the market with the lower price. However, in this chapter, we ignore movements of
firms or consumers between the two places to focus on the role that might be played
by the shipment of commodities.
Before I do this, what about comparative statics in Model 4A? Here, I present
results for Place 1 only; the story would be the same for Place 2. In this autarky
model, there are five givens (α, β, C1 , δ 1 , and N1 ) and two outcomes (P1 and Q1 ).
Remember here the implicit role of prices everywhere : consumer income and the
prices of consumer commodities shape α and β; input prices help shape the local
supply curve, hence C1 and δ 1 .13 As in Model 2A, comparative statics are easy
to do because I have explicit solutions for P1 and Q1 . See (4.1.3) and (4.1.4) and
Table 4.2.
C1 If C1 is increased, the supply curve shifts upward. Therefore, market equilib-
rium runs up the demand curve: P1 increases and Q1 decreases.
N1 When N1 is increased, the aggregate demand curve becomes flatter, that is,
sweeps counterclockwise about (0, α). Therefore, market equilibrium runs up
the supply curve: P1 and Q1 both increase.
13 To the extent that population adjusts so as enable each resident to be best off, N may also be
1
shaped by prices. I do not pursue that idea further here. See Chapters 11 and 12.
4.3 Model 4B: Integrated Market Solution: Zero Shipping Cost 107
In autarky, the price of a product may vary from one local market to the next. As
a thought experiment, imagine that the local demand and local supply curves are
identical in the two markets, so that the equilibrium prices are the same. Now imag-
ine a shift in either the demand or supply curve in one of the two markets. That shift
alone would generally be sufficient to cause the equilibrium price in that market to
differ from the other market.
If the difference in prices is sufficiently large, arbitrageurs will be encouraged
to purchase in the lower price market for resale in the other market. Commonly, we
think of an arbitrageur as a professional who trades only in a commodity: i.e., neither
produces the commodity nor purchases it for own consumption or use. However,
when producers look for a higher priced market in which to sell their commodity,
or consumers scour markets looking for the lowest price, their behavior is similar in
effect to that of the professional trader. In keeping with the treatment in Chapter 2,
I use arbitrage to refer to behavior by traders, producers, and consumers that leads
to the flow of a commodity from one local market to another.
As in Chapter 22, assume Place 1 is x km away from Place 2, and the cost of ship-
ping a unit of product a distance of 1 km is a constant and directionally symmetric s
dollars. Therefore, the unit shipping cost from Place 1 to Place 2, or vice versa, is sx
108 4 The Struggling Masses
per unit shipped. In keeping with the treatment of cost elsewhere in this book, I take
the cost of shipping to include any normal profit associated with arbitrage. In this
chapter, and throughout the book, I assume for simplicity that the cost of arbitrage
is the same for every potential trader.14
For the moment, assume an integrated market.15 In effect, the action of arbi-
trageurs is to convert two markets into a single unified market on the assumption that
sx = 0. To see the consequences of this supposition, ignore the arbitrageurs for the
moment, and instead simply aggregate first the demand curves and then the supply
curves for the two markets by summing horizontally. In fact, this is the consequence
of arbitrage. See Fig. 4.4.
In the case of the two demand curves, horizontal aggregation is relatively easy
since I have assumed that the demand curves in the two markets have the same
Y-intercept (i.e., α). Rearranging (4.1.1) under the assumption that P1 = P yields
Q1 = aN1 /β + N1 P/β. A similar rearrangement of the demand at Place 2 yields
Q2 = aN2 /β + N2 P/β. Summing these two equations yields Q1 + Q2 =
α(N1 + N2 )/β + (N1 + N2 )P/β. Upon rearrangement, the combined demand at
Places 1 and 2 is given by (4.3.1). See Table 4.3. In Fig. 4.4, the two local demand
curves (AB1 from Fig. 4.2 and AB2 from Fig. 4.3) sum to the aggregate demand
curve ABt .
The two supply curves can also be aggregated horizontally. However, computa-
tional complexity arises here because the two local supply curves need not have the
same Y-intercept. In general, the aggregate supply curve will be kinked. Without loss
of generality, let Place 1 be such that C1 < C2 . The aggregate supply curve can now
be written as (4.3.2), (4.3.3), and (4.3.4) depending on the level of P. Under (4.3.2),
there is no production since P is not high enough for production even among the
lower cost suppliers at Place 1. If P is high enough to enable production at Place
1, but not high enough for production at Place 2, (4.3.3) ensues. Equation (4.3.4)
applies when it is feasible to produce at both Places 1 and 2. Of these, only (4.3.3)
and (4.3.4) need concern us here. In Fig. 4.4, the two local supply curves (C1 D1
from Fig. 4.2 and C2 D2 from Fig. 4.3) sum to the aggregate kinked supply curve
C1 Ct Dt .
The integrated market equilibrium price corresponding to (4.3.1) and (4.3.3)—
using only suppliers from Place 1—is given by (4.3.5) and that corresponding to
(4.3.1) and (4.3.4)—using suppliers from both markets—is given by (4.3.6). In
Fig. 4.4, equilibrium quantity is OFt and equilibrium price is OGt . Local suppliers
at Place 1 supply OH units and local customers there demand OI units.
Consider a numerical example wherein the two markets differ only in that
C2 > C1 : namely, α = 15, β = 1, N1 = N2 = 1,000, C1 = 10, C2 = 11, and
δ1 = δ2 = 0.001. See first the autarky solutions presented in Table 4.5. From
14 See Anderson and Ginsburgh (1999) for an analysis of the case where the cost of arbitrage is
different between firms and consumers.
15 A condition of two places arising when the cost of shipping product from one market to the other
is zero.
4.3 Model 4B: Integrated Market Solution: Zero Shipping Cost 109
Gt Et
C2 Ct
C1
0
I H Ft B2 B1 Quantity Bt
Fig. 4.4 Model 4B: equilibrium in integrated market with zero shipping cost.
Notes: This example assumes same market parameters as Figs. 4.2 and 4.3. At equilibrium in
integrated market, Pc = 10.00 and Qc = 9,500. Of this total, consumers at Place 1 demand 5,000
units (D1c ), and firms at Place 1 supply 8,000 units (S1c ); the remainder is demanded or supplied
at Place 2. Therefore, 3,000 units are shipped from the lower priced Place 1 to the higher priced
Place 2. Horizontal axis scaled from 0 to 30,000; vertical from 0 to 22
(4.1.3), the autarky price at Place 1 is 12.50. From (4.3.6), the integrated market
price (P) is 12.75; since P here is larger than C2 , production occurs at both places.
In the integrated market, as calculated from (4.1.2), firms at Place 1 produce 22,750
units: up from the 22,500 units produced under the autarky price. They (including
any arbitrageurs who buy from them) are better off because they can now sell their
product in the higher price market. Consumers at Place 1 are worse off in the inte-
grated market since they now pay 12.75 a unit compared to 12.50 previously. As
calculated from (4.1.1) their demand in the integrated market is only 2,250 units:
down from 2,500 units under autarky. The excess supply16 at Place 1 (the amount
16 In a market at a given price P, excess supply is the amount if any by which local supply exceeds
local demand. Local here excludes demand or supply by arbitrageurs.
110 4 The Struggling Masses
Table 4.3 Model 4B: competitive equilibrium in integrated markets (zero shipping costs)
assuming C1 < C2
Notes: Rationale for localization (see Appendix A): Z3—Implicit unit price advantage at some
locales. Givens (parameter or exogenous): Ci —Intercept of inverse supply curve at place i;
Ni —Population of place i; α—Intercept of individual linear inverse demand curve; β—Slope
of individual inverse demand curve; δ i —Slope of inverse supply curve at Place i. Outcomes
(endogenous): P—Price (same at both places); Qi —Equilibrium quantity supplied by Place i;
Q—Aggregate quantity.
by which local supply exceeds local demand, local here excluding demand or supply
by arbitrageurs) is 2,750 − 2,250 = 500 units. This is the amount that arbitrageurs
purchase at Place 1 for resale at Place 2. At the more costly Place 2, the autarky
price is 13.00. In the integrated market, the firms of Place 2 produce 1,750 units:
down from 2,000 units under autarky. They are worse off in the integrated market
compared to autarky because they have lost customers to the more efficient pro-
ducers from Place 1. At the same time, consumers at Place 2 benefit because, for
them, the integrated market price is lower than in autarky. Consumers now demand
2,250 units: up from 2,000 units under autarky. The excess demand17 at Place 2 (the
amount by which local demand exceeds local supply, local here excluding demand
or supply by arbitrageurs), 2,250 − 1,750 = 500, of course, is exactly the amount
reallocated from Place 1 to Place 2.
I label demand by consumers at a place as local demand in Table 4.5; supply
by producers at that market is called local supply. This serves to differentiate such
demand and supply from the (nonlocal) demand and supply attributable to arbi-
trageurs. It is interesting to note that I have drawn neither a demand curve for
17 Ina market at a given price P, excess demand is the amount, if any, by which local demand
exceeds local supply. Local here excludes demand or supply by arbitrageurs.
4.3 Model 4B: Integrated Market Solution: Zero Shipping Cost 111
arbitrageurs in the market with the low autarky price (here Place 1) nor a supply
curve for them in the market with the high autarky price (here Place 2). This is
because the demand (and supply) of arbitrageurs is a derived demand (and supply);
its level can be solved after netting out local demand and local supply at each place.
Before leaving the integrated market solution, let me make a final comment about
the kink in the supply curve noted above. To some, including many of my mathe-
matical friends, kinks are ugly. At the least, having to solve both (4.3.5) and (4.3.6)
before I can figure out the market price is tedious. In the standard versions of most of
the basic models in Economics, kinks are nowhere to be found. So, what makes the
integrated model so different? At the essence of the kink is the notion that geogra-
phy here is both distinct and differentiated. Being punctiform, it is distinct; there are
two markets between which consumers and producers themselves do not relocate.
For example, the number of consumers at each place is assumed fixed (N1 and N2 )
even though you might think consumers would be motivated to reside in a market
with a lower price, other things being equal. Second, geography is differentiated on
the supply side here; the supply curve parameters at Place 1 are different from those
at Place 2. Why should there be a difference? What prevents firms from carrying
production technologies and efficiencies from one market to the next even though
they can freely ship the commodity between markets?
To help spur thinking here, assume for the moment that every producer in both
markets was equally efficient. We can get as much output as we want in either mar-
ket at the same unit cost: just as I did in the two-factory solution in Chapter 2.
Under these circumstances, the supply curve at each place would be horizontal and
the equilibrium price at each place would be the same regardless of the population
in that market. In contrast, however, this chapter makes two different assumptions
about the local supply curve.
1. Upward-sloped local supply curve at each place. As a result of congestion, the
marginal unit becomes more costly to produce as quantity is produced in total
locally. One aspect of this is an inputs argument that follows from the idea that
a product is manufactured from material inputs. When the level of output is low,
those material inputs can be obtained nearby. When the level of output is higher,
the material inputs must be purchased (for some reason) from further afield,
hence a higher unit cost. Implicit here is the idea that inputs are not ubiquitous.
Another is an entrepreneur argument based on the idea that each unit of output is
produced by an entrepreneur and that some entrepreneurs are more skilled than
others. When quantity produced is small, only the most efficient entrepreneurs
enter the market, and unit cost is low. As the demand for the commodity rises,
less efficient entrepreneurs are drawn into the market, driving up marginal cost
and hence price.
2. Supply curve intercepts, C1 and C2 , are not necessarily equal. Put differently, the
most efficient firm at Place 1 and the most efficient firm at Place 2 have different
unit costs. Perhaps this is because of the inputs or entrepreneur arguments above.
In the integrated market version of the problem, these assumptions are not par-
ticularly bothersome. After all, in the absence of shipping costs on the commodity,
112 4 The Struggling Masses
why should it matter where or how a firm produces? However, as we will see shortly,
vexing questions come to the fore when shipping costs are nonzero.
Before I do this, what about the comparative statics of Model 4B. In the inte-
grated market solution, there are eight givens (α, β, C1 , C2 , δ1 , δ2 , N1 , and N2 ) and
two endogenous variables (P and Q). Let us look here at the comparative statics
only in the cases where C1 < P < C2 and where P > C2 ; in other words, I will
ignore knife-edge cases (wherein P shifts from below C2 to above C2 ). The explicit
solutions for P and Q in (4.3.3), (4.3.4), and (4.3.6) make it relatively easy to do
comparative statics here. See Table 4.4.
C1 + −
C2 + −
N1 + +
N2 + +
α + +
β − −
δ1 + −
δ2 + −
C1 If C1 is increased, the supply curve shifts upward parallel and market equi-
librium runs up the demand curve. Price rises, while the quantity transacted
drops. The results are the same whether C1 < P < C2 or P > C2 .
C2 If C1 < P < C2 , an increase in C2 has no effect on the outcome of the model,
since production facilities at Place 2 are not being used. On the other hand,
suppose. If C2 is increased here, the supply curve shifts upward parallel and
market equilibrium runs up the demand curve. Price rises, while the quantity
transacted drops.
N1 When N1 is increased, the aggregate demand curve becomes flatter; that is,
sweeps counterclockwise about (0, α). It traces out the supply curve. Both P1
and Q1 increase.
N2 When N2 is increased, the aggregate demand curve becomes flatter—that is,
sweeps counterclockwise about (0, α). It traces out the supply curve. Both P1
and Q1 increase.
α If α is increased, the demand curve shifts upward and market equilibrium runs
up the supply curve. Price rises; so too does the quantity transacted. The results
are the same whether C1 < P < C2 or P > C2 .
4.4 Model 4C: Spatial Price Equilibrium with Shipping Costs 113
18 As
in Chapter 2, I assume here no congestion over the shipping network. The firm can ship as
much, or as little, as it likes for the same unit cost sx.
114 4 The Struggling Masses
of the shipping rate. Indeed that is the case. See the second numerical example
presented in Fig. 4.5. The vertical difference between the excess supply curve for
Place 1 and the excess demand curve for Place 2 is called a price difference curve19
(curve ABCD in Fig. 4.5), because it shows the quantity that must be traded by
arbitrageurs to result in a given difference in prices between the two markets. For
example, where the shipping rate in Fig. 4.5 is zero, the integrated market solution
obtains. This would result in arbitrageurs purchasing 1,353 units at Place 1 (amount
OJ in Fig. 4.5) for resale at Place 2. Such activity would cause the price at Place 1
to rise (because of increased demand) from OS to OL in Fig. 4.5 and cause the price
at Place 2 to fall (because of increased supply) from OR to OL in Fig. 4.5 such that
the price in both markets becomes 10.
The linear nature of the price difference curve is easily confirmed by the
following procedure
1. Derive the excess supply schedule for Place 1. See (4.6.1) in Table 4.6 and
the polyline labeled E1 F1 G1 H1 in Fig. 4.5. This schedule shows the amount by
19 In a two-region model of trade, the Price Difference Curve shows the amount of the good shipped
from the lower priced to the higher priced region that would result in a given difference in prices
between the two regions. Calculated as the vertical difference between the excess supply curve in
the lower price region and the excess demand curve in the higher price region. In this text, I refer
to this derivation as the Samuelson Model. There are variants of this approach that are essentially
the same: see Siebert (1969, pp. 85–87) or Takayama and Judge (1971, pp. 135–137).
4.4 Model 4C: Spatial Price Equilibrium with Shipping Costs 115
$ ABCD
E1F1G1H1
Price difference curve
Excess supply curve for Place 1
E2F2G2H2 Excess demand curve for Place 2
OI Quantity shipped from Place 1 to Place 2
OJ Quantity shipped in integrated market (Model 4B)
OL Price in integrated market (Model 4B)
OM Equilibrium price at Place 1
ON Equilibrium price at Place 2
E2 OP Unit shipping cost
H1
F2 Q
G1
R
M
L
N
B S
G2
T
V
F1
U
P
E1 0 IJ H2 Shipment
Fig. 4.5 Model 4C: excess supply (or demand) curve and price difference curve.
Notes: α = 15, β = 1, δ1 = 0.001, δ2 = 0.002, C1 = 2, C2 = 7, N1 = 1,000, N2 = 900,
s = 0.013, x = 100. With arbitrage, P1 = 9.46, P2 = 10.76, and quantity shipped from Place 1 to
Place 2 is 1,929. Horizontal axis scaled from –15,000 to 25,000; vertical from –8 to 22
which local supply exceeds demand (negative, if local demand larger than sup-
ply) at any given price. It is calculated by reexpressing Q1 as D1 in the demand
equation (4.1.1), Q1 as S1 in the supply equation (4.1.2), rearranging the two
equations to put quantity as a function of price, and then subtracting the demand
equation from the supply equation.
2. For Place 2, calculate the excess demand schedule in a similar manner. See
(4.6.5) and the polyline labeled E2 F2 G2 H2 in Fig. 4.5.
3. Rearrange the excess supply curve at Place 1 and excess demand curve at Place
2 each to put price on the left and excess supply (or demand) on the right. See
(4.6.2) and (4.6.6).
4. Take (4.6.2) minus (4.6.6) under the assumption that ES1 = ED2 to get the price
difference as a linear function of excess supply at Place 1. The result is (4.6.9).
The price difference (P2 − P1 ) is now a linear function of excess supply (ES1 ).
116 4 The Struggling Masses
Table 4.6 Model 4C: price difference curve (assuming C1 < P1 < α and C2 < P2 < α and
product shipped only from Place 2 to Place 1
Where
a2 = (C2 β + αδ2 N2 )/(β + δ2 N2 ) (4.6.3)
b2 = (βδ2 )/(β + δ2 N2 ) (4.6.4)
Excess demand at Place 1
ED1 = (C1 /δ1 + αN1 /β) − (1/δ1 + N1 /β)P1 (4.6.5)
Re-arranging (4.6.5) yields
P1 = a1 − b1 ED1 (4.6.6)
Where
a1 = (C1 β + αδ1 N1 )/(β + δ1 N1 ) (4.6.7)
b1 = (βδ1 )/(β + δ1 N1 ) (4.6.8)
Equation (4.6.2) minus equation (4.6.6) yields
P1 − P2 = (a1 − a2 ) − (b1 + b2 )ES1 (4.6.9)
Re-arranging (4.6.9) yields price difference curve
ES1 = (a1 − a2 )/(b1 + b2 ) − (1/(b1 + b2 ))(P1 − P2 ) (4.6.10)
Notes: Rationale for localization (see Appendix A): Z3—Implicit unit price advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): Ci —Intercept of
inverse supply curve at Place i; Ni —Population of Place i; α—Intercept of individual linear inverse
demand curve; β—Slope of individual inverse demand curve; δ i —Slope of inverse supply curve
at Place i. Outcomes (endogenous): EDi —Excess demand at Place i; ESj —Excess supply at Place
j; Pi —Price at Place i; Qi —Equilibrium quantity transacted at Place i.
In Fig. 4.5, note some basic characteristics of the excess supply, excess demand,
and price difference curves.
• Since I have assumed a common value for α, the excess supply and demand
curves have a kink in common at P = α, namely points F2 and G1 in Fig. 4.5.
• The excess demand curve for place 1 has a kink at P = C1 : namely point F1 .
Similarly, the excess supply curve for place 2 has a kink at P = C2 : namely point
G2 .
• Every kink in an excess demand (or excess supply) curve matches up vertically
with a kink in the price difference curve: e.g., point F2 with point B and point G2
with point C in Fig. 4.5. In Fig. 4.5, we do not see the kinks in the price difference
curve corresponding to points F1 and G1 simply because these are outside the area
graphed.
4.4 Model 4C: Spatial Price Equilibrium with Shipping Costs 117
We are now able to predict the shipment, if any, between the two places. First,
plot the unit shipping cost on the vertical axis of Fig. 4.5 (see point P): I use
sx = 1.30 to illustrate here. Move horizontally to the right until you cross the price
difference curve; if no intersection, then sx is too high to permit trade. Then, drop
vertically down Fig. 4.5 to the point (I) where we cross the horizontal axis. This
gives the amount (OI) shipped from Place 1 to Place 2. Then, extend the same ver-
tical line upwards until it crosses the Excess Demand Place 2 curve or the Excess
Supply Place 1 curve and move horizontally to the left until you reach the vertical
axis; this gives the spatial equilibrium price, P2 (corresponding to OM in Fig. 4.5)
or P1 (corresponding to ON), respectively. Equivalently, we can use (4.6.2) and
(4.6.6) to predict the price at each place. Calculated in this way, these prices, there-
fore, differ at most by the unit shipping cost: amounts MN and OP are the same in
Fig.4.5.
Let us return to the example in Table 4.5. There, I show in columns [4] and [5]
the solution to the problem when sx = 1.30, not zero as in the integrated market
solution, but at least smaller than the sx that would bring on autarky. In this case,
1,929 units are shipped, and the equilibrium prices are P1 = 9.46 and P2 = 10.76.
Compared to autarky, consumers at Place 1 are worse off while local producers are
better off. Compared to autarky, consumers at Place 2 are better off while local
producers are worse off.
However, there is an important caveat here. The price difference curve, as shown
in Fig. 4.5, is itself kinked, forming a polyline of up to 5 possible linear segments.20
These segments represent all feasible combinations of three levels of P1 —low
(P1 < C1 ), medium (C1 < P1 < α), and high (P1 > α) —and three levels of P2 —
low (P2 < C2 ), medium (C2 < P2 < α), and high (P2 > α). What is infeasible are
the combinations of (1) P1 and P2 simultaneously low since there is no production
anywhere, (2) P1 and P2 simultaneously high since there is no demand anywhere, or
any other combination where p1 < C1 since we have assumed C1 < C2 . Table 4.6
illustrates the calculations for just one of five possible cases.
20 Early writers largely ignored this. Enke (1951, p. 42), for example, assumes that excess supply
will always be linear in price, not piecewise linear as argued here. Samuelson (1952, pp. 286
and 288) draws excess supply curves and a price difference curve that are also kinked. However,
Samuelson also draws local demand and supply curves that are kinked without any explanation
and does not draw the kinks in the excess supply curves or price difference curves to correspond
to situations, where either local demand or local supply have been driven to zero. Takayama and
Judge (1971, p. 135) do incorporate corner solutions that arise because of kinks, but refer to such
solutions as irregular.
118 4 The Struggling Masses
Table 4.7 Model 4C: price difference curve (assuming product shipped only from Place 1 to
Place 2)
What about comparative statics here? Because Table 4.7 does not permit explicit
solutions for endogenous variables, I cannot derive general solutions for the com-
parative statics here. What I can do, however, is to take a base case and then show the
magnitudes of changes in each of endogenous variables when I change a particular
given. Here, I start from the following base case: α = 15, β = 1, C1 = 2, C2 = 7,
δ1 = 0.001, δ2 = 0.002, N1 = 1,000, N2 = 900, and s = 0.02, and x = 100. In
the base case, the equilibrium price in each market (P1 = 9.18 and P2 = 11.18) is
above C1 and C2 , respectively, which implies that there will be local production at
both places. In equilibrium, a shipment occurs from Place 1 to Place 2.
120 4 The Struggling Masses
G2
B
M
L
N
S C
P
F1
E1 0 I J H2 Shipment
Fig. 4.6 Model 4C: a second example showing demand by arbitrageurs at Place 1 for resale at
Place 2.
Note: α = 15, β = 1, δ2 = 0.0003, δ2 = 0.002, C1 = 2, C2 = 11, N1 = 1,000, N2 = 900,
s = 0.04, x = 100. With arbitrage, P1 = 6.03, P2 = 10.03, zero quantity produced at Place 2, and
quantity shipped from Place 1 to Place 2 is 7,452. Horizontal axis scaled from –15,000 to 25,000;
vertical from –8 to 22
In Table 4.8, I show the comparative statics results. Each row is the result of
changing one given value. In the first row, for example, I show what happens when
α is increased to 15.1.
Table 4.8 Model 4C: relative change in endogenous variable as percentage ratio of relative change
in given
C1 2.1 + + − − − + −
C2 7.1 + + − − + − +
N1 1,100 + + + − + + −
N2 1,000 + + − + + + +
sx 0.024 − + + − − + −
α 15.1 + + + + + + +
β 1.01 − − − − − − +
δ1 0.0011 + + − − − + −
δ2 0.0021 + + − − + − +
Notes: See also Table 4.6. Calculated relative to base case where α = 15, β = 1, C1 = 7,
C2 = 7, δ1 = 0.001, δ2 = 0.002, N1 = 1,000, N2 = 900, and s = 0.02. Calculations by author.
curve: price rises, while the quantity demanded drops. The same thing hap-
pens at Place 2 because part of its supply comes from producers at Place 1.
Shipments from Place 1 to Place 2 fall off because Place 1 now has a lesser
cost advantage. The level of production drops at Place 1, but Place 2 producers,
now relatively less inefficient, produce more.
C2 If C2 is increased, the supply curve shifts upward parallel among producers at
the less-efficient Place 2. At Place 2, market equilibrium runs up the demand
curve: price rises, while the quantity demanded drops. Because firms at Place
1 are now comparatively even more efficient, shipment from Place 1 rises,
pushing up the price at Place 1 and cutting into the quantity demanded locally.
N1 When N1 is increased, local price increases at both places. The aggregate
quantity demanded falls at Place 2 as a result; however, demand goes up
at Place 1, despite the higher price, because of its now-larger population.
Production rises at both places, and shipment falls off because Place 1 has
a smaller price advantage than was the case before.
N2 When N2 is increased, local price increases at both places. The aggregate
quantity demanded falls at Place 1 as a result; however, demand goes up
at Place 2, despite the higher price, because of its now-larger population.
Production rises at both places, and shipment increases because the more effi-
cient firms that are at Place 1 can take advantage of the larger population now
at Place 2.
s When s is increased, it becomes less attractive to ship the commodity from
Place 1 to Place 2. This pushes up the price at Place 2 (less supply in total) and
down at Place 1 (more supply). The reduced shipment means also that firms
at Place 2 produce more, while firms at Place 1 produce less. The changes
in price also imply that the quantity demanded will rise at Place 1 and fall at
Place 2.
122 4 The Struggling Masses
α When α is increased, the demand curve at each of Places 1 and 2 shifts upward.
Therefore, ignoring shipments, market equilibrium runs up the local supply
curve. Therefore, the equilibrium price rises at each local market as does the
quantity demanded locally and the quantity supplied locally. Since Place 1
producers are more efficient at any scale (i.e., C1 < C2 and δ1 < δ2 ), the
shipment from Place 1 to Place 2 increases.
β When β is increased, the individual inverse demand curve sweeps clock-
wise about (0, α), and, again ignoring shipments for the moment, market
equilibrium runs down the supply curve; hence lower price and lower quan-
tity demanded locally and lower quantity supplied locally at each place. The
decline in demand is felt mainly by the less-efficient producers at Place 2, and
this is exacerbated by an increase in shipments from Place 1.
δ1 If δ 1 is increased, the supply curve becomes steeper among producers at the
more efficient Place 1. The effect here is similar to an increase in C1 . At Place
1, price rises, while local demand and local production drop. At Place 2, price
rises and local demand drops; however, local production goes up. Shipments
from Place 1 to Place 2 fall off.
δ2 If δ 2 is increased, the supply curve becomes steeper among producers at the
less-efficient Place 2. The effect here is similar to an increase in C2 . At Place
2, price rises while local demand and local production drop. At Place 1, price
rises and local demand drops; however, local production goes up. Shipments
from Place 1 to Place 2 go up.
In this chapter, the principal model has been 4C. I included Models 4A and 4B to
help readers better understand aspects of Model 4C. In Table 4.9, I summarize the
assumptions that underlie Models 4A–4C. Many assumptions are in common to all
models: see panel (a) of Table 4.9.
In Model 4A, Places 1 and 2 are each in autarky. Each forms its own market for
the commodity populated only by local suppliers and local demanders. The equi-
librium price at Place 1 may or may not be equal to the equilibrium price at Place
2. What separates the two markets is the idea that—under autarky—a shift in any
of the parameters in the market at Place 1 has no effect on the equilibrium price at
Place 2 and vice versa. Further, since local production is just sufficient to meet local
demand, there is no localization of production here. There are no shipments.
In Model 4B, Places 1 and 2 now form a single market with a single equilibrium
price. As a single market, a change in any parameter of local demand or local supply
at one Place can affect equilibrium price at both Places 1 and 2. Where production
of the commodity at one place is particularly advantageous, we see localization of
production at that place. Model 4B says nothing about shipments. Since it is costless,
there may be any amount of shipping; cross hauling is possible here. The extent of
localization will depend on differences between the parameters of the local supply
4.5 Final Comments 123
4A 4B 4C
Assumptions [1] [2] [3]
curves at the two places. Given C1 < C2 , production of the commodity will be solely
at Place 1 if δ 1 —the slope of the local supply curve there—is small enough. See
(4.3.7) and (4.3.8). Since there is only one market here, there is still no Walrasian
process at work to link prices between markets.
In Model 4C, Places 1 and 2 form two submarkets. As in Chapter 2, the place
more costly to serve will have a price premium. Shipping occurs only up the price
gradient. In this chapter, the price premium will be equal to the unit shipping cost.
The price premium will be between zero (Model 4B solution) and the autarky price
difference (Model 4A solution). As the price premium approaches the Model 4A
solution, the extent of localization approaches zero.
Model 4C exemplifies a Walrasian linkage of the prices at two places. Anything
that causes the price to rise at the exporting place causes price to rise by the same
amount in the importing place. Anything—other than a change in unit shipping
cost—that causes the price to rise in the importing place will similarly cause the
price to rise in the exporting place. And, when the unit shipping cost rises, price will
increase in the importing place and drop in the exporting place. Note the difference
in effect here vis-à-vis Model 2D; there, a constant marginal cost of production
meant that a shift in demand at Place 1 had no effect on price at Place 2 and vice
versa. At the same time, this Walrasian process is restricted; it looks at simultaneity
only across two places; it does not consider the simultaneity between price of the
commodity and the prices of other goods and services.
In Chapter 1, I argue that prices are important in shaping the location of firms. In
this chapter, the prices of inputs used by the firms in local production are implicit in
the intercept and slope of the local supply curve. Arbitrageurs face input prices that
are implicit in s. As in Chapter 2, consumers have income constraints and face prices
for other goods and services that determine the intercept and slope for their demand
124 4 The Struggling Masses
curve. As in Chapter 2, these other prices are all determined in markets outside the
scope of the models in Chapter 4. Once again, I think Walras would have argued
that the analysis has been only partial in the sense that we have not looked explicitly
at the simultaneity among prices in these markets. Once again, we must wait until a
later chapter for the opportunity to do that.
What is the source of localization in Model 4C? Presumably, it is the differences
among places: i.e., in Ci , δ i , α, β, or Ni . A combination of these can make the autarky
price sufficiently lower at Place i to give arbitrageurs the incentive to participate.
However, Model 4C is silent on the reasons for this advantage. While it is true, for
example, that economies of scale or division of labor might account for a lower
Ci at Place i, there are other possible explanations: e.g., variations in input prices.
The models in this chapter represent the outcome of a competitive market in
which no one buyer or seller is able to influence the price at which the commodity
is sold. As such, these models do not take into account the kind of monopoly power
that arises in geographic space where, for example, competitors elsewhere find unit
shipping cost makes it more costly for them to compete for customers near a given
firm. I consider such effects later in this book.
Finally, the models in this chapter do not say anything about the distribution of
income in society among units of labor, capital, and land. As in Chapter 2, the firm
is envisaged to incur costs, but these are not related explicitly to the hiring of labor
or the rental of land. In any case, all income gains that can arise (e.g., because of
an increase in α or a decrease in C1 , C2 , or s) accrue to firm owners in the form of
increased profit. At the same time, any adjustment of market price affects the well-
being of consumers; in this chapter, such changes are being measured by consumer
surplus.
Chapter 5
Arbitrage in the Grand Scheme
Perfect Competition at Many Places
(Samuelson–Takayama–Judge Problem)
Imagine N places, isolated from the rest of the world. Assume perfect competi-
tion; no one supplier or demander can affect market price. If unit shipping costs
are prohibitive, all places with both local demanders and local suppliers will be in
autarky. However, if at least one unit shipping cost is sufficiently low, arbitrageurs
will purchase the commodity where price is low for resale elsewhere. How does
the amount shipped and the resulting prices at the N places depend on unit ship-
ping costs? Following a conjecture by Samuelson as implemented by Takayama
and Judge, Model 5A is a quadratic program in which Global Net Social Welfare—
consumer benefit (under a linear demand curve) minus producer cost (under a linear
supply curve) minus shipping costs—is maximized. This chapter builds on models
presented in Chapters 3 and 4. Chapter 4 solves a similar problem graphically using
price difference curves but that method limits us to two places. Model 5A solves the
same problem for any number of places. Chapter 3 solves a logistics problem using
linear programming where the quantities demanded and supplied at each place are
fixed. Model 5A makes the quantities endogenous to the model and solves for the
equilibrium prices (one for each place) that give rise to these outcomes.
have more than two places.1 At this point, it is also instructive to ask why, in this
book, I want to look at the case of three or more places sharing a common fiat money
economy. At the outset, I stated that the purpose of this book is reinterpretation.
What motivates me in this chapter is the impact of geography on market formation
in particular and on the insights this gives about economic reasoning in general.
A spatial price equilibrium is just one example of a model of economic equilib-
rium.2 Harker and Pang (1990, pp. 161–162) argue that there have been three main
approaches to computation of equilibrium: (1) the fixed point (homotropy-based)
approach, (2) the variational inequalities—including complementarity—approach,
and (3) the optimization approach. In my experience, the first two approaches are
useful in computation but shed little added light on the nature and implications of
spatial price equilibrium.3 To me, the first two approaches are also unsatisfactory;
they assume spatial price equilibrium rather than have that equilibrium arise as an
outcome of the model. I intend no offence to proponents of the other two approaches
here; I am merely laying out my own preferences.
In contrast, I find the optimization approach gives insights into the impact of
geography on market formation and economic thinking. In an important and influ-
ential step forward, Samuelson (1952) invokes an optimization approach in thinking
about spatial price equilibrium. Let me now describe Samuelson’s approach (the
Samuelson conjecture) in more detail. He argued intuitively that economists would
want to conceptualize trading activity as leading to the maximization of something
at the level of the entire society. He conjectured that if shipment was to be desirable,
it had to improve what he labeled net social payoff (NSP) in aggregate across the
places in the same sense that an autarky solution maximizes well-being at one place.
Specifically, he proposed (p. 288) a net social payoff with three components: Social
Payoff in region 1 + Social Payoff in region 2 – Shipping Cost. Social payoff for
a region is the algebraic area under its excess-demand curve. Samuelson (p. 288)
was guarded in his use of terms here: being careful to say that social payoff was not
the same as consumer surplus.4 At the same time, however reasonable Samuelson’s
approach may seem to economists, it is not clear to noneconomists why markets
should necessarily behave in this way. I return to this matter shortly.
Samuelson had two hunches about this problem. First, he argued (p. 90) that
one could experimentally vary exports, by trial and error, so as to maximize net
social payoff. A number of early studies apparently took this advice.5 Fox (1953,
p. 548) confirms Samuelson’s hunch: In this book, I do not pursue this trial-and-
error process further because it does not contribute to the book’s focus on economic
reasoning.
His second hunch was based on the argument that part of the spatial price equi-
librium problem was subsumed in the Hitchcock–Koopmans Problem (discussed
in Chapter 3). The Hitchcock–Koopmans model is a linear (mathematical) pro-
gram that solves for shipments given quantities demanded and supplied at each
place and, therefore, does not solve for market prices.6 However, the Hitchcock–
Koopmans model does solve for the opportunity costs (termed the shadow prices)
of supply restrictions and demand requirements at each place. Samuelson argued
that there should be a relationship between shadow price and market price, and sev-
eral scholars use variants of that idea to solve the spatial price equilibrium problem.7
Another variant of this approach involves iterative techniques which move back and
forth between predictions of shipment flows and predictions of prices until we find
mutually sustainable solutions.8
In this chapter, I use a notion similar to Samuelson’s net social payoff that I label
global net social welfare.9 I present and interpret the method used by Takayama
and Judge (1971) to solve spatial price equilibrium and shipments when there are
more than two places and that allows for the kinks that can arise when price locally
becomes (1) low enough that there are no longer any local suppliers at one or
more places, or (2) high enough that there is no longer any local demand for the
commodity at one or more places.10
Because this chapter helps us think about an industry that could conceivably span
nations or even the globe, I will discuss the possibilities and limitations of looking
5 See, for example, Fox (1953) and Fox and Taeuber (1955). Another early approach was to further
simplify the problem by assuming that supply at each place was a given quantity (not a schedule
varying by price). See Judge and Wallace (1958).
6 A similar approach is successive approximations each time using a linear objective function. See
Marcotte, Marquis, and Zubieta (1992).
7 See Baker (1961), Henderson (1955), Judge and Wallace (1958), and Daniel and Goldberg (1981).
8 See the applications of reactive programming in Tramel and Seale (1959), Boyd (1983a, 1983b)
and Seale, Seale, and Leng (2004).
9 Takayama and Judge (1971) prefer the term “net quasi-welfare”.
10 The model presented in this chapter is the simplest of the models presented in Takayama and
Judge (1971). That book considers the case of (1) spatial price equilibrium across multiple com-
modities, rather than the spatial equilibrium for just one good considered here and (2) changes in
inventory holding between periods so that demand need not equal supply each period as assumed in
this chapter. Guise (1979) notes two limitations in the Takayama and Judge formulation: (1) com-
modities must be stored in the region where they are produced; and (2) a good cannot be stored
for more than one time period. These are not applicable to this chapter because I ignore inventory
holding altogether.
128 5 Arbitrage in the Grand Scheme
5.2 Model 5A
11 A tax charged on an imported good: often ad-valorem (i.e., a percentage of the value of the
good).
12 A limitation, other than a tariff, such as a quota or other regulation that makes importing less
attractive or feasible.
13 I leave aside here the debate—dating from Arrow (1951)—over whether a community, society, or
nation can have a collective social welfare function by which members can be thought to rationally
choose one outcome over another.
5.2 Model 5A 129
the monopolist in Model 2A. The firm’s revenue is P1 Q1 and Its semi-net revenue
is (P1 − C)Q1 . How does the entry of our firm affect national income? If I assume,
for simplicity, that there are no labor costs included in C, then semi-net revenue is
the amount that the firm itself contributes to national income. If the firm then finds
it profitable to serve customers at Place 2 also, either from the factory at Place 1 or
from a new factory at Place 2, the firm’s semi net revenue would rise and its contri-
bution to national income would increase.14 However, the firm’s own contribution
to national income is not the entire story. When our firm opts to produce, consumers
at Place 1 (and possibly Place 2) now start to purchase the commodity. As they do,
they presumably switch from purchasing other commodities that they had been con-
suming before. In other words, other firms may experience a loss of revenue and
semi-net revenue because of the entry of our firm. In practice, the story is com-
plicated by the extent of substitutability—or complementarity—among consumer
commodities. As well, there are income effects here; a consumer able to purchase
the commodity at a lower price than before is now better off in real terms, and their
consumption of commodities and services changes accordingly. In short, to fully
assess the impacts of the entry of our firm on national income requires considerable
data about all economic activity nationally.
That has led some economists to look at approximations (short-cuts) to the mea-
surement of economic well-being. That brings us to the concepts of consumer
surplus, consumer benefit, producer surplus, producer cost, monopoly excess profit,
and social welfare.15
Marshall is generally credited with introducing the term consumer surplus,
although Dupuit had earlier labeled it relative utility.16 Consumer surplus is a money
measure of the area below the demand curve and above the market price to the left
of market quantity. Put differently, consumer surplus is the hypothetical amount
that consumers might be thought to be willing to pay over and above the market
price rather than go without the commodity. Among proponents, consumer surplus
is thought to measure the benefit to consumers from participating in the market in
excess of the amount P that they pay for each unit consumed. We can think of an
aggregate demand curve as the ordering of consumers in descending order of the
price they are willing to pay; in this sense, consumer surplus identifies the value
placed on consumption by the most enthusiastic purchasers (i.e., those willing to
pay the most for the commodity) relative to the marginal consumer. A related con-
cept, consumer benefit, measures the entire area under the demand curve up to the
14 In the alternative case where C includes labor costs, the firm’s contribution to National Income
would be larger than its semi-net revenue by the wages paid (i.e., the amount of wage bill).
15 I do not consider other measures of well-being here. In general, economists have concerns
with the use of consumer surplus as a measure of social benefit. See Winch (1965), Schmalensee
(1972), Burns (1973), Willig (1976), Chipman and Moore (1980), Turnovsky, Shalit, and Schmitz
(1980), Hausmann (1981), Blackorby, Donaldson, and Moloney (1984), and Hanemann (1991).
Alternatives include the compensating variation and equivalent variation proposed by Hicks (1956,
pp. 69–94).
16 Ekelund and Hébert (1999, p. 15).
130 5 Arbitrage in the Grand Scheme
quantity demanded. Consumer surplus (CS) is equal to consumer benefit (CB) net
of consumer expenditure (PQ) on the commodity: i.e., CS = CB − PQ.
To fix ideas here, let me assume an individual consumer with an inverse demand
curve as in Chapters 2 and 4: P = α − βq. At market price P, this consumer
demands q = (α − P)/β. The consumer values the marginal unit as just worth
P. Diminishing marginal utility leads us to expect that additional units of the com-
modity consumed are less and less valuable to the consumer. Put differently, if the
consumer were restricted to a smaller quantity than q, the value of the marginal unit
would be larger than P. Viewed in this way, the slope of the demand curve tells
us something about the utility gained from additional units of consumption: going
say from 0 to q. However, such a conclusion is problematic in at least two respects.
First, consumer demand is typically thought to depend on income, the prices of other
commodities, and tastes, as well as on the price of the commodity itself. When we
write P = α − βq, we hide the effects of income, prices of other commodities,
and tastes in α and β. This makes it difficult to interpret movement along an individ-
ual’s demand curve as corresponding to any particular change in utility (well-being).
Second, when we move from an individual to an aggregate demand curve, it is not
clear how to separate the effects of the distribution of income (the idea that different
consumers have different incomes) on an aggregate measure of consumer surplus.
In a similar way, producer surplus is a money measure of benefit to efficient
producers (i.e., for all but the marginal unit supplied) from supplying to the market
over and above their marginal cost of production.17 It is measured as the area under
the market price and above the supply curve to the left of the market equilibrium.
Producer surplus is the amount received by efficient producers over and above what
is needed to secure their participation in the market. Producer surplus is a measure
of monopoly excess profit that arises because some producers are more efficient than
the marginal producer. Presumably, this profit accrues to the benefit of the owners
of these efficient firms. It is in that sense that both producer surplus and consumer
surplus can be thought to be the benefits that arise from having a competitive mar-
ket that clears at price P. Related to this is the notion of producer cost. This is a
money measure of the area under the supply curve to the left of quantity supplied.
Producer surplus (PS) is consumer expenditure (PQ) net of producer cost (PC):
PS = PQ − PC.
In a competitive market, the sum of consumer surplus and producer surplus is
a money measure labeled social welfare (SW): SW = CS + PS. Equivalently,
social welfare is the amount by which consumer benefit exceeds producer cost:
SW = CB + PC. However, in the case of a monopolist firm as in Chapter 2,
we need to account also for the monopoly excess profit (MP) that arises because
the firm exploits a downward sloping demand curve. In a monopoly, social wel-
fare includes consumer surplus, producer surplus, and monopoly excess profit:
SW = CS + PS + MP. This conceptualization of social welfare is simple minded;
17 Economists in general have the same uneasiness about producer surplus as they do about
consumer surplus. See Martin and Alston (1997).
5.2 Model 5A 131
Notes: Assumptions (see Appendix A): A1—Closed regional market economy; A3—Punctiform
landscape; B1—Exchange of soap for money; B2—Upwardly sloped local supply curve each sup-
ply place; B4—Local demand at each customer place; C4—Identical customers; C5—Identical
linear demand; E4—Unit shipping costs symmetric. Rationale for localization (see Appendix A):
Z3—Implicit unit price advantage at some locales; Z8—Limitation of shipping cost. Givens
(parameter or exogenous): Ci —Intercept of inverse supply curve at Place i; Ni —Population of
Place i; α—Intercept of individual linear inverse demand curve; β—Slope of individual inverse
demand curve; δ i —Slope of inverse supply curve at Place i. Outcomes (endogenous): Pi —Price at
Place i; Qi —Equilibrium quantity transacted at Place i; Qij —Amount produced at Place i for sale
to local demand at Place j.
A D
A D
I G H
H G1
Price
Price
I C
F
C
0 0
E Quantity B E F Quantity B
Fig. 5.1 Consumer surplus, producer surplus, and social welfare at Place 1
up until the quantity where the price that the marginal consumer is willing to pay
is just equal to the marginal cost of the last unit supplied by local producers. See
(5.1.6). Of course, this is nothing more than the autarky solution: i.e., equilibrium
of local demand and local supply. If we were to consider a still larger quantity, SW
would in fact decline. In this sense, where the local demand curve and local supply
curve intersect is a best solution for society in the case of autarky.
I use here global net social welfare (GNSW), which I measure as the sum of social
welfare at all places less shipping costs. Unit shipping cost, which I am viewing here
as exogenous, can be envisaged as a loss that reduces the benefit to society overall:
just as it lowered the profit to the firm in Chapter 2. Ignored here, for the sake of
5.4 Net Social Payoff and Global Net Social Welfare 135
simplicity, is the concept of an endogenous price of shipping, that is, the possibility
of a shipping sector that generates earnings for shipping firm employees, profits
for shippers, and other factor income. To me, the literature appears confused as to
whether NSP (area under excess demand curves net of shipping cost) and GNSW
(area under demand curve less area under supply curve net of shipping cost) proxy
the same thing.19 To some, net social payoff has come to be seen as equal to social
welfare net of shipping costs.20 What is clear is that NSP is zero in the absence of
shipments (i.e., autarky) but that GNSW is generally positive even in autarky. Put
differently, GNSW may proxy NSP in the sense that the shipments that maximize
NSP also maximize GNSW, but GNSW is not equal to NSP.
The implementation of a GNSW approach to spatial price equilibrium is evi-
denced in Table 4.5 from the previous chapter. There, I calculated producer and
consumer surplus for each of two places (Places 1 and 2) in autarky. The two autarky
prices were 12.50 and 13.00. The Table sums the two surpluses to get net social wel-
fare at each place in autarky. Then, GNSW is calculated as the sum of these (I can
ignore shipping costs here because there are no shipments in autarky). In contrast,
when the shipping rate is only 0.10, the Table predicts 400 units shipped and a higher
overall GNSW (10,290, up from 10,250 in autarky). If shipping rate had been zero
(the integrated market solution), GNSW would have been even larger (10,326).
I have already raised the question as to why a market economy would act to
maximize GNSW and asserted that, to a non-economist, the underlying faith in the
efficiency of markets might seem unwarranted.21 More interesting to me, however,
is the question of whether maximizing GNSW will lead to a spatial price equilibrium
of the kind that we have already seen in Chapter 4. That it does indeed do this further
supports the idea that the maximization of GNSW is intuitively reasonable.
The maximization of GNSW is relatively easy to implement numerically. Let Qij
be the amount of output produced at Place i to meet a local demand at Place j. As a
result Si = j Qij and Dj = i Qij . We can then write the local inverse demand and
supply curves for Place i as (5.2.1) and (5.2.2). See Table 5.2. For Place i, the area
under the demand curve is now given by (5.2.3), the area under the supply curve
by (5.2.4), and social welfare by (5.2.5). Note that SWi is a quadratic function of
quantities shipped (Qij ). We then get the GNSW in (5.2.6) by first aggregating social
welfare for each place and then subtracting total shipping costs. I can then find
the maximum value of GNSW (5.2.7) subject to the identities in (5.2.8) and (5.2.9)
and the nonnegativity condition on every shipment (5.2.10). Since the constraints
19 Takayama and Judge (1971, pp. 111–112) was among the earliest to model GNSW directly.
20 In a widely cited paper Currie, Murphy, and Schmitz (1971, p. 780) states that maximizing NSP
is equivalent to maximizing net welfare gains. However, this argument is based on a comparison
of areas on a graph—Currie, Murphy, and Schmitz (1971, Fig. 14)—that I find difficult to compare
precisely.
21 Smith (1963) reinterprets Samuelson’s model as the minimization of global economic rent. This
is an attractive idea because it suggests that the role of the arbitrageur is to reduce the consumer
and producer surplus that would arise in the absence of shipments. For a discussion of the use of
spatial price equilibrium models in development planning, see Luna (1978).
136 5 Arbitrage in the Grand Scheme
(5.2.8) and (5.2.9) are linear in shipments, this formulation exemplifies a quadratic
programming22 problem. In recent decades, commercial software that solves vari-
ous kinds of optimization problems, including quadratic programming, has become
widely available.23 Quadratic programs are solved numerically in a set of repeated
steps (each step is called an iteration): such a procedure is termed an algorithm.
As in Chapter 3, I cannot derive an explicit (algebraic) solution as a general rule
here; instead, I have to rely on numerical methods to solve specific problems of
this type.
In the version presented above, I have assumed a set of n places with local
suppliers and local demanders in each. However, the model can be solved in a sim-
ilar manner if some of these have suppliers only or demanders only. Interestingly,
the GNSW approach can also be extended to cover situations where there is price
rigidity (e.g., price controls) at one or more places that mean a local market is in
disequilibrium.24
To begin, assume n = 3 places each with a horizontal supply curve. In other words,
there is as much supply forthcoming at Place 1, at a price of C1 as consumers might
demand. At Places 2 and 3, suppliers similarly supply as much as needed at a price
of C2 or C3 , respectively. Put differently, δ1 = δ2 = δ3 = 0. Note the asymmetry
here; firms are identical in terms of efficiency at each place but may differ otherwise
from one place to the next.
Assume initially that shipping rates are sufficiently high so as to deter any ship-
ment of the commodity. In this autarky condition, local supply equals local demand
at each place. In effect, there are three separate markets here. Since δi = 0 at each
Place i and all markets are competitive, we get Pi = Cj . The quantity demanded
at Place i is given by (5.3.3). See Table 5.3. Consumer surplus is given by (5.3.5).
Since δ1 = δ2 = δ3 = 0, every producer at Place i is equally efficient. Therefore,
no producer earns any excess profit and producer surplus is zero: see (5.3.7). Since
Price at Place i
Pi = Ci in autarky, (5.3.1)
P2 = C1 +s12 with shipment from 1 to 2 only (5.3.2)
Demand at Place i
Di = (α − Ci )Ni /β in autarky, (5.3.3)
D2 = (α − C1 − s12 )N2 /β with shipment from 1 to 2 only (5.3.4)
Consumer surplus at Place i
(α − Ci )2 Ni /(2β) in autarky, (5.3.5)
(α − minj=i [Cj + sji ])2 Ni /(2β) with shipment from 1 to 2 only (5.3.6)
Producer surplus at Place i
0 in every region under autarky (5.3.7)
0 in every region even with shipment from 1 to 2 only (5.3.8)
GNSW
i (α − Ci )2 Ni /(2β) in autarky (5.3.9)
((α − C1 )2 N1 + {(α − C1 − s12 )2 − 2s12 (α − C1 − s12 )}N2 + (α − C3 )2 N3 )/(2β)
with shipment from 1 to 2 only (5.3.10)
Net gain in GNSW with shipment
(α − C1 − s12 )2 − 2s12 (α − C1 − s12 ) (5.3.11)
there are no shipping costs and producer surplus is everywhere zero, GNSW is just
the sum of consumer surplus at the three places.
Now, let us undertake a thought experiment in which we reduce shipping rates
everywhere proportionally until shipment now becomes attractive for just one pair
of places. Without any loss of generality, assume shipping rates are low enough to
attract shipment from Place 1 to Place 2. This happens when the shipping rate for
one unit of output shipped from Place 1 to Place 2, s12 , satisfies C1 + s12 ≤ C2 .
At this point, local production of the commodity at Place 2 ceases, and the firms
at Place 1, being more efficient, now satisfy all the demand in both Places 1 and 2.
At Place 1, consumer surplus is unchanged. As the shipping rate is still too high,
Place 3 remains in autarky. Therefore, consumer surplus at Place 3 is unchanged.
However, at Place 2, consumers are better off. The effective price drops: from a
level of C2 in autarky to C1 + s12 . Consumers there are better off now that Place
2 is supplied from Place 1: GNSW is now larger. This is not surprising; after all, if
shipments had not improved well-being why would anyone ship the commodity?
In this special case where δ1 = δ2 = δ3 = 0, consumers—but not producers—
benefit from shipments and only in those places where local supply is displaced by
a lower cost supplier elsewhere. Producers do not benefit because all the producers
at a given place are identical; no one is more efficient, and no one has access to a
monopoly profit.
Now, consider Example Problem 1 as detailed in Table 5.4. Here there are n = 3
places as before. The difference now is that the local supply curve at each place
is upward sloped: no longer horizontal; for some reason, as in Chapter 4, there is
congestion in production. The three places are identical except that N2 > N3 >
N1 . Here, at any given price, the quantity demanded at Place 2 is greater than at
Place 3, which in turn is greater than at Place 1. To coax that larger supply locally,
the autarky price at Place 2 would have to be larger than at Place 3 and in turn
higher than at Place 1: see (5.4.14). Where prices differ between places in autarky,
the possibility of shipments arises. Given that shipping costs are positive, shipment
of the commodity will occur only from a place with a lower price in autarky to
a place with a higher price. In the case of Table 5.4, the shipping rates are too
high to facilitate shipment. In (5.4.11), (5.4.12), and (5.4.13), producers ship only
to local consumers, and therefore the prices (5.4.14) are autarky prices. For any pair
of places, the difference in autarky prices is less than the corresponding shipping
rate: (5.4.8), (5.4.9), and (5.4.10).
Now, let us look at the components of GNSW in autarky. The absence of ship-
ments means that shipping costs are zero: see (5.4.17). Since autarky price is highest
there, the assumption that the supply curves are similar at the three places means that
producer surplus is highest at Place 2: see (5.4.16). Despite the high autarky price at
Place 2, the larger N there means that consumer surplus there—being an aggregate
measure—is the greatest of the three places: see (5.4.15).
5.6 Three Examples of Multiregional Shipment 139
Now, consider Example Problem 2 described in Table 5.5. This is the same as
Problem 1 except that unit shipping cost is everywhere now 1.00 instead of 2.00.
Now, the solution that maximizes GNSW is one wherein suppliers at Place 1 supply
all of the local demand at Place 1 and part of the local demand at Place 2. Price rises
at Place 1 and falls at Place 2: compare (5.5.14) with (5.4.14). The price difference
between the two places is 11.85 – 10.85, which is equal to the unit shipping cost;
remember this is an endogenous outcome of maximizing GNSW, not an assumption
of the model. Local suppliers at Place 2 supply the rest of demand of Place 2. Local
suppliers at Place 3 supply all the local demand at Place 3; the price at Place 3
is therefore unchanged. We have now two markets: Places 1 and 2 form a single
market; price may differ between the two places but only by an amount related to
shipping cost. Place 3 is the other market. As in Chapter 4, market formation is now
endogenous to the model.
What happens to GNSW here? Shipments from Place 1 to Place 2 now create
a shipping cost: see (5.5.17). At the same time, producer surplus rises at Place 1,
drops at Place 2, and is unaffected at Place 3. This is not surprising. When unit
shipping costs are lower, it becomes efficient to supply some of Place 2 from more
efficient producers at Place 1 rather than the less-efficient producers at Place 2.
This is because, even though the local supply curve is the same at both places, the
larger number of consumers at Place 2 means that the marginal supplier at Place 2 in
140 5 Arbitrage in the Grand Scheme
autarky is less efficient than the marginal supplier at Place 1. The lower unit shipping
cost strips away the protection afforded to the less-efficient marginal producer at
Place 2. In contrast, consumer surplus drops at Place 1 (since price rises relative
to autarky), rises at Place 2 (since price has dropped), and stays the same at Place
3 (since price stays the same). So, the lower shipping rates in Example Problem 2
increase consumer benefit at Place 2 and increase in producer surplus at Place 1.
Intuitively, GNSW is higher in Example Problem 2 because lower shipping rates
allow us to better match demand and supply across places.
Let us turn now to Example Problem 3 as described in Table 5.6. This is similar
to Problems 1 and 2 except that unit shipping costs are everywhere now 0.19. Now,
the solution that maximizes GNSW is one wherein suppliers at Place 1 supply all of
the local demand at Place 1 and part of the local demand at each of Place 2 and Place
3. Price rises at Place 1 still further and falls at Places 2 and 3: compare (5.6.14) with
(5.5.14) and (5.4.14). Because suppliers at Place 1 ship to both Place 2 and Place 3
and the unit shipping cost is 0.19 to both, the equilibrium prices in Places 2 and 3
are 0.19 higher than the price at Place 1; again, this is an endogenous outcome of a
model that maximizes GNSW. We now have just one market, and it incorporates all
three places.
From Table 5.4, 5.5, and 5.6, unit shipping costs have decreased. As the cost
of shipping declines, GNSW goes up. Looking back at Table 3.4, we saw a similar
5.6 Three Examples of Multiregional Shipment 141
result in the two-place case as we moved from autarky (high shipping rate) to the
integrated market (zero shipping rate) solution. However, as I noted in Chapter 4,
not everyone becomes better off when shipping costs are lower. Here, consumers
at Place 1 are worse off (lower CS) when the shipping rate is lower because local
suppliers are now drawn upon to supply other places as well, and this pushes up the
marginal cost curve at Place 1. On the other hand, consumers in the larger Place
2 (and even Place 3 if shipping rate is low enough) benefit (higher CS) from a
lower price. It is the opposite pattern among producers; those at Place 1 benefit
from lower unit shipping costs, while those at Places 2 and 3 are worse off. Why
these results? When unit shipping costs are high enough to result in autarky all
around, consumers at Place 1 benefit because only they have access to the low cost
producers there. When shipping rate is lower, those consumers at Place 1 lose the
advantage of location and now pay a higher price for the commodity. The opposite
is true for consumers at Places 2 and 3.
What I have undertaken here can be thought of as a kind of sensitivity analysis.
In other words, how and in what ways are shipments and local prices sensitive to
shipping rates? Of course, unit shipping costs are just one set of parameters in this
model. We could also look at the sensitivity of shipments and local prices to other
parameters in the model: α, β, Ni , Cj , and δ j . However, I have already done this in
the case of two places in Chapter 4; and the story is broadly similar in this chapter.
142 5 Arbitrage in the Grand Scheme
In this chapter, we have assumed that consumers everywhere have the same α
and β. Why is that? Are consumer incomes and the prices of other commodities
the same everywhere? In reality, we might expect incomes and the prices of other
commodities to vary from place to place. This model in that sense is just a simpli-
fication of reality that helps us see the role of suppliers and customers in what is an
endogenously defined market.
5.7 Application
geographic regions wherein demand varies seasonally. Models of both storage facil-
ities and a pipeline network with 79 links are incorporated. Spatial price equilibrium
models like these raise the question of whether, in practice, economic actors—be
they buyers or sellers—succeed in integrating markets so as to give prices, which
differ between places by no more than unit shipping cost. Baulch (1997a, p. 477)
argues that market integration is key to market liberalization and parastatal reform
in developing countries. Without integration, price signals are not correctly trans-
mitted from sector to sector, agricultural producers do not specialize efficiently, and
gains from trade are not realized.
In practice, it is not easy to implement a spatial price equilibrium analysis. There
are at least four principal problems.
27 Baulch (1997), for example, looks at the integration in the wholesale rice market around
metropolitan Manila and its hinterland.
144 5 Arbitrage in the Grand Scheme
Boyd, Doroodian, and Abdul-Latif (1993) present a case study of the soft-
wood lumber industry in North America when the North American Free Trade
Agreement (NAFTA) first came into effect.30 The study divides America and
Mexico broadly into 10 lumber-consuming regions centered in the following cities:
Boston, New York City, Chicago, Minneapolis, Washington (DC), Dallas, Nashville,
Los Angeles, Missoula (MT), and Mexico City. Note that this study area covers
three currency areas and national sets of tariffs and nontariff barriers plus a vari-
ety of regional, state, and provincial regulations governing the logging and lumber
industries.
28 A market condition in which there only a few vendors. No vendor has a monopoly, but none of
the vendors is strictly a price-taker.
29 See, for example, Boyd (1983a, 1983b), Buongiorno and Uusivuori (1992), Jung and Doroodian
(1994), De Vany and Walls (1996), and Bukenya and Labys (2005).
30 Aspects of the market for softwood lumber in North America are the subject of Buongiorno and
Uusivuori (1992), Gilligan (1992), Jung and Doroodian (1994), Montgomery, Brown, and Adams
(1994), Mogus, Stennes, and van Kooten (2006).
5.8 Case Study 145
The study estimates a linear demand for lumber for each of these 10 regions: each
treated as a place. The demand for lumber in Canada is ignored here because there
were no reliable estimates of the price elasticity of demand31 for lumber there at the
time. The demand equations in this study are linear and link the quantity of lumber
demanded to the quantity of construction activity (the principal use of lumber) and
to the price of lumber. Two kinds of lumber are modeled here: pine and fir. The
study assumes that all pine lumber is perfectly substitutable, all fir lumber is per-
fectly substitutable, and that pine and fir are highly but not perfectly substitutable.
In that study, unlike earlier in this chapter, demand per capita is not assumed to be
necessarily the same in each region. For example, the demand for pine in region
j (Dpj ) takes the form Dpj = αpj − βpj Ppj + βfj Pfj , where Pp and Pf are the prices
of pine and fir, respectively; the intercept, own price coefficient, and cross price
coefficients can each vary from one region to the next.
On the supply side, the study identifies 18 regions (again treated as geographic
points) that produce lumber: 12 in America, 5 in Canada, and 1 in Mexico. The
study estimates a linear supply curve for each of the 18 supplier places. There are
28 places in total: 18 for customers and 10 for suppliers. The study, unlike earlier in
this chapter, does not include any places that are both demanders and suppliers. It is
not clear here whether the level of demand or supply is actually zero in each case,
or just small enough to ignore.
The study then combines estimated regional supply and demand equations with
interregional freight and tariff rates to solve for a competitive spatial price equi-
librium as of 1988. Discrepancies between actual and predicted shipments are
presumed to be the result of nontariff barriers. As the model incorporates places
in three distinct currency areas, presumably the demand functions, supply func-
tions, and tariff and shipping rates were adjusted by prevailing exchange rates. The
procedure is repeated under three alternative tariff regimes that might arise under
NAFTA. Scenario (a) assumes status quo tariffs; scenario (b) assumes that the then-
current 6.51% duty on Canadian softwood coming into the American market had
been eliminated; scenario (c) includes also elimination of the 15% Mexican tariff
on American softwood. The purpose of the study is to measure impacts of these
three alternatives on shipment and economic welfare.32
The shipments of pine (in billions of board feet) that are predicted under each
of the tariff regimes are summarized in Table 5.7.33 The scenario (a) projections
31 Price elasticity of demand (e) is the percentage change in quantity demanded (q) for associated
with a 1% change in price (p): e = (p/dp). Since dq/dp is typically negative, most treatments take
the absolute value of e. However, for the benefit of non-economists, I leave the negative sign to
remind the reader that when price goes up, we expect quantity demanded to drop.
32 The study solves the model in each scenario using reactive programming.
33 Surprisingly, in Table 5.7, total demand for pine in North America declines when tariffs are
reduced. We might have expected the price of pine to drop and therefore the quantity demanded to
rise when tariffs are reduced. Why that did not happen here is unclear from the study. My guess
that the results also reflect a drop in the price of fir lumber; the strong substitutability between pine
and fir may have caused the quantity of pine consumed to drop even though its price had decreased.
146 5 Arbitrage in the Grand Scheme
Table 5.7 Shipments of softwood lumber (pine) in Bbf under three scenarios
Supplying region
Total
US North US South US West Mexico Canada demand∗
[1] [2] [3] [4] [5] [6]
should correspond to actual shipments of lumber in 1988. However, the report does
not say whether these scenario (a) projections were close to the shipments actually
observed in that year. The projections under scenarios (b) and (c) suggest the fol-
lowing interpretation. Suppliers in America North (Northeast and Central United
States) are largely unaffected by the tariff reductions. However, suppliers in the US
South see a reduction in shipments to the US North in scenario (b) compared to
(a). Suppliers in the US West also see erosion in shipments with a reduced tariff
on Canadian lumber. However, the elimination of the Mexican tariff in scenario (c)
has no discernable effects on shipments anywhere in North America. Because the
amount of lumber consumed changed by only a small amount in scenario (b) vs. (a),
the study finds that overall change in economic well-being is small. The implication
here is that NAFTA, as exemplified by scenarios (b) and (c), would have relatively
little effect on well-being in the three countries.34
This study exemplifies all of the problems of application discussed above. First,
the need to model both pine and fir shipments tells us that even a commodity as
34 The report describes the history of the trade dispute between Canada and America over softwood
lumber before the introduction of NAFTA. That dispute has continued to fester long after NAFTA
came into effect. Considering the mild effects of NAFTA predicted by this study, it is hard to
understand why this trade dispute has been so protracted.
5.9 Final Comments 147
Spatial price equilibrium solves for quantities produced (hence localization) as well
as for prices. In the three example problems in Tables 5.4, 5.5, and 5.6 produc-
tion always occurs in all three regions because we have assumed that there was no
particular advantage to producing in any one region. In fact, in the example prob-
lems, the only reason why prices differ is because local demand (population) at
Place 2 has pushed production further up the marginal cost curve for local produc-
ers than is the case for producers at Places 1 and 3. However, it is not difficult to
envisage a different set of conditions wherein local production does not happen in
some locales because those places are not efficient enough. In those locales, we can
think that firms are not choosing to locate there. In that sense, localization can be an
outcome of spatial price equilibrium models.
Chapters 4 and 5 of this book include a discussion of the kinks that arise when
equilibrium price is either too low to encourage local production or too high to
permit local demand.36 These kinks are important in helping us understand the
localization of production. Not surprisingly, this chapter reconfirms conclusions
drawn in Chapter 4. The discussion in Section 5.3—where I assume horizontal local
supply curves—illustrates the role of unit shipping costs in determining markets,
submarkets, and the extent of localization. When unit shipping costs are everywhere
prohibitive, each place is in autarky (i.e., its own market) even if autarky price dif-
fers from one place to the next and there is no localization. When unit shipping cost
becomes low enough for the difference in autarky prices to induce shipment, pro-
duction ceases at the place where it is costly and localizes where it is less costly.
The place where production was costly ceases to be a market on its own and pro-
ducers at the less costly place gain another submarket. The discussion in Section 5.4
extends this story by allowing local supply curves to be upward-sloped; we get local-
ization but no longer necessarily the full cessation of production at a more costly
place; depending on the slopes of the local supply curves, there may continue to be
some local production even in the presence of shipments from less costly producers
elsewhere.
Can we use this kind of model to answer the question of where production might
go in the future? The answer here depends on how we think of a local supply curve:
i.e., on our rationale for localization. This chapter envisages that each place has its
own unique place-intrinsic supply curve. Its supply curve is invariant with respect
to the quantity supplied at any other place. This might be attributable to any of a
number of factors that give advantage to production at that place: e.g., a favorable
climate or proximity to inputs needed for production. In that case, if demand in
total were to grow over time, upward-sloped supply curves imply that there would
come a time when firms would start to produce at a place that had previously been
inefficient. The problem here is that when we observe the production costs incurred
by a given set of producers at their current locations, it is difficult to know whether
new producers at those sites in the future would incur similar costs. What explicitly
is the rationale for localization here?
In Chapter 1, I argue that prices are important in shaping the location of firms.
As in Chapter 4, I continue to assume that the prices of inputs used by the firms
36 Takayama and Judge (1971, p. 136), perhaps unaware of their significance, refer to these simply
as “irregular cases”.
5.9 Final Comments 149
in local production are implicit in the intercept and slope of local supply curves.
Arbitrageurs face input prices that are implicit in s. As in Chapter 4, consumers
have income constraints and face prices for other goods and services that determine
the intercept and slope for their demand curve. As in Chapter 4, these other prices
are all determined in markets outside the scope of the models in Chapter 5. What we
have achieved in this chapter is the ability to analyze a simultaneity among multiple
submarkets. This is an important step forward. However, I think Walras would have
argued that the analysis has been only partial in the sense that we have not looked
explicitly at the simultaneity among prices of distinct commodities. Because this
chapter, like Chapter 4, looks at competitive markets only, it does not consider the
issue of local monopoly in space arising from unit shipping costs. Once again, we
must wait until a later chapter for opportunities to address these matters.
As in Chapter 4, the models in this chapter do not say anything about the distri-
bution of income in society among units of labor, capital, and land. As in Chapter 2,
the firm incurs costs, but these are not related explicitly to labor or land inputs.
All income gains that can arise (e.g., because of an increase in α or a decrease in
Ci or s) accrue to firm owners in the form of increased profit. As in Chapter 4, any
adjustment of market price affects the well-being of consumers; in this chapter, such
changes are being measured by consumer surplus.
Chapter 6
Ferrying Inputs and Outputs
Factory Location in a Non-ubiquitous World
(Weber–Launhardt Problem)
A firm is considering where to locate a factory. The firm employs a given tech-
nology to make its product. The firm incurs a cost to ship product to a customer
place. The firm uses two non-ubiquitous inputs—each available only at a specified
location—that it then ships to the production site. All shipping costs are propor-
tional to distance. All other inputs are ubiquitous. What location for the factory
will maximize the firm’s unit profit? In Model 6B, geography takes the form of
a rectangular plane. Model 6A (location on a line) is a simplification of Model
6B that illustrates important ideas about localization (here, clustering with a cus-
tomer or supplier). Model 6A is also helpful in thinking about location on a network
(Model 6F). Model 6C introduces substitutability of inputs and the possibility of
economies of scale. Model 6D makes the prices of output and inputs endogenous.
Model 6E expands on the number of inputs and the number of customer places. This
chapter builds on Chapters 3 through 5. In Chapters 4 and 5, a local supply curve was
assumed at each place that could have a different intercept. Chapter 3 made a similar
assumption about unit production cost varying by location. Chapter 6 explains such
differences in part as a result of the variation in the effective prices of non-ubiquitous
inputs from place to place.
In Chapters 4 and 5, I assumed a local supply curve that could vary from one place
to the next. However, I did not ask why the supply curve differed among places.
At the same time, we have seen how, in the market for an output, shipments arise
that bring prices into a spatial equilibrium. Since the outputs of some firms are the
inputs of others, the implication here is that the cost of production will vary from
place to place for the firms’ customers.
In Chapter 2, I assumed that the firm had the same costs wherever it chose to put
up its factory. I presented there a model in which the firm chooses where to build
a factory (or factories). That model compares the unit shipping cost to a market
from elsewhere with the opportunity cost of the capital required to build a separate
factory there. One of the assumptions of that model is that other costs of production
are the same wherever we construct a factory. However, costs of production may
well vary from place to place: e.g., local tax rates may differ as may labor costs or
energy costs. In this chapter, I focus on the role of shipping costs on inputs that the
firm uses to produce its output. Sometimes, the firm can purchase a unit of the input
at the same cost wherever they locate. Such inputs are ubiquitous.1 The model in
Chapter 2 assumes that all inputs are like this. In this chapter, I assume some inputs
are not ubiquitous; they are available only from particular places(e.g., supply points
or basing points), and the firm incurs a unit shipping rate in transporting those inputs
to the production site. In that case, under the assumption that the factory builds
only one factory, what is the best location for the factory? Will there be conditions
under which the firm would cluster with (i.e., choose a location beside) a supplier
or customer?
This chapter is also a first point to begin thinking about the role of substitutability.
After all, if the prices of non-ubiquitous commodities vary from one site to the
next, the firm may well want to substitute in favor of an input that is less costly.
In the early models presented in this chapter, I assume—for simplicity—a Leontief
technology wherein substitution is not possible. Later, in the chapter, I introduce a
Cobb-Douglas production function2 that does allow for substitution.
This chapter is also an appropriate spot to introduce ideas about conservation
and sustainability. In this chapter, the firm gathers inputs from various places and
uses them to manufacture a commodity then sold (possibly elsewhere). In so doing,
the firm incurs shipping costs. One simple way to think about unit shipping cost is
that it depends on the weight of material being shipped.3 If a manufacturing process
creates a lot of waste material (i.e., the weight of the input materials greatly exceeds
the weight of the output that the firm sells), then the firm might well locate the
factory closer to the site of inputs. I have structured the models in this chapter to
facilitate consideration of such matters.
In thinking about the effect of shipping costs here, I assume in this chapter, for
each commodity, the shipping rate per kilometer per unit shipped is constant; there
are neither economies of the large haul nor economies of the long haul. To illustrate
this point, consider the use of an input inventory by the firm. It is not uncommon for
firms to find that, the larger the size of shipment, the cheaper it is to ship per unit of
a commodity: i.e., economies of the large haul. The firm must, therefore, consider
4 I assume here that all the costs of inventory are borne by the firm. Under certain circumstances, a
supplier may have an incentive to get the customer firm to keep a larger inventory. In the context of
inventory holding by retail outlets and resale price maintenance by a manufacturer, see Deneckere,
Marvel, and Peck (1996).
5 Krzyzanowski (1927, pp. 281–284) is an early and thoughtful overview of Weber’s contributions
to thought in this area.
6 A cwt (hundredweight) is 1/20 of a ton: in the American measurement system, it is 100 lb
(45.4 kg); in the metric system, it is 50 kg.
7 The Leontief coefficients (a ) are here treated as givens; it is as though there is no other way to
i
produce the good other than through this fixed combination of inputs. Such an approach ignores
the idea that, elsewhere, the same (or complementary) good may be produced using a different
combination of inputs available locally. Further, the notion that a good is produced using other
commodities as inputs raises the prospect that the output itself might be an input to the production
of yet another good. This raises important questions about production chains: the sequencing of
production of various components in an industrial economy. See McCann (1995).
8 An extensive literature that reworks the Weber–Launhardt model to incorporate substitutability of
inputs is generally thought to have started with Moses (1958).
9 In this book, a representation of the world as though it were simply a two dimensional surface.
Ignored here, for simplicity of exposition, are (1) spherical properties of a globe and (2) differences
in elevation. On a rectangular plane, a geographic point can be represented simply by a pair of
Cartesian coordinates.
10 Representation of a point in two-dimensional geographic space by a pair of distance coordinates:
(x, y). An example of such a planar representations are is given by Universal Transverse Mercator
(UTM) coordinates which take the form of easting and northing coordinates. Cartesian coordinates
154 6 Ferrying Inputs and Outputs
are an approximation to location in that they ignore both elevation and the fact that the Earth
is a sphere. The advantage of Cartesian coordinates is that distance calculations are simplified
compared to spherical coordinates.
11 The square of the length of the hypotenuse of a right angle triangle in two-dimensional space is
the sum of the squares of the other two sides.
12 In two-dimensional space, the straight-line distance between two points with Cartesian coordi-
√
nates (x,y) and (x2 ,y2 ) is given by ((x1 − x2 )2 + (y1 − y2 )2 ).
13 As there is no transportation network here, we don’t have to worry about congestion on that
network.
14 A condition of the demand curve facing a firm or an industry whereby consumers willingly
demand any quantity at a given price but are unwilling to pay any more. Also known as a horizontal
demand curve.
15 For the reader who wants a taste of the effects of uncertainty, see Asami and Isard (1989),
Cromley (1982), Dean and Carroll (1977), Hsu and Tan (1999), Mai (1984, 1987), Mai, Yeh, and
Suwanakul (1993b), Mathur (1983, 1985), and Stahl and Varaiya (1978).
6.2 Model 6A: I = 2 Input Places, J = 1 Output Place; Location on a Line 155
profitable at some places compared to others.16 To the restless reader, such assump-
tions may seem unrealistic. After all you might say, as in Chapter 2, is it reasonable
to assume a firm at a given location can produce any quantity at all at the same
unit cost? Doesn’t congestion set in when firms become large? Isn’t it possible that
managerial inefficiency creeps in depending on the scale of production? When the
firm becomes sufficiently large, isn’t there a scale where it is no longer a price taker
in markets for its inputs and perhaps also for its output? The answer may well be
yes to all of the above. In my defense, however, this chapter is an extension of the
model presented in Chapter 2 (where all inputs were assumed ubiquitous). In this
chapter, I explore the role of non-ubiquitous assumptions. In later chapters, I will
release other assumptions. As I do, the analysis becomes more difficult. The reason
for holding these assumptions at the moment is to enable a better understanding of
the role played by the particular assumption that has just been released.
Finally, this is the first chapter where I need to be specific about the geometry
of geographic space. In earlier chapters, I refer to the distance between two places
but have not been specific about how this distance might be calculated or measured.
That changes in this chapter because I need to measure how a change in the firm’s
location affects distances, and hence shipping costs, to its suppliers and customers.
The impatient reader might say “why assume the world is flat as is done here when
in fact the world is an oblate spheroid?” Good question! The reasons are twofold.
First, it is much easier to solve the Weber–Launhardt problem using rectangular
distances than spherical distances. Second, provided that the places overall are not
too far apart, rectangular distances are a good approximation to spherical distances.
Consider the case where the firm uses 2 inputs each purchased from a single place
to produce a commodity sold in a market located at a third place. The firm seeks a
site for its factory that maximizes profit per unit output. See (6.1.1) in Table 6.1. The
ubiquitous inputs, which might include labor and capital as well as normal profit,
cost the firm a fixed f dollars per unit output regardless of location. Only inputs 1
and 2 are non-ubiquitous.
In this section, I consider the case where the three places lie along a straight line
with the market place xa kilometers east of input Place 1 and xb kilometers west of
input Place 2.17 See the map in Fig. 6.1. Initially assume the firm considers putting
its factory only at one of these three places. The firm’s profit per unit produced
would then be as shown in (6.2.1), (6.2.2), and (6.2.3). See Table 6.2 wherein I
summarize equations, assumptions, notation, and rationale for localization used in
Model 6A. Between Places 1 and M, Place 1 is the more profitable if freight savings
16 For an extension that incorporates external economies, see Mai and Hwang (1997).
17 Sakashita (1968) also used a line market in his modeling of Weber’s location problem.
156 6 Ferrying Inputs and Outputs
on input 1 is larger than the additional shipping cost for input 2 and the output: see
(6.2.4). Now compare the profit if the factory locates at Place 2 vs. M. Place 2 is
more profitable if the freight savings on input 2 exceed the additional shipping costs
for input 1 and the output: see (6.2.5). Finally, compare Place 1 and 2. Place 2 is
the more profitable if the freight savings on input 2 vis-à-vis Place 1 are larger than
the additional shipping costs on input 1 taking into account the change (positive or
negative) in output shipping cost: see (6.2.6).
So far, I have considered a factory location only at Places 1, M, or 2. Might
another place be even more profitable?
• We can exclude sites off the line passing through places 1, M, and 2 because any
such site would involve longer distances and a smaller unit profit than some point
on the line 1M2.
• By a similar argument, the firm would never choose a location on a line extended
west of input Place 1 or east of input Place 2 because the firm could always
improve on profit by choosing instead input Place 1 or 2, respectively. In other
6.2 Model 6A: I = 2 Input Places, J = 1 Output Place; Location on a Line 157
Notes: See Fig. 6.1. The three places lie along a straight line with the market place xa kilo-
meters east of input Place 1 and xb kilometer. Rationale for localization (see Appendix A):
Z8—Limitation of shipping cost or travel delay; Z9—Shipping cost on nonubiquitous inputs.
Givens (parameter or exogenous): ai —Amount of input i (cwt) needed per cwt of output; f—Cost
per unit output of ubiquitous inputs; P—Price received per cwt of output delivered to the market
place; Pi —F.o.b. price paid per cwt of input i; s—Cost to ship cwt of output a distance of 1 unit;
si —Cost to ship cwt of input i a distance of 1 unit; xa —Distance east of input Place 1 to market;
xb —Distance west of input Place 2 to market.
words, the most profitable location always has to be somewhere along the line
segment from Place 1 to Place 2.
• Finally, the firm would never find an intermediate location along the line segment
more profitable than locating at one of 1, M, or 2. To see why, imagine that we
calculate the unit profit for any intermediate location along the part of the line
segment between Places 1 and M. Now imagine that we calculate the profit for a
place 1 km further to the west but still on the line segment between Places 1 and
M. The freight savings for input 1 per unit output each kilometer closer to Place
1 is a1 s1 , for input 2 per each kilometer closer to Place 2 is a2 s2 , and per unit
output each kilometer closer to Place M is s. See (6.2.7), (6.2.8), and (6.2.9).18
18 An Isodapane is the name given to the locus of all geographic points on a map that have the same
total transport cost. See Isard (1951), Lindberg (1953), Pred (1965), Smith (1966), Koropeckyj
(1967), Feldman (1976), Ohta and Thisse (1993), and Puu (2003).
158 6 Ferrying Inputs and Outputs
Since the move 1 km west means that we are 1 km closer to Place 1, we would
save a1 s1 dollars on freight for input 1. Since we would then be 1 km further away
from Places M and 2, we incur an additional cost of a2 s2 + s on freight for input
2 and the output, respectively. If a1 s1 > a2 s2 + s, we earn more profit moving
1 km closer to Place 1. However, since this would then be true for each and every
kilometer, we would want to move the factory site even further west until we
reach Place 1. If, on the other hand a1 s1 < a2 s2 + s, by a similar argument, we
would keep moving east until we reached Place M. Where a1 s1 = a2 s2 + s,
any location along the line segment from Place 1 to Place M is equally profitable;
however, even here no intermediate location is more profitable than Places 1 or
M. By analogy, the same argument works for factory sites on the segment from
Place M to Place 2.
Put differently, because unit profit is linear in distance along the line segment
between 1 and M as well as the line segment from M to 2, I need look at only
three sites (1, M, and 2) to find the maximum unit profit. That we can exclude any
intermediate place is sometimes called the Exclusion Theorem.19 The unit profit
is maximized at Place 1 if a1 s1 > a2 s2 + s; it is maximized at Place 2 if a2 s2 >
a1 s1 + s; else (s > |a1 s1 − a2 s2 |) it is maximized at Place M. So far, I have assumed
that Place M is between Places 1 and 2; by an analogous argument, conditions can
be derived for location on a line when either Place 1 or Place 2 is the middle point.
These results are summarized in Table 6.3.
Let me add a brief aside about the special case where the cost of shipping a cwt
is the same for all commodities: i.e., s = s1 = s2 . In this case, we need only
look at the weight of commodity 1 (a1 ) and of commodity 2 (a2 ) compared to the
1 unit of output they produce. The production process is said to be weight-losing
(presumably involving solid waste) if a1 + a2 > 1 or weight gaining—as, for
example, when ubiquitous local water is added to other inputs to make an output—
when a1 + a2 < 1. Here (6.3.1) reduces to a1 > a2 + 1 which means that the
factory is located at Place 1 when the weight of input 1 used to produce one unit
of output is relatively large. Similarly, (6.3.2) implies that the factory is located at
Place 2 when the weight of input 2 is relatively large. In either case, this is generally
consistent with the idea the production process is weight-losing. In contrast, (6.3.3)
implies that when the weight of output (1 cwt) is large relative to the other two,
the factory is sited at the market; this is consistent with the idea that the production
process is weight-gaining.
What about comparative statics here? How do variations in the givens (a1 , a2 ,
xa , xb , s1 , s2 , and s) affect the choice of most profitable location? From the previous
paragraph, xa and xb have no effect at all on location. From Table 6.3, we see that a
small change in a1 , a2 , s1 , s2 , or s has no effect on the efficient location of a factory.
Put differently, these givens have no effect on the efficient location unless the change
is sufficient to cross one of the thresholds in Table 6.3.
Now, consider the case where Places 1, M, and 2 do not lie on a line. Instead, they
are three points on a rectangular plane. In this case, the firm seeks a place for its
factory—namely a Cartesian coordinate pair (X, Y)—that maximizes profit per unit
output. Location (X, Y) determines distances to the market, input 1, and input 2:
xm , x1 , and x2 , respectively. See (6.1.2), (6.1.3), and (6.1.4). These three places in
general form what is referred to as the Weber–Launhardt map triangle.20 Without
loss of generality, I assign Cartesian coordinates (0, 0) to Place 1, (X2 , 0) to Place 2,
and (Xm , Ym ) to the market place.
A simple and intuitive graphical analysis helps us narrow down the area of the
map within which the firm might locate. See Fig. 6.2, which shows a map containing
the map triangle M12M. Consider a place outside the map triangle as represented by
point A. The shortest path21 from A to input Place 1 is the straight line A1, from A
to input Place 2 is A2, and from A to the market is AM. Now, using M as the center,
draw a circle that passes through A: i.e., has radius AM. As drawn, the circle crosses
the map triangle at Place B. Place B, by construction, is the same distance from M
as is Place A. However, Place B is closer to both input places, 1 and 2, than is A.
Therefore, by (6.1.1), a factory at B has to be more profitable than a factory at A. Of
course, a site inside the map triangle may be even more profitable. In conclusion,
a site along the edge or inside the map triangle is always to be preferred to a site
outside the map triangle.
However, where within or on the map triangle is the most profitable site for our
factory? To keep matters simple for the moment, assume the most profitable site is
inside the map triangle (i.e., not at vertex 1, M, or 2). In that case, I can maximize
20 In the I = 2, J = 1 version of the Weber–Launhardt Model, the triangle formed when three
map points on a rectangular plane—each with a pair of Cartesian coordinates—are drawn on a map.
21 Two definitions here: (i) on a two-dimensional plane, the Euclidean distance between two points;
(ii) on a network, the path from one vertex to another that is the shortest.
160 6 Ferrying Inputs and Outputs
1 2
0 50 100
km km km
unit profit in (6.1.1) by partially differentiating with respect to the factory’s coordi-
nates, X and Y, and then set the partial derivatives to zero. See (6.4.1) and (6.4.2) in
Table 6.4. However, there are two problems with this approach. First, since (6.4.1)
and (6.4.2) do not yield explicit equations for X and Y, it is hard to draw conclu-
sions about how the factory site might shift in response to a given change in any one
parameter. Second, (6.4.1) and (6.4.2) presume that the most profitable site is not at
any of 1, M, or 2 but do not say anything about when this might happen.
Table 6.4 Model 6B: first-order conditions to the Weber–Launhardt factory location problem
assuming interior solution (i.e., factory not at 1, M, or 2)
Notes: Rationale for localization (see Appendix A): Z8—Limitation of shipping cost or travel
delay; Z9—Shipping cost on nonubiquitous inputs. Givens (parameter or exogenous): ai —Amount
of input i needed per unit of output; s—Cost to ship cwt of output a distance of 1 unit; si —Cost
to ship cwt of input i a distance of 1 unit; Xi —X-coordinate of Place i; Ym —Y-coordinate of Place
M. Outcomes (endogenous): X—X-coordinate of factory; xi —Distance from factory to Place i;
Y—Y-coordinate of factory. See also Table 6.1.
6.3 Model 6B: I = 2 Input Places, J = 1 Output Place 161
β3
a2s2 a1s1
22 In this, he collaborated with Georg A. Pick, a professor of Mathematics at the German University
of Prague at the time. Figures like Fig. 5.4 are therefore sometimes called “Pick’s diagram.” See
Ellison (1991).
23 In the I = 2, J = 1 Weber–Launhardt location problem, there are four points on a two-
dimensional map; the output point, the production point, and two input points. The marginal
shipping cost is the extra profit per unit output that could be earned if only the production point
were 1 km closer to another point. Marginal shipping costs are w1 , w2 , and wm .
24 In Mathematics, this is also known as the Steiner problem.
25 In the I = 2, J = 1 version of the Weber–Launhardt Model, the “equilibrium of forces” triangle
created using the weights w1 , w2 , and wm . See Marginal Shipping Cost.
162 6 Ferrying Inputs and Outputs
wm
γ2 P γ1
γ3
w1 w2
1 2
0 50 100
km km km
Fig. 6.4 Model 6B: interior solution for sample Weber–Launhardt factory location problem.
Notes: See also Fig. 6.3. γ 1 , Angle (g1 in diagram) formed at interior production place opposite
input site 1. Angles γ 1 and β 1 are supplements: γ1 + β1 = 180◦ , γ 2 Angle (g2 in diagram) formed
at interior production place opposite input site 2. Angles γ 2 and β 2 are supplements: γ2 + β2 =
180◦ γ 3 , Angle (g3 in diagram) formed at interior production place opposite market site M. Angles
γ 3 and β 3 are supplements: γ3 + β3 = 180◦
How are we to find the site that forms such angles? A geometric solution by Pick
is one possibility. Referring to Fig. 6.5 to illustrate, the steps are as follows
1. Scale the weight triangle until the side of length s corresponds to the distance
from input Place 1 to input Place 2. See (6.5.1) in Table 6.5. Draw the weight
triangle upside down so that it hangs below the line joining input places 1 and 2
and such that the a1 s1 side is opposite input Place 1 and the a2 s2 side is opposite
input Place 2. Label the bottom vertex of this weight triangle as W. In Fig. 6.5,
the weight triangle is now the dotted W12W.
2. Draw a circle that passes through all three vertices of triangle W12. The easiest
way to do this (thinking back to high school geometry) is to draw the perpendicu-
lar bisector to each of two sides of the triangle. The place where the two bisectors
intersect is the center of this circle. The center of the circle is C in Fig. 6.5. See
(6.5.4), (6.5.5), and (6.5.6)
3. Draw a straight line from W to the market site M. If there is an interior solution,
it is the place where this straight line crosses the circle drawn in step 2. The most
profitable production site is indicated by a P in Fig. 6.5. See (6.5.7) through
(6.5.13)
6.3 Model 6B: I = 2 Input Places, J = 1 Output Place 163
γ2 P γ
1
γm
1 2
β1 wm β2
w2 w1
βm
0 50 100
km km km
In general, these give the coordinates of the most profitable location as long as
the solution is interior to the Weber–Launhardt map triangle.
Why does this work? In essence, the geometric construction ensures that the
angles formed about Place P are the supplements of the angles in the weight trian-
gle. Consider ∠1P2. Since
P12 and
W12 have the same circumcircle and share
the same chord 12 with P to one side and W to the other, ∠1P2 and ∠1W2 are sup-
plements. Since ∠1W2 = βm , ∠1P2 = γm . As a consequence of this circle, the
other two angles around P, ∠1PM, and ∠MP2 must be supplementary to β 2 and β 1 ,
respectively, and the Places M, P, and W must be collinear.
However, under certain conditions, the most profitable location is at a vertex of
the map triangle and not interior to it. Such cases constitute corner solutions.26 What
are those conditions? Let ϕ1 = ∠M12 on the map triangle; let ϕ2 = ∠12M and let
ϕ3 = ∠1M2. These are geographic angles because they each depend only on the
locations of the three map points. In Fig. 6.4, γ1, is larger than the geographic angle
at the Place 1 vertex. However, were I to change the weight triangle in such a way
that the most profitable location is now closer to Place 1 (by increasing w1 relative
to the other two weights), then γ1 would shrink. In fact, if w1 were to be increased
enough relative to the other two weights, the most profitable location would be at
Table 6.5 Model 6B: algebraic solution to the I = 2, J = 1 Weber–Launhardt factory location
problem assuming interior location
Table 6.6 Model 6B: conditions under which factory is located at vertex in the I = 2, J = 1
Weber–Launhardt factory location problem
shipping cost associated with that vertex is sufficiently large relative to the shipping
costs associated with other vertices. It is relatively easy to show at most that one
of these conditions can be true. I refer above to relative geography. Note (6.6.1),
(6.6.2), and (6.6.3) use the three map angles but take no account of the distances
between places. The localization conditions do not depend on whether the places are
1 km or 1,000 km apart. This is different from the model in Chapter 2 where, given
a shipping rate, there was a certain minimum distance beyond which the firm finds
it profitable to construct a second factory (in that freight savings are larger than the
opportunity cost of the capital needed to build the second factory). In this chapter,
however, the conditions for localization depend on map angles rather than distances.
In the preceding section of this chapter, I considered the case where Places 1, M,
and 2 formed a line in geographic space. In that case, ϕ1 = 0◦ so cos [ϕ1 ] = 1;
ϕ2 = 0◦ so cos [ϕ2 ] = 1, and ϕ3 = 180◦ so cos [ϕ3 ] = − 1. In this situation, the
localization conditions (6.6.1), (6.6.2), and (6.6.3) reduce to the following. Locate
at Place 1 if w1 ≥ w2 + wm ; at Place 2 if w2 ≥ w1 + wm ; at Place M if wm ≥
|w1 − w2 |. These are, of course, the same conditions I had derived earlier.
What about comparative statics in this model? There are two possibilities here.
One is when the least cost solution is at a vertex of the Map Triangle. In that case, a
small change in any of the givens results in no change in the optimal location. The
other possibility is that the least cost solution is an interior place within the Map
Triangle. Here, an increase in any one given draws the least cost solution closer to
the relevant vertex. An increase in a1 or s1 draws the firm closer to input Place 1; an
increase in a2 or s2 draws it closer to Place 2; an increase in s draws it closer to M.
To this point, I have envisaged a firm as making two decisions: (1) location of its fac-
tory and (2) scale of production there. As noted above, I set up the Weber–Launhardt
model in such a way that scale itself is indeterminate. The linearity of a Leontief
production technology separates those two decisions in the sense that the firm can
decide where to locate irrespective of how much it will produce there. Students often
say to me that this separation between the effects of location on marginal cost and
quantity supplied on P (price) seems strange. Would it not be more appropriate, they
say, to have a model where location and scale of production are jointly determined?
To illustrate, let us consider here the case of a firm that is about to increase
substantially its output. Two plausible situations come to mind here, wherein the
firm’s choices of location and quantity will be interdependent.
• One is that the firm will now find it profitable to alter the mix of its non-ubiquitous
inputs.27 For some reason, it is more profitable to produce the firm’s commodity
27 Nijkamp and Paelinck (1973) propose a grid search method to investigate the impact of
substitutability among inputs.
166 6 Ferrying Inputs and Outputs
in a different way when the quantity being produced is larger. Put differently,
the firm’s production function does not yield a linear expansion path. Typically,
this is because of an indivisibility in production. If the scale of output is too
small, the firm can’t take advantage of the indivisibility and so has to produce the
commodity as best it can using some other mix of inputs. In this case, as the scale
of production is increased, the firm seeks to locate closer to sources of inputs that
will now be more important.
• The other is that the amount of each non-ubiquitous input needed per unit of
output shrinks when output is increased. This kind of economy of scale typically
happens when the firm is better able to make use of its materials (i.e., reduce
S wastage) the larger the scale of output. In this case, as scale of production is
increased, the firm finds its optimal location now less dependent on the cost of
shipping inputs and relatively more dependent on the cost of shipping product to
market.
These two situations illustrate the idea that separability of scale and location
arises because I have so far assumed (1) linear expansion paths and (2) constant
returns to scale in production.
In part, what has driven location theorists here is the desire to integrate their
theory with the neoclassical theory of the firm wherein input substitution and
economies of scale are important. In neoclassical theory, much of the emphasis
here has been on the substitution between labor and capital in production. However,
we need to remember here that a Weberian model is necessarily focused on non-
ubiquitous inputs. If labor and capital are ubiquitous inputs, the substitution between
them in itself will not affect the outcome of a Weberian model. Location is affected
only when there is substitution between non-ubiquitous inputs. For that reason, here,
I will not pursue further the implications of nonlinear expansion paths. However, the
possibility of reduced wastage with larger production levels makes it useful to think
about incorporating economies of scale into a Weberian model.
Modern attempts to incorporate production functions other than a Leontief tech-
nology started with Moses (1958) and found a fuller statement in Khalili, Mathur,
and Bodenhorn (1974).28 Let me here use a model (Model 6C) that emphasizes the
role of economies of scale. As this model can be solved only numerically in gen-
eral, I will use the numerical values shown in (6.7.7) through (6.7.12) of Table 6.7.
Unless otherwise stated, I retain all assumptions from Model 6B. Assume here that
there are N consumers at Place M and that each demands q units of the commodity
at a given price P. The firm incurs an annual fixed cost (F). The firm’s total cost,
to be minimized, is given by (6.7.1). Now, assume that the firm chooses an input
combination (Q1 , Q2 ) sufficient—in terms of its Cobb-Douglas production function
28 See also Sakashita (1968), Bradfield (1971), Hurter and Wendell (1972), Tellier (1972). Emerson
(1973), Woodward (1973), Thisse and Perreur (1977), Miller and Jensen (1978), Mathur (1979),
Eswaran, Kanemoto, and Ryan (1981), Alperovich and Katz (1983), Mai (1984), Hsu and Mai
(1984), Kusumoto (1984, 1985), and Kilkenny and Thisse (1999).
6.4 Model 6C: Substitutability, Scale, and Location 167
Table 6.7 Model 6C: cost minimization in the I = 2, J = 1 Weber–Launhardt factory location
problem: Cobb–Douglas production function
Notes: Rationale for localization (see Appendix A): Z8—Limitation of shipping cost or travel
delay; Z9—Shipping cost on nonubiquitous inputs. Givens (parameter or exogenous): A—Scalar
in production (A > 0); f—Cost per unit output of ubiquitous inputs; F—Firm’s fixed cost; N—
Number of consumers at market place; P—Price received per unit of output delivered to the market
place; Pi —F.o.b. price per unit of nonubiquitous input i; q—Amount of commodity demanded by
each consumer; s—Cost of shipping 1 unit of output a distance of 1 km; si —Cost of shipping
1 unit of input i a distance of 1 km; X2 —X-coordinate of Place 2; Xm —X-coordinate of market
place; Ym —Y-coordinate of market place; γ —Exponent of input 2 in production function; ε—
Exponent of input 1 in production function. Outcomes (endogenous): Q—Firm’s output (scale of
production); Qi —Quantity of input i used in production; X—X-coordinate of factory; xi —Distance
to supplier of input i; xm —Distance to market; Y—Y-coordinate of factory.
to—meet demand: see (6.7.2). Cost minimization means that it chooses a combina-
tion that satisfies (6.7.4). Finally, as in Model 6B, assume the firm also chooses a
location for its factory to minimize total cost: see (6.7.3).
The least cost site for production can be the same as in Model 6B. For a given
quantity of output needed (Q), the solution to Model 6C consists of input levels
(Q1 , Q2 ) and a factory location (X, Y). From these, estimate the Leontief production
168 6 Ferrying Inputs and Outputs
Input 2
G
I F
0
H Input 1 B
we can solve even the case where αβ > 1 here because we have assumed total quantity
29 Here
demanded (i.e., Nq) is fixed.
6.4 Model 6C: Substitutability, Scale, and Location 169
M
C
1 2
0 50 100
km km km
30 Mai and Hwang (1997) look at Cournot-Nash competition among firms in the presence of exter-
nal economies. When external economies prevail, constant returns to scale at the firm level is not
a sufficient condition for ensuring the separability of location and scale. Moreover, they show that
when free entry is allowed and the production function exhibits decreasing returns to scale, but
170 6 Ferrying Inputs and Outputs
with strong external economies, the optimal location moves toward or away from the market as
demand increases according to whether the demand function is convex or concave.
31 Of course, I assume here that the profit is sufficiently large; otherwise a firm would choose no
factory at all.
32 Before leaving Model 6D, it is instructive to think about how competition might shape the down-
ward sloping demand curve faced by firm. I am thinking here of the effect of entry by competitors
producing the same product.32 Another aspect concerns the effect of competition on the price and
location of a firm. The key here is the conjectural variation. Under some assumptions of how a
competitor might react to the presence of another, it is possible to get different locations for each
competitor. See Mai and Hwang (1994).
6.6 Model 6E: More Than 2 Input Places and/or More Than 1 Output Place 171
Table 6.8 Model 6D: the I = 2, J = 1 Weber–Launhardt factory with a downward-sloping demand
curve and upward-sloping supply curves for its inputs
Profit
(P − C)Q − F (6.8.1)
Demand
P = α − βQ/N (6.8.2)
Unit cost
C = f + a1 (P1 + s1 x1 ) + a2 (P2 + s2 x2 ) + sxm (6.8.3)
Inverse supply curve for input 1
P1 = C1 + δ1 (a1 Q) (6.8.4)
Inverse supply curve for input 2
P2 = C2 + δ2 (a2 Q) (6.8.5)
Distances from factory to input points 1 and 2 and market place M
See (6.1.2), (6.1.3), and (6.1.4) (6.8.6)
Notes: See also Table 6.1. Rationale for localization (see Appendix A): Z8—Limitation of shipping
cost or travel delay; Z9—Shipping cost on nonubiquitous inputs. Givens (parameter or exogenous):
α—Intercept of individual demand curve; β—Slope of individual demand curve; ai —Amount of
input i needed per unit of output; C—Unit cost (production and shipping); Ci —Intercept of supply
curve for input i; f—Cost per unit output of ubiquitous inputs; F—Fixed cost; N—Number of cus-
tomers at Place M; s—Cost of shipping 1 unit of output a distance of 1 km; si —Cost of shipping
1 unit of input i a distance of 1 km; Xm —X-coordinate of market place; X2 —X-coordinate of Place
2; Ym —Y-coordinate of market place; δ i —Slope of supply curve for input i. Outcomes (endoge-
nous): C—Unit cost (production and shipping); P—Price received per unit of output delivered to
the market place; Pi—F.o.b. price of input i; Q—Quantity of output produced; X—X-coordinate
of factory; xm —Distance from factory to market place; xi —Distance from factory to Place i;
Y—Y-coordinate of factory.
6.6 Model 6E: More Than 2 Input Places and/or More Than 1
Output Place
Another extension is to consider more than 2 input places and/or more than one out-
put place. In such a case, none of the methods developed above is of use. However,
the Kuhn-Kuenne algorithm33 first published in 1962 solves this problem numer-
ically. Where we have more than two input places, but just 1 output place, the
problem remains one of minimizing unit cost. When more than 1 output place is
involved, the problem becomes one in which we maximize unit profit. See Table 6.9.
Maximize (6.9.1) subject to (6.9.2) and (6.9.3). For each input Place i, the algorithm
requires knowledge of the Leontief coefficient, ai , and the shipping rate per kilome-
ter, si . For each market place j, the algorithm requires the shipping rate, s, as well as
the quantity to be supplied to that market, Qj . The same problem can now generally
be solved also by standard optimization software.
Table 6.9 Model 6E: the Weber–Launhardt factory location problem with I input places and J
output places
Profit
j (Pmj − sxmj )Qj − {f + i ai (Pi + si xi )}j Qj − F (6.9.1)
Euclidean distance from factory site to market place J
√
xmj = ((X − Xmj )2 + (Y − Ymj )2 ) (6.9.2)
Notes: Rationale for localization (see Appendix A): Z8—Limitation of shipping cost or travel
delay; Z9—Shipping cost on nonubiquitous inputs. Givens (parameter or exogenous): ai —Amount
of input i needed per unit of output; f—Cost per unit output of ubiquitous inputs; F—Fixed cost;
Pmj —Price received per unit output delivered to market place j; Qj —Quantity sold at market place
j; Pi —F.o.b. price paid per unit of input i; s—Cost to ship unit of output a distance of 1 unit;
si —Cost to ship unit of input i a distance of 1 unit; Xi —X-coordinate of input place i; Yi—Y-
coordinate of input place i; Xmj —X-coordinate of market place j; Ymj —Y-coordinate of market
place j. Outcomes (endogenous): X—X coordinate of factory site; xi —Distance from input place i
to factory.; xmj —Distance from factory to market place j; Y—Y-coordinate of factory site.
In this extension, the firm will still locate near or at an input or output place if the
unit cost saving by moving a kilometer closer to that place is substantially greater
than the saving from moving toward any other place. Further, the comparative statics
in this model are the same as in Model 6B. When the least-cost solution is not at an
input or output place, a small change in any of the givens results in no change in the
optimal location. When the least-cost solution is not at an input or output place, an
increase in any one given draws the least-cost location closer to the relevant place.
An increase in ai or si draws the firm closer to input place i; an increase in s draws
it closer to an output place. As before, an increase in total output—specifically an
increase spread proportionally among existing demand places—has no effect on
location. However, the introduction of demand at a new place, or a relative shift in
demands at existing places, can affect the most profitable site for the firm.
transportation network
103
99
80
98
H
G 61 I
F 100 91 J
131
94
91
87
91
91
14
9
M
L
K 100 91 161 N
0 50 100
km km km
Now, let us examine any one link on that transportation network: say for the sake
of illustration the east–west link GH in Fig. 6.8. At either end of that link is a vertex.
Imagine a production site at an intermediate location along that link. All inputs and
outputs must be transported along part of the link because they all come from (or
through) the vertex to the east (H) or the vertex to the west (G). We can readily
calculate the profit per unit possible at this site from (6.1.1). Let us then consider a
second possible production site a little further west along this link. Recalculate the
profit per unit output at this new location. There are three possibilities here: profit
has (1) increased, (2) stayed the same, or (3) gone down. In fact, because (6.1.1) is
linear in distances, each kilometer I move the factory from east to west changes the
profit per unit output by exactly the same amount.
The implication is that the profit per unit product is never higher in mid-link than
it is at one or the other vertex (or possibly both). This is known in location theory as
the Hakimi Theorem.37 To find the most profitable location on a network, therefore,
I need look only at the profit at each vertex. In Table 6.10, I show an example using
the network from Fig. 6.8. Assume I = 2 inputs: one available from vertex A, the
other from C. Assume 1 market place; in this case, vertex N. At the weights (w1 , w2 ,
w3 ) given in Table 6.10, the most profitable site is vertex C since total shipping costs
there are the lowest among all vertices. The calculation of unit profit site-by-site for
the purpose of choosing the best location has long been known in location theory as
Table 6.10 Model 6F: total shipping cost per unit output at each vertex for location on network
where I = 2 and J = 1
i wi Shipping cost
Vertex [1] [2] [3]
A 1 2.00 957
B 857
C 2 2.50 702
D 1,224
E 1,862
F 1,337
G 997
H 923
I 1,084
J 1,645
K 1,663
L 1,369
M 1,290
N Market 1.00 1,689
Notes: Vertices and distances shown in Figure 6.8. Rationale for localization (see Appendix A):
Z8—Limitation of shipping cost or travel delay; Z9—Shipping cost on nonubiquitous inputs.
Givens (parameter or exogenous): None used. Outcomes (endogenous): None used.
comparative cost analysis.38 The advantage created by the Hakimi Theorem is that
I need do comparative cost analysis only at places that are vertices.
What about comparative statics here? From the previous paragraph, distance
along any one link has no effect at all on location unless we start from an initial
condition where total cost is the same everywhere along the link. Except in that
knife-edge situation, we see now as well that a small change in a given has no effect
on the efficient location of factory.
6A 6B 6C 6D 6E 6F
Assumptions [1] [2] [3] [4] [5] [6]
38 A technique for finding the least cost location for a firm by enumerating possible locations and
then calculating the unit cost of production at each.
176 6 Ferrying Inputs and Outputs
(Marshall–Lentnek–MacPherson–Phillips Problem)
A firm has a machine that breaks down periodically. In Model 7A, the firm does
repairs in-house. The firm minimizes overall cost by balancing an inventory of prod-
uct (to meet customer demand during downtime) and an inventory of repair staff. In
Model 7B, the firm outsources repairs to a contractor as needed. There may now
be a delay in starting repairs awaiting the arrival of the contractor’s repair crew.
Outsourcing becomes attractive because of an insurance principle. If the client is
in an industry using similar machines that break down stochastically over time, a
contractor, by redeploying repair staff from client to client, may achieve lower unit
costs than a firm doing repairs in-house. A clustering of clients with the same kind
of machine enables the repair contractor to be more efficient. In Model 7C, the firm
uses cost minimization to choose between in-house repair and outsourcing. Model
7D examines the implications of profit maximization by a repair contractor servic-
ing client firms. This chapter builds on Chapters 4 and 5. In those chapters, a local
supply curve was assumed at each place that could have a different intercept. In
Chapter 6, I explained that difference in part as a result of the variation in the effec-
tive prices of non-ubiquitous inputs. Here in Chapter 7, it is the clustering of clients
with the same kind of machine that enables an efficient repair contractor and thereby
reduces unit cost for client firms. Model 7D parallels Chapter 6 in terms of how a
firm and its supplier (in this case a repair contractor) co-locate. In this chapter, the
repair contractor is a metaphor for any kind of producer service that can be out-
sourced, and the chapter suggests how we might similarly think about shippers and
the endogeneity of the price of shipping and unit shipping cost.
1 See
McCann (1995).
2 Forempirical evidence on localization economies, see Black and Henderson (1999) and
Ó hUallacháin and Reid (1991).
7.1 The Marshall–Lentnek–MacPherson–Phillips Problem 179
3 A consequence on one economic actor (e.g., a firm) arising from the behavior of another that is
not priced.
4 This argument suggests a simple principle. Because of the efficiencies made possible by indivisi-
bilities, a firm will purchase inputs directly or indirectly from a supplier whose scale of production
is much larger than their own and sell to firms whose scale of production is much smaller than
their own.
5 See also Phillips, MacPherson, and Lentnek (1998).
180 7 What the Firm Does On-Site
Firm’s profit
π = pqT − (f + cq)T − piIT − wnT (7.1.1)
Cost to be minimized
C = piIT + wnT (7.1.2)
Average product inventory
I = (1/2)qΔT (7.1.3)
Downtime (fraction of T)
Δ=δ (7.1.4)
Repair time (fraction of T)
δ = a + k/n where 1 > a > 0 and k > 0 (7.1.5)
Marginal effect of repair staff on cost
dC/dn = −pi(1/2)q(k/n2 )T 2 + wT (7.1.6)
Cost-minimizing
√ repair staff (where dC/dn = 0)
n = (kqpiT/(2w)) (7.1.7)
Minimum feasible repair staff (1=1)
k/(1 − a) (7.1.8)
Maximum wage if marginal effect of repair staff on cost is negative at minimum
feasible repair staff
w ≤ pi(1/2)q(1 − a)2 T/k (7.1.9)
Notes: Rationale for localization (see Appendix A): Z4—Risk spreading and insurance. Givens
(parameter or exogenous): a—Minimum repair time (fraction of production cycle); c—Unit vari-
able production cost; f—Daily fixed cost; i—Interest rate (daily); k—Responsiveness of repair time
to size of repair staff; p—Selling price of unit of output; q—Units sold daily; T—Length of produc-
tion cycle (days) including downtime; w—Daily wage of repair worker. Outcomes (endogenous):
C—Repair plus inventory cost to be minimized; l—Average stock of product inventory held-over
period; n—Number of repair workers; δ—Time (fraction of T) required for repair; Δ—Downtime
(fraction of T); π—Profit.
7.2 Inventory Models in Management 181
(exogenous) daily demand of q units must be met; the model does not allow for
unsatisfied demand. Assume also that the fixed cost (f) and marginal cost (c) of pro-
duction are fixed; therefore the firm’s total cost of production is fixed. The model
assumes that the price (p) at which the firms sell its output is given; hence its
sales revenue is also given. The LMP Model instances a classic inventory model.
It assumes that the firm operates production machinery subject to periodic break-
down. The model assumes the machinery breaks down regularly each T days, ΔT
days are required to repair it, and the remaining (1 −
)T days is the length of the
ensuing production period. Each time while the machinery is being repaired, the
firm supplies customers from its inventory of product. Over the repair period, inven-
tory is envisaged to drop steadily to zero see curve AB in Fig. 7.1. When repairs are
complete, production resumes and output is used both to satisfy current customer
demand and to rebuild inventory in anticipation of the next breakdown: see curve
BC in Fig. 7.1. By the end of the production period (OT in Fig. 7.1), inventory has
been returned to its original level, and the firm is ready for its next production cycle.
As all other costs of production are fixed by assumption, it is only the inventory
cost and the repair cost that the firm can vary. The two costs sum to an amount C as
shown in (7.1.2).
A C
Inventory
0
B T
Time
In early studies of inventory within the field of Management, the problem was to
derive a policy for inventory to enhance profit. Holding inventory is seen to be
182 7 What the Firm Does On-Site
costly; typically in terms of storage costs (including rent, lighting, heating, cleaning)
and of the working capital tied up (i.e., imputed interest cost) in holding inventory.
To continue an earlier example, assume an efficient firm that manufactures bicy-
cles. The firm purchases inputs: tires and valves, wheel rims and spokes, axles and
bearings, fenders, seat, brake and gear systems, pedals and grips, and tubular steel.
Suppose the firm fabricates the frame and handlebar from the steel and then assem-
bles bicycles using purchased components. Suppose also that the firm uses a “just
in time” production process wherein suppliers arrive at one factory door with an
input exactly when and as needed, and a shipper arrives at another door to take the
newly manufactured bicycle immediately to the firm’s customer. In this way, the
firm eliminates inventory, minimizes its floor space requirements, and keeps its unit
cost low.
Compared to this “just in time” production process, why might a firm choose
instead to order in advance and then hold inventory of its inputs and possibly of
manufactured bicycles as well? Two arguments come to mind.
One argument is that the fixed costs of placing an order are substantial so
that the larger the order the marginal cost per unit ordered decreases. In an early
study in the Harvard Business Review, Wilson (1934) emphasized the role of the
ordering amount—an expenditure to restock inventory also known as the economic
order quantity (EOQ)—and the ordering point (minimum inventory that triggers a
replenishment order).6 Wilson apparently had in mind an inventory cost that
summed two components: LA/C + IC/2 where A is annual consumption, C is the
amount ordered each time (EOQ), L is the cost of placing an order, C/2 is the average
inventory (which runs from C down to zero), and I is the cost of interest, insurance
and other loss risks, depreciation, rent, lighting, cleaning, heating and other inven-
tory holding costs per dollar of inventory held.7 The first component decreases as
the firm increases its EOQ; the second
√ component increases. To minimize inventory
cost then, Wilson then set C = (2LA/I).8 However, this raises a problem for this
book. If indeed the cost of placing an order is substantial, it is equivalent to say-
ing that there are economies of the large haul. However, I have ignored economies
of the large haul for simplicity of exposition elsewhere in this book. Nonetheless,
to whoever bears the shipping cost (be it the bicycle manufacturer or its supplier),
where I is small relative to L in Wilson’s model, economies of the large haul create
6 The confusion here over how something is labeled (in this case a quantity) and what is meant (in
this case, an expenditure) is regrettable. To an economist, expenditure is price times quantity. EOQ
predicts expenditure, not quantity.
7 In addition, the firm faces potential costs when customers are unable to get the product when and
as needed: e.g., loss of business. Interestingly here, Wilson does not take into account the costs
that can arise to a firm because a customer is not able to get the product as needed. Kahn (1992)
presents evidence from the retailing of automobiles in the United States to support the argument
that stockout avoidance is a major factor in retail inventory decision making.
8 Baumol (1952, pp. 545–547) uses a similar model to look at the cash holdings of households in a
fiat money economy.
7.3 Model 7A: The Firm Doing Repairs In-House 183
an incentive to store product somewhere until there is an amount that can be shipped
economically.
A second argument has to do with uncertainty. The firm might be uncertain about
the demand for its bicycles. If demand tomorrow were to suddenly rise, the firm
might find its inventory (of fenders, say) gets depleted before the replenishing ship-
ment arrives. A second uncertainty concerns how long it will take the fender supplier
to deliver. In other words, the firm may run out fenders in inventory simply because
the supplier takes longer to deliver than expected. Whenever the firm runs out of
inventory, it cannot satisfy any more demand for bicycles until it gets a supply of
the needed parts. Perhaps, customers will just patiently wait for their order to be
completed. Perhaps, however, they will start purchasing from another bicycle man-
ufacturer. In such cases, there is a cost to our firm: the present and future profit
foregone because of the shortfall in inventory. The firm may choose from among
several strategies for dealing with such uncertainty. One of these is to maintain larger
inventories of fenders, other bicycle components, and assembled bicycles.9
The LMP model assumes that the firm has staff dedicated to equipment repair. They
do nothing else: they sit idle while the machinery is functioning. The firm incurs
two costs related to this cycle of repair and production. One is the cost of financing
the commodity inventory necessary to bridge the downtime. The average amount
of inventory is shown in (7.1.3). The other is the cost of maintaining the necessary
repair staff. In this model, repair staff are paid at a daily wage w. As well, they are
hired for the entire interval, T, even though they are needed only for the downtime
interval ΔT. In the LMP Model, repair and production staff are apparently special-
ized so that neither can do the other’s job: production staff sit idle while repair staff
are working and vice versa. In (7.1.4), the fraction of the time the machinery is down
is entirely spent in repair (δ). In a sense, the choice of the firm is between two kinds
of inventory—an inventory of product and an inventory of repair labor—each with
its own cost. All other costs (e.g., wages for production staff) are assumed to be
fixed (i.e., remain the same) regardless of how repairs are handled.
The LMP model assumes repair time in (7.1.5) is inversely related to the number
of repair workers (n). Here, it is assumed also that n is infinitely divisible: i.e., in
addition to n = 1 or n = 2 for example, we can have fractional assignments like
n = 0.1 or n = 1.7.10 Note that (7.1.5) is a production function. As such, it has
two properties of interest here. First, the more repair staff are employed, the faster
9 Another strategy is to try to smooth out the demand for bicycles: e.g., by contracting regular deliv-
eries to the principal customers in advance. Still another is to negotiate with current or prospective
fender suppliers and shippers to ensure a more orderly flow or to share the risks (e.g., consignment).
A third alternative is to negotiate consignment or right-to-return clauses in purchasing inventory.
10 Taken in conjunction with (7.1.2), this assumes no possibility that the firm might pay overtime
wages to have a job completed more quickly.
184 7 What the Firm Does On-Site
the repair. Second, each repair worker added reduces the time to repair by a smaller
amount: i.e., (7.1.5) exhibits diminishing marginal productivity or congestion.
The effect of the assumptions above is to make C U-shaped in n for the firm.
After substituting (7.1.3), (7.1.4), and (7.1.5) back into (7.1.2), I get an expression
for cost (C) in terms of the level of repair staff (n). Because of the decreasing returns
to scale in repairs (7.1.5) in n, the marginal saving in inventory cost (from having one
more repair worker) decreases as n is increased. On the other hand, repair cost rises
steadily (proportionally) with n. The implication is that, when n is small enough,
C will initially fall as n is increased because the marginal saving in inventory cost
exceeds the marginal cost of a repair worker (wT); at a large enough n however, C
will begin to rise because the marginal saving in inventory cost is now smaller than
wT. C is U-shaped.11 See curve AB in Fig. 7.2. The same argument can be seen
by differentiating C with respect to n as shown in (7.1.6): in Fig. 7.2, total cost is
minimized at a repair staff of OG.
In-house repair
AB Total cost of repair
CD Cost of inventory
EF Cost of repair labor
OG Amount of repair labor that
minimizes total cost of repair
OI Efficient cost of repair
A B
H F
I
C
Cost
E D
0
G Repair staff
Therefore, there is a cost-minimizing level of repair staff (n). We can see this as
the bottom of the curve (C) labeled AB in Fig. 7.2. Using calculus to minimize C, I
set the first derivative equal to zero and solve for n (see OG in Fig. 7.2); the result
11 A neat trick has been performed here. In Wilson (1934), it takes an economy of the large haul
to produce a U-shaped inventory cost curve. In the LMP model, the same result obtains simply by
assuming diminishing marginal productivity to repair labor.
7.3 Model 7A: The Firm Doing Repairs In-House 185
is the firm’s demand equation for repair staff: (7.1.7). In this, the level of repair
staff varies directly with k, q, p, i, and T and inversely with the wage w. The price
elasticity of demand for repair staff here is 0.5; therefore, the demand for in-house
repair staff is price-inelastic.
0
C Repair labor (n)
In the above, I presume that the amount of repair staff is sufficiently large that
downtime is smaller than T (otherwise the firm could not rebuild its inventory before
the next breakdown). Since this means δ < 1, therefore, no > k/(1 − a) is the
minimum feasible repair staff: see (7.1.8) and amount OC in Fig. 7.3. For the mini-
mum cost solution to be a labor input larger than k/(1 − a), C must be declining at
n = k/(1 − a) ; otherwise least cost is the corner solution n = k/(1 − a) . The first
derivative in (7.1.6) must be negative at no = k/(1 − a). See panel (a) of Fig. 7.4. In
contrast, panel (b) there shows a corner solution. Since the firm’s price elasticity of
demand for labor is negative, the derivative will be negative (i.e., no corner solution)
provided that the wage (w) is small enough: see (7.1.9). Under these circumstances,
as n rises above the minimum feasible repair staff, C initially falls before beginning
to rise. Hence, given (7.1.9), the cost-minimizing repair staff is given by (7.1.7).
The LMP Model to this point has 7 givens (a, i, k, p, q, T, w); the cost-minimizing
n depends on 6 of them.
What about comparative statics here? Let us focus here on the impacts on three
outcomes: n, δT, and I. See Table 7.2. Assuming (7.1.9), I get:
a Where repair time is larger for any given n, repair staff is unaffected.
However, repair time and inventory rise.
186 7 What the Firm Does On-Site
A B
D
A
D
Cost
Cost
0 0
C Repair staff C Repair staff
(a) No corner solution (b) Corner solution
Fig. 7.4 Model 7A: corner solutions in the case of in-house repairs.
Notes: AB is total cost of repair. OC is the minimum amount of repair labor: that is, k/(1 − a). In
panel (a), (7.1.9) is satisfied; in panel (b), it is not
Outcome
n δT I
Given [1] [2] [3]
a 0 + +
i + – –
k + + +
p + – –
q + – −
T + + +
w − + +
i Where it is more expensive to hold inventory, the firm increases the amount
of repair staff, while repair time and inventory shrink.
k Where k is larger, the firm increases the amount of repair staff and downtime
and inventory rise.
7.3 Model 7A: The Firm Doing Repairs In-House 187
p Where the commodity sold is more valuable, the firm increases the amount of
repair staff, repair time is reduced, and inventory shrinks.
q Where a greater quantity of the commodity is sold daily, the firm increases
the amount of repair staff and repair time and inventory are reduced.
T Where the production period is longer, the firm increases the amount of repair
staff and repair time and inventory rise.
w Where each unit of repair staff is more costly, the firm reduces the amount of
repair staff. This causes repair time and inventory to rise.
Students sometimes voice concerns about the LMP model. In what respects does
the LMP Model seem strange to them?
• If a firm knew with certainty the machinery was going to break down every T
days, why not keep inventory at zero until the last day of the production period,
then rapidly restock inventory just before the breakdown? Wouldn’t that be less
costly than slowly rebuilding inventory as assumed here? The reason is that, as
is typical of inventory models generally, the LMP Model is a non-stochastic
representation of what is essentially a stochastic problem. It is the randomness
associated with breakdown that keeps the firm slowly rebuilding inventory; it
simply does not know when the next breakdown will occur.
• The LMP Model attempts to describe behavior in the face of uncertainty, where
one might expect a risk aversion strategy rather than the simple cost minimization
strategy used here.
• The LMP model considers only a few options for the firm. The model assumes
that the cost associated with delaying customer demand is infinite. While it may
be true that customers might seek out another supplier if the firm is slow to
deliver, it may also be the case that the firm has room to maneuver on the deliv-
ery schedule. The firm is assumed as well to have the option only of a specialized
repair team and not be able to make use of overtime pay to get the repairs done
sooner.
• The firm can use other strategies to mitigate the risk associated with machine
breakdown. One possibility is to engage in preventive maintenance. A second is
to have multiple production lanes in the factory so that a machine breakdown in
one lane does not stop production altogether. Finally, as seen in Chapter 3, the
firm might have spare capacity in the form of multiple factories, some perhaps
more efficient than others, that give it the possibility of still meeting customer
needs.
These concerns remind us that the LMP model is not without its shortcomings.
However, the value of the LMP model is the way that it allows us to get a glimpse
of an important role for agglomeration economies.
Does Model 7A allow us to say something about the effects of division of labor?
Suppose that a firm operates at a scale that, to maximize profit, requires 1 man-
ufacturing machine, 20 production staff, and 2 repair staff. Suppose the firm then
experiences a doubling of the demand for its product and finds that this is most
188 7 What the Firm Does On-Site
profitably done with 2 machines, and 40 production staff. If the 2 machines break
down independently, and downtime is a relatively small fraction of the production
period, the firm might find the most profitable size of repair staff was still only 2 per-
sons. In that case, the firm is operating more efficiently overall (because it did not
need to double its repair staff). However, I think this is better seen as a consequence
of the indivisibility of repair (wherein the worker is idle until the next breakdown)
rather than a division of labor wherein the firm gets more productivity from workers
by having them do more specialized work.
Now, introduce the possibility that the firm can outsource this repair activity. The
specific problem addressed by the firm in the LMP Model is the choice between (1)
maintaining one’s own in-house repair staff as discussed above and (2) outsourcing
of repair work. In the model, the choice is either-or; the firm chooses one and only
one of them. This assumption simplifies the analysis. In reality, we can imagine a
range of strategies. These might include (1) regular manufacturing employees also
trained to do repairs, (2) having a small core repair staff that are complemented with
outsourced workers when needed, and (3) renting repair staff to others at times when
idle at the firm.
Now, consider the firm as outsourcer purchasing repair services at a given daily
rate per worker (wo ) from a contractor. See Table 7.3. I use a subscript “o” to denote
variable and parameter values specific to the firm that outsources. The cost to be
minimized, Co , is shown in (7.3.1). The average product inventory is shown in
(7.3.2), the downtime in (7.3.3), and the repair time in (7.3.4). Assume here the
same repair time function as for in-house repair staff. Taken in conjunction with
(7.3.1), this assumes that the firm can choose how many repair workers to call up.
Further, with outsourcing, the firm now finds that it has to wait ωT days for the
contractor to arrive before repairs commence. This wait time is like a unit shipping
cost; it is a direct cost of not having the repair staff in-house. In addition, the daily
rate per repair worker (wo ) is presumably higher than the daily wage (w) paid by
a firm doing in-house repairs. Presumably, these two disadvantages of outsourcing
are offset by one major advantage: the outsourcing firm pays for repair labor only
for the length of the repair (δ o T), not the whole production cycle (T) as does the firm
doing repairs in-house.
As in the case of the firm doing repairs in-house, I can derive an expression for
the outsourcing firm’s repair plus inventory costs. After substituting (7.3.2), (7.3.3),
and (7.3.4) back into (7.3.1), I get an expression for cost (Co ) in terms of the level of
repair labor (no ). Co is u-shaped in no . The outsourcing firm finds that its repair costs
rise, and its inventory costs fall, with the amount of repair services used daily (no ).
See curves CD and EF Fig. 7.5. Because of the decreasing returns to scale (7.3.4)
in no , eventually Co begins to rise following repair labor costs: see curve AB. Put
differently, I can use calculus to differentiate Co with respect to no , yielding (7.3.5).
Because the slope (dCo /dno ) is negative for small no and becomes positive for large
7.4 Model 7B: Outsourced Repairs 189
Cost to be minimized
Co = piIo T + wo no δo T (7.3.1)
Downtime (proportion of T)
Δo = w + δo (7.3.3)
Repair time
δo = a + k/no where 1 > a > 0 and k > 0 (7.3.4)
Marginal effect of repair labor on cost
dCo /dno = −pi(1/2)qT 2 (k/n2o ) + wo Ta (7.3.5)
Cost-minimizing
√ repair labor (where dCo /dno = 0)
no = (kqpiT/(2wo a)) (7.3.6)
Minimum feasible repair labor (Δ = 1)
k/(1 − a − w) (7.3.7)
Maximum daily rate if marginal effect of repair labor on cost is negative at minimum
feasible repair labor
wo < pi(1/2)qT(1 − a − w)2 /(ak) (7.3.8)
Notes: Rationale for localization (see Appendix A): Z4—Risk spreading and insurance; Z8—
Limitation of travel delay. Givens (parameter or exogenous): a—Minimum repair time (fraction
of T); i—Interest rate (daily); k—Responsiveness of repair time to size of staff; p—Selling price
of unit of output; q—Units sold daily; T—Length of production cycle (days) including downtime;
wo —Daily fee per repair worker contracted; ω—Wait time until arrival of repair contractor (frac-
tion of T). Outcomes (endogenous): Co —Repair plus inventory cost to be minimized; lo —Average
stock of product inventory held-over period; no —Number of repair workers; δ o —Time (fraction
of T) required for repair; Δo —Downtime (fraction of T).
no , Co is U-shaped. To minimize Co , I set the first derivative equal to zero and solve
for no ; the result is the firm’s demand equation for repair labor: (7.3.6). In this, repair
staffing varies directly with k, q, p, i, and T, and inversely with wo and a. The price
elasticity of demand for repair labor here is 0.5: therefore, demand for outsourced
repair labor is price inelastic.
It is interesting to compare the least cost use of repair labor under outsourcing in
(7.3.6) with the least cost use of labor for the in-house firm (7.1.7). The numerators
are the same in the two equations. However, the denominator in (7.3.6) incorporates
wo a as opposed to w in (7.1.7). This arises because w is what the firm pays daily for
in-house repair workers; in contrast wo a is the minimum amount the firm pays daily
on average for each outsourced worker.
In the above, assume the amount of repair labor is sufficiently large that down-
time is smaller than T. Since this means δ < 1, therefore, no > k/(1−a−ω) is the
minimum feasible repair staff: see (7.3.7). For the minimum cost solution to be a
labor input larger than k/(1−a−ω)—that is, no corner solution—the first derivative
in (7.3.5) must be negative at no = k/(1−a−ω) . Since the firm’s price elasticity of
190 7 What the Firm Does On-Site
Outsourced repair
AB Total cost of repair
CD Cost of inventory
EF Cost of repair services
OG Amount of repair labor that
A minimizes total cost of repair
Cost OI Efficient cost of repair
H B
I
F
C
0
G Repair labor
Table 7.4 Model 7C: in-house versus outsourced repairs assuming no corner solutions
Notes: See also Table 7.1. Rationale for localization (see Appendix A): Z4—Risk-spreading and
insurance; Z8—Limitation of travel delay. Givens (parameter or exogenous): a—Minimum repair
time (fraction of T); c—Unit variable production cost; f—Daily fixed cost; i—Interest rate (daily);
k—Responsiveness of repair time to size of staff; p—Selling price of unit of output; q—Units
sold daily; T—Length of production cycle (days) including downtime; w—Daily wage of repair
worker (in-house); wo —Daily fee per repair worker contracted (outsourced); ω—Wait time until
arrival of repair contractor (fraction of production cycle). Outcomes (endogenous): C—Repair
plus inventory cost to be minimized (in-house); Co —Repair plus inventory cost to be minimized
(outsourced); l—Average stock of product inventory held-over period; n—Number of repair work-
ers; Δ—Downtime (fraction of T); δ—Time (fraction of T) required for repair; Δo —Downtime
(fraction of T); δ o —Time (fraction of T) required for repair.
One way to think about C − Co > 0 is that for the in-house option to be rejected
the daily wage for in-house staff (w) must be relatively high. I can rearrange (7.4.3)
to show that C −Co > 0 implies (7.4.4). The tradeoff is illustrated in Fig. 7.6, which
shows combinations of in-house daily wage (w) and outsourced daily rate (wo ). In
the area below the curve AB are the combinations of w and wo where in-house repair
is chosen because it is less costly (assuming wait time is zero); in the area above the
curve, outsourcing is less costly. In Fig. 7.6, the curve CD shows the corresponding
tradeoff between w and wo when wait time is 10% of T. In fact, there is a family of
curves here: one for each possible value of ω. The greater the wait time, the lower
the outsource rate (wo ) tolerable for any given ω.
A caveat is in order here about the treatment of uncertainty in this model. As
noted above, the LMP Model uses a cost-minimization approach. However, I have
already argued above that the LMP model is an attempt to think about the firm
192 7 What the Firm Does On-Site
In-house chosen
F
G Outsourcing chosen
A
C
0
E In-house daily wage (w)
coping with uncertainty. How might our analysis have differed if I had assumed
instead a strategy of risk aversion? In the LMP model, the only factor that the firm
can vary is the size of the repair staff/labor. The firm presumably might be expected
to choose a repair staff/labor to offset stochastic variation in any part of this problem
(e.g., the time between breakdowns, the quantity demanded daily, or the time to
complete repairs). Given a loss attached to each possible outcome, a risk-averse
firm with either in-house repair staff or outsourcing might want to think about the
relative losses associated with overshooting as opposed to undershooting.
In this regard, it is instructive to compare Fig. 7.5 with Fig. 7.2. There are similar-
ities. In both diagrams, inventory cost falls off quickly, asymptotically approaching
zero, as repair staff/labor is increased. In both cases, total cost, be it C or Co ,
initially decreases with an increase in repair staff/labor; then begins to increase.
However, there is one important difference here. In Fig. 7.2, as I increase or decrease
in-house repair staff (n), the repair cost (wnT) changes proportionally. In Fig. 7.5,
as I increase or decrease the amount of outsourced repair labor (no ), the repair cost
(wo no δo T) changes only modestly. This is because an increase in no is offset in part
by a decrease in δ o . There is no corresponding effect in the in-house repair wage
bill. Although Co is U-shaped like C, it has a relatively modest upturn for repair
labor above the least cost solution (7.3.6).
7.6 Model 7D: The Advantage of Agglomeration 193
Now, let us focus on the choice of n facing an in-house firm. Suppose the in-
house firm chooses a level of n using (7.1.7): based on expectations about time
between breakdowns, quantity demand, time to repair, and so on. Suppose that one
or more of these expectations turns out to be incorrect so that, in fact, some other
level of repair staff is cost-minimizing. In its chosen n, the firm may incur a loss—a
loss by either undershooting or overshooting the correct cost-minimizing n. In terms
of Fig. 7.2, the chosen n lies either to the left or the right of the cost-minimizing n.
Either way, the firm finds itself with a loss: spending more than it needs for inventory
and repair costs. However, since C rises somewhat less quickly to the right of the
cost-minimizing n than it does to the left, a risk-averse firm might prefer to err on
the side of overshooting (as opposed to undershooting) n.
Let us now look at the outsourcing firm. Compare Fig. 7.2 with Fig. 7.5. From
Fig. 7.2, the loss from overshooting—that is, the additional cost arising because I
had more than the efficient size of repair staff—is much higher than the loss from
overshooting in Fig. 7.5. In Fig. 7.5, there is only a small increase in cost with having
repair labor in excess of the cost-efficient amount. If the cost of outsourcing is only
modestly greater than the cost of in-house repairs, the risk-averse firm would prefer
outsourcing for this reason.
In this section, I have referred to the ideas in terms of outsourcing. Cast differ-
ently, the ideas equivalently are about in-house production. When a firm decides
to do something in-house, rather than use a contractor, it can be said to instance
vertical integration. In the great industrial restructuring, downsizing, and globaliza-
tion of the past three decades or so, large firms have looked to outsourcing to help
them both to reduce costs or risks and focus on their core (most profitable) business
activities.12
Suppose our firm is the only business enterprise in the region and that there is no
contractor from whom repair services can be purchased as needed. In this case, the
firm presumably must maintain its own in-house repair staff. Suppose now the man-
ager of the repair team approaches the firm to suggest that the firm spin off the
repair team and contract it to provide repair services when and as needed. But, the
firm might reasonably ask, what is the advantage of this? If the repair team is effi-
ciently managed within the firm, how might the firm be better off with a contractor
here? “Ah”, students often say, “what about labor savings?” After all, if the firm
needs to pay only for repair services when needed, won’t it save money? The prob-
lem with this argument is that the contractor would not be able to retain repair staff
unless it was able to pay them appropriately. If the repair staff needed only 20% of
12 Atthe same time, vertical integration thins markets for inputs. Every firm that chooses an in-
house repair staff risks undermining the efficient provision of contractor repairs for firms that
prefer to outsource. See, for example, McLaren (2000).
194 7 What the Firm Does On-Site
the time, for example, then the daily wage rate charged by the contractor would have
to be at least 5 times the daily wage paid to in-house repair workers. You might be
tempted to argue that the contractor might instead seek out part-time workers, peo-
ple who would be happy to work only 20% of the time at the standard wage. Would
it not then be feasible for a contractor to be more efficient than the firm at doing
repair work? Of course, the answer here is “no”; after all, if a contractor can find
part-time workers then what is to stop the firm from hiring part-time workers itself.
The contractor has no inherent advantage here.
Now assume that, in the vicinity, additional identical firms each require the same
repair services under identical circumstances: m firms in all. Suppose further the
machinery at each firm breaks down at a schedule uncorrelated with breakdowns at
other firms. In this case, an efficiency arises to the extent that a repair contractor can
better utilize the available resource staff than can any of the client firms in isolation.
Rather than remain idle when not serving the first client, the contractor can now
use staff to repair machinery at another client. Effectively, this means an efficiency
advantage in that, for the repair contractor, the cost of its repair staff is being shared
among the client firms. Under the assumption that C−Co > 0, the demand for repair
services by each firm is given by (7.5.1): see Table 7.5. The aggregate demand for
the m client firms is given by (7.5.2). Total contractor revenue can then be calculated
using (7.5.3).
Table 7.5 Model 7D: the contractor with m identical cost-minimizing customers
Notes: See also Table 7.1. Rationale for localization (see Appendix A): Z4—Risk-spreading and
insurance; Z8—Limitation of travel delay. Givens (parameter or exogenous): a—Minimum repair
time (fraction of production cycle); i—Interest rate (daily); k—Responsiveness of repair time to
size of staff; m—Number of client firms; p—Selling price of unit of output; q—Units sold daily
by client firm; T—Length of production cycle (days) including downtime; w—Daily wage rate;
γ —Congestion parameter for repair firm. Outcomes (endogenous): no —Number of repair work-
ers; R—Total revenue of repair contractor; wo —Daily fee per repair worker contracted; δ o —Time
(fraction of T) required for repair; ω—Wait time until arrival of repair contractor (fraction of T).
7.6 Model 7D: The Advantage of Agglomeration 195
13 In(7.5.5), I have assumed that the repair contractor has no fixed costs. I had assumed the same
thing in the case of the firm doing repairs in-house. However, suppose instead that the repair con-
tractor does incur a fixed cost. In that case, there would be a minimum number (m∗ ) of client firms
below which contracting would not be profitable.
196 7 What the Firm Does On-Site
0
C B
Contractor's workforce (N)
profit is best illustrated by setting γ =0 locally; i.e., where an increase in the num-
ber of client firms has no effect on wait time. The repair contractor provides the
kind of insurance function Alfred Marshall envisaged and constitutes a kind of
agglomeration economy.
It is not necessary to have a contractor to realize this kind of efficiency. Instead,
one firm might hire its own repair team and then rent out its team to other firms
as available and needed. The important notion here is that the efficiency can arise
regardless of the organization of production. The contractor merely exemplifies how
the efficiency might be implemented.
As well, note that the economies of scale implicit in (7.5.4) gradually run out.
What happens then as the number (m) of client firms in the vicinity continues to
rise: i.e., as the urban agglomeration grows in size. The repair contractor would like
to retain all these client firms, but the possibility of a new contractor entering the
market must now be considered. Presumably, a new contractor entering the market
7.8 Final Comments 197
reduces the profit of an existing contractor by stripping away customers and push-
ing down the price of repair services. It is at this stage that client firms begin to
experience the advantages of agglomeration: i.e., a lower price for repair services. I
consider further the impact of competitors on price in Chapter 8.
In the model here, the contractor becomes ever more efficient. In that sense, the
contractor is like the monopolist in Chapter 2 who experiences only a decline in
average cost as output is increased. Presumably, however, some kind of congestion
sets in above a particular level of output, and this opens the way for competitors to
enter the business. I do not explore this idea further, as my objective is principally
to show the tradeoff between in-house and outsourced production.
In this chapter, the principal models have been 7C (the client firm choosing in-house
versus outsourcing) and 7D (the repair contractor). I included Models 7A and 7B
to help readers better understand aspects of Model 7C which in turn helps us to
understand 7D. In Table 7.6, I summarize the assumptions that underlie Models
7A through 7D. Many assumptions are in common to all models: see panel (a) of
Table 7.6.
14 An interesting possibility here would be the approach of Dana and Spier (2001) in their study of
vertical control in the video rental industry.
198 7 What the Firm Does On-Site
7A 7B 7C 7D
Assumptions [1] [2] [3] [4]
In practice, a firm may use various kinds of contractors. Repair services are an
interesting example, but just one of the possibilities here. Another possibility for
example is the independent shipper. While some large firms do have their own in-
house fleets (be they planes, ships, rail, or trucks) to ship their own materials and
commodities, many firms in practice rely on contractors in the shipping business.
As the density of firms in need of shipping services increases within a particular
geographic locale, presumably the typical shipper becomes more efficient for the
same reasons as we saw above in the case of repair services. The shipper looks for
the most profitable amount of transportation equipment and staff (e.g., drivers) and
in so doing trades off the economies possible because customer demands are not
perfectly correlated with one another against the cost of any delay experienced by
customer firms. If the need for transportation services is sufficiently large to foster
the establishment of a number of shippers in the vicinity, presumably, the economies
achieved by the shippers will, as a result of competition among shippers, lead to
lower shipping rates and form an urbanization economy for the client firms.15
In Chapter 1, I argue that prices are important in shaping localization. As in
earlier chapters, I continue to assume that the prices of inputs (including shipping
services) used by the firm are exogenous; input prices are all determined in markets
outside the scope of the models in this chapter. As well, the models in this chapter
are silent on prices in output markets. However, in beginning to think explicitly
15 Let me add a cautionary note here. The advantage of focusing on the repair contractor in this
chapter is that this is a clear example of an application of the insurance principle. In the case
of shippers, there is some risk in shipping; however, my guess is that there are also substantial
indivisibilities and economies of scale to the shipping business.
7.8 Final Comments 199
about the nature of contracting, I have introduced the idea that a sufficient number
of client firms nearby makes it profitable for a contractor to locate there. In so doing,
the contractor may well affect the price of repair services for firms. For the first time
in this book, there is the possibility of simultaneous determination in the prices
of two distinct commodities: the good manufactured by the firm and the service
provided by the repair contractor.
This is also the first chapter in the book to include specifically a (repair) labor
input. In principle, this could mean that the chapter might be able to say something
about the distribution of income in society among units of labor and capital, if not
land. However, in several of the models presented in this chapter, the wage rate for
repair staff is exogenous. Whether the firm decides to employ many repair staff or
just a few, the wage paid remains the same. Therefore, all cost savings that can arise
(e.g., because of a decrease in a or k) accrues to the firm owner: implicitly increasing
profit. As in Chapters 3 and 5, no adjustment of market price is envisaged, and there
is no change in the well-being of consumers.
Also, this is the first chapter to address the question of what the firm does
and where. Under certain circumstances, the firm finds it efficient to have its own
repair staff; under other circumstances it does not. What gets done on-site is now
endogenous. I characterized it as a tradeoff between the daily wage paid to in-house
repair staff and the daily rate charged by the repair contractor. However, in impor-
tant ways it is more than this. Eliminating in-house staff also means savings in floor
space, in training costs, and in management of the repair staff. Nonetheless, the
models in this chapter are helpful in thinking about issues confronting firms as they
decide which activities to do in-house.
Chapter 8
Staking Out the Firm’s Market
Price and the Geometry of Competition
A firm typically has a market area: a geographic area wherein the firm dominates
and does much of its sales. Within its market area, a firm is able to affect the
price received to the extent that the effective prices from other suppliers make them
uncompetitive. What determines the size and shape of a market area? How might
the presence of a market area affect firm behavior? Model 8A considers market
area when a firm and its competitor sell at the same f.o.b. price. Model 8B looks at
market area when firms have different f.o.b. prices. In 8C, the firm sets a price that
maximizes its profit assuming that competitors do not react. Model 8D studies how
the firm’s market area boundary adjusts to capacity constraints. Model 8E shows
how the firm’s market area boundary varies when it sells a different but perfectly
substitutable good. Models 8F, 8G, and 8H introduce differences among consumers
as well as imperfectly substitutable goods. Chapter 1 argues that there are impor-
tant linkages between prices and localization. I illustrated that idea in Model 2D
wherein price and localization were joint outcomes for the monopolist. However,
since that chapter, the models in this book have been concerned only with how prices
affect localization. This chapter considers how a firm sets its price in response to the
proximity of competitors and the prices they set. This chapter explores how, as a
consequence, localization and price are jointly determined.
1 As this book does not deal with optimal location theory, I exclude optimal market areas here. See,
for example, Hsu (1997).
2 On contributions to market areas by economists, see Bacon (1992), Eswaran and Ware (1986),
Fetter (1924), Greenhut (1952a, 1952b), Greenhut and Ohta (1975), Hartwick (1973a), Hyson
and Hyson (1950), and Stern (1972a). On the modeling of destination choice, see Ghosh and
McLafferty (1987), Lo (1990, 1991a, 1991b, 1992), and Miron and Lo (1997).
3 Geographers and other regional scientists who have written about the analysis of market area
include Batty (1978), Boots (1980), Daly and Webber (1973). Epping (1982), Erickson (1989),
Fotheringham (1988), Gillen and Guccione (1993), Golledge and Amedeo (1968), Golledge (1967,
1970, 1996), Huff and Jenks (1968), Lentnek, Harwitz, and Narula (1988), Lo (1990, 1991a, 1991b,
1992), McLafferty and Ghosh (1986), Miron and Lo (1997), Mu (2004), Mulligan (1982), O’Kelly
and Miller (1989), Parr (1995a, 1995b, 1997a, 1997b), Pitts and Boardman (1998), Pred (1964),
Rushton (1971a), Sheppard, Haining, and Plummer (1992), Solomon and Pyrdol (1986), Stimson
(1981), and Wilson (1967).
4 See, for example, Pate and Loomis (1997).
8.1 The Market Area Problem 203
and localization. The literature in this part of location theory has a history that traces
back to Hotelling (1929) and Lösch (1954) and that has benefited from advances in
game theory.5
Models of the market area for a firm—in the presence of competitors—come
in two varieties. In one variety (as in Models 8A through 8C below), the model
assumes we know precisely which customers a firm will serve; in a simple case,
the firm captures all potential customers within its market area and none outside.
Where, for example, competitors price f.o.b. and lowest effective price is the sole
determinant of customer choice, the firm’s market area is the geometric shape within
which no other competitor is able to sell their product. In the other variety (as in
Models 8D through 8G), the customer chooses a firm (supplier) based at least in
part on considerations outside the realm of the model. Taking a model in which the
only explanatory variable is the firm’s f.o.b. price, Hotelling (1929, p. 41) argues
that (1) in spite of moderate differences of price, some purchasers of a commodity
buy from one seller, some from another, and (2) If the supplier of a good gradually
increases his price while his rivals keep theirs fixed, the diminution in his volume of
sales will take place continuously rather than abruptly.
In either variety, a range of models of competitive behavior is possible here. In a
simple model, we might assume a firm and its competitors—with geographic loca-
tions given—each offering a given good (or variety of goods) at a given f.o.b. price
and a given level of service. Customers who bear the unit shipping cost then choose
from among the firms. In more sophisticated models, we might imagine the firm
and its competitors adjusting their prices, geographic locations, the kinds of goods
they offer for sale, and the quality of service they provide. Hotelling (1929, p. 44)
argues that when a seller increases price he will only gradually lose business to his
rivals. Some customers will still prefer to trade with him because he is more conve-
nient, provides better service, or sells goods they desire. In Hotelling’s view, such
customers make every vendor a monopolist within a limited geographic area, and
there is no monopoly that is not confined to a limited geographic area. In any of
5 Important early work in the field also includes Lerner and Singer (1937) and Smithies (1941).
Other work that largely predates the application of game theory include Capozza and Van Order
(1978), Devletoglou (1965), Eaton and Lipsey (1975, 1979), Gannon (1972, 1973), Harker (1986),
Mills and Lav (1964), Salop (1979), and Webber (1974). For game theory applications, see
Anderson, de Palma, and Hong (1992), Anderson, Goeree, and Ramer (1997), Basu (1993, chap. 8),
Benassi and Chirco (2008), Bester, de Palma, Leininger, Thomas, and von Thadden (1996), Boyer,
Laffont, Mahene, and Moreaux (1995), Braid (2008), Damania (1994), De Frutos, Hamoudi,
and Jarque (2002), Dorta-González, Santos-Peñate, and Suarez-Vega (2005), Economides (1993a,
1993b), Gabszewicz and Thisse (1986b, 1992), Goettler and Shachar (2001), Gupta, Pal, and
Sarkar (1997), Gupta, Lai, Pal, Sarkar, and Yu (2004), Hamilton and Thisse (1992), Hay (1976),
Huang (2009), Huck, Müller, and Vriend (2002), Irmen and Thisse (1998), Isard and Smith
(1967), Iyer and Seetharaman (2008), Lambertini (1997), Lederer (1994), Lederer and Hurter
(1986), Matsumura, Ohkawa, and Shimizu (2005), Osborne and Pitchik (1987), Pires (2005, 2009),
Prescott and Visscher (1977), Seim (2006), Selten and Apesteguia (2005), Shaked and Sutton
(1982), Stuart (2004), Tabuchi (1994), Tabuchi and Thisse (1995), and Zhu and Singh (2009).
204 8 Staking Out the Firm’s Market
these models, imagine a spatial equilibrium exists in which no firm has an incen-
tive to change its behavior. In the presence of unit shipping costs, effective price
might vary across the landscape; so too might localization (the density of competi-
tors nearby) as well as the kind of goods and level of service locally. In all of these
models, the firm is competing in a world of differentiated goods, even if the differ-
entiation is solely on the basis of unit shipping cost. Such ideas help us think about
what might be causing Hotelling’s stability of competition.
To exemplify stability, Hotelling uses a bounded linear market along which cus-
tomers are spread at a given uniform density. Two competitors at given locations
along this line compete for customers by each setting a f.o.b. price. In Hotelling’s
model, there was no inherent reason for customers to patronize one particular store
other than a lower effective price. However, in a linear market each competitor has
a protected flank: the customers between him and the end of the line away from his
competitor. Unless the competitor’s f.o.b. price is so low as to eliminate our firm’s
market completely, the firm can always count on its protected flank. Critics might
say that, in practice, bounded linear markets are uncommon in an unbounded rect-
angular plane—again assuming a uniform density of customers everywhere—the
notion of a protected flank vanishes. However, I think that Hotelling would simply
argue here that a linear market was just a representation that illustrates a source of
stability.
The purpose of this chapter is to give readers the flavor of this line of research
through a sequence of models that emphasize aspects of the problem. Sometimes,
these models are best cast in the context of a firm purchasing inputs from a supplier.
Other times, these models are best cast in terms of a customer choosing a retail outlet
(store) from which to purchase a commodity. Models 8A through 8H are presented
in sequence.
8A Market areas when competitors charge the same f.o.b. price, shipping cost is
everywhere proportional to distance, customers are identical, uniformly spread
across geographic space, and purchase where shipping cost is lowest (Thiessen
Polygons).
8B Market areas when competitors charge different f.o.b. prices for their com-
modity. Other assumptions remain the same as Model 8A.
8C Market areas when there is price competition among firms selling the same
product.
8D Market areas when each establishment has a different capacity to supply the
commodity.
8E Market areas when competitors supply different but perfectly substitutable
commodities. Other assumptions remain the same as in Model 8B.
8F Market areas when customers are of two different types. This builds on
Model 8E.
8G Market areas when competitors supply unrelated commodities (zero cross
price elasticity).
8.2 Range and Geographic Size of Market 205
Why these seven models? These seven are not exhaustive but do illustrate factors
that shape market area, a kind of spatial price equilibrium, and localization.6
Following these seven models, I briefly consider market areas when destination
choice can be thought to be subject to uncertainty (discrete choice modeling). Each
place can be thought of as a store selling the commodity at an f.o.b. price. In sim-
ple versions of this model, each store sells the same commodity and is similar
in other respects such as store attractiveness, purchase warranty, and level of ser-
vice. The stores are each just a place from which the commodity can be obtained.
The only relevant consideration is the cost of purchasing a unit of the commod-
ity: f.o.b. price plus any shipping cost. More realistic models can be envisaged in
which some of these assumptions are relaxed. Ignored here are shopping safaris
where a customer purchases commodities while traveling from place to place.7
Ignored also is the practice of bulk buying where the customer acquires a substan-
tial inventory (i.e., stocks up on purchases) so as to minimize the effect of trip cost.
Instead, the customer is presumed to incur a fixed shipping cost for each unit of the
commodity consumed. Finally, the consumer (or analyst) may well have imperfect
information, that is, does not know the price and availability of the commodity at all
places.
In this chapter, as elsewhere so far in this book, I take the locations of consumers
as given. I do not ask why consumers have come to be where they are. Nor do I ask
why they might not respond to differences in price by changing their location. Such
considerations are left to later chapters in this book. I also assume that the firm, its
customers, and competitors share a common fiat market economy. As in Chapter 2
and Chapter 3, price differences among places open up the question of the happiness
of customers. Other things being equal, we might expect customers at places where
the effective price is higher to want to relocate to places where the effective price is
lower. In this chapter, I continue to ignore such inclinations.
6 I do not, for example, include models based on Marxist perspectives. See Plummer, Sheppard,
and Haining (1998).
7 See, for example, McLafferty and Ghosh (1986).
206 8 Staking Out the Firm’s Market
each customer has a linear inverse demand curve. For a customer at distance x from
the monopolist, the effective price is P + sx, and the individual demand curve is
given by (8.1.1): see Table 8.1. Here total trip cost for a customer is, therefore,
proportional to consumption. There is no notion here of the consumer having an
Table 8.1 The monopolist’s market in two-dimensional space using f.o.b. pricing
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): C—Unit cost of pro-
duction; F—Fixed cost; g—Density of consumers everywhere; P—F.o.b. price set by firm; s—Unit
shipping rate; x—Distance from customer to firm; α—Intercept of individual inverse demand
curve; β—Slope of individual inverse demand curve. Outcomes (endogenous): Q—Aggregate
amount demanded; q—Quantity demanded by a consumer; X—Range of commodity.
8.2 Range and Geographic Size of Market 207
inventory of the commodity at home and making a tradeoff between trip frequency
(trip cost) and household inventory (inventory cost).8
Lösch characterized models like (8.1.1) as giving rise to a demand cone9 over
the rectangular plane.10 This implies that, given an f.o.b. price P, there is a maxi-
mum distance, range (X), beyond which customers will demand zero: see (8.1.2).
Of course, this is also the radius (geographic size) of the market for this monopolist.
The higher the firm sets its f.o.b. price P, the smaller the geographic area of the
market. Aggregate quantity demanded is given by (8.1.3). After rearranging (8.1.1),
substituting, and integrating, we get (8.1.4); I show this intermediate step to help
the reader with the derivation. Finally, after substitution from (8.1.2), the aggre-
gate quantity demanded reduces to an expression for the demand cone (8.1.5). As
one might expect, quantity demanded is an increasing function of α and g, and a
decreasing function of β, P, and s. This is the so-called free spatial demand curve;
“free” in the sense that the firm does not here consider competitors.
The aggregate demand equation (8.1.5) is not linear in price even though we have
assumed each customer individually has a linear demand. This is a consequence of
the fact that the firm’s demand curve is aggregated over customers: some close by
and paying a low effective price, others further away and paying a high effective
price. This is illustrated in Fig. 8.1 where I assume the market contains just 3 cus-
tomers, each with the same linear inverse demand curve. Customer 1, nearest the
firm, has the lowest effective price; Customer 2 is further away, and Customer 3 still
more so. At a high f.o.b. price, the firm would see demand only from Customer 1.
As it lowers its f.o.b. price, it eventually attracts demand from Customer 2 and at a
still lower price from Customer 3. That the aggregate demand curve here is kinked, a
polyline, is a result of assuming just three customers. However, (8.1.5) assumes cus-
tomers are spread evenly across the market; thus, (8.1.5)—cubic function of P that
it is—can be thought of as the continuous equivalent of a kinked aggregate demand
curve. Put differently, when geographic space is continuous, so too is the aggregate
demand curve; when geographic space composed of discrete places (punctuated),
the aggregate demand curve is kinked.
Now, let us look at the behavior of a firm that maximizes profit. Assume the
firm has a fixed cost (F) and a marginal cost (C); profit is given by (8.1.7).
After substituting from (8.1.5), we can find the first-order condition for profit
maximization: marginal revenue equals marginal cost. See (8.1.6). This yields the
f.o.b. price: (8.1.8). In Chapter 2, I argue it is helpful to think of C (marginal cost of
production) as a minimum price to complement the idea of α as the maximum price.
The profit-maximizing price is now one-quarter of the way from C to α. This is a
lower price than is charged by the firm in the simplest version of the two-market
8 See, for example, Bacon (1992). Here, inventory models of the kind used in Chapter 7 allow us to
look at the economics of buying in bulk and such phenomena as a buying safari or buying spree.
9 A term characterizing the tendency for individual quantity demanded to fall off with increasing
distance from a supplier pricing f.o.b. because of the rise in effective price.
10 See Long (1971) for further discussion of the properties of a spatial demand curve.
208 8 Staking Out the Firm’s Market
A2 C
Price
A3 D
0
B3 B2 B1 Q E
Q3 Q2 Q1 Quantity
Fig. 8.1 Aggregate demand curve in market with three customers, each with same linear inverse
demand curve but paying different effective prices
model in Chapter 2; there, the firm sets its price halfway between C and α. This
difference arises because I here assume f.o.b. pricing; in Chapter 2, I had assumed
that the firm discriminates in pricing between customers at Place 1 and customers at
Place 2.
How large is the geographic area of the firm’s market here? After substituting
(8.1.8) into (8.1.2), we get (8.1.9). From (8.1.2), we see that, were the firm to price
at C, the range for its commodity would be (α − C)/s. This is the maximum possible
range; therefore, the maximum possible geographic area is π ((α −C)/s)2 . However,
because the firm maximizes profit, it sets the higher f.o.b. price in (8.1.8) that gen-
erates a range 3/4 the size of the maximum possible range. Put differently, the firm
foregoes serving customers at radiuses from beyond (3/4)(α − C)/s to (α − C)/s
because these remote customers are not profitable enough.11 I presume here that the
market radius given by (8.1.9) is sufficiently large to make the firm profitable. As
shown in (8.1.10), α and g must be sufficiently large and/or β, C, F, and s sufficiently
small for this to happen.
Finally, even though this is the market for the monopolist’s product and even
though a single f.o.b. price is determined there, the effective price is different for
customers depending on how far they are from the firm. Up until now, we have
• Compare the delivered price in (8.2.6) with the f.o.b. price in (8.1.8). For cus-
tomers nearby (small x), delivered price is higher than f.o.b. price; for customers
further away, it is smaller. Were a firm to switch from f.o.b. pricing to delivered
pricing, customers nearby would become worse off while more remote customers
would become better off.
• In Chapter 2, we saw the half-freight rule. Here, (8.2.6) instances it again.
• Profit under delivered pricing, now given by (8.2.11), is higher than it was for
f.o.b. pricing: compare (8.2.11) with (8.1.10).
• The geographic size of market under delivered pricing—see (8.2.7)—is the
largest possible; it is larger than it would be, for instance, under f.o.b. pricing.
Table 8.2 The monopolist’s market using delivered pricing assuming P = P0 + 0.5sx, hence
X = (α − P0 )/(0.5s)
Aggregate demand
(4/3)πg(α − P0 )3 /(βs2 ) (8.2.1)
Total revenue
Q(α + P0 )/2 (8.2.2)
Marginal Revenue
0.5(a + P0 ) − (1/6 ){(3/4)βs2 Q/(π g)}1/3 (8.2.3)
Total cost
F + Qd [C + α − P0 ] (8.2.4)
Marginal cost equals marginal revenue for customer at distance x
C + sx = α − 2βq (8.2.5)
Profit maximizing price for customer at distance x
P = 0.5α + 0.5(C + sx) (8.2.6)
Radius of market (where P = α)
X = (α − C)/s (8.2.7)
Quantity at profit-maximizing price
(1/6)πg(α − C)3 /(βs2 ) (8.2.8)
Revenue
x of firm
0 2πgxqPdx (8.2.9)
Profit
x
of firm defined
0 2πgx(P − C − sx)qdx − F (8.2.10)
Maximized profit
(1/24)(π g)(α − C)4 /(βs2 ) − F (8.2.11)
Consumer benefit (CB)
αQ − (3βs2 /(πg))1/3 (3/4)Q4/3 (8.2.12)
Producer cost including F (PC)
CQ + F (8.2.13)
Consumer surplus (CS)
CB − PQ (8.2.14)
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): C—Unit cost of
production; F—Fixed cost; g—Density of consumers everywhere; P—Delivered price set by
firm; s—Unit shipping rate; x—Distance from customer to firm; α—Intercept of individual
inverse demand curve; β—Slope of individual inverse demand curve. Outcomes (endogenous):
P0 —Price set by firm for an adjacent customer; P0 —Price set by firm for an adjacent cus-
tomer; Q—Aggregate amount demanded; q—Quantity demanded by a consumer; X—Range of
commodity.
8.3 Trade Area and Market Area in Retailing 211
B D
G
J I
Price
0
Distance from firm H E F
the market area represents the core of your current customers. There are two crit-
ical numbers involved in this definition: (i) the firm’s share of customer purchases
locally and (ii) sales within the market area as a percentage of the firm’s total sales.
We can imagine that, in the retail sector containing a set of stores, there is a market.
In a simple case, the trade area would be partitioned into a set of market areas, one
for each store. In that sense, any one firm’s market area is a geographic subset of the
market (i.e., the trade area) in which it and its competitors compete.
We can characterize a market area of a firm as (u, v) where u is proportion of sales
(or, alternatively, customers) by all firms in the area that accrue to the firm, and v
is the proportion of the firm’s total sales (or customers) that originate in that area.
For example, in a (60, 80) market area the firm captures 60% of purchases made
by customers from there, and the market area accounts for 80% of the firm’s total
sales. If a trade area contains just four firms, all clustered at a central location, and
doing equal sales, each would share the same (25, 100) market area. If the absence
of competitors means a firm’s market area is limited only by range, we could find
the (100, 100) market area for this firm. In practice, there are, in general, many ways
of drawing a (u, v) market area for any one firm. Usually, analysts choose a com-
pact area: e.g., a circle around the firm’s site with a minimum radius that satisfies
u and v.
Table 8.3 Model 8A: two suppliers on rectangular plane selling commodity at same f.o.b. price
(P)
Location of supplier 1
(0, 0) (8.3.1)
Location of supplier 2
(d, 0) (8.3.2)
Location of customer on boundary
(X, Y) (8.3.3)
Notes: Rationale for localization (see Appendix A): Z3—Implicit unit price advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): d—Distance (in
kilometers) between firm and competitor; P—F.o.b. price set each by firm and by competitor;
s—Unit shipping rate. Outcomes (endogenous): X—X-coordinate of customer on boundary
between the two firms; x1 —Distance from customer to firm; x2 —Distance from customer to
competitor; Y—Y-coordinate of customer on boundary between the two firms.
places. Therefore, viewed geometrically, the boundary between the two market areas
is the perpendicular bisector between the two firms.12 See the line CE on the map in
Fig. 8.3.
Because of competition among firms, we might expect that the firm will not be
able to realize the monopoly profit given in (8.1.10). After all, if the firm were to
earn substantial profits, other firms would be attracted into the industry. Where they
choose locations with a market radius that overlaps that of our firm, they cut into the
profit and the market area of our firm. Imagine the following thought experiment.
Two competitors—each pricing f.o.b. and producing the same product—locate d
kilometers apart on a rectangular plane. Assume initially the competitors are too
far apart to impinge on each other’s market: i.e., d > 2X where X is defined in
(8.1.9). They, therefore, each earn the profit given by (8.1.10) and have a circular
market whose radius is given by (8.1.9). Now, imagine we push the two competi-
tors closer together until d < 2X. At this stage, the two markets intersect, and the
12 To draw a boundary between the two suppliers, simply draw a straight line from Place s1 to 2
and then find a line that is perpendicular and bisects it. One method, familiar from high school
mathematics, is to use a compass to draw a circle of radius r (where r > d/2) centered at Place
1, repeat at Place 2, then draw a straight line (the perpendicular bisector through the points of
intersection of these two circles.
214 8 Staking Out the Firm’s Market
K H E
A(1) J GD B(2)
I F C
0 50 100
km km km
Fig. 8.3 Model 8B: map of boundary between market areas of two suppliers
13 Alfred Henry Thiessen (born 1872), an American climatologist, defined polygons around
individual rainfall stations to estimate total rainfall across a region.
14 A map polygon formed on a rectangular plane by constructing perpendicular bisectors to the
straight lines joining a point (typically a store) to similar points nearby. The partitioning of a map
in this way is called a Voronoi Diagram.
15 Georgy Fedoseevich Voronoi (born 1868), a Russian mathematician worked on polygonal par-
titioning of a two-dimensional plane. This follows on earlier work by Rene Descartes (writing
around 1644) and Johann Peter Gustav Lejeune Dirchlet (writing around 1850).
16 See Boots (1980), Byers (1996), Graham and Yao (1990), Miles and Maillardet (1982), and
Sibson (1980).
8.5 Model 8B: Market Area Boundary Between Two Firms Selling 215
Under the assumptions made and ignoring knife-edge cases (customer straddles
the boundary), a Thiessen polygon is a (100, 100) market area. That is because there
would be no incentive for a customer inside a polygon to purchase the commodity
from a firm elsewhere, or for a customer outside the polygon to purchase from this
firm.
So far, we have assumed the firms each charge the same f.o.b. price for their
input. Is that reasonable? Presumably, to the extent each firm can affect the price
they receive, they would want to set a price where marginal revenue equals marginal
cost as we saw earlier in (8.1.6). At the same time, one might imagine competi-
tion between the two firms should drive their prices to be similar if not identical.
However, the two firms each have their own market area. This implies the firms are
not perfectly competitive; is there any reason, therefore, why competition should
lead to the same price for each firm?
Retain all of the assumptions above, except now assume the firm at Place 1 on a
rectangular plane sells the commodity at one f.o.b. price (P1 ), while the firm at
Place 2, located d kilometers from Place 1, sells for another, P2 .17 For a purchaser
located x1 kilometers from Place 1 and x2 kilometers from Place 2, the effective
price of a unit is, therefore, P1 + sx1 purchased from Place 1 or P2 + sx2 purchased
from Place 2. See Table 8.4. For the moment, let us not ask why the two firms have
different f.o.b. prices.
Under these assumptions, a purchaser will be on the boundary if x1 − x2 =
(P2 − P1 )/s. The implication here is that the boundary is the set of all places such
that the difference in distance to two fixed places is a fixed amount (i.e., (P2 − P1 )/s.
However, this is also the definition of a hyperbola. In the special case where the
difference in distance is zero (i.e., P1 = P2 ), the hyperbola reduces to a straight line
(the perpendicular bisector) as we have already seen.
Under the assumptions made and again ignoring knife-edge cases where cus-
tomers straddle a boundary, a market area constructed in this way is (100, 100). As
in the case of the Thiessen polygon, there would be no incentive for a customer
inside the market area to purchase the commodity from a firm elsewhere or for a
customer outside the polygon to purchase from this firm.
Imagine now a thought experiment in which the two competitors initially charge
the same f.o.b. price. Assume the firms are close enough that the boundary between
them incorporates a perpendicular bisector. Now, suppose the firm at Place 2 were
to lower its price. The relevant portion of the market boundary would now become
a hyperbola. It would generally lie closer to Place 1 than does the perpendicular
bisector. See the boundaries FH (when P1 = P2 + 1) and IK (when P1 = P2 + 2)
Table 8.4 Model 8B: two suppliers on rectangular plane selling commodity at different f.o.b.
prices (P1 and P2 )
Location of supplier 1
(0, 0) (8.4.1)
Location of supplier 2
(d, 0) (8.4.2)
Location of customer on boundary
(X, Y) (8.4.3)
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): d—Distance (in
kilometers) between firm and competitor; P1 —F.o.b. price set by firm; P2 —F.o.b. price set by com-
petitor; s—Unit shipping rate. Outcomes (endogenous): X—X-coordinate of customer on boundary
between the two firms; x1 —Distance from customer to firm; x2 —Distance from customer to
competitor; Y—Y-coordinate of customer on boundary between the two firms.
in Fig. 8.3. As hyperbolae, these each bend back toward Place 1. Put intuitively, a
firm holds onto to customers best when they are close by or behind it. There is an
upper limit to the price the firm at Place 1 can set if it wants to have any market area
at all: P1 < P2 + sd. Of course, if positions were reversed and firm 1 had the lower
price, there would be a minimum price charged by the firm at Place 1 below which
the firm at Place 2 would lose its markets: P1 + sd < P2 .
I am not saying anything about the amount any one purchaser might demand;
I am saying only that the purchaser will or will not patronize a particular firm.
Nonetheless, there is an interesting insight here into the demand curve faced by
a firm. To see this, suppose each purchaser has a price-inelastic demand for just
one unit of the commodity per time period. Suppose further N purchasers in total.
If firm 1 sets its price too high (P1 > P2 + sd), the demand for its product is zero.
If firm 1 sets it price sufficiently low (P1 < P2 − sd) to undercut its competitor, its
demand is N. If firm 1 sets a price anywhere in between these two, it gets a point
on a negatively sloped demand curve. In other words, firm 1 has a demand curve
kinked at both P1 = P2 − sd and P1 = P2 + sd. Imagine a thought experiment in
which we push firms 1 and 2 closer together. As a polar case, d = 0 when Places
8.6 Model 8C: Why Do Prices Differ Among Firms? 217
1 and 2 are adjacent. Now, the demand curve for either firm reduces to the familiar
perfect competition case: a horizontal line. In that sense, the kink is not inconsistent
with neoclassical theory. However, we again have to ask why the kink arises. The
answer once again is tied up inextricably with the discreteness of geography. It is a
consequence of the idea that firm 2 is where it is, neither closer nor further away.
The purpose of this model is to better understand when and why competitors
might charge different f.o.b. prices for the same product. Introducing the possibility
that firms sell at different prices raises a question. What determines price in such
instances? Making the assumptions that firms are price takers and everyone faces
the same price is consistent with perfect competition. However, the notion implicit
in a market area is that with f.o.b. pricing customers face different prices depending
on location. Add to this the possibility that different firms set different prices, and
the notion of market equilibrium is called further into question.
To see how prices might be set, consider a simple geography wherein customers
are spread uniformly along an east−west line of infinite length and that firms are
also evenly spread out (one firm each d kilometers).18 Consider one firm with a
competitor to the west (labeled w) and another to the east (labeled e). Assume all
firms price f.o.b. Let 0 < x < d be the location (kilometers) of a customer to the
west of the firm. For that customer, effective price of a unit purchased from the firm
is P+sx and the effective price from competitor w is Pw +s(d − x). Assume Pw ≤ Pe
and that w does not price e out of the market: i.e., Pw + 2sd > Pe (remembering w
and e are 2d kilometers apart). Further, assume here P < Pw + sd and P + sd > Pe
so the three firms coexist. At the boundary, Xw , between competitor w and the firm,
we get (8.5.1). See Table 8.5. This yields a formula for the western boundary—see
(8.5.2)—wherein our firm gets half the market to its west (d/2) adjusted for any
difference in f.o.b. price with respect to its competitor there. Similar results apply in
the market to the east of our firm: see (8.5.3). Therefore, the total length of market,
X = Xw + Xe , for our firm is given by (8.5.4).
Suppose further customers are spread uniformly along this line at density g and
that each customer demands q units of the commodity regardless of price. In this
case, aggregate demand for the firm’s product is given by (8.5.5). If the firm has a
marginal cost of production, C, and a fixed cost of production, F, then its profit is
given by (8.5.6). The firm then sets a price so as to maximize profit: see (8.5.7).
Assume here the firm is myopic; it does not expect either firm w or firm e to change
their price in response to the price chosen. This yields the profit-maximizing price
shown in (8.5.8): halfway between the average of Pw and Pe on the one hand and
a cost (C + sd) on the other hand. Note C + sd is greater than the actual cost
(C + sd/2) of serving someone at a place midway between any pair of firms.
18 Other geographies have also been explored. See, for example, Sarkar, Gupta, and Pal (1997).
218 8 Staking Out the Firm’s Market
Table 8.5 Model 8C: customers spread uniformly along a line market (zero price elasticity) with
supplier every d kilometers and f.o.b. pricing
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): C—Unit cost of pro-
duction; d—Distance (in kilometers) between firm and competitor to East or West; F—Fixed
cost; g—Density of consumers (assumed constant along line); Pe —Price of competitor to east;
Pw —Price of competitor to west; q—Quantity demanded by a consumer; s—Unit shipping rate.
Outcomes (endogenous): P—F.o.b. price set by firm; Q—Aggregate amount demanded; X—Total
length of firm’s market; Xe —Distance to customer on boundary to east of firm; Xw —Distance to
customer on boundary to west of firm.
There is something special about C + sd here. If all three firms were to each have
an f.o.b. price of C+sd, there would be no incentive for our firm or, under symmetric
conditions, for either competitor to change price. If the two competitors each set a
price lower than C + sd, there would be an incentive for our firm to set its price
higher than them and closer to C + sd. On the other hand, if the two competitors
each set a price higher than C + sd, there would be an incentive for our firm to set
its price lower and closer to C + sd. Therefore, C + sd is a kind of equilibrium price
for all firms. I do not want to make too much of this idea, however, because of the
following conundrum. If all firms in the market, not just the three examined here,
were to conspire, they would want to set the price as high as possible since we have
assumed demand is insensitive to price. However, in (8.5.8), there is no mechanism
that would drive prices up in this way. Of course, you might say, this is a competitive
market so why would we expect prices to be driven up? If it is indeed competitive,
why isn’t profit being driven to zero? At price equal to C + sd each firm has the
potential to some excess profit unless fixed cost is prohibitive. Why? Presumably,
this is because the market is rich enough to support firms at a density at least equal
8.6 Model 8C: Why Do Prices Differ Among Firms? 219
to one firm per d kilometers. This is the so-called Löschian equilibrium problem in
which we imagine free entry of new firms, and the rearrangement of these firms in
geographic space, and a shrinking of d until the excess profit earned by the marginal
firm is effectively zero.19
As in Chapters 3 and 6, this chapter assumes that customers each demand a fixed
amount of output, q units, regardless of price. Can we release this assumption? Is it
possible, for instance, to imagine a linear inverse demand curve for each customer
here? The firm’s aggregate demand Q on this linear market would then be defined by
(8.6.1): see Table 8.6. Profit would then be given by the identity (8.6.3). Assuming
again the firm is myopic in that it sets a price for its product ignoring the possi-
bility that the two competitors might react by changing their prices implies (8.6.4).
To solve this equation, we need to differentiate (8.6.2): see (8.6.5). As well, we
need to remember Xw , Xe , and X here are determined by (8.5.2), (8.5.3), and (8.5.4).
Unfortunately, substituting these equations back into (8.6.4) gives a quadratic equa-
tion (8.6.6) from which it is difficult to obtain insights into the determinants of P.
That is why I presented the zero price elasticity of demand solution in Table 8.5; it
is easier to solve and interpret.
Table 8.6 Model 8C: customers spread uniformly along a line market (nonzero price elasticity)
with supplier every d kilometers
Demand
x for firm’s product x
Q = o w (α − (P + sx))(g/b)dx + o e (α − (P + sa))(g/b)dx (8.6.1)
Which yields
Q = (g/β)((α − P)X − (s/2)(Xw2 + Xe2 )) (8.6.2)
Profit
Z = (P − C)Q − F (8.6.3)
Myopic first-order condition
∂Z/∂P = Q + (P − C)∂Q/∂P = 0 (8.6.4)
where
∂Q/∂P = (g/β)(−X/2 − (α − P)/s) (8.6.5)
Yields a quadratic in P
a1 P2 + b1 P + c1 = 0 (8.6.6)
where
a1 = 6 − 3s (8.6.7)
b1 = 2s(α + C) − 2(2α + C) − 3(Pw + Pe + 2ds) (8.6.8)
c1 = (2α + C) (Pw + Pe + 2ds) + s(Pw + ds) (Pe + 2ds)− (8.6.9)
(s/2) (Pw + Pe + 2ds)2 − 2saC
19 Inan industry where firms produce the same commodity and compete by choosing geographic
locations, a long-run equilibrium wherein firms no longer have the incentive to enter or leave the
industry. Every firm earns normal profit only. See, for example, Mai and Hwang (1994).
220 8 Staking Out the Firm’s Market
that solution did not give us much of an idea as to the geography of market areas.
Now, we know that the boundary between two suppliers on a rectangular plane will
be either a perpendicular bisector or a hyperbola bending back toward the supplier
with the higher unit cost or the capacity constraint. So, the conclusion here is that
a capacity constraint affects the market boundary exactly the same manner as does
a price difference. It causes the market boundary to become hyperbolic and to bend
back in the direction of the factory with the binding capacity constraint.
Under the assumptions made (see Table 8.7) and once again ignoring knife-edge
cases of customers straddling a boundary, a market area constructed in this way
is (100, 100). As in the case of Models 8A and 8B, there would be no incen-
tive for a customer inside the market area to purchase the commodity from a
supplier elsewhere or for a customer outside the polygon to purchase from this
supplier.
8.8 Model 8E: Market Area Boundary Between Two Firms with
Different, but Perfectly Substitutable, Commodities
Now assume the two firms each use f.o.b. pricing. See Table 8.8 for the list of
assumptions and rationale for localization. Suppose that customers see the two firms
as different but fully substitutable; each customer would purchase a quantity of only
one of these two commodities. If the two firms are otherwise identical, the cus-
tomer’s choice then depends strictly on the price they have to have to pay, inclusive
of shipping costs. If customers see no inherent difference between the two com-
modities, they would choose the less-expensive commodity; Model 8E reduces to
Model 8A if the commodities have the same f.o.b. price or Model 8B if their f.o.b.
prices differ. On the other hand, suppose the commodities are inherently different
in a way that customers prefer commodity from supplier 2 when effective prices are
222 8 Staking Out the Firm’s Market
the same. Presumably, there is a difference in effective price at which the customer
is indifferent between the two commodities. That difference therefore defines the
boundary between the two firms. That difference also gives rise to a hyperbolic
boundary as we have already seen in Model 8B. Under the assumptions made (see
Table 8.8) and again ignoring straddling consumers, a market area constructed in
this way is (100, 100). As in the case of Models 8A, 8B, and 8D, there would be no
incentive for a customer inside the market area to purchase the commodity from a
firm elsewhere.
8.9 Model 8F: Market Area Boundary Between Two Firms with
Different, but Perfectly Substitutable, Commodities When
Customers Are of Two Types
Until now, we have assumed that customers are everywhere identical. How does cus-
tomer heterogeneity add complexity to the determination of market areas? Assume
two different kinds of customers: types 1 and 2. See the list of assumptions and
rationale for localization in Table 8.9. Every customer of type 1 is identical. So too
is every customer of type 2. Assume here that the two firms sell different but per-
fectly substitutable commodities and price f.o.b. See Table 8.9. It is then possible
to imagine a market area for any given firm among customers of type 1 based on
Model 8E above. Similarly, we can imagine a market area for the same firm among
customers of type 2. These two market areas may well differ; for example, cus-
tomers of type 1 may prefer the product of the first firm (where effective price is
the same), whereas customers of type 2 prefer the second. In such cases, it would
not be possible in general to construct a single (100, 100) market area that covers
both types of customers. Here then is a situation where, no matter how we draw the
market boundary we might expect some customers inside the market area for firm
1 to purchase from firm 2, and some customers outside the market area of firm 1 to
purchase from that firm nonetheless.
8.10 Model 8G: Market Area Boundary Between Two Firms 223
This analysis generalizes from Model 8E. At this point in the book, we need some-
thing more general than a demand curve for a single product; we need to model how
a consumer substitutes between two or more commodities. The economist does this
by means of a utility function, U = f (q1 ,q2 ) that shows how the well-being (U) of
the consumer varies depending on the amounts of commodities 1 and 2 consumed:
q1 and q2 , respectively. An indifference curve traces combinations of quantities
of commodity 1 and commodity 2 that generate the same level of well-being for
the consumer. Maximizing utility subject to a budget constraint in turn yields the
individual demand curves that we have used to this point in the book.
In a world of differentiated commodities, it may seem strange to be focused here
on the case where two commodities are perfectly substitutable. Is it not reasonable
to expect that a given customer would purchase some amount of the commodity
from firm 1 and some amount from supplier 2? Assume two suppliers on the map: d
kilometers apart. They each sell one product. Each uses f.o.b. pricing. The prices
set by the two firms are p1 and p2 , respectively. Consumers are identical. They
each consume only the commodities provided by the two suppliers. Where q1 is
the amount purchased by a customer from the first store monthly, and q2 is the
amount purchased from the second store, each is rational; he or she maximizes the
log-linearutility function given in (8.10.1).20 See Table 8.10. For ease of exposition,
Table 8.10 Model 8G: two stores selling different commodities; consumers with log-linear or
CES utility functions
Log-linear utility
U = qν1 q1−ν
2 (8.10.1)
Budget constraint
Y = (p1 + sx1 )q1 + (p2 + sx2 )q2 (8.10.2)
Notes: Rationale for localization (see Appendix A): Z7—Variation in goods; Z8—Limitation
of shipping cost. Givens (parameter or exogenous): pi —F.o.b. price charged by firm i; s—Unit
shipping rate; Y—Income of customer; ν—Relative preference for commodity 1: log-linear;
d—Relative preference for commodity 1: CES; r—Substitutability of. Outcomes (endogenous):
qi —Individual consumption of commodity 1; xi —Distance from firm i; X—Fraction of distance
from Supplier 1 to Supplier 2; U—Utility of individual.
these are the only goods consumed and therefore exhaust consumer income. Assume
also each consumer incurs a shipping cost of s dollars per kilometer for each unit of
a commodity consumed. Consumers each have the same monthly income (Y). For
a customer at distance x1 from one store and distance x2 from the other store, the
budget constraint is (8.10.2).
A utility-maximizing consumer can be shown to have demands for the two com-
modities as described by (8.10.3) and (8.10.4). These do not give the linear inverse
demand curve that I have used elsewhere in the book so far. The log-linear utility
function (8.10.1) used here has the implication that the consumer always spends
fixed fractions (ν and 1 − ν) of income on the two goods.21 The demand for either
21 The literature includes some experimentation with alternative utility functions. See, for example,
Lo (1990, 1991a, 1991b, and 1992).
8.10 Model 8G: Market Area Boundary Between Two Firms 225
commodity does not depend on the price of the other commodity. This zero cross-
price elasticity is another feature of the log-linear utility function (8.10.1) used
here. While consumers exhaust their income on spending for the two commodi-
ties, (p1 + sx1 )q1 and (p2 + sx2 )q2 , respectively, the revenues received by each store
from the consumer are only p1 q1 and p2 q2 , respectively. Therefore, the expenditure
ratio for a customer at any given location is given by (8.10.5). I define the boundary
of the market area to be the locations where (p1 q1 )/(p2 q2 ) = 1. Under the boundary
condition (8.10.6), x2 is a linear function of x1 . Note here restrictions on (8.10.6):
x1 ≥ 0, x2 ≥ 0, and x1 + x2 ≥ d.
What does the market boundary look like? In the special case where ν = 0.5
and p1 = p2 , (8.10.5) reduces to x2 = x1 which means that the boundary is the
perpendicular bisector. For other parameter values, the boundary is of an elliptical
form. Under the assumptions made, a market area constructed in this way is never
(100, 100). No matter how the market area for supplier 1 is drawn, customers inside
it will still purchase some amount from supplier 2, and customers outside it will
purchase some amount from supplier 1.
As noted elsewhere in this book, a log-linear utility function implicitly assumes
that the cross-price elasticity—e.g., dln[q1 ]/dln[p2 ]—is always zero and the elas-
ticity of substitution
is always −1. The CES utility function shown in (8.10.7) is a more general utility
function; it has two parameters (δ and ρ) where δ (like ν), constrained to lie between
0 and 1, is the relative preference for good 1 and ρ,constrained to lie between −1
and 0 is tied to the elasticity of substitution. From (8.10.8),
To illustrate ideas, consider now a customer located somewhere along the straight
line joining Suppliers 1 and 2 on a rectangular plane. See Fig. 8.4. Let x denote the
fraction of the way from Suppliers 1 to 2: x therefore ranges from 0 (customer
located adjacent to Supplier 1) to 1 (customer located adjacent to supplier 2). When
x is near 0, p1 + sx1 is relatively small and p2 + sx2 is relatively large; the oppo-
site holds when x is near 1. In the case of a log-linear utility function, the budget
share for a commodity is the same for customers everywhere along this line; how-
ever, for customers closer to Supplier 1, this in turn implies more quantity of that
commodity. In the case of a CES utility function, both the budget share and quan-
tity of commodity 1 increase as x approaches 0. As we increase x—move closer to
Supplier 2—we find customers with a CES utility function reallocate budget share
in favor of commodity 2. In fact, there is a distance (a fraction X of the length from
Suppliers 1 to 2) where the budget share is the same regardless of substitutability
(ρ). See (8.10.9) and distance OK in Fig. 8.4. We also see that as the elasticity of
226 8 Staking Out the Firm’s Market
A B
0
K I
Fraction of distance from Firm 1 to Firm 2
Fig. 8.4 Model 8G: budget share schedule along line from Suppliers 1 to 2 using log-linear and
CES utility functions.
Notes: Horizontal axis scaled from 0 to 1; vertical from 0 to 1.2. Givens also include p1 = 11,
p2 = 12, s = 10
substitution becomes larger (in absolute value) the budget share schedule transi-
tions from AB to CD to FG to HI. The logical implication is that as the two goods
approach perfect substitutes, the budget share approaches a Z shape (HJKI) where
budget share is 1.0 when x < X, 0.0 for x > X, and in effect a vertical line at x = X.
In such a case, Model 8H reduces to Model 8A (if p1 = p2 ) or 8B (if p1 = p2 .)
22 Astatistical model used to predict consumer choice from among a set of discrete and
denumerable alternatives. This includes the Multinomial Logit Model and Nested Logit Model.
8.12 Final Comments 227
to have begun with Reilly (1931).23 In its simplest form, discrete choice modeling
imagines that from among all possible stores the consumer formulates a choice set.
The consumer is then assumed to collect information about stores in the choice set
and then choose one store to make the purchase.24 In so doing, there may be dif-
ferences among consumers that lead them to formulate different choice sets and/or
make different choices. The analyst is assumed to know only about some (not all) of
the differences among consumers and some (again not all) of the differences among
stores. Because of this imperfect information, we can never be sure that we have
properly identified the consumer’s choice set nor the full set of factors that shape
choice from within the choice set.
Analysts typically use a variant of a multinomial logit model to predict shop-
ping behavior. In a logit model, the log odds (reflecting the analyst’s uncertainty)
of choosing a particular alternative are seen to gradually rise (fall) as conditions for
that choice become more (less) favorable. By varying the magnitudes of the slope
coefficients in the model, we can make consumers more sensitive (or less sensitive)
to a particular factor. Earlier in this chapter, I presented several models in which
there is a critical distance at which the customer suddenly and completely shifts
from one supplier to another. In principle, it is possible to make the multinomial
logit model reproduce a sudden and complete shift, but the model is most appropri-
ate when change is gradual. What is not clear here, however, is just when and why
change ought to be gradual. This brings us back to Hotelling’s thoughts on stability
in competition.
In this chapter, I have explored ways of defining the market area boundary for a
firm. In Table 8.11, I summarize the assumptions that underlie selected models from
8A through 8G. Many assumptions are in common to all these models: see the list
in panel (a) of Table 8.11. In 8A, the firm and its competitor sell the same com-
modity at the same given f.o.b. price. In 8B, the competitor sets an f.o.b. price that
is different but given. In 8C, the firm gets to set a price that maximizes its profit
assuming that competitors do not react. In 8D, I study how the firm’s market area
boundary adjusts to capacity constraints. In 8E, I show how the firm’s market area
boundary varies when it sells a different, but perfect substitutable good. In terms
23 Other early work in the area includes Carrothers (1956), Brown (1957), and Huff (1963, 1964).
Important contributions include Davis (2006), de Palma, Lindsey, von Hohenbalken, and West
(1994), Iyer and Seetharaman (2008), Lee and Pace (2005), Miron and Lo (1997), Sheppard,
Haining, and Plummer (1992), and Slade (2005). O’Kelly and Miller (1989) and Parr (1997b)
are good summaries of work in this area. See also Berry, Parr, Epstein, Ghosh, and Smith (1988,
Chap. 7).
24 Dudey (1990) considers the impact of consumer search behavior on the location choices of
retailers. Also see Schulz and Stahl (1996).
228 8 Staking Out the Firm’s Market
8A 8B 8C 8D 8E 8F 8G
Assumptions [1] [2] [3] [4] [5] [6] [7]
of the geometric shape of market areas, Models 8B, 8D, and 8E give similar out-
comes. Models 8F, 8G, and 8H introduce differences among consumers, as well as
imperfectly substitutable goods.
In Chapter 1, I argue that there are important linkages between prices and local-
ization. I illustrated that idea in Model 2D wherein price and localization were joint
outcomes for the monopolist. In this chapter, I consider how a firm sets its price in
response to the prices set by its competitors nearby and hence is affected by the geo-
graphic density of firms. The idea here has been to explore how, as a consequence,
localization and price are jointly determined. Much more could be done here. There
is an extensive game-theoretic literature in which the locations of firms and their
prices are reactions to the actions of their competitors. By itself, that is the subject
of another book. My modest aim in this chapter has been to provide pointers in that
direction.
8.12 Final Comments 229
(Economides–Siow Problem)
1I first considered density of customers in Chapter 7 and then again in Chapter 8 in more detail.
9.2 The Barter Market 233
are beyond the farm’s control. To maximize its utility, a farm whose harvest con-
sists mainly of grain production might want to exchange with other farmers whose
harvest was mainly other kinds of desirable produce: e.g., eggs, milk, meat, or veg-
etables. As a supplier, you decide the type and quantity of produce to bring to market
without knowing in advance what will be brought by other farmers. As a consumer,
in deciding whether to attend a local market, you cannot be sure in advance about
the availability, quality, and exchange rate (price) of various produce. In the interest
of simplicity, assume buyers and sellers have perfect information once they arrive
at the local market and that this translates into a single Walrasian market-clearing
exchange rate for the day; all those who want to exchange produce at that rate are
able to do so. Under what conditions will a farm participate in a particular local
market? If a farm has a choice between participating in a local market nearby or a
larger market further away, which will it choose and why?
This chapter is inspired by a pioneering spatial model of liquidity and market size
in Economides and Siow (1988).2 At the same time, the E&S model is different from
the one presented here. How? First, the E&S model assumes that actors maximize
expected utility. Here, however, I assume behavior based on risk-return. A second
difference from the E&S model is that I start with a version where geography plays
no role. Later in the chapter, I introduce into this model a geography different from
the E&S model. I will provide more detail on these and other differences later in
this exposition.
2 Economides and Siow (1988) are primarily concerned with the existence and size of financial mar-
kets. However, at the outset, that paper describes a simple locational model of trading by farmers,
which is the focus of this chapter. Others who have made use of Economides and Siow (1988) to
look at questions of location include Camacho and Persky (1990), Casella (2001), Gehrig (1998),
Glazer, Gradstein, and Ranjan (2003), and Henkel, Stahl, and Walz (2000).
3 I use barter here in the economic sense of an exchange—a trade of some amount of one good
in return solely for an amount of another good with no money involved—that takes place in the
context of a market. This is seen here strictly as a matter of business; I exclude here any exchange
(e.g., an exchange of gifts) where the motivation is, at least in part, something else.
234 9 The Cautious Farmer and the Local Market
commodity. Among the reasons why people demand money is that it is a store of
value; by a further transaction, you can readily get back most or all of what you
gave up for it. Other commodities, such as gold, are also thought to be like money
in the sense that they too are a store of value. If indeed money is a commodity, then
in effect even exchanges in a fiat money economy are barter.
While economic actors can meet up to match their needs on a pairwise basis, it is
not always clear how an exchange rate (price ratio) gets established between every
pair of commodities in a barter economy. Suppose, for example, we have a simple
barter market that includes only two farmers. For the sake of argument, assume
Farmer X arrives with an endowment of 0.4 units of wheat (commodity 1) and 0.6
units of corn (commodity 2): and that Farmer Y has an endowment of 0.60 units of
wheat and 0.40 units of corn: qX1 = 0.40, qY2 = 0.40, qY1 = 0.60, and qX2 = 0.60.
In Fig. 9.1, I draw a diagram—called an Edgeworth Box—for this problem. There,
A5
A3
A
C3
C5
de f
Good 2
c
b
Walrasian
exchange
rate
B5
B3
E
B
D
D3
D5
0
Good 1
Fig. 9.1 Barter and Walrasian price setting in market with two farmers (X and Y).
Notes: ν = 0.3. Initial endowments are qX1 = 0.40, qX2 = 0.60, qY1 = 0.60, and qY2 = 0.40.
Indifference curve reached in absence of barter: AbEB for person X; AfED for person Y. Walrasian
exchange rate shown as dotted line passing through point A on the vertical axis and point d along
the Marshallian Contract Curve bcdef. Horizontal axis scaled from 0.40 to 0.75; vertical axis from
0.40 to 0.65
9.2 The Barter Market 235
point A represents the endowments before any barter. Through point A, I have drawn
Person X’s indifference curve (AbEB) before barter. The idea of an Edgeworth Box
is to show the indifference curves also for person Y; since total endowments are
fixed, we can draw Y’s indifference maps from an origin at the upper right-hand
corner of Fig. 9.1. In that case, Y’s indifference curve before barter would be AfED.
The eye-shaped area above AbE and below AfE contains all the barter outcomes
that would leave each of X and Y better off (or at least no worse off) than without
barter.
Where within this eye does bartering lead us? There are two answers to this
question: Marshallian and Walrasian.
For simplicity of exposition, I adopt the Walrasian view in the remainder of this
chapter and find the exchange rate that will leave neither farmer wanting to trade
any more.
4 Implicit in this description is an assertion that such a point exists and is unique. A determination
of the conditions under which this assertion is valid is beyond the scope of this book.
5 The slope of the Walrasian exchange rate line is the negative of the exchange rate.
236 9 The Cautious Farmer and the Local Market
• A large number of potential buyers and a large number of potential sellers. Put
differently, the asset is widely desired and widely held.
• The asset is held at zero (storage) cost. However, buyers (sellers) incur transaction
costs to acquire (dispose) of the asset.
• Potential buyers and potential sellers each form a statistical population of individ-
uals. Overall, these populations are both of the view that the market price of the
asset is not expected to change in the future. However, individuals within either
population randomly deviate from this. On any given market day, each potential
buyer (seller) has a broad sense of (1) the current price and expectations about
the future price of the commodity—expectations that differ randomly from per-
son to person and from day to day—and of (2) the transaction costs associated
with their participation in the market.
• At the outset of any given market day, a subset of the potential buyers (call
them market buyers) and a subset of the potential sellers (call them market
sellers) engage in the market. By engage, I mean they undertake one or more
of the following activities: search, gather, and analyze information; make con-
tacts and establish relationships; negotiate price and terms; and acquire/dispose.
Why only a subset? In my view, time and effort are required (i.e., transaction
costs are incurred) to do these things. To acquire or dispose of a commod-
ity, for example, these would—as noted earlier—include costs related to bank
transactions and credit authorization, freight and transfer, storage and inventory,
agency and brokerage fees, cost of insurance and other loss risks, installation
and removal, warranty and service, and taxes and tariffs. As the day progresses,
some market buyers/sellers will transact; others will—on the basis of the infor-
mation obtained—choose not to transact. For simplicity, I assume that decisions
to engage the market in previous days do not affect the transaction costs to be
incurred today.
• Other potential buyers and sellers do not participate in the market. I assume here
that the perceived transaction costs are large enough relative to the gain expected
keep such potential buyers or sellers out of the market that day.
• Except insofar as transaction costs and price expectations in part vary randomly
from one person to the next, market buyers and market sellers are no different
6 A market condition in which there are many buyers and sellers. From a search-theoretic perspec-
tive on markets, a seller in a thick market does not have to wait long to get a fair price for their
good.
7 A market condition in which there are few buyers and sellers. From a search-theoretic perspective
on markets, a seller in a thin market typically must wait longer to get a fair price for their goods.
9.3 Uncertainty and Rationality 237
statistically from other potential buyers and sellers; they can be thought of as just
an independently drawn random sample.
• Market sellers in aggregate have an upward-sloping supply schedule showing the
amount they would sell at any given exchange rate.
• Market buyers in aggregate have a downward-sloping demand schedule showing
the amount they would purchase at any given exchange rate.
• Following the Walrasian perspective, the market exchange rate that day settles at
a level such that no market seller leaves with product that they would prefer to
have sold at that exchange rate and no market buyer leaves without product they
would have preferred to have bought at that exchange rate. Put differently, the
market clears for market buyers and sellers that day. If the market did not clear,
then either market suppliers would be left with inventory or market demanders
would be left with unsatisfied demands.
For an asset market that can be characterized this way, we would therefore expect
to see some variation in market price from day to day even when no one expects
price to change over the longer run. We also expect to see the number of market
sellers or buyers rise one day and then perhaps fall the next on a random basis. I
define the market to be thin when the numbers of market buyers and market sell-
ers are small: thick when the numbers are relatively large. We might expect that the
price of the asset in a thick market would be about the same from day to day because
of the many participants. In a thin market under similar circumstances, however, we
expect the exchange rate of a commodity to vary more from day to day. Put differ-
ently, there is more price risk8 in a thin market than in a thick market; the vendor
in a thin market, for example, might get less than, or more than, either potential
suppliers or potential buyers think is the expected price for the asset.
Why engage in a market at all? In general, choosing a market can be seen as a
means of reducing or spreading price risk. As a farmer, you do not necessarily need
to participate in a weekly farmer’s market. You might have, for example, estab-
lished relationships with one or more customers who travel to your farm weekly to
purchase commodities. Why bother with the inconvenience of shipping to market if
you can get a good exchange rate at the farm gate? However, if you do not attend the
market, it is hard to know whether you are getting a good exchange rate; customers
too might want to know if they are paying too much. For both farmer and customer,
the market provides a means of assessing whether the exchange rate for a given
transaction is fair. Even a thin market can be helpful here. However, the thicker the
market, the less the price risk.
8 A loss (or increase in cost) arising because of an unforeseen change in market conditions that
causes price to change over the short term, price risk is associated with price volatility. In a search-
theoretic perspective, sellers hold an asset until the price bid by a potential purchaser exceeds the
vendor’s reservation price. Here, a distinction can be drawn between price risk and liquidity risk.
Liquidity risk is the loss arising because of the delay in obtaining a bid at or above the reservation
price. In practice, it is difficult to distinguish between price risk and liquidity risk. The approach in
this book is to treat liquidity risk as simply an element of price risk.
238 9 The Cautious Farmer and the Local Market
To implement the notion of price risk, we need to think about what it means to
be rational under uncertainty.9 So far in the book, we have used the term rationality
only in regard to choices made under certainty. On the one hand, we have assumed
the firm maximizes profit or rate of return. On the other hand, we have assumed the
individual maximizes utility, which in turn has been based on the consumption of
commodities. In both situations, economists might argue that it is straightforward to
model choice in the absence of uncertainty.10
In 1738, Daniel Bernoulli (a Swiss mathematician) developed a theory on the
measurement of risk that set the stage for modern approaches.11 He starts from the
notion of an expected value E[X]. For a discrete random variable (X), this is the
sum of all possible occurrences of X each multiplied by the probability, p[X], of that
occurrence: E[X] = x x(p[X = x]).12 However, he then assumes that
1. the value someone places on an outcome depends not on the expected money
gain from a gamble but rather on the utility it yields;
2. the added utility, U[X], from a given money gain, [X], is more for a pauper than
for someone who is rich; and
3. the expected value of the gain in utility, E[U] = x U[x]p[X = x], is what
motivates individuals in a gamble.
Note the implication here, drawn out by Bernoulli himself (p. 29), that no one
would therefore rationally gamble in a fair game—one in which a dollar gain was
as likely as a dollar loss—because the loss has a greater change in utility attached
to it than does the gain.
Bernoulli’s assertion (2) above is problematic with respect to the ordinality of
utility in two regards. First, in effect, he assumes that individuals have a diminishing
marginal utility of income. Why might this be problematic? Diminishing marginal
utility of income itself need not be surprising since we commonly assume diminish-
ing marginal utility in commodities consumption. However, in practice, we assume
that a utility function is unique up to a monotonic transformation. For example, the
utility functions f [x, y] = xb y1−b where 1 < b < 0 and g[x, y] = axb yc where
b > 0, c > 0, and b + c < 1, calculated at consumption of x units of wheat and
y units of corn, generate the same rank ordering: i.e., g[x, y] is a monotonic trans-
formation of f [x, y]. The easiest way to think about diminishing marginal utility of
9 Important work in the area of utility and decision making under uncertainty includes von
Neumann and Morgenstern (1947), Marschak (1950), Hurwicz (1953), Simon (1955, 1959), Koo
(1959), Bishop (1963), Harsanyi (1965, 1966), Loomes and Sugden (1982), and Sugden (1991).
10 For the interested reader, Sugden (1991) discusses the philosophical limitations of neoclassical
perspectives on rational choice.
11 Bernoulli (1954) is an English translation of that paper.
12 For example, if we toss a fair coin twice and let X be the number of times a head obtains. X can
take on the values 0, 1, and 2. From the Binomial Theorem, we know that probabilities are 1/4,
1/2, and 1/4, respectively. Therefore, E(X) = 0(1/4) + (1/2)(1) + (1/4)2 = 1.0.
9.3 Uncertainty and Rationality 239
income is that b + c < 1, but then how would this differ from a monotonic transfor-
mation of b + c = 1? Second, Bernoulli assumes that the utility levels of consumers
(the pauper and the rich person) are comparable which again violates the ordinality
of utility.
Although risk in the context of investment had long been of concern in
Economics, it is Von Neumann and Morgenstern’s path-breaking book, Theory of
Games and Economic Behavior, first published in 1944, that is widely credited with
spawning the focus in Economics in general on game theory and in particular on
the nature of rational behavior in the presence of uncertainty.13 That book builds
on Bernoulli’s idea that economic actors maximize the expected value of the gain
in utility. However, there are problems with expected utility maximization. (1) In
practice, how do we find the required probabilities, especially when these may be
subjective (Bayesian) in nature? (2) Do individuals have a taste for risk or an aver-
sion to risk that leads them to prefer one gamble to another even when two gambles
have the same expected utility? (3) More generally, why assume that rationality
necessarily requires expected utility maximization?
An alternative to analyze rational decision making under uncertainty is through
a mean-variance (or, alternatively, risk-return) approach that originated with
Markowitz (1952) and Sharpe (1963, 1964). Under this approach, we calculate
two measures: (1) the expected utility ε—otherwise known as the mean or as the
return—as per Bernoulli and (2) the variance ν—otherwise known as the risk—
in utility.14 If two choices have the same return but different risks, the individual
is thought to prefer the choice with the lower risk. If two choices have the same
risk but different returns, the individual is thought to prefer the alternative with the
higher return. If the two choices have different returns and different risks, then we
need some way to measure the tradeoff between return and risk. Typically, this is
done using what is termed a beta analysis.15
In this chapter, such a risk-return approach is used to characterize rational choice
under uncertainty. Here, I distinguish between sub-utility and utility. Sub-utility
is the level of happiness that arises from a choice when uncertainty is, or can
be, ignored. Utility is the level of happiness after uncertainty has been taken into
account; in a conventional beta analysis, utility is given by (9.1.1). Beta here is a
parameter that measures the aversion of the individual to risk; when β = 0, the
individual is indifferent to risk, for larger β, the individual is increasingly averse to
choices with substantial risk.16
13 In my view, Georgescu-Roegen (1954, p. 503) is correct in pointing out that mathematicians and
statisticians dating back to Daniel Bernoulli and Gabriel Cramer had worked on similar ideas much
before this. Harsanyi (1956) points out the similarities of game theory to earlier work by Zeuthen
(1930).
14 That is, ε = U[x]p[X = x] and ν = (U[x] − ε)2 p[X = x].
x x
15 Beta is the increase in mean (return) required to offset a unit increase in variance if two
alternatives are to be thought to be equally preferable.
16 This is an approach initially suggested by Markowitz (1976). See also Levy and Markowitz
(1979) and Kroll, Levy, and Markowitz (1984).
240 9 The Cautious Farmer and the Local Market
Many advances have been made using a risk-return approach in an area now
known broadly as financial engineering. However, the approach is not without its
critics. Among these are the following:
• Some economists are not fond of the risk-return approach; they prefer an
approach better grounded in neoclassical utility theory. To simplify the invest-
ment problem here, imagine an individual choosing between a risky investment
with a higher expected return and a risk-free investment with a lower return.
In effect, by declining the risky investment, individuals forego an amount (the
amount by which the expected return is higher in the risky investment) to guar-
antee their wealth (the principal invested) at a future date. As such, deciding
to invest in the risk-free alternative is like buying insurance and should be
analyzable in that way.
• The risk-return approach does not directly incorporate an asymmetry
(skewedness) of gains and losses as proposed by Bernoulli.
• The risk-return approach—as usually applied in investment analysis—assumes a
continuity across investment choices: i.e., the ability to blend investments at dif-
fering levels of risk. The location problem that I consider in this chapter exhibits
a kind of lumpiness that needs to be addressed specifically.
On the other hand, students tell me that they, or their parents, deal with financial
advisors who regularly cast investment portfolio choices in terms of risk versus
return. Therefore, I find it helpful pedagogically to cast this problem using a risk-
return approach.
17 Economides and Siow (1988) also look at the case where the utility function is Constant
Elasticity of Substitution (CES).
9.4 Model 9A: Non-spatial Market 241
Table 9.1 Model 9A: farmers and the market in a non-spatial economy
Utility of farmer
U − βV (9.1.1)
Sub-utility of farmer
U = q1ν q21 − ν (9.1.2)
Initial endowment of farmer
P[1, 0] = 1 − φ and P[0, 1] = φ (9.1.3)
Mix of farmers in market
N = N1 + N2 (9.1.4)
Probability of getting N1 wheat farmers in market
P[N1 ] = CNN1 (1 − φ)N1 φ N−N1 where CN1 N = N! / (N !(N − N )!)
1 1 (9.1.5)
Expected number of wheat and corn farmers in
market
E[N1 ] = (1 − φ)N E[N2 ] = φN (9.1.6)
Wheat and corn expected to be offered in market
E[Q1 ] = (1 − ν)(1 − φ)N E[Q2 ] = νφN (9.1.7)
Ratio of expected offers (wheat per unit of corn)
E[Q1 ] / E[Q2 ] = (1 − ν)(1 − φ) / (νφ) (9.1.8)
Notes: C—Combination symbol: Cab = b!/(a!(b − a)!); P—Probability. Rationale for localization
(see Appendix A): Z4—Risk-spreading and insurance; Z6—Differences among consumers; Z7—
Variation in goods. Givens (parameter or exogenous): N—Number of farmers in market; β—Risk
aversion parameter; ν—Exponent of wheat in utility function. Relative preference for wheat; φ—
Probability farmer has (0,1) endowment. Outcomes (endogenous): N1 —Number of wheat farmers
in market; N2 —Number of corn farmers in market; Q1 —Aggregate quantity of wheat offered in
market; q1 —Quantity of wheat consumed by farmer; Q2 —Aggregate quantity of corn offered in
market; q2 —Quantity of corn consumed by farmer; U—Sub-utility of farmer; V—Variance.
that each farmer is efficient and that those who harvest corn (as well as those who
harvest wheat) produce an amount of 1 unit of the commodity net of costs including
the opportunity cost of land (rent). Further, assume that these initial endowments
obtain as though outcomes were random and statistically independent and that for
each farmer there is a probability φ that he or she will be endowed with corn, and
therefore 1 − φ probability of being endowed with wheat. See (9.1.3).
In illustrating Model 9A (and again in Model 9B and Model 9C that follow),
I use particular values for ν and φ. I assume ν = 0.3, which implies consumers
prefer to consume relatively more wheat than corn. I assume φ = 0.4 which means
that farmers are more likely to be endowed with wheat than corn. Together, these
two values describe a world in which 60% of farmers are endowed with wheat, but
where each farmer wants to spend only 30% of his or her endowment on wheat
consumption. In that sense, our farmers would be happier if endowments of corn
were more commonplace (in other words, if 1 − φ were closer to ν) and less happy
otherwise. This further contributes to the imperative to trade. If they do participate
in a market, they give up a portion of their initial endowment to get some amount of
the other commodity to consume.
242 9 The Cautious Farmer and the Local Market
Suppose N farmers constitute a market for this purpose: see (9.1.4). To simplify
the subsequent analysis, I assume that farmers decide on their size of market in
advance of knowing either their endowment or that of anyone around them (i.e., the
endowments of other farmers who might be in the same market). This assumption
may seem strange. After all, why not let farmers choose their market later. However,
this assumption will make more sense later in this chapter when we introduce
geography into the model. Having decided on a size of market and having sub-
sequently harvested and realized their endowment, I assume that the farmers then
meet in this market, transact as best they can, and get the utility that arises to all
farmers with their endowment at the conclusion of the market. As stated above, the
endowment for each farmer among these N is stochastic and independent of the
endowment of any other farmer. Then, we can think of the mix of endowments at
the marketplace as the outcome of a Bernoulli process18 : i.e., consists of N inde-
pendent trials (one for each farmer) wherein there is a fixed probability φ that a
particular outcome—a (0, 1) endowment—occurs.19 In that case, the market out-
comes, as measured by N1 —the number of occurrences of a wheat farmer—follow
a Binomial probability distribution: see (9.1.5). The number of corn farmers in this
market is N2 = N − N1 and the fraction that they make of the market is k = N2 /N.
Because N1 is a stochastic variable that is binomially distributed, it has a known
expected value, E[N1 ]: see (9.1.6). Because N2 is a linear function of N1 , it too is a
stochastic variable and has a binomial distribution with a calculable expected value.
In a market of size N, we expect (on average) that farmers with a wheat endow-
ment and farmers with a corn endowment will each offer in total the amounts of
wheat and corn shown in (9.1.7) in exchange for the other crop. The ratio of these—
the exchange rate of corn in terms of wheat—is shown in (9.1.8).20 However, the
actual exchange rate will differ from the ratio of expected offers because k can (and
often does) differ from φ. If we have N1 wheat farmers and N2 corn farmers in the
market, wheat farmers will offer a total of (1 − ν)N1 units of wheat, corn farmers
will offer a total of νN2 units of corn, and on average an equilibrium exchange rate
of (1 − ν)(1 − k)/(νk) will therefore result.
However, we do not always get this average exchange rate: as when farmers
arrive at the market to discover to their chagrin that k = 0 or k = 1.21 Consider a
18 A Bernoulli trial is a statistical experiment which can result in only one of two possible real-
izations. An experiment consisting of a series of independent Bernoulli trials is called a Bernoulli
process.
19 We would not have a Bernoulli process if each individual could wait until harvest time to see his
endowment and that of his or her neighbors before deciding in which local market to participate.
20 Economists usually say in this case that wheat is numéraire which means that other goods (corn
in this case) are valued in units of wheat.
21 In an earlier footnote, I raised the question of whether a Walrasian outcome existed and was
unique. In the case of a log-linear utility function, the answer intuitively is straightforward. Each
farmer maximizes utility by allocating income so that the proportions spent on wheat and corn
are α and 1 − a, respectively. At the equilibrium exchange rate, (1 − α)(1 − k)/(ak), a Walrasian
solution exist; the market clears and farmers of each endowment are as well off as possible. The
Walrasian solution is also unique; no other exchange rate clears the market.
9.4 Model 9A: Non-spatial Market 243
Table 9.2 Possible realizations in market of size 2 wherein ν = 0.3 and φ = 0.4
N2 N1 k P[k] q1 q2 U1 q1 q2 U2 W[k] Ex
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]
Notes: Calculations by author. See also Table 9.1. In panel (b), mean unit shipping cost is 0.0689.
–, Indicates no one of that type present. See also Table 9.1.
Ex, Exchange rate: units of wheat per unit of corn
Endogenous: U[N], Utility of being in market of size N; V[N] Variance in utility in market of size
N; W, Average utility weighted by number of farmers of each type.
market of just two farmers (N = 2). Remember here that I assume farmers choose
a market in advance of knowing their endowment. There are three possible realiza-
tions for N1 : 0, 1, or 2. When φ = 0.4 and ν = 0.3, the ratio of expected offers
(9.1.8) is 3.50. For each possible outcome of N1 , the corresponding binomial prob-
ability is shown in panel (a) of Table 9.2. In the event, N1 = 0 or N1 = 2 here, the
market consists entirely of corn endowments or wheat endowments, respectively,
and the utility for each farmer in the market is therefore zero. When N = 2, there
is a probability of 0.5222 that the market participants will have a zero utility: i.e.,
return home without any of the other commodity. Suppose instead N1 = 1, which
means we have one farmer of each type in the market. The farmer with a wheat
endowment consumes 30% of his or her endowment of wheat, and trades the remain-
ing 70% for corn. The farmer with a corn endowment consumes 70% and trades
the remaining 30% for wheat. There are two possible utilities. With a probability
(1 − k), the farmer will have an endowment of (1, 0), a consumption bundle
22 0.36 + 0.16.
244 9 The Cautious Farmer and the Local Market
(0.3, 0.3), and a utility (U1 ) of 0.3. With a probability k, the farmer will have an
endowment of (0, 1), a consumption bundle (0.7, 0.7), and a utility (U2 ) of 0.7. In
this situation, the farmer with the corn endowment is better off after trade than the
farmer with the wheat endowment: not surprising given that the corn is the commod-
ity more strongly preferred. Put differently, the farmer is here guaranteed a utility
of 0.3 with a 50% chance of getting 0.7 instead. By calculating W[k] = 0.5,23 I am
simply taking an average of the two possible utilities weighted by the probabilities
of the two outcomes. Considered over all the possible realizations of k, the weighted
average utility of being in a market of N = 2, U[N], from columns [4] and [11] in
panel (a) of Table 9.2, is 0.240.24,25 The variance in this utility, V[N], is also now
calculable; V[N] = 0.082.26
What is the expected exchange rate across the three possible realizations of N1
from 0 to 2? We cannot calculate an exchange rate when k = 0 or k = 1 because no
trade happens. In column [4] of panel (a) in Table 9.2, we see that the probability that
k = 0 or k = 1 is 0.52. From column [12], we see that at the only other possibility,
k = 0.50, the exchange rate is 2.33 units of wheat per unit of corn. I label this the
Conditional Expected Exchange Rate (CEER); that is, the exchange rate we expect
on average on the condition that k is neither 0 nor 1. This is different from the ratio
of offers expected, (9.1.8), which incorporates the amounts offered when k = 0 and
k = 1. In comparison, the ratio of offers expected is 3.50 when ν = 0.3 and φ = 0.4
as noted above.
To begin thinking about what might happen if N were larger than 2, let us do simi-
lar calculations for a market of N = 8 participants. As before, I continue to illustrate
using the case where ν = 0.3 and φ = 0.4. See panel (a) of Table 9.3. Consider first
the case where N1 = 3. The three farmers with a wheat endowment each offer
0.7 units of wheat in exchange for corn. The remaining five farmers have a corn
endowment; each offers 0.3 units of corn in exchange for wheat. The equilibrium
exchange rate here is therefore 1.4027 units of wheat per unit of corn. Each farmer
with a wheat endowment therefore consumes 0.3 units of wheat, and 0.528 units of
corn for a utility of 0.43. Each farmer with a corn endowment consumes 0.7 units of
corn and trades away the remaining 0.3 units in exchange for 0.4229 units of wheat
to achieve a utility of 0.60. Therefore, the weighted average of utilities, W[k], is now
0.536.30 Considered over all the possible realizations of k as shown in Table 9.3, the
weighted average utility of being in a market of N = 8, U[N], from columns [4] and
[11] in panel (a) of Table 9.3, is 0.423 and the variance V[N] is now 0.087.
23 0.5(0.3) + 0.5(0.7).
24 0.52(0.00) + 0.48(0.50).
25 This is another place where ordinalists might well cringe. If utility is indeed ordinal, what does
it mean to take a linear combination of utilities as we do when we calculated U[N].
26 0.52(0.00 − 0.24)2 + 0.48(0.5(0.30 − 0.24)2 + 0.5(0.70 − 0.24)2 ).
27 (3(0.7)/(5(0.3)).
28 0.7/1.40.
29 (0.3)(1.40).
30 (3/8)(0.43) + (5/8)(0.60).
9.4 Model 9A: Non-spatial Market 245
Table 9.3 Possible realizations in market of size 8 wherein ν = 0.3 and φ = 0.4
N2 N1 k P[k] q1 q2 U1 q1 q2 U2 W[k] Ex
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]
Note: Calculations by author. See also Table 9.2. In panel (b), mean unit shipping cost is 0.12.
What is CEER when N = 8? Here we see from column [4] in panel (a) of
Table 9.3 that the probability that k = 0 or k = 1 is 0.018, and from column [12]
that, for k in between, the exchange rate varies from 16.33 down to 0.33 which yields
CEER = 4.84. This is higher than the 3.50 we noted above for the ratio of expected
offers: the opposite of what we had found when N = 2.
In this model, the farmer chooses the size of market in which to trade. This means
the farmer will compare the combination of return, U[N], and risk, V[N], with those
246 9 The Cautious Farmer and the Local Market
achievable at other sizes of market.31 We are now able to compare the case of N = 2
and N = 8. When N = 8, the return is larger (0.423 vs. 0.240) but so too is the risk
(0.087 vs. 0.082) compared to N = 2. In a risk-return analysis, the farmer would
therefore prefer N = 8 over N = 2 if β is sufficiently small and prefer N = 2
otherwise.
To understand what is happening to CEER, suppose we let N vary from 2 to
30 and calculate CEER at each N. The resulting estimates of CEER are plotted in
Fig. 9.2. The dotted line is the ratio of expected offers (9.1.8); it is horizontal because
this ratio is the same at every N.32 The solid curve shows CEER as a function of
N. CEER is below the ratio of expected offers when N = 2, rises quickly, and
peaks well above the ratio of expected offers at about N = 8, then begins to fall off
asymptotically to the ratio of expected offers as N becomes large. We are now ready
to answer some questions.
• Why is the equilibrium exchange rate low at small N? This is because (1) the
exchange rate is a declining function of k, (2) the equilibrium exchange rate
cannot be calculated at k = 0 or k = 1, and (3) P(k = 0) is larger than P(k = 1)
D
A B
Exchange rate
• Why is CEER then above the ratio of expected offers for N sufficiently large
(above N = 4 in Fig. 9.2)? This happens because of the asymmetry of an
exchange rate. To see this, suppose the quantities of the two commodities offered
in the market are identical; the exchange rate here is 1.00. Now consider increas-
ing the quantity of either the numerator or denominator. As we increase the
denominator quantity, the exchange rate can drop from 1 to as low as 0. As we
increase the numerator quantity, the exchange rate rises from 1 without limit. This
asymmetry means that, ignoring the effect of the exclusion of k = 0 and k = 1
offers, CEER should be systematically higher than the ratio of offers expected in
(9.1.8). However, this bias dissipates as the size of market becomes larger.
Of course, CEER is not the key variable here. To understand how farmers
choose markets, we must look at how market size affects utility. In Fig. 9.3, I
use a solid line—the risk-return curve—to connect combinations of U[N] and V[N]
attainable—for each level of N, again from N = 2 to N = 30. The attainable combi-
nation at each N is a black dot on this curve. Further, I have labeled the size of market
at selected dots in Fig. 9.3. Here, we see that, as N is increased, the mean increases
and the variance starts dropping above N = 4. In fact, the locus of points on this
curve from N = 4 through N = 30 suggests that U[N] will continue to increase,
and V[N] will decrease albeit both ever more slowly as N becomes still larger. In a
beta analysis, we assume that the farmer is willing to accept a higher risk associated
with a higher reward. One such tradeoff curve is shown as a dotted line in Fig. 9.3.
We can imagine a family of such dotted lines, all parallel, such that the farmer is
30
8
6
turn tradeoff 5
Risk-re 4
3
Risk: U[N]
happier the higher and to the left the tradeoff curve lies. Given the risk-return curve
is negatively sloped above N = 4, the farmer would prefer to be in as large a market
as possible. This is not surprising. After all, there is no disincentive here to join a
larger market, and the larger the market the higher U[N] and the lower V[N].
In Fig. 9.3, I also include the case of autarky (N = 1) wherein farmers do not
assemble into a market. In autarky under either endowment, the farmer gets a
sub-utility of 0 with certainty: i.e., a zero variance. Autarky corresponds to the
origin—that is, the intersection of the horizontal and vertical axes—in Fig. 9.3 since
U[N] = 0 and V[N] = 0 there.33 If we were to draw two risk-return tradeoff lines,
each parallel to the dotted line in Fig. 9.3, one drawn through the point where N = 2
and the other through the autarky point (the origin), we see that the utility of being
in a market of size 2 exceeds the utility of not being in autarky. A similar analysis
would lead us to conclude that utility is higher at N = 3 and still higher at N = 4.
Put another way, the slope of the risk-return tradeoff drawn in Fig. 9.3 is suffi-
ciently small (i.e., the risk premium, β, was low enough) to initiate an agglomeration
process.
Does the reverse argument hold true? Is there a β sufficiently large that farms
might never switch from isolation (N = 1) to a market where N = 2, or from there
to move to markets of N = 3 or N = 4? Consider the line in Fig. 9.3 joining autarky
to N = 2. If β were sufficiently large to make the risk-return tradeoff line steeper
than this, the utility of autarky would appear to be greater than the utility of N = 2.
Suppose alternatively the risk-return tradeoff line passing through the origin cuts
the line segment joining N = 2 and N = 3. Then the utility of being in a market
of N = 3 would be greater than the utility of autarky. If so, does that imply that
farms would never get to a market size of 3, even though it is advantageous, because
no one would first have the incentive to form a market of 2? If so, this would be
a disturbing feature of the model because it would imply that planners would have
to do what a market could not: i.e., push myopic farmers, unable to see the benefits
from further agglomeration, from autarky into a market of N = 2 to make it possible
for farmers to then form a market of 3 or more.
However, I think the problem here, specifically any reluctance to move from
autarky to N = 2, when β is large, is actually indicative of a limitation of risk-
return analysis. In autarky, the farmer receives a sub-utility of zero with certainty.
When N = 2, the farmer receives a sub-utility of zero if both farmers have the same
endowment, and a larger sub-utility if they do not. In other words, at the worst, the
farmer in an N = 2 market does as well as in autarky. In my view, the farmer in
an N = 2 market is always therefore at least as well off as in autarky and has the
possibility of being better off. Even if we assume that the farmer is myopic about the
prospect of more farmers joining the agglomeration and pushing utility even higher,
there is an incentive here to form a market of N = 2.34
35 See McGuire (1972) on economic models of club formation. For other modeling of cooperation
in a geographic context, see Jayet (1997) and Soubeyran and Weber (2002).
36 Here I implicitly assume contingent shipping rates. That is, the cost of shipping wheat a kilome-
ter is s units of wheat, and the cost of shipping corn a kilometer is s units of corn. Similarly, the
coop fee is contingent; it is f units of wheat if the farmer has a wheat endowment, and f units of
corn if a corn endowment. There is no adjustment here for the exchange rate between wheat and
corn that will be obtained in the market. For the storyteller, the advantage of this scheme is that it
simplifies decision making for the farmer who is still in anticipation of the harvest and does not yet
know his or her endowment.
250 9 The Cautious Farmer and the Local Market
Mid radius
√ for farm at ring i
mi = ((i − 0.5)/(πg))for1 ≤ i ≤ N (9.4.2)
Consumption of wheat and corn and sub-utility of farm with wheat endowment
q11 = ν(1 − f ) q12 = (1 − ν))(1 − f ) / P U1 = (1 − f )ν ν (1 − ν)(1 − ν) / P1 − ν
(9.4.4)
Consumption of wheat and corn and sub-utility of farm with corn endowment
q21 = ν(1 − f )P q22 = (1 − ν)(1 − f ) U2 = (1 − f )ν ν (1 − ν)(1 − ν) Pν
(9.4.5)
Coop’s balanced budget
fN = i smi (9.4.6)
Notes: See also (9.1.1) through (9.1.6) I and Table 9.1. Rationale for localization (see Appendix A):
Z4—Risk spreading and insurance; Z6—Differences among consumers; Z7—Variation in goods;
Z8—Limitation of shipping cost. Givens (parameter or exogenous): g—Density of farms (farms
per square kilometer); N—Number of farmers in market; r—Opportunity cost of land (assumed
zero); s—Unit shipping rate; β—Risk aversion parameter; ν—Exponent of wheat in utility func-
tion. Relative preference for wheat; φ—Probability farmer has (0,1) endowment. Outcomes
(endogenous): f—Co-operative fee; mi —Mid-radius of farm i; P—Market exchange rate; Q1 —
Aggregate quantity of wheat offered in market; q1i —Quantity of wheat consumed by farmer of
type i; Q2 —Aggregate quantity of corn offered in market; q2i —Quantity of corn consumed by
farmer of type i; U—Sub-utility of farmer; V—Variance; xi —Outer boundary of farm i.
Fig. 9.4 Models 9B and 9C: Maps of farms as rings in market of N = 6, N = 4, and N = 2
farm.37 The rationale for making each farm annular is that this shape minimizes the
cost of shipping their endowments to the market. I ignore here other considerations
that might shape the farm in a geographic sense.38 In effect, the market is at the
center of the innermost circle: the center of farm 1. Since each farm occupies the
same amount of land, the shipping cost associated with the marginal farmer (i.e., the
farmer furthest from the market) and the fee (average cost) paid by each farmer in
a coop increases at a decreasing rate with the number of farmers in the market. See
Fig. 9.5. As I have already assumed land is plentiful, I do not need to worry about
the possibility of clubs with overlapping market areas; a club would simply move
to an unoccupied area so that it can achieve the same low total shipping cost as any
other club of the same size.
In this model, farmers may differ from one another in the following ways: (1)
a randomly determined endowment not known in advance of club formation; (2)
a given tradeoff between risk and return; and (3) a location vis-à-vis the market
that the farmer can choose. Otherwise, I assume farmers are identical: same prefer-
ences for wheat and corn and same fee to join a given coop. Therefore, a club will
be formed by farmers with similar tradeoffs between risk and return: i.e., similar
βs. Once in a club, the farmer is indifferent to location because the club pays the
marginal cost of shipping from that site to the market.
Assume each farmer uses a fixed amount of land, 1/g, in agricultural produc-
tion.39 Assume that land is not used for any other purpose (we ignore here any need
for land for transportation, for a market site, for housing, or for the production of
37 Mid-radius here is the distance which divides the farm into two equal areas.
38 For instance, while a ring might be the most efficient shape for getting the agricultural com-
modity to market, it may be inefficient for the daily chores of the farmer throughout the growing
season. Thünen (1966, chaps. 11 and 13) discusses aspects of this problem.
39 This might be because each farmer has a Leontief technology that requires all inputs be in fixed
proportion; however, the model to this stage is silent on other inputs to production.
252 9 The Cautious Farmer and the Local Market
C
A
0
Size of Market (N)
40 Economides and Siow (1988) model the case where farms are spread out along a line in one-
dimensional space.
41 (0.40)(0.0728).
42 (0.40)(0.1262).
43 Since each farmer has an endowment of either (1, 0) or (0, 1), if I assume that each farmer carries
his or her entire endowment to the market to trade (not unreasonable given that the farmer does not
know the exchange rate that might be established), the shipping cost associated with each farmer
is fixed whether measured per unit shipped or per farmer.
9.5 Model 9B: Cooperation in a Spatial Market 253
recovery: i.e., total shipping cost incurred by the coop is split equally among the N
farmers in the coop: see (9.4.6).
A feature of the log-linear utility function is that consumption has a linear expan-
sion path.44 Even though the farmer’s income available for consumption of wheat
and corn is now net of the membership fee, consumption changes proportionally
with income; the marginal farmer still spends the same proportion (ν) of his or her
net income on wheat and the remainder on corn.
Here in Model 9B, I assume farmers cooperate by sharing equally the total ship-
ping costs of all farmers in the market. If market size (N) is just 2 farms, the co-op
fee (f) borne by each farmer would be 0.0398.45 Suppose N1 = 1. One farmer has
an initial endowment (income net of production cost and land rent) of (1, 0). Since
ν = 0.3, this person prefers to consume 30% of his or her endowment net of coop
fee in wheat itself, and trade the remaining 70% away for corn. See panel (b) of
Table 9.2. The other farmer has an initial endowment of (0, 1) of which he or she
prefers to consume 70% (again net of coop fee) and trade away the remaining 30%
for wheat. See (9.4.3), (9.4.4), and (9.4.5). Since each farmer pays a coop fee equal
to the average shipping cost, that exchange ratio between wheat and corn in this
market is 2.3346 units of wheat per unit of corn just as in Model 9A. If the N = 2
market, there are two possible utilities when k = 0.5: a (1, 0) endowment that yields
a consumption bundle (0.29, 0.29) and a utility (U1 ) of 0.29 (compared to 0.30 in
Model 9A) with a probability (k) of 0.5 or a consumption bundle (0.67, 0.67) and
a utility (U2 ) of 0.67 (compared with 0.70 in Model 9A) with a probability (1 − k)
of 0.5. Then, W[k] = 0.48047 (compared to 0.50 in Model 9A). The introduction of
shipping cost here leaves the exchange rate unchanged but reduces the sub-utility in
every outcome, not surprisingly given that the membership fee reduces the amount
of wheat and corn available for consumption. As a result, U[N], from columns [4]
and [11] in panel (b) of Table 9.2 is 0.230,48 down from 0.240 in Model 9A. Further,
the variance in this utility, V[N], is also now calculable; V[N] = 0.07549 down from
0.082 in Model 9A. To conclude, in the case of N = 2, the introduction of shipping
costs reduces both return and risk compared to Model 9A.
Now let us do similar calculations for a market of N = 8 participants: see panel
(b) of Table 9.3. CEER remains the same as in Model 9A. However, compared to
Model 9A in panel (a), we find—as when N = 2—that U1 , U2 , and W[k] drop for
any 0 < k < 1. As a result, U[N] = 0.390 is smaller than for Model 9A. V[N] too
44 As used here, a condition of the utility function wherein, if as income is increased by a fixed
proportion holding prices of commodities constant, the rational consumer purchases the same pro-
portion more of each good. Put differently, each good has an income elasticity of +1.0. Such a
utility function is also said to exhibit homotheticity.
45 (0.0505 + 0.0291)/2.
46 0.70(1−0.0398)/(0.30(1−0.0398)).
47 0.5(0.29) + 0.5(0.67).
48 0.52(0.00) + 0.48(0.480).
49 0.52(0.00 − 0.23)2 + 0.48(0.5(0.29 − 0.23)2 + 0.5(0.67 − 0.23)2 )
254 9 The Cautious Farmer and the Local Market
is smaller than in Model 9A. As in N = 2, return and risk are both smaller once we
take shipping cost into account.
We can then compare this combination of return and risk with those achievable
at other sizes of market. I repeated the same process of calculating U[N] and V[N],
as described above, for all market sizes from N = 2 to N = 30 in the spatial case.
See Fig. 9.6. There, I show the risk-return curve as a faint line joining achievable
combination at each market size in Model 9A (reproduced from Fig. 9.3) and as
a solid line joining dots (labeled Model 9B).50 I have also reproduced the risk-
return tradeoff from Fig. 9.3. For the risk-return curves in Models 9A and 9B, I have
labeled selected market sizes from 2 to 30. Here, we see the effects of introducing
shipping cost.
Model A
10
5
U[N] Expected sub-utility
deoff 1
turn tra 14 10
Fig. 9.6 Model 9B: U[N], Risk-re
Model B
V[N], and size of market in a 10 5
deoff 2
spatial market: α = 0.30, Risk-re
turn tra
5
φ = 0.40, g = 30, and
Model C
s = 0.40.
Notes: Market size, N, shown
as dots for selected N from 2
to 30. Gray line is the
U[N] − V[N] curve when
shipping costs are zero. Solid
curve is the U[N] − V[N]
curve for a marginal
individual taking into account
higher shipping costs to reach
larger market. Horizontal axis
scaled from 0.04 to 0.10;
vertical axis scaled from 0.2 0
to 0.5 V[N] Variance in sub-utility
50 In Model B, the shape of the risk-return curve is sensitive to the unit shipping rate, s.
As s approaches zero, the risk-return curve approaches that for Model A in Fig. 9.6. On the other
hand, as s is made larger, the risk-return curve for Model B is pulled even further back and down
at larger N.
9.5 Model 9B: Cooperation in a Spatial Market 255
In the hypothetical example shown in Fig. 9.6, in the presence of shipping costs,
the marginal farmer with a β associated with the risk-return tradeoff line displayed
would find it best to participate in a cooperative of 14 farmers. In this way, shipping
costs help explain why market size is not unlimited. Model 9A suggests farmers
have an unlimited appetite for participating in large markets. However, shipping
costs in the form of membership fees curbs this appetite. Size of market reflects the
offsetting influences of return and risk on the one hand and average shipping costs
on the other. If farmers have a higher β—that is, a greater aversion to risk—the
risk-return tradeoff line in Fig. 9.6 would be steeper, and farmers would then opt
for a coop with a larger N. See (9.4.7) and (9.4.8). In other words, farmers would
then prefer a lower level of risk even though that might involve a substantially lower
return.
Earlier, I had said that farmers with a similar β would form a club. As is evident
from Fig. 9.6, that is not always the case. If s is sufficiently small, every farmer
would prefer an infinitely large market, and we might therefore find a great mix of βs
among farmers in any club so formed. It is more correct to say that if s is sufficiently
large, the risk-return curve under Model 9B will consist of a set of distinct market
sizes each of which will make that size best for farmers within an interval of β. In
that sense, Model 9B is lumpy. I draw the risk-return curve as a polyline in both
Figs. 9.3 and 9.6; however, this is just for ease of exposition. In fact, the risk =
return curve is just a set of combinations of U[N] and V[N]: one for each integer
size of market; the line segments joining them have no particular meaning. When,
in a beta analysis, we draw a risk-return tradeoff as an upward sloped line, we are
asking simply which combination of U[N] and V[N] on the risk-return curve allows
the farmer to reach the highest risk-return tradeoff. In the example shown in Fig. 9.7
for instance, farmers whose β is above 0.0873 would choose at least N = 12 (since
that is the slope of risk and return joining N = 12 to N = 11); farmers whose β
is below 0.3168 would choose no more than N = 12 (since that is the slope of risk
and return joining N = 12 to N = 13). In Fig. 9.8, I show a step function from
which we can read, for any given β, the appropriate size of market in this example;
for instance, at β = 1.4, the farmer chooses a market of 20 in Model 9B. Others
prefer to tell economic stories without such step functions (lumpiness); they would
like something akin to Model 9B but wherein size of market, N, was a continuous
256 9 The Cautious Farmer and the Local Market
14
15
0
V[N] Variance in sub-utility
variable that could be analyzed more easily using calculus or other methods that rely
on continuity.51 However, I like Model 9B from a pedagogical perspective because
it clarifies just how a farm might decide in practice whether to join a given coop.
What about comparative statics in this model? First note that compared to models
elsewhere in this book Model 9B has relatively few parameters: ν, β, φ, g, and s.
51 For example, Models A and B here are built on binomial probabilities that arise because we
have characterized market formation as a Bernoulli Process. It is well known that the Normal
Distribution (which is continuous) can approximate a Binomial distribution as the number of trials
in the Bernoulli process (in this case, the size of market) becomes sufficiently large.
9.5 Model 9B: Cooperation in a Spatial Market 257
Fig. 9.8 Risk aversion and Models 9B and 9C: Size of market and risk aversion
AB Risk aversion and size of market in Model C
size of market in Models 9B CD Risk aversion and size of market in Model B
and 9C.
Notes: α = 0.30, φ = 0.40, B
g = 30, and s = 0.40. Here, b
and c are the lower and upper
a
b
c
0
A C Size of market (N)
12 14
travel further to benefit from a larger market, and for a sufficiently small s
and thereby f, there will be a single global market.
β If β is increased, farmers become more risk averse. They are willing to spend
more to join a larger coop because they attach more importance to reducing
risk. Provided s is sufficiently large to make the risk-return curve bend back
down enough so that a risk-return tradeoff line can be tangential to it, an
increase in b causes farmers to prefer a larger market.
ν If ν is increased, each farmer prefers more wheat relative to corn. This causes
CEER to fall since farmers now see wheat as more valuable. However, it has
no effect on the efficient size of market, N.
φ If φ is increased, a farmer is more likely to have a corn endowment. Effects
on the model depend on the size of φ relative to 1 − ν. If φ is smaller than
1 − ν, an increase in φ brings the ratio of corn to wheat endowments closer
to what consumers would prefer. If φ is larger than 1 − ν, an increase in φ
pushes the ratio of corn to wheat endowments higher than what consumers
would prefer. The size of market is unaffected.
Model 9B is the first place in this book to make use directly of the idea of a
club. Upon reflection, it might seem strange to introduce the idea of a club—whose
members cooperate—into a book on the effects of competition (often thought to
be the antithesis of cooperation). Why do farmers here seek to cooperate? What
is driving them here is the risk associated with the randomness of the harvest—
i.e., their endowment. To form a club, as is done in Model 9B, is just one possible
response to that risk. What might be some other possibilities?
258 9 The Cautious Farmer and the Local Market
• Crop insurance. If farmers need to consume both corn and wheat, and plant both
crops, it would be reasonable for them to purchase insurance against crop failure.
Model 9B does not include this possibility, so the alternative is to band together
with other farmers hoping that there will be sufficient amounts of both wheat and
corn at the market to enable a good trade. Presumably, crop insurance is struc-
tured to give farmers (every farmer in this case) an amount of the commodity
they fail to harvest. Presumably the insurer collects a premium from each farmer,
awards every farmer an insurance benefit, and still manages to keep something
for himself or herself (e.g., profit and wages for the insurer and his or her work-
ers). Under Model 9B, each farm pays its share of shipping costs (which is like
an insurance premium). Each farm also receives back some amount of the com-
modity that they did not harvest (except when k = 0 or k = 1). In that sense,
the farmers in Model 9B have greater risks than they would with crop insurance
(since they could still arrive at a market where k was near 0 or near 1 but less
risk than they would have in the absence of both a commodity market and crop
insurance).
• Arbitrage. Model 9B assumes there is no way to purchase commodities other
than by traveling to a market. As envisaged here, there would potentially be a
large number of local markets each with its own exchange rate between wheat
and corn. Where exchange rates differ between two local markets by more than
the unit shipping cost between markets, presumably arbitrageurs would have an
incentive to enter. If the farmer knows that there are competitive traders lurking
in every market looking for such opportunities, they will have an incentive to
choose a smaller market and save on the cost of shipping to the extent that they
know traders will act in a way that keeps the local exchange rate from becoming
too unfavorable.
At the outset of this book, I suggested that the organization of the firm was itself
endogenous to locational competition. We have seen illustrations of that idea earlier
in the book, most recently in the outsourcing decision in Chapter 7. If we think of
farmers as firms, and the coop as an extension of the activities of a firm, Model
9B offers a new insight in this matter. With the creation of the coop, each farmer
is agreeing to be bound by conditions and restrictions of that group. In effect, the
farm’s pursuit of its own well-being is now spread across two establishments: one
at the farm site and one at the level of the cooperative (market). In effect, the coop
helps each farmer achieve a better market outcome than they might otherwise have.
In the Marshallian sense, can Model 9B be thought of as a localization economy?
I think the answer to that is yes. In Model 9B, the coop spreads (shipping) cost
among participants and clusters farms. The process is endogenous; farmers partici-
pate in a cluster (or not) not by fiat but on the basis of their own assessment of costs
and benefits. Here although farmers close to the market point pay a greater co-op fee
than they recoup in shipping cost reimbursed, there is no incentive to leave the coop.
Even though they are cross-subsidizing the more remote farmers, they know that the
coop has pushed those farmers into concentric rings that are the most efficient and
that they could not get the efficiencies of a market of this size without the coop.
9.6 Model 9C: Competition for Land in a Spatial Market 259
Let me expand on the idea of a club here. As presented, the coop internalizes
the externality of market formation by sharing shipping cost and ensuring efficient
location. In a more general sense, local government generally can be thought to (1)
address local externalities through infrastructure investment, service provision, and
activity regulation and to (2) redistribute costs among municipal revenue sources
such as property tax, local sales tax, and other sources in addition to user fees. In this
sense, a club can be thought of as a metaphor for local government. Put differently,
coops serve to separate farmers by risk category; local governments can perform a
similar function.
Table 9.5 Model 9C: competitive farmers and the market in a spatial economy where r is
endogenous
Notes: See also (9.1.1) through (9.1.6) I and Table 9.1. Rationale for localization (see Appendix A):
Z4—Risk spreading and insurance; Z6—Differences among consumers; Z7—Variation in goods;
Z8—Limitations of shipping cost. Givens (parameter or exogenous): g—Density of farms; i—
Position of farm: rings away from the market; N—Number of farmers in market. Outcomes
(endogenous): mi —Mid-radius of farm i; r—Market rent; xi —Outer boundary of farm i.
260 9 The Cautious Farmer and the Local Market
established in any local market. In Model 9C, I retain the assumption that farms each
form a concentric ring around the market point. After all, shipping cost now gives
the farm an incentive to want to be near the market point. Between any pair of farm
sites (rings), competitive bidding for land means the difference in rent in equilibrium
must be just enough to make the farmer indifferent between the two sites. Rent for
the farmer at the boundary of the market area is zero; see (9.5.4). Further, since I
assume that the amount of land used by a farmer is fixed, the equilibrium difference
in rent (per harvest) between the two sites must exactly offset any savings in the
cost of shipping. That rent plus shipping cost must therefore total to a constant is
the so-called Wingo condition.52 Rent for the farmer in ring i is given by (9.5.5).
Once scarcity rents reach this level, there is no incentive for a farmer to prefer any
one ring to another within a given market. However, there will still be differences
among markets of different sizes. The implication of (9.5.5) is that the larger the N,
the higher the rent in ring 1.
Given a market of size N, there are important differences between Models 9C and
9B. In Model
√ 9C, the farmer at ring N loses from his or her endowment a shipping
cost of s ((N − 0.5)/(πg)), zero rent, and (needless to say here in Model 9C) there
is no coop fee. In Model 9B, the √same farmer incurs zero shipping cost, zero scarcity
rent, and a coop fee of (1/N)i s ((i − 0.5)/(π g)). At ring N, the loss to endowment
is smaller in Model 9B than in Model 9C. At any location closer to the market, the
loss to endowment for the farmer in Model 9B stays the same and so too does the
loss to endowment for the farmer in Model 9C since rent there increases to offset
any savings in shipping cost. Therefore, the farmer in Model 9B is better off than
the farmer in Model 9C by the same amount regardless of location in a market of a
given size. Why is this? It is because absentee landlords are collecting scarcity rents
in Model 9C that do not appear in Model 9B.
To begin an interpretation of Model 9C, consider panel (c) of Table 9.2 wherein
N = 2. Suppose N1 = 1. The wheat farmer has an initial endowment (income) of
(1, 0). Continuing the assumption that ν = 0.3, this person prefers to consume 30%
of their endowment (after rent and shipping cost) of wheat and trade the remaining
70% away for corn. The corn farmer has an initial endowment of (0, 1) of which
he or she prefers to consume 70% (again after rent and shipping cost) and trade
away the remaining 30% for wheat. Since each farmer, regardless of location, pays
the same total of rent and shipping cost, that exchange ratio between wheat and
corn in this market is still 2.33 units of wheat per unit of corn just as in Models
9A and 9B. There are two possible utilities when k = 0.5: a (1, 0) endowment that
yields a consumption bundle (0.28, 0.28) and a utility (U1 ) of 0.28 (compared to
0.29 and 0.30 in Models 9B and 9A respectively) with a probability (k) of 0.5 or
a consumption bundle (0.66, 0.66) and a utility (U2 ) of 0.66 (compared with 0.67
52 Wingo (1961) originated the idea that land rent offsets transportation cost savings. Later, Alonso
(1964) argued that the relationship between land rent and transportation cost savings was also
affected by the elasticity of substitution between land and other commodities. Since I have here
assumed that the amount of land used by each farmer is fixed, I do not have to take elasticity of
substitution into account.
9.6 Model 9C: Competition for Land in a Spatial Market 261
53 0.5(0.27)+ 0.5(0.66).
54 0.52(0.00)+ 0.48(0.475).
55 0.52(0.00 − 0.228)2 + 0.48(0.5(0.28 − 0.228)2 + 0.5(0.66 − 0.228)2 ).
262 9 The Cautious Farmer and the Local Market
therefore chooses a smaller size of market at any given level of risk aversion. In
effect, the mechanism of land rent serves to separate farmers by risk category just
as do coops.
It is typically argued that competitive markets are efficient. Is that always the case
here? The answer—perhaps surprisingly—is no. Specifically, a farm with a β corre-
sponding to the risk-return tradeoff shown in Fig. 9.6 would be better off in Model
9B than in Model 9C. Put differently, the outcomes in the cooperative scheme in
Model 9B appear to be more efficient than in the competitive scheme in Model 9C.
Why is this? Two related answers come to mind. First, Model 9B may only appear
to be more efficient because it ignores the costs of organizing cooperatives. If such
costs were substantial, they could drag the risk-return curve for Model 9B below that
for Model 9C, making the competitive solution the more efficient. Second, under the
cooperative scheme, farms organize themselves into a larger unit to reap the advan-
tages brought about by the reduction of risk. In this sense, the Coasian view would
be that the firm is merely internalizing something (here, a reduced risk) that might
otherwise be unavailable or costly to provide in a competitive market. Put differ-
ently, belonging to a collective is integral to firm organization: allowing the firm
to be efficient. A similar idea arose in the modeling of repair services within or
outside the firm back in Chapter 7. There too the firm sought to deal with risk (in
that case, machine failure) efficiently through firm organization: be it in-house or
through outsourcing.
Models 9B and 9C say something new about why individuals participate in a
local market. In earlier chapters, I presented the notion of range of good wherein
size of market was limited only by the notion that a commodity was expend-
able. In Chapter 8, we saw a market boundary defined by the criterion that the
presence of a competitor nearby limited the ability of a firm to attract customers.
In this chapter, the farmer can be thought to make a choice between a smaller
market and a larger market in terms of the of risk and return associated with
each.
This is the first model in the book wherein land rents vary by location within the
regional economy. We can think here of a farm at the edge of its market area (i.e.,
the Nth distant farmer in a market of N farms) as a marginal farm. Compared to the
marginal farm, the endowments of farms that are closer to the market (i.e., farms
1 through N − 1) now can be seen to include an excess utility that arises because
of a savings in shipping cost. In Model 9C, the excess utility gets bid away as land
rents so that the first N − 1 farmers end up only as well off as the marginal farmer in
their market. Put differently, these land rents do not arise because of something that
landlords have done to make their properties individually more attractive to tenants.
Instead, they arise because of the clustering of N farms. To farms, they are a loss in
consumption (a leakage to the farm economy) that happens because farms compete
in the land market. The model is silent on how farms produce their endowments.
However, we could readily imagine the farm as an enterprise that uses capital, labor,
and land to produce its wheat or corn. If so, Model 9C hints at a fundamental under-
lying relationship among factor payments to these inputs. In Model 9C, the gain
in utility from participating in a larger market is divvied up between (1) implicit
9.6 Model 9C: Competition for Land in a Spatial Market 263
returns to labor and capital in the farm enterprise and (2) the return to land. I present
models in later chapters of this book that further explore this idea.
What about the comparative statics of Model 9C? Once we take into account
that the coop fee in Model 9B is now replaced by shipping cost and land rent, the
comparative statics are much the same as in Model 9B. Why is that? In part, it is
because Models 9B and 9C have the same parameters: ν, β, φ, g, and s. In Model
9B, the farmer pays a fixed cost (the coop fee) regardless of location that increases
with the size of market. In Model 9C, the farmer also pays a fixed total cost (for
shipping plus land rent) that increases with the size of the market.
g In Model 9C, as g is decreased, the density of farms declines and the marginal
farmer has to travel further to participate in a market of size N, and thereby
shipping cost is increased. Put differently, the loss associated with shipping
and/or rent increases when g is decreased.
s If s is near zero, farmers in Model 9C form a single global market because
the risk-return curve bends up. If s is large enough to cause the risk-return
curve to have a positive slope (bend back down) for sufficiently large N, we
observe farmers choosing to form a local (smaller) club. Put differently, if
we decrease s, the farm has an incentive to travel further to benefit from a
larger market, and for a sufficiently small s, and there will be a single global
market.
β If β is increased, farmers become more risk averse. They are willing to spend
more to join a larger coop because they attach more importance to reducing
risk. Provided s is sufficiently large to make the risk-return curve bend back
down enough so that a risk-return tradeoff line can be tangential to it, an
increase in β causes farmers to prefer a larger market.
ν If ν is increased, each farmer prefers more wheat relative to corn. This causes
CEER to fall since farmers now see wheat as more valuable. However, it has
no effect on the efficient size of market, N.
φ If φ is increased, a farmer is more likely to have a corn endowment. Effects
on the model depend on the size of φ relative to 1 − ν. If φ is smaller
than 1 − ν, an increase in φ brings the ratio of corn to wheat endowments
closer to what consumers would prefer. If φ is larger than 1 − ν, an increase
in φ pushes the ratio of corn to wheat endowments still higher than what
consumers would prefer. The size of market is unaffected.
Before leaving this section, let me add one thought about landlords. As envisaged
here, there will be a mix of markets across the landscape; some small and some
large. Corresponding to these will be a mix of landlords; some will receive a high
scarcity rent for their site because it happens to be near the center of a large market;
others will receive only a rent equal to the opportunity cost of the land: i.e., a zero
scarcity rent. If landlords themselves were competitive, why would a landlord with
a zero scarcity rent not seek to entice farmers to relocate more advantageously? In
the absence of collusion and given enough landlords (who each own only a small
amount of land), there is no mechanism by which this could be achieved. At the
264 9 The Cautious Farmer and the Local Market
same time, the potential for a higher land rent might in practice create an incentive
for landlords to collude locally to ensure that they attract and hold the scarcity rents
associated with a larger market.
In this chapter, the principal model has been 9C. I included Models 9A and 9B to
help readers better understand aspects of Model 9C. In Table 9.6, I summarize the
assumptions that underlie Model 9A through 9C. Many assumptions are common
to all these models: see the list in panel (a) of Table 9.6. The models differ in that
(i) Model 9A assumes shipping costs are zero, (ii) Model 9B assumes that farms
address differences in shipping costs with location by forming cooperatives, and
(iii) Model 9C assumes that farms address differences in shipping costs by bidding
up the price (rent) for land at advantageous locations.
Assumptions 9A 9B 9C
[1] [2] [3]
a size of market or location). At the same time, there is nothing to guarantee that the
farmer will always get a better price in a larger market. In these respects, choosing
a larger market is like searching. The farmer in a larger market is getting to see a
wider cross section of market participants. This is like (but not the same as) a con-
sumer gathering information from different suppliers before deciding whether and
from whom to purchase. The reader might therefore be tempted to apply this model
to help in understanding search behavior. However, there are important differences
between market formation as modeled here and conventional thought about search
behavior. Principal among these in my mind is the idea that search is often seen as
a sequential process. In my view, the consumer gathers information about another
supplier or commodity, then makes a decision about whether to continue searching,
to purchase, or give up on the idea of purchasing such a commodity. In contrast, the
farmer in Models 9B and 9C is simply making a gamble among sizes of market;
there is no sequential process at work here.
Second, the farmer in this chapter is always assumed to use the same amount
of land in production. However, in Model 9C, land becomes relatively more costly
to rent, the closer the farm is to the market. Presumably, when an input like land
becomes more costly, we might expect the farmer to change the way in which he
or she produces agricultural commodities. One simple way to do this is to switch
between production that is less labor intensive (when land rents are low) and pro-
duction that is more labor intensive (when land rents are high). To do this, our model
of the farm would have to include a production function that enables a substitution
between land and labor. The models in this chapter assume a fixed amount of land
per farm and are silent on labor input. From my perspective, the easiest way to think
about the farmer in this chapter is that he or she constitutes one unit of labor and
that the farm has a Leontief production technology that requires exactly 1 unit of
labor and 1/g units of land to produce stochastically an endowment of either (1, 0)
or (0, 1). What happens if we make amount of land used by a farm endogenous to
the model? I return to this question over the next three chapters.
Third, we began Model 9A with the apparently innocent assumption that each
farm had an independently drawn random endowment of 1 unit of either wheat or
corn. Implicit in this assumption is the idea that the scale of a farm is somehow
fixed. Imagine how much different the model might look if we started with the
assumption that the farm had an independently drawn random endowment of 1 unit
of either wheat or corn for each unit of land farmed and was free to hire labor and
rent land as needed. In that case, a farm could largely (if not entirely) eliminate the
need to travel to a market by being of sufficient size to render insignificant the risk
of being without suitable quantities of the two goods. In effect, the farm uses its
own size to self-insure rather than rely on what is essentially an insurance function
performed by the market. The gain to the farmer here is the higher level of utility
now possible in the absence of scarcity rents.
Fourth, I am able to solve the models in this chapter because I took a Walrasian
perspective on exchange rates: i.e., I found the exchange rate that leaves neither
kind of farmer feeling that there is still some amount of one commodity that they
would prefer to trade for the other. However, as noted at the outset of the chapter,
266 9 The Cautious Farmer and the Local Market
the Walrasian solution is just one point on a Marshallian Contract Curve. At one
level, one might argue that this indeterminacy of outcome makes the Marshallian
approach less useful than Walrasian approach. One might also argue that, as the
number of market participants becomes large, the Marshallian solution converges
on the Walrasian solution. Nonetheless, it seems to me unreasonable to believe that
farmers will entirely ignore the prospect that in a small market their bargaining
power (however derived) may leave them at an effective exchange rate different
from that envisaged by Walras. The models in the chapter do not consider this.
Fifth, what about Walrasian equilibrium across markets? In Model 9A, there is
only one market: a global market for the exchange of wheat for corn. Walrasian
equilibrium is irrelevant here. In Model 9B, unit shipping costs drive farms to orga-
nize themselves into local markets for the exchange of wheat for corn. Because
the exchange rate realized in each local market is subject to random variation, we
can say only that this particular process of choosing local markets leaves individ-
ual farms best off viewed over a longer term of repeated trials. The same is true of
Model 9C with a twist; in 9C, there is a market for land as well as local markets for
the exchange of wheat for soap.
Sixth, Chapter 9 gives us some hints about the regional economy. However, much
remains to be done to flesh this out. We need to know more about the process by
which endowments of wheat and corn are generated. It would be helpful for example
to know about the use of labor, capital, and land inputs in the production process.
That would make it possible for us to better understand the regional economy and
the impacts of changes in givens on regional well-being.
Chapter 10
Farming for Cash
Market Participation and Demand
(Thünen–Lee–Averous Problem)
An isolated place (the city) has aggregate demand schedules for up to two crops.
Tenant farmers—each renting the same amount of land—supply the city: absent
the uncertainties central to Chapter 9. This chapter considers landscapes that are
either an unbounded ribbon or rectangular plane. A two-dimensional plane means
that the local supply curve is a quadratic and the model becomes more cumbersome
to solve without adding much in economic insight. Farms incur costs of production
and shipping. This generates an endogenous supply curve, an equilibrium price and
quantity in the market, and an outer radius for farm production. At the outer radius,
the cost of production plus shipment is just equal to the price that the farmer receives
for a unit of product. In a competitive market, for land these profits accrue to owners
of the property; tenant farmers everywhere, in competing for the best sites, bid up the
market rent for land until excess profits everywhere are driven to zero. Model 10A
describes a regional economy wherein farms, spread along a line, produce one crop
for the market. Model 10B considers the same problem except that farms locate in
two-dimensional space. In Model 10C, I introduce demand for a second crop when
farms are spread along a line. Even though Model 10C assumes a zero cross-price
elasticity between the two crops, prices of the two are linked because both kinds of
farmers compete for land around the demand place. Model 10D considers the two-
crop case when farms are in two-dimensional space. For the first time in this book,
a model solves jointly for prices of commodities in different markets (the markets
for crops and the market for land).
As will be seen in this chapter, the land market is central to contributions that
location theory makes to Walrasian analysis. This should not be surprising. After
all, if firms compete by choosing locations for their activities, it should affect the
market rent for land and—since this can be an important component of cost overall
for businesses—it should therefore also affect the prices at which they sell their
products. The remainder of this book is devoted to that idea. In this, the second of
four chapters that look at models of agricultural markets and a regional economy, I
consider how shipping cost shapes the supply curve for an agricultural commodity
and how the freight savings from being closer to a market result in competition for
land which in turn impacts the prices of commodities.
How land rents get determined in a market economy is a question that has long
interested economists. In his definitive history of economic analysis, Schumpeter
(1954, pp. 209–223) discusses early contributions to ideas of locational rent by
William Petty (born 1625), Richard Cantillon (born around 1680), and Adam Smith
(born 1723). However, I think many economists would agree that it is David Ricardo
(born 1772) who is the first important figure here. Ricardo saw his Principles of
Political Economy and Taxation first published in 1817.1 This work predated the
development of neoclassical economics upon which modern location theory is built.
Nonetheless, Ricardo presents a novel and relevant argument in his book about how
income produced in an agricultural society gets distributed (attributed to factors of
production). To Ricardo, the income generated by economic production is divided
among three classes in the regional economy: rent to owners of land, profit to own-
ers of capital, and wages to the laborers by whose industry it is cultivated. Ricardo
argued that, in different societies, the proportions allotted to each of these classes
would differ, depending mainly on the fertility of the soil, the accumulation of cap-
ital, and the skill, ingenuity, and instruments of its workforce. In Ricardo’s view,
to determine the laws which regulate this distribution is the principal problem in
Political Economy.
Some economists argue that Ricardo’s emphasis is misplaced; Economics, they
say, focuses on questions of efficiency rather than on questions of who gets what. In
my view, Ricardo was trying to show how market competition (put differently, an
efficient market) leads to a particular distribution of income. He builds an argument
about the distribution of income on the notion that the excess profits that might
otherwise be earned by firms get distributed as factor payments in a competitive
economy. His argument assumes that:
• Firms (be they farms or other economic enterprises) are numerous;
• Firms require factors (i.e., labor, capital, and land) inputs to produce their outputs;
• The quality of factors varies exogenously (e.g., some parcels of land are more
fertile than others);
• The amount of each factor is bounded (i.e., not inexhaustible) at any given level of
quality (Ricardo treated the amount of each factor available as exogenous; he did
1 Second and third editions followed in 1819 and 1821, respectively. Ricardo died shortly thereafter
(1823).
10.1 The Thünen–Lee–Averous Problem 269
not discuss how the owners of a factor of production might increase or decrease
their holdings of a factor) in response to conditions in the factor market;
• Firms compete in factor markets for labor, capital, and land for the purpose of
production;
• Factor owners, being numerous, are price takers too: they have no ability to
extract factor payments over and above those determined in a competitive
market.
Ricardo (1821, pp. 33–45) argues that, in what later would come to be seen as
part of a Walrasian equilibrium, firms bid up factor prices (payments) until all excess
profit disappears (in this sense, factor payments reflect the scarcity of input of that
quality). Put differently, the owners of the factors of production—be they workers,
landlords, or capitalists—have the potential to earn scarcity rents; now known as
Ricardian rents. Ricardo implicitly recognized the importance of location here,2 but
presented neither a mathematical nor a graphical model to illustrate or accompany
his discussion of rent and its impact on factor payments. Indeed, Ricardo’s focus was
on the differences in fertility of soil from one site to the next. In his mind, the owner
of a plot of land more fertile than the marginal plot would earn a corresponding rent.
Model 9C—wherein farmers compete for farm sites—could also be thought of
as similar to the process envisaged by Ricardo. Farm sites are not all the same.
However, this is not because of differences in fertility. Rather, it is because some
sites closer to the market have a lower shipping cost than do sites further away. In
Model 9C, we assumed that farmers competed for sites closer to the market and,
in the process, bid up the rent on land there until the advantage (the excess profit
earnable by a farmer locating near the market) was reduced to zero. No farmer is
better off than any other farmer in their market, and some landlords benefit to the
extent that shipping costs make their sites more attractive.
Model 9C follows more closely the ideas of Thünen than Ricardo. From his farm,
Tellow, in Mecklenburg (Germany), Johann Heinrich von Thünen (born 1783)—
I refer to him as Thünen, but he is also popularly labeled “Von Thünen” or “von
Thünen”—advanced Ricardo’s thinking by marrying scientific experimentation in
agricultural practice with economic theory and mathematical modeling. At the time,
the leading scholar on agricultural practice in Germany, Albrecht Daniel Thaer
(born 1752), was promoting the adoption of the then-modern English agricultural
practices. Thünen took Ricardo’s conceptualization of rent, integrated it into the
neoclassical approach that he helped devise and came up with a locational model
to show the working and significance of rent. In so doing, he challenged some of
Thaer’s key conclusions.3 This chapter, as well as the next two chapters, is based on
analyses that originate with Thünen.
2 Ricardo (1821, p. 35) argues that no charge could be made for the use of land, if land every-
where had the same properties and were unlimited in quantity unless there were differences in the
situation of that land.
3 See Murata (1959, pp. 38–56)
270 10 Farming for Cash
To set the chapters up, let me explain here briefly his principal ideas. Clark (1967,
pp. 370–371) argues that (1) Thünen’s complaint against Ricardo is that Ricardo
developed his theory of rent in terms of an undifferentiated agricultural product,
and (2) Thünen’s great contribution was to see transport costs as the cause, and
rents the consequence, of important differentiations of agricultural, dairy, and forest
production. Note here that Clark is not asking how shipping services come to be
priced. Like Thünen, he treats transport cost simply as a resource use. Thünen’s
model of the city and its rural hinterland makes the following assumptions:
• A large city early in the 19th century—neither growing nor declining and viewed
simply as a point in space4 —demands agricultural commodities and exchanges
city-made commodities (including manure) for them;
• The city is surrounded by limitless land of uniform quality (fertility) in which
agricultural commodities can be produced;
• No other use for land;
• A competitive market for land with large numbers of landlords and tenant
farmers, each a price taker;
• A given unit shipping rate for each commodity (higher for perishable and heavy
commodities, lower for nonperishable and lighter commodities) regardless of
location;
• Other inputs to agricultural production, notably labor, are ubiquitous and build-
ings, timber, fences, and other capital assets on the farm can be thought to be
rented at a going rate of interest;
• Farms experience diminishing returns to labor as they attempt to use farm land
more intensively.
Thünen analyzed price determination for agricultural commodities, land rent, the
efficiency of various farming systems, labor costs, manure production in city and
countryside and its use in farming, rings of production, and the impact of taxes.
What Thünen brought to bear on these subjects was a development of marginal
analysis (marginal productivity of labor and capital) that helped usher in neoclassi-
cal economics. He also emphasized the role of competition; for production of any
one commodity to take place, the land rent bid for land by such farmers some-
where must exceed the rent bid by farmers producing other commodities. In the
case of each commodity, he argues that the price must be sufficiently high to ensure
adequate production. In Part I of The Isolated State, Thünen is interested in the
most profitable use of land by farms. Organized by the relative cost of shipping, he
envisaged six concentric rings around the city. From inner to outer, these were as
follows:
4 In contrast, in a city that is growing, landlords and farmers just outside the current built-up area
anticipate that land will become more valuable for urban uses in the near future and therefore
reasonably make different decisions about investment in land improvements (e.g., soil drainage,
fencing, barns, and other improvement to agricultural productivity) compared to landlords and
farmers further away.
10.1 The Thünen–Lee–Averous Problem 271
Ring Description
1 Cash cropping: delicate food (fruit and vegetable) production closest to the
city; milk further away (largely from stall-fed cows), hay and straw furthest
away5 : town manure is available and used only in this ring6 ;
2 Forestry for timber and fuel needs
3 Crop alternation system: regular alternation between grain crop and non-
grain crop over time.
4 Improved system (Koppelwirtschaft): cereal crops and a short grass ley7 on
enclosed fields.
5 Three-field system (spring grain, winter grain, fallow)
6 Stock farming
It is noteworthy here that Thünen constructed his model without any specific
reference to the demand for agricultural products in the city.8
In Part II (Section I) of The Isolated State—see Dempsey (1960, pp. 185–367)—
Thünen, like Ricardo, was interested fundamentally in distribution (the allocation
of society’s product between wages and profit including interest and rent) in an
agrarian society.9 What determines how much income accrues to each factor of
production? Thünen added two new insights into this question. He assumed away
the part of the problem dealing with the return to capital by assuming a competitive
market in which the price of (return on) capital is exogenous. In terms of the return
to labor, Thünen assumed diminishing marginal productivity. In terms of the return
to land, Thünen assumed that unit shipping rates limited the amount of agricultural
commodity that could be produced as this would require more distant farm sites to
add to production.
More generally, the question of distribution—who gets what out of society’s
production—had long been a concern of classical economics. A focus on what
Thünen referred to as the natural wage (w), in work first published in 1850, was a
part of this. Thünen imagined here a farmer—newly arrived at the frontier—whose
activity generates annually product in the amount p. Of this, the worker is paid w,
and in turn spends a on consumption and invests the remainder as capital in the
farm. To maximize workers’ income from investment, Thünen finds that the natural
5 Thünen refers to this as “free” cropping in the sense that no crop is cultivated simply to nurse the
soil by means of a bare fallow.
6 Of Ring 1, Clark (1967, p. 371) argues that, in the time of Thünen, much agricultural area had
to be set aside to produce feed for horse. At the time, every town produced much horse manure
(not to mention the cow manure from urban dairies) that had to be disposed of without incurring
excessive shipping costs. This necessitated a small belt of market garden and milk-producing land
in the immediate neighborhood of each town.
7 An arable field used temporarily as a pasture for grazing.
8 See Dickinson (1969, pp. 895–896).
9 See also Leigh (1946).
272 10 Farming for Cash
of land (say, one hectare) is the highest. Garrison and Marble (1957) is among the
first to provide a graphical integration of intensity, land rent, and shipping cost.
Thünen’s work has also been extended to look at land uses within urban regions;
Ullman (1941) is an early example. Indeed, I think of farm here simply as any firm
that requires land as an input into its production process.
In this chapter, I start the process of recasting Thünen’s analysis from a modern
perspective. I build on the idea, first presented in Chapter 2, of an isolated market
that can be thought to be a place. In that chapter, we began with a single monopolist
located adjacent to the market. The place was where demand occurs. The customer
point was thought to contain N consumers, each with the same linear inverse demand
curve for a crop that I will call wheat. See (10.1.1) in Table 10.1, wherein I summa-
rize equations, assumptions, notation, and rationale for localization in Model 10A.
Therefore, the aggregate inverse demand curve for wheat is given by (10.1.2). In
this chapter, let us retain the same assumptions about a customer point.
So far in the book, we have assumed similarly an exogenous local supply curve.
In Chapter 2 for example, we assumed that the firm had a variable cost and a fixed
cost of production. Since the firm was a monopolist, it set its price where marginal
revenue equals marginal cost of production. In effect here, the supply curve in this
local market is the firm’s marginal cost curve even though the industry is not in a
competitive equilibrium: i.e., not producing where demand equals supply. However,
we might well want to understand how potential suppliers make decisions about
when to supply wheat to this market. We saw elements of this in Chapters 2, 4,
and 5 wherein producers elsewhere, or agents engaging in arbitrage, contribute to
supply at a place by shipping product there when price makes it attractive.
Is it possible to build a simple model to show when and how much producers
spread across geographic space will supply to a given market? The models presented
in this chapter do just that. They are inspired by a simple and insightful model in Lee
and Averous (1973). However, I begin with an even simpler variant to illustrate the
key mechanisms. The models presented in this chapter have assumptions in com-
mon. They each assume a common fiat money economy for all participants. They
each assume consumers at a place demand wheat. The models assume the suppliers
of wheat are farmers nearby and that farmers are all identical. The models assume
the farm produces wheat at a fixed rate (yield) per unit land: to produce more wheat,
the farm needs proportionally more land. In later chapters of this book, I introduce
the possibility of varying labor intensity and thus yield; I assume a fixed yield in
this chapter for simplicity of exposition. Since yield is fixed, we can measure the
price of wheat, P, as the amount paid in dollars for the amount of wheat that can be
produced on one unit of land. Assume each farm incurs no fixed cost of production
(i.e., in the notation of earlier chapters, this means f = 0) and has a marginal cost of
production cost, C, per unit of land farmed. As before, assume C includes normal
274 10 Farming for Cash
Profit for farmer at distance x from customer point, also rent in competitive
market for land
R[x] = P − C − sx (10.1.3)
Radius of market
X = (P − C) / s if P ≥ C, 0 otherwise (10.1.4)
Aggregate supply of wheat
Q = 2X = 2(P − C) / s if P ≥ C, 0 otherwise (10.1.5)
Equilibrium price
P = (2β/ (2β + Ns))C + (Ns / (2β + Ns))α (10.1.6)
Equilibrium quantity
Q = 2 N(α − C) / (2β + Ns) (10.1.7)
All farms’ revenue from wheat net of production costs
2(P − C)2 / s (10.1.8)
All farms’ payments for shipping
(P − C)2 / s (10.1.9)
All farms’ payment for rent
(P − C)2 s (10.1.10)
Producer surplus: Farms’ revenue net of production cost, shipping, and rent (PS)
0 (10.1.11)
Consumer benefit
2 N(α − C)(βC + αβ + αNs))/(2β + Ns))2 (10.1.12)
Producer cost (including shipping)
N(4αβC − 4βC2 − C2 Ns + α 2 Ns) / (2β + Ns)2 (10.1.13)
Consumer surplus (CS)
2
2βN(α − C)2 / (2β + Ns) (10.1.14)
Social welfare (SW)
N(α − C)2 / (2β + Ns) (10.1.15)
Total market expenditure (PQ)
2 N(2βC + Nsα)(α − C) / (2β + Ns)2 (10.1.16)
Notes: A unit of the commodity is the amount yielded, assumed constant, by a unit of land.
Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some locales;
Z8—Limitation of shipping cost. Givens (parameter or exogenous): C—Production cost per
unit of land; s—Unit shipping rate ($/kilometer) for wheat; N—Number of consumers; x—
Distance (kilometers); α—Intercept of individual linear inverse demand curve: maximum price;
β—Negative of slope of individual linear inverse demand curve: marginal effect of quantity on
price received. Outcomes (endogenous): CS—Consumer surplus; P—Price of wheat (at market);
PS—Producer surplus; Q—Quantity of wheat (square kilometers of yield); R[x]—Land rent bid
by tenant farmers that leaves them with zero (excess) profit; SW—Social welfare; X—Radius of
area supplying wheat (kilometers).
10.2 Model 10A: Farms Producing Wheat Along a Line 275
profit. The models assume further that the farmer bears the cost of shipping wheat to
the market and that the shipping cost per unit shipped per kilometer, s, is the same
for every farmer. I treat this shipping cost simply as a loss to the local economy: i.e.,
as resources used up (e.g., maintenance and fuel costs for the vehicle used to ship)
and not as a source of factor payments (e.g., profit for shipping firms or wages paid
to its workers).
There is no uncertainty here. Put differently, the farmer knows exactly what
amount of crop will be harvested and what price will be received.13 The idea,
as in Chapter 9, that somehow there is a randomness associated with climate or
environment is ignored.14
Assume a one-dimensional line (say, a road) of indefinite length down the center
of a ribbon of land 1 km wide. The customer point lies on this road. All land in
the ribbon is available to farmers. There is no alternative use for the land. Farmers
incur no shipping cost in getting wheat from the field to the road itself; they incur a
unit shipping rate only in getting wheat from there down the road to the customer
point. Consider a farmer at a distance of x kilometers from the customer point.15
The farmer here earns an excess profit per unit land given by (10.1.3). By the def-
inition of normal profit, the existence of excess profit means the farmer is earning
more on his or her equity, skills, and so on, than would be the case in the next best
alternative use. So, presumably, the farmer would want to engage in wheat produc-
tion. The boundary of wheat production would then be the distance, X, at which
P − C − sX = 0 which implies (10.1.4). As does Thünen, I assume here a frontier
area sufficiently far away from the market wherein land is waste (i.e., unused) and
therefore earns a rent of zero. In effect the farmed area surrounding the market is
itself surrounded by wilderness in which there is no economic activity.
The supply of wheat within radius X is simply 2X, taking into account land to
either side of the customer point. Therefore, we get (10.1.5). Aggregate demand,
derived from (10.1.2), can then be computed, and demand and supply equated.
Upon simplification, this gives the equilibrium price in (10.1.6). Note here P is once
again a weighted average of the “minimum price” (C) and the maximum price (α).
As either s or N becomes large or β small, P approaches α. On the other hand,
as β becomes large (i.e., price sensitive to demand) P approaches C. All other
endogenous variables now can be solved by back substitution. For example, substi-
tuting this price back into (10.1.5) gives an expression for the equilibrium quantity
transacted: (10.1.7).
What about the well-being of producers and consumers in this market for wheat?
The equilibrium quantity and price can be used to solve for farmers’ revenue
net of production cost (i.e., (P − C)Q): see 10.1.8). Half of this net revenue is
13 Thünen models have been extended to look at cases where the farm faces uncertainty: see Okabe
and Kume (1983).
14 See Cromley (1982) for a discussion of the role of uncertainty in the Thünen model.
15 I measure distance here to be the length of the trip to ship the commodity to the market. I usually
characterize this as the mid-distance in the sense that half of the farm’s parcel of land is closer to
the customer point than this, and half is further away.
276 10 Farming for Cash
expended on shipping: see (10.1.9). The other half is expended on land rent: see
(10.1.10). Here land rent is not a cost (i.e., a use of resources) but rather a real-
location of income from farmer to landlord. Because of competition, the farmers
have no revenue left over after production costs, shipping, and land rents have
been deducted. Consumer benefit, (1/2) (P + α)Q, can also be determined by back
substitution: see (10.1.12). Producer cost is given by (10.1.13). Consumer surplus,
(1/2)(α − P)Q, reduces to (10.1.14), and social welfare is given by (10.1.5).
A diagrammatic solution to market equilibrium is shown in Fig. 10.1. Shown
there are aggregate demand curves Av Bv (where v = α); each of these is the same
as the aggregate demand curve first introduced in Chapter 2. For the purposes of
exposition, I have emphasized the demand curve where α = 4.5. Also shown is
the supply curve (CD), which, while apparently similar to the conventional upward-
sloped linear supply curve first used in Chapter 4, is in fact something different.
Remember we have assumed all tenant farmers are equally efficient. In a non-spatial
model, this would lead to a horizontal supply curve. However, CD in Fig. 10.1 is
upward sloped because congestion arises; to produce more wheat, farms need to use
land further away from the customer point and thereby incur greater shipping costs.
Pa Ea
A2.2
Fa
A0.8
0
B0.85 Qa B2.20 B4.50 B6.00 B7.64
Quantity
Fig. 10.1 Model 10A: equilibrium in the market for wheat on a ribbon landscape.
Notes: α = 4.5, β = 12, N = 1,000, C = 1, and s = 0.05. Equilibrium solution is P = 3.36,
Q = 95. Price axis scaled from 0 to 8; quantity axis from 0 to 700
10.2 Model 10A: Farms Producing Wheat Along a Line 277
From (10.1.5), we can see the supply curve approaches a horizontal line—that is,
congestion disappears—at P = C as the unit shipping rate (s) approaches zero.
This has implications for the market rent for land. Assume tenant farmers bid
for land for wheat production and that there are no other users for this land. The
marginal farmer at distance X earns zero excess profit. However, farmers closer
to the customer point have the potential to earn excess profits, compared to the
marginal farmer, because they spend less on shipping. However, this excess profit is
not arising because of some unique managerial skill or other monopoly advantage
attributable to the farmer; instead, it arises because of the proximity of the site. The
notion of Ricardian Rent is that, with competition, such an excess profit will accrue
to the owner of the asset: in this case, the landowner. This is because the existence of
excess profit implies other tenant farmers will try to bid away that land; the pressure
to outbid other users continues right up to the point where excess profit drops to
zero at every location. By this mechanism, Ricardian rent is translated into market
rent and in this sense generates a spatial equilibrium. See Fig. 10.2.
Ga
Ra Ia
Rent
Outcome
P Q X R[0] R[X]
Given [1] [2] [3] [4] [5]
C + − − − 0
N + + + + 0
s + − − + 0
α + + + + 0
β − − − − 0
the market point, (2) how much the farm is spending on inputs (other than ship-
ping services and rent) and how much wheat revenue is income in the farmer’s
hands, and (3) what the farm uses its income for. Nonetheless, undertaking such
a description is useful in starting to address Thünen’s focus on the distribution of
regional income. As an example, suppose α = 4.5, β = 12, N = 1,000, C = 1,
and s = 0.05. These are the values I used to generate Figs. 10.1 and 10.2. As
shown in the notes to those two charts, the outcomes are P = 3.36, Q = 95,
X = 47.30, and R[0] = 2.36. Column [1] of Table 10.4 provides more detail about
this regional economy under these conditions. This example portrays a regional
economy in which shipping costs are substantial. Land rent payments to absentee
landlords are a significant drain on the regional economy. Out of an expenditure
of $318 by demanders at the market point, farmers receive only $95 to purchase
their other inputs (e.g., seed, fertilizer, workers) and to cover the normal profit on
their own otherwise unpriced labor, management, and capital inputs. Anything that
reduces the unit shipping rate (s) will reduce the leakages due to shipping costs and
rents and thereby reduce price and increase the quantity of wheat transacted. We
will also see shortly that constraining farms to a line (ribbon landscape) causes the
price of the commodity to be substantially higher than it would be on a rectangular
plane.
Model 10A is useful in starting to think about Thünen. However, we need to
remember here that Model 10A is simpler than the regional economy envisaged by
Thünen. In Model 10A, we have considered only one crop, produced using fixed
yields, on a strip of land; Thünen was thinking of many crops—produced on a rect-
angular plane—for each of which yield was linked to labor intensity. Model 10A,
absent a two-dimensional geography and multiple crops, has no rings. Model 10A
also does not include any direct discussion of labor or capital inputs.16 And, unlike
Thünen’s derivation of nonlinear functions for transportation costs, crop prices, and
land rents, the solution to Model 10A involves only linear functions. At the same
time, what Model 10A adds that is new from the perspective of Thünen is a direct
connection to demand theory so that the impact of a shift in demand on crop price
and land rents can be assessed.
16 From Thünen’s perspective, Model 10A might also be unsatisfactory because it does not consider
crop rotation.
280 10 Farming for Cash
Supply of wheat
π ((P − C) / s)2 if P ≥ C, 0 otherwise (10.3.1)
Aggregate demand
(αN / β) − (N / β)P (10.3.2)
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): C—Production cost
per unit of land; s—Unit shipping rate ($/kilometer) for wheat; N—Number of consumers; x—
Distance (kilometers); α—Intercept of individual linear inverse demand curve: maximum price;
β—Negative of slope of individual linear inverse demand curve: marginal effect of quantity on
price received. Outcomes (endogenous): P—Price of wheat (at market); Q—Quantity of wheat
(square kilometers of yield); R[x]—Land rent bid by tenant farmers that leaves them with zero
(excess) profit; X—Radius of area supplying wheat (kilometers).
10.3 Model 10B 281
the quantity supplied. Since quantity supplied is equal to the land area under culti-
vation and since all land within distance X is used to produce wheat, the quantity
supplied is given by (10.3.1). See Table 10.3. The supply curve is also drawn as
CEb Db in Fig. 10.3.
A4.5
Price
Pa Ea
A2.2
Eb Db
Pb
Fa
C Fb
A0.8
0
B0.85 B2.20 Qb B4.50 B6.00 B7.64
Qa
Quantity
Fig. 10.3 Model 10B: equilibrium in the market for wheat on a rectangular plane.
Notes: α = 4.5, β = 12, N = 1,000, C = 1, and s = 0.05. Equilibrium solution is
P = 1.45, Q = 254. Price axis scaled from 0 to 8; quantity axis from 0 to 700
In an important respect, however, CEb Db is not the usual supply curve. Typically,
for ease of exposition in this book, I draw supply curves as straight lines. This
was the case for the supply curve (10.1.5) in the one-dimensional model: repro-
duced as CD in Fig. 10.3. However, the supply curve in two-dimensional space is
the upper half of a quadratic: the half where P is above C.17 Quantity supplied is
zero when the price is less than or equal to C, then rises quickly at higher prices.
Therefore, CEb Db in Fig. 10.3 lies below and to the right of the supply curve (CD) in
Model 10A.
This shift in the supply curve compared to Model 10A means that the two-
dimensional geography of Model 10B has immediate implications in terms of
equilibrium price. Graphically, we can see one of these in Fig. 10.3; from Models
10A to 10B, equilibrium price drops substantially (from OPa to OPb when α =
4.5), and quantity increases (from OQa to OQb ). The nonlinear supply curve in
(10.3.1) also implies that the equilibrium price in Model 10B is more complicated
to solve. Rearranging the aggregate inverse demand equation in (10.1.2) yields
aggregate demand (10.3.2). Equating demand (10.3.2) and supply (10.3.1) yields an
expression for price that is quadratic in price. Ordinarily, a quadratic has two solu-
tions (roots). However, one of these solutions is when P < C, and we can ignore
this case since it is uneconomic. Therefore, the only solution to this problem is the
larger root. That root is given by (10.3.4): see also (10.3.5).
So far, you might think this is a simple model of two-dimensional geography.
Yet, even this simple model generates a solution for price (10.3.4) that is messy to
analyze. Why? The problem here appears to be one of geography since, in Model
10A, the solution for P (a weighted average of C and α) was simple by comparison.
Nonetheless, the solution for price in Model 10B is similar to Model 10A in at
least two important respects. We can see from (10.3.5) that since the term under the
square root sign is always no smaller than s2 N 2 , v > 0 and therefore P must always
be larger than C. This is similar to what we found in (10.1.6) for P in Model 10A.
Further in 10A, we found that P was larger for larger α, C, N, or s, or smaller β.
Again, the same results hold in Model 10B. So, although (10.3.5) is unwieldy, it
does generate solutions for P similar to those in Model 10A.
What happens to market rent? Market radius (X) is determined by (10.3.7), and
market rent at any location out to X is given by (10.3.8). Market rent has to decline
linearly with distance, just as it did in Model 10A. The rent at distance 0 is R[0] =
P − C, also just as in Model 10A. In other respects too, Model 10B is similar
to Model 10A. Model 10B also solves for equilibrium in two markets at the same
time: the market for wheat and the market for land. With Model 10B, we can also
partially envisage a regional economy in the sense of identifying aggregate farm
receipts and disbursements. Finally, Model 10B too is based on just one agricultural
commodity.
17 According to Backhaus (2002, p. 437), Launhardt was perhaps the first person to identify this
relationship.
10.3 Model 10B 283
Rent
distance axis scaled from 0 to
50. See also Fig. 10.3
Gb
Rb Ib
0
K Hb Ha
Distance from demand point
Despite the similarities, Model 10B describes a different world from that of
Model 10A. I use the same parameter values to construct Figs. 10.3 and 10.4 in
Model 10B as I had used for Figs. 10.1 and 10.2 in Model 10A. The only differ-
ence between these two pairs of charts is 1 dimension (ribbon) versus 2 (rectangular
plane). In Model 10B (Figs. 10.3 and 10.4), the market draws out 254 km2 of wheat
production within a radius of 9 km; that enables a market equilibrium price for
wheat of just $1.45. In Model 10A (Figs. 10.1 and 10.2) by contrast, the market
had to extend 47.3 km to generate just 95 km2 of wheat production, and the market
equilibrium price there had to more than double ($3.36). In Fig. 10.3, the shift in
supply curve (from CD in Model 10A to CEb Db in Model 10B) means that we slide
from point Ea on the demand curve Av Bv all the way to point Eb . We can think of
the ribbon of Model 10A and the rectangular plane of Model 10B as two extremes.
In reality, we might expect a fraction of the land around a customer point to be
unavailable or unsuitable for crop production; in that case, the supply curve in prac-
tice might well lie somewhere between CD and CEb Db . The regional economy is
different in Model 10B compared to Model 10A. In Model 10B, production costs a
more substantial proportion of farm receipts; shipping costs in total are much lower;
rents are lower still. See column [2] of Table 10.4.
How about comparative statics here? As the only difference between Models 10B
and 10A here is the nonlinearity of the supply curve, the comparative statics are the
same as for Model 10A above. To illustrate, imagine a thought experiment in which
we decrease the unit shipping rate per kilometer, s, in Model 10B. As we do, the
supply curve starts to flatten. Equilibrium price travels down the demand curve, and
equilibrium quantity increases. Given the assumption of fixed yields, the land area
284 10 Farming for Cash
Farm Receipts
Wheat 318 369 206 391
Corn 0 0 278 398
Total 318 369 484 789
Farm Disbursements
Production cost
Wheat 95 254 53 238
Corn 0 0 78 232
Total 95 254 131 470
Shipping cost
Wheat 112 76 36 69
Corn 0 0 141 143
Total 112 76 177 213
Land rent
Wheat 112 38 117 84
Corn 0 0 59 23
Total 113 38 176 106
Excess profit
Wheat 0 0 0 0
Corn 0 0 0 0
Total 0 0 0 0
As percentage of total
farm disbursements
Shipping cost 36 21 36 27
Land rent 36 10 36 13
under cultivation (πX 2 ) must rise, hence market radius must increase. Since market
rent falls linearly with distance at a slope of -s, the decline in s causes the market
rent curve in Fig. 10.2 to flatten. As a limiting case when s approaches 0, the supply
curve becomes a horizontal line of the form P = C. As we saw before in Model
10A, the case of s = 0 corresponds to a non-spatial world wherein the customer
point can get as much of the product as it wants at a price corresponding to marginal
cost. The same result would also hold in the case of the one-dimensional supply
curve (10.1.5) as we let s approach zero. The comparative statics are summarized in
Table 10.5.
10.4 Model 10C 285
Outcome
P Q X R[0] R[X]
Given [1] [2] [3] [4] [5]
C + − − − 0
N + + + + 0
s + − − + 0
α + + + + 0
β − − − − 0
Notes: See also Table 10.3. From (10.3.1), Q and X always move
in same direction; +, Effect on outcome of change in given is
positive;−, Effect on outcome of change in given is negative; 0,
Change in given has no effect on outcome.
18 For any two commodities, labeled 1 and 2, respectively, cross-price elasticity (c) is the percent
change in quantity of good 1 demanded (q1 ) given a one percent change in the price of good 2
(p2 ):c = (p2 /q1 )(dq1 /dp2 ).
286 10 Farming for Cash
Table 10.6 Model 10C: wheat and corn produced on ribbon landscape assuming s1 > s2 and both
crops produced
Aggregate inverse independent demand curves for the two crops
P1 = α1 − β1 Q1 / N and P2 = α2 − β2 Q2 / N (10.6.1)
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): Ci —Production cost
per unit of land for crop i; si —Unit shipping rate ($/kilometer) for crop i; N—Number of con-
sumers; x—Distance (kilometers); α i —Intercept of individual linear inverse demand curve for
crop i: maximum price; β i —Negative of slope of individual linear inverse demand curve for crop
i: marginal effect of quantity on price received. Outcomes (endogenous): Pi —Price of crop i (at
market); Qi —Quantity of wheat (square kilometers of yield); Ri [x]—Bid rent by tenant farmers
producing crop i that leaves them with zero (excess) profit; Xi —Radius of area supplying crop i
(kilometers). A unit of either commodity is the amount yielded, assumed constant, by a unit of
land.
of land users, wheat and corn farmers, bidding for land: each willing to bid up to
the amount of the excess profit per unit land (i.e., their bid rent). Let R1 [x] be bid
rent by farmers producing crop 1 for a unit of land at distance x from the market:
R2 [x] for farmers producing crop 2. Assume that land is awarded to the highest
bidder. Under the assumption of a competitive market for land, market rent, R[x],
10.4 Model 10C 287
will be the envelope curve of these bid rents: i.e., R[x] = Max{R1 [x],R2 [x]}. At
each location, there are potentially both wheat farmers and corn farmers vying for
land, and landlords accept the higher of the two bid rents.
Without specifying the supply sides for the moment, let us imagine that an equi-
librium price results for each product. From that, we can then imagine rents bid by
the two kinds of farmers that look like (10.6.2); that is, a linear function of distance
from the customer point with a slope of minus the shipping rate per kilometer for
wheat (i.e., − s1 or − s2 ). There are two possibilities in thinking about the market
rent that emerges here. One possibility, a corner solution, is that one kind of farmer
(say crop 1 farmers to illustrate the idea here) outbids the other kind of farmer
everywhere. This would be the case for example if, for crop 2 farmers, C2 > α2 .
The other possibility is that both kinds of farmers are able to bid successfully for
land somewhere. In this case, since s1 > s2 , farmers producing crop 1 would bid
higher near the customer point while farmers producing crop 2 bid higher further
away. To understand this important idea, see Fig. 10.5. There, curve G1 IH1 is the
bid rent curve of crop 1 farmers, curve G2 IH2 is the bid rent curve of crop 2 farmers,
and curve G1 IH2 is the resulting market rent (envelope) curve. As drawn here, curve
G2
J I
0
K H1 H2
Distance from demand point
Fig. 10.5 Models 10C and 10D: Equilibrium in the market for land when two crops are produced.
Notes: P1 − C1 = 10; P2 − C2 = 6; s1 = 2.0; s2 = 0.60. Distance axis scaled from 0.0 to
12; rent axis scaled from 0 to 12
288 10 Farming for Cash
G1 IH1 is steeper than curve G2 IH2 . If both crops are to be produced, there must be
a point of intersection (I). However, since curve G1 IH1 is steeper, it must lie above
curve G2 IH2 to the left of I and below curve G2 IH2 to the right. Therefore, since
s1 > s2 , crop 1 is produced close to the customer point, while crop 2 is produced
further away. Market rent, as the higher of the two bid rents at any location, is the
envelope curve G1 IH2 .19
Ignoring for the moment the possibility of a corner solution, we are now ready
to model supply and market equilibrium. Crop 1 is produced from distance 0 to dis-
tance X1 , while crop 2 is produced from distance X1 to distance X2 : see (10.6.3).
The total land area under cultivation (and, given fixed yields, the amount of each
crop produced in total) is given by (10.6.4). Equating aggregate demand and supply
for crop 1 then yields (10.6.5) which reduces to a linear equation (10.6.6), wherein
the price of crop 1 depends on exogenous parameters but also on the price of crop
2. Similarly, equating aggregate demand and supply for crop 2 then yields (10.6.9)
which reduces to a linear equation (10.6.10), wherein the price of crop 2 depends
on exogenous parameters but also on the price of crop 1. Note that this depen-
dence on the price of the other crop is not arising because consumers substitute
one for the other but rather because the two crops compete for land. These two lin-
ear equations—(10.6.6) and (10.6.10)—in two unknowns (P1 and P2 ) can then be
solved. In (10.6.13), I show the resulting solution for P2 ; P1 can then be solved
by back-substitution. At this point, the reader might look longingly back at Model
10A where equilibrium price was familiarly a weighted average of C and α for that
crop alone. In contrast, (10.6.13) tells us that, in Model 10C, equilibrium price for
each crop depends on parameters for both crops. The computational complexity in
(10.6.13) should not be surprising. After all, this is the most sophisticated model of
Walrasian equilibrium to this point in the book. Model 10C solves simultaneously
for equilibria in the markets for wheat and corn, as well as in the market for rented
land.
Let us now return to the question of drawing supply curves for the two crops.
The difficulty here is that the supply of each crop now depends on the prices of both
crops because these affect bid rents and the amount of land available for production
of a crop. Normally, when we draw a supply curve—because a graph is limited to
two dimensions—we assume the quantity supplied is a function of own price only.
However, in Model 10C, because of the competition for land, the amount supplied
of one crop depends also on the price of the other crop. Therefore, when we switch
from a one-crop model to a two-crop model we lose the ability to draw a simple
supply curve to go with our aggregate demand curve for each product. However,
despite that limitation, we have already seen (in Table 10.6) that we are able to
solve the model algebraically in the absence of a corner solution.
How about comparative statics here? See Table 10.7. In the following, I continue
to presume an absence of corner solutions.
19 Stevens (1968) uses a mathematical programming approach to determine rents at a given location
in a regional economy with two crops.
10.4 Model 10C 289
Table 10.7 Model 10C: comparative statics of an increase in exogenous variable (assuming
s1 > s2 ))
Outcome
C1 + − − − + − + − 0
C2 − + − − − + − − 0
N + + + + + + + + 0
s1 + − − + + − + − 0
s2 − + − 0 − + − − 0
α1 + + + + − + − + 0
α2 + − − + + + + + 0
β1 − − − − + − + − 0
β2 − + − − − − − − 0
Notes: See also Table 10.6; +, Effect on outcome of change in given is positive; −, Effect on
outcome of change in given is negative; 0, Change in given has no effect on outcome.
Cι When Cι is increased, the supply curve for crop i shifts upward parallel. As
it does, own price (Pι ) increases and quantity (Q) decreases in equilibrium.
Because Pι does not increase as quickly as Cι , market rent, Rι [x], falls every-
where by the same amount; this downward shift in the market rent curve in
Fig. 10.2 has the implication that the radius of the market (Xι ) must also
decrease. On net, this increases the area available for production of the other
crop j and causes its price (Pj ) to fall.
N When N is increased, each demand curve becomes flatter: sweeps counter-
clockwise around the quantity–price combination (0, αι ). As it does, both price
(Pι ) and quantity (Qι ) increase in equilibrium. As a result of the increase in
P, market rent, Rι [x], rises everywhere by the same amount; this upward shift
in the market rent curve in Fig. 10.2 has the implication that the radius of each
production area (X1 and X2 ) must also increase.
sι When sι is increased, the supply curve for crop i becomes steeper; it sweeps
counterclockwise about (0, Ci ). As it does, price (Pi ) increases and quan-
tity (Qi ) decreases in equilibrium. As a result of the increase in Pi , bid rent
at the origin, Ri [0], rises. However, a larger si means also the slope of the
market rent curve in Fig. 10.2 has increased. Since Qi has decreased, the
radius of the market (Xι ) must also decrease. On net, this increases the
area available for production of the other crop j and causes its price (Pj )
to fall.
αι When α ι is increased, the demand curve for crop i shifts upward parallel. As
it does, both own price (Pι ) and quantity (Qι ) increase in equilibrium. As a
result of the increase in Pι , market rent, Rι [x], rises everywhere by the same
amount; this upward shift in the market rent curve in Fig. 10.2 has the implica-
tion that the radius of the market (Xι ) must also increase. On net, this reduces
the area available for production of the other crop j and causes its price (Pj )
to rise.
290 10 Farming for Cash
βι When β ι is increased, the demand curve for crop i becomes steeper: sweeps
clockwise around the quantity-price combination (0, αi ). As it does, both own
price (Pι ) and quantity (Qι ) decrease in equilibrium. As a result of the
decrease in Pι , market rent, Rι [x], falls everywhere by the same amount; this
downward shift in the market rent curve in Fig. 10.2 has the implication the
radius of the market (Xι ) must also decrease. On net, this increases the area
available for production of the other crop j and causes its price (Pj ) to fall.
What about the possibility of corner solutions (kinks) here? Because the only
difference between Models 10C and 10A is the presence or absence of crop 2, it
is interesting to compare outcomes here in the two models as we shift the demand
curve for crop 1 (by changing α 1 ). In Model 10A, where there is no production
of crop 2, equilibrium price and quantity increase linearly with α 1 above the level
required to make production profitable (i.e., α1 > C1 ). For Model 10A, I show a
supply curve—the locus of equilibrium price and quantity traced out by parallel
shifts in the demand curve—as the straight line CFD.20 At α1 = 7.64, for example,
the equilibrium price for crop 1 would be P∗ and quantity supplied would be Q∗ in
Fig. 10.6. In Model 10C, producers of crop 1 compete for land with producers of
crop 2. Where crop 2 is profitable, a higher price for crop 1 is required to generate
P∗ F
A4.50
Pc Ec
crop 1 of up to 2 crops on a
ribbon landscape. A2.20
Notes: α2 = 4.5,
β1 = β2 = 12, N = 1,000, Pb Eb
Db
C1 = C2 = 1, s1 = 0.05, C
and s2 = 0.039. In A0.85
constructing ADB and CDB,
α 1 is varied from 2.0 to 200. 0
Price axis scaled from 0 to 8; B0.85 Qa B2.20 Qb B4.50 B6.00 B7.64
Qc Q∗ Quantity
quantity axis from 0 to 700
20 Normally, we think of a supply curve as the relationship between quantity and own price with
other prices held constant. However, along the quasi supply curves CFD and Cc FD, the price of
crop 2 is constantly changing.
10.4 Model 10C 291
a given quantity of crop 1 than would be the case in Model 10A. As shown in
Fig. 10.6, this is a price of Pa in the case of Model 10A versus Pc in 10C. In Model
10C, I can draw something I call a quasi supply curve: the locus of equilibrium P1
and Q1 traced out by parallel shifts in the demand curve. The quasi supply curve
for crop 1 is different from the supply curve for crop 1 in Model 10A because the
latter assumes all other prices are held constant, whereas other prices (specifically
the price of crop 2) may vary along the quasi supply curve. The quasi supply curve
is given by a polyline Cc FD that consists of two straight-line segments Cc F and FD.
In the segment FD, α 1 is so large that production of crop 2 is nowhere profitable.
Put differently, in segment FD, Model 10C reduces to 10A. However, for α 1 < 7.64,
it is still profitable to produce crop 2. So, in the segment Cc F, P1 and Q1 increase
with α 1 , but the slope of Cc F is less than the slope of CF in Model 10A. In Model
10C, we can therefore envisage two critical values for α 1 that produce kinks in the
relationship between P1 and Q1 ; one is when P1 is too low to encourage production
of crop 1 (that is, P1 is below the amount OCc in Fig. 10.6); the other is the P1 above
which there is no production of crop 2 (that is, α1 ≥ 7.64 in Fig. 10.6). In Model
10C, therefore, the supply curve of crop 1 is the polyline OCc FD with kinks at Cc
and F; in Model 10A, it is OCFD with a kink at C only.
The kink at point F complicates the interpretation of Model 10C. Let me now
discuss a numerical method for solving Model 10C that specifically allows for kinks.
The method is an adaptation of Samuelson’s programming method for finding a
spatial price equilibrium discussed in Chapter 5. See Table 10.8. First, set the width
of each ring: see (10.8.1). That in turn fixes the radii in (10.8.2), the supply of each
Table 10.8 Model 10C: numerical solution method wherein either or both crops produced
Consumer benefit
CB1 = 0.5(P1 + α1 )Q1 and CB2 = 0.5(P2 + α2 )Q2 (10.8.6)
Production plus shipping cost (excludes land rent)
PC1 = Q1 (C1 + 0.5 X1 s1 ) and PC2 = Q2 , (C2 + 0.5(X1 + X2 )s2 ) (10.8.7)
Table 10.9 Model 10C: examples of numerical solution wherein NSW maximized
Notes: Calculations by author. See also Table 10.6. α2 = 4.5, β1 = β2 = 12, N = 1,000,
C1 = C2 = 1, s1 = 0.05, and s2 = 0.039. CB consumer benefit; PC producer cost; NSW net social
welfare (CB1 + CB2 − PC1 − PC2 ); – Not applicable. Neither production nor consumption
of that crop. Numerical solution method is to find d1 and d2 that maximize Net Social Welfare
(NSW).
crop in (10.8.3), and the prices that consumers are then willing to pay for the crops
in (10.8.5). From that, we can then calculate consumer benefit (10.8.6), producer
cost (10.8.7), and net social welfare (10.7.8). To summarize, the objective here is
to find the d1 and d2 , each zero or positive, that maximizes net social welfare. As
this is just a quadratic programming problem, it can be solved numerically just as
we solved the Samuelson spatial price equilibrium problem. In Table 10.9, I show
the solutions to Model 10C as α 1 is varied from 1.00 to 11.00 with other parameters
given in the Note to that Table—the same numerical conditions as underlie Fig. 10.6.
From Fig. 10.6, there is neither production nor consumption of crop 1 at α1 < 3.10
and neither production nor consumption of crop 2 at α1 > 7.64. Since α 2 is fixed
here, consumer benefit regarding crop 2 (CB2 ) declines above α1 = 3.10 since
increases in α 1 then cause P2 to increase.
What happens when s1 = s2 or when s1 < s2 . We don’t have to worry about
s1 < s2 because this means simply that we should relabel the crops so that the crop
with the higher unit shipping rate is crop 1. However, we do have a special case when
s1 = s2 . In this case, there are two possibilities. In one P1 − C1 = P2 − C2 ; here
both types of farmers are able to bid the same amount for land at any given site.
There are no rings here and the location of any one type of crop is indeterminate. In
the other case, P1 − C1 = P2 − C2 . In this case, only the farmers producing the
most profitable crop are able to bid successfully for land; there is no production of
the crop for the lower of P1 − C1 or P2 − C2 .
10.5 Model 10D 293
What happens to the regional economy? See column [3] of Table 10.4. As in
Model 10A, the ribbon landscape in Model 10C means that shipping cost and land
rents will be substantial.
Finally, consider the case of tenant farmers producing two crops on a rectangular
plane. Other than this, I use the same assumptions as in Model 10C. The model is
summarized in Table 10.10. The two aggregate demand curves are shown in
(10.10.1). The rent bid by each crop at a given location is shown in (10.10.2). The
radius (X1 ) within which crop 1 is produced and the outer limit (X2 ) on crop 2
production are shown in (10.10.3). These are all the same as in Model 10C.
Table 10.10 Model 10D: wheat and corn on rectangular plane: assuming s1 > s2
New to this model are the supply curves for the two crops. Crop 1 is produced
in the inner ring as before, but the total land area here now is the area of a circle of
radius X1 . Crop 2 is produced in the outer ring, from X1 to X2 kilometers from the
customer point. The two quantities of crop production are shown in (10.10.4).
The market equilibrium prices for the two crops, P1 and P2 , are those that make
demand in (10.10.1) equal to the supply in (10.10.4) taking into account the radii in
(10.10.3). This involves solving two simultaneous quadratic equations. As in Model
10B, there are two roots (solutions) each for P1 and P2 . One solution for Pi is smaller
than Ci and can therefore be ignored. Therefore the solution is the larger root for
each of P1 and P2 .
To solve these two prices, I adapted the Net Social Welfare algorithm that I earlier
used to solve Model 10C. See Table 10.11. To summarize, the algorithm invokes
mathematical programming to maximize NSW. In each step of the algorithm carry
out the following:
1. Assign a value each for d1 and d2 , consistent with (10.11.1);
2. Calculate the two radii in (10.11.2);
3. Calculate the supply of each crop from (10.11.3);
4. Calculate the price forthcoming for each crop from (10.11.5);
294 10 Farming for Cash
Table 10.11 Model 10D: numerical solution method wherein either or both crops produced
Radii
X1 = d1 X2 = d1 + d2 (10.11.2)
Supply
Q1 = πd12 Q2 = π (X22 − d12 ) (10.11.3)
Rent
R[X2 ] = 0 R[X1 ] = s2 (X2 − X1 ) R[0] = (10.11.4)
s2 (X − X1 ) + s1 X1
Price
P1 = α1 − β1 Q1 ,/ N (10.11.5)
P2 = α2 − β2 Q2 ,/ N
Consumer benefit
CB1 = 0.5(P1 + α1 ) Q1 (10.11.6)
CB2 = 0.5(P2 + α2 ) Q2
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): Ci —Production
cost per unit of land for crop i; si —Unit shipping rate ($/kilometer) for crop i; N—Number of
consumers; x—Distance (kilometers); α i —Intercept of individual linear inverse demand curve
for crop i: maximum price; β i —Negative of slope of individual linear inverse demand curve
for crop i: marginal effect of quantity on price received. Outcomes (endogenous): Pi —Price
of crop i (at market); Qi —Quantity of wheat (square kilometers of yield); Ri [x]—Bid rent by
tenant farmers producing crop i that leaves them with zero (excess) profit; Xi —Radius of area
supplying crop i (kilometers).
Table 10.12 Model 10D: examples of numerical solution wherein NSW maximized
Notes: Calculations by author. See also Tables 10.11. α2 = 4.5, β1 = beta2 = 12, N =
1,000, C1 = C2 = 1, s1 = 0.05, and s2 = 0.039. Numerical solution method is to find d1
and d2 that maximize Net Social Welfare (NSW).
crop introduced the possibility of a kinked supply curve Cc FD. In Model 10D, how-
ever, there is no similar kink. It is true that Cd Ed Dd in Model 10D and CEb Db
in Model 10B do approach one another asymptotically as α 1 becomes sufficiently
large. However, there is no kink in the supply curve Cd Ed Dd in the case of a rect-
angular plane. Given the important role that kinks play in competitive location—as
evidenced throughout this book—the asymmetry between ribbon and rectangular
plane geographies is an interesting oddity.
What about comparative statics in Model 10D? In essence, Model 10D differs
from Model 10C only in that the supply curve is now quadratic in distance instead
of linear. However, despite the presence of a kink, that does not fundamentally affect
the comparative statics; these remain the same as in Model 10C above.
What happens to the regional economy? See Column [4] of Table 10.4. As with
Model 10B, the farm economy on a rectangular plane in Model 10D spends rela-
tively less on shipping cost and land rent than does the farm economy under either
ribbon geography (Models 10A or 10C). Farms spend more on shipping and land
rent in Model 10D than in Model 10B, presumably because of the competition
between crops.
A6.00
P∗ F
A4.50
Price
Pc Ec
Pa Ea
Cc
A2.20
Dd
Pd Ed
Cd Db
Pb Eb
C
A0.8
0
B0.85 B2.20 Qd B4.50 B6.00 B7.64
Qc Qa Q∗ Qb Quantity
Fig. 10.7 Model 10D: equilibrium in the market for crop 1 of up to 2 crops on a rectangular plane.
Notes: α2 = 4.5, β1 = β2 = 12, N = 1,000, C1 = C2 = 1, s1 = 0.05, and s2 = 0.039.
In constructing ADB and CDB, α 1 is varied from 2.0 to 200. Price axis scaled from 0 to 8; quantity
axis from 0 to 700
10A and 10C to show the same results on a ribbon landscape. In Table 10.13, I
summarize the assumptions that underlie Model 10A through 10D. Many assump-
tions are in common to all these models: see the list in panel (a) of Table
10.13.21
Given all other parameters, and allowing α or α 1 to vary gives us the opportu-
nity to trace out a kind of supply curve for crop 1. The four models give distinctive
(quasi) supply curves—CD, Cc FD, CEb Db , and Cd Ed Dd —depending on the com-
bination of geography and number of crops. If we compare Cc FD with Cd Ed Dd in
Fig. 10.7, it appears that geography does play a role in shaping the supply curve for
21 Inthe models presented in this chapter, there is an exogenous demand for the agricultural prod-
uct in the city, which I envisage as being for local consumption. Another alternative is to think
of the demand at the city being by a firm that processes the agricultural commodity for resale to
consumers including those living on farms. See Hsu (1997) for an extension of this type. Another
extension might be to look at the setting of shipping rates by a profit-maximizer supplier of ship-
ping services. See Harley (1988), Harris (1977), Hummels (2007), Macdonell (1891), and North
(1958) for examples of work in this area.
10.6 Final Comments 297
a crop. In the ribbon geography, there is a kink not found when geography takes the
form of a rectangular plane. The various combinations of geography and number
of crops also result in distinctively different farm economies. However, in all four
models, land rents, the price of crop 1, and (in Models 10C and 10D) the price of
crop 2 are jointly determined in Walrasian equilibrium.
Early on in the chapter, I argued that the farm here can be thought of as any firm
(factory) producing a good that requires land for production. There are parallels here
between factory and farm in the sense that both typically use land, labor, and capital
to produce an output that is then sold. The complication here is generally thought to
be that the use of capital is more pervasive in the factory compared to farm. The par-
allels between factory and farm suggest that the notion of a Walrasian equilibrium
based on competition in land, labor, and capital markets should be generalizable to
a broad range of firms and production processes.
Chapter 9 emphasized the importance of uncertainty in market formation. On
the other hand, the models in Chapter 10 ignore uncertainty altogether. On the one
hand, this is not unreasonable; Chapter 9 was concerned with the uncertainty that
arises when farmers arrive to trade agricultural commodities among themselves,
while Chapter 10 concerns farms supplying a city with a commodity for which
the demand is known. On the other hand, Chapter 10 ignores plausible risks: e.g.,
the risk of crop loss or the risk of a sudden shift in demand. Let me wrap up this
argument by saying simply that Chapter 9 gives some ideas about how to extend the
models of Chapter 10 to incorporate uncertainty.
While uncertainty is not a focus of this book, the simultaneity of prices and the
linkage to localization are. What do the models in Chapter 10 tell us about Walrasian
equilibrium and localization? In Models 10A and 10B, the price of the crop (wheat)
and the rent gradient for land are jointly determined. Within the market radius, there
298 10 Farming for Cash
(Thünen–Beckmann–Samuelson Problem)
A competitive industry in the city produces one commodity (soap) demanded by city
dwellers and farmers alike. Farmers produce a second commodity (wheat) similarly
demanded. Farmers ship wheat to the city in exchange for soap: again absent the
uncertainties central to Chapter 9. That market is in competitive equilibrium. Tenant
farmers bid up the market rent for land for sites close to the city until the Ricardian
rent associated with proximity is entirely spent. Labor moves freely between city
(soap production) and farm (wheat production). Labor market equilibrium is estab-
lished wherein Model 11E incorporates land and labor as factors in agricultural
production and therefore allows us to look at factor incomes. I include a non-spatial
version, Model 11D, to show the consequences of adding unit shipping rates. I also
included three versions of the simpler Beckmann Model because it too incorporates
substitutability between land and labor in crop production: 11A (without shipping
cost), 11B (one crop with shipping costs), and 11C (two crops with shipping costs).
Chapter 11 extends the models in Chapter 10 first by incorporating labor and a
production function for the agricultural commodity. In that respect, it can also be
seen as akin to Chapter 7 with its repair labor and repair production function. By
also introducing a good produced in the city and exchanged for the farm good,
Chapter 11 builds on Chapters 9 and 10. As farms vary the intensity of land use, they
also alter the geographic density of customers (farm workers) for the city’s product.
In this way, the chapter considers how prices in commodity and input markets and
the localization of farms are joint outcomes in a competitive market.
The models in Chapter 10 are instructive in thinking about how geography shapes a
supply curve and how excess profits get attributed to landowners. In that chapter—
with free entry—farms at different locations pay different rents for the use of land:
rents that exhaust excess profits. However, the models in Chapter 10 have obvious
shortcomings. This is the third of four chapters that look at models of agricultural
markets. In this chapter, I introduce three new ideas to address shortcomings of the
models in Chapter 10.
First, we might reasonably expect the farm to use land more intensively where it
is relatively costly. In Chapter 10, I assumed the farmer always has a fixed yield per
unit of land area. However, in Chapter 10, I also showed land rent would increase as
we move closer to the customer point. In this chapter, I consider how the ability of
farmers to substitute between land and labor in production affects the market for an
agricultural product. Presumably, farmers use relatively less land if it is expensive
and relatively more of other inputs (e.g., labor). This in turn affects rents as well as
the market price required to elicit any given quantity of supply. To incorporate such
substitution, this chapter moves beyond models that assume a fixed yield.
Second, I look at how the allocation of labor between city production and farm
production can be made endogenous in a model of the regional economy. The mod-
els in Chapter 10 do not say anything about how or why demand arises at a place.
To think more about regional development and economic growth, we need to build
a better understanding about what is generating demand here. I make a link in this
chapter between individuals as participants in the labor market and as participants
in the market for the exchange of commodities. In the later models in this chapter,
I assume that city workers produce a good (soap) that they then trade for another
good (wheat) produced on the farm. Farm and city then engage in trade because it
is beneficial to both communities; demand becomes endogenous to the model.
Third, the models in Chapter 10 do not say much about the distribution of income.
Once we incorporate labor, we can begin to look at the individual wage rate and total
wages paid to workers as part of the regional economy. In so doing, we get to link
outcomes across markets in commodities, land, and labor. This is different from
the approach used in Chapter 4, where we had assumed that consumption could be
examined in the absence of labor force impacts. In the spatial price equilibrium
model presented in Chapter 4, I argued that geography was treated as distinct and
that it was important to think about situations in which consumers might be moti-
vated to change their place of residence. The present chapter does that. I also argued
that the spatial price equilibrium model in Chapter 4 differentiated locations on the
supply side. For this chapter, I will continue to differentiate producers in that one
group of firms (soap makers) cannot produce the commodity made by the other
group of firms (wheat farmers).
These three ideas are interconnected. Wheat workers who are further away from
the city incur a greater shipping cost in consuming soap; presumably, they then
need a higher wage to reach the same level of utility as farm workers closer to the
city. At the same time, they substitute between soap and wheat as the exchange
rate varies. As we move away from the city, we see land rents fall and wages rise;
two trends that make the ability of farms to substitute between land and labor in
production important. Equilibrium outcomes in the commodity and factor markets
are then simultaneous and determined in part by the extent of substitution in both
production and consumption.
This chapter goes beyond Chapter 9 in two important respects. First, in Chapter 9,
each individual had an endowment of one commodity and then seeks to trade it for
another; there was no production there. There was a similar treatment in Chapter
10 where we assumed that the farmer had a fixed level of output (the yield or
11.1 The Thünen–Beckmann–Samuelson Problem 301
equivalently the endowment) for each square kilometer of land farmed. In this
chapter, production is explicitly considered. Second, by introducing a labor mar-
ket into a Thünen economy, we can get a fuller picture of the allocation of regional
income among workers, capitalists, and landlords and the resulting interconnections
among prices.
In this chapter, I look in sequence at two interesting reformulations of Thünen:
first, I present three models (11A, 11B, and 11C) inspired by Beckmann (1972) and
then two (11D and 11E) based on Samuelson (1983).1 Both sets of reformulations
help us to progressively better understand the operation of a regional labor market
and its impacts on land and commodity markets.
Before turning to Model 11A, let me comment briefly on the notion of a dual
economy. Thünen models divvy up a regional economy into rural and urban sectors.
Notions of dualization have permeated thinking in development economics since
at least 1950.2 In thinking about the transition from underdevelopment nations to
developed nations, many economists saw the importance of switching from an econ-
omy based on a traditional (often agrarian) sector to a modern (often technological,
industrial, or manufacturing) sector in the national economy. In a classic in the field,
Jorgenson (1961) describes the typical ingredients in a model of the dual economy
in a closed region as follows: (1) output of the traditional sector depends on land and
labor alone with diminishing returns and no capital accumulation; (2) the supply of
land is fixed; (3) output of the modern sector depends on capital and labor alone
with constant returns to scale and no land used; (4) production technology improves
in both traditional and modern sectors over time; (5) population growth depends
on the supply of food per capita; (6) the availability of labor for the modern sec-
tors varies directly with agricultural surplus; and (7) wage rates are not necessarily
the same in the two sectors. With such a model, development economists sought to
explain differences in economic development around the globe. Related to this was
research on the process of migration from rural to urban areas based on Harris and
Todaro (1970).3 This chapter culminates in Model 11E wherein, unlike the standard
dual economy model, the regional labor market is always in equilibrium. In that
important respect, Model 11E does not fit into the dual economy literature.
Nonetheless, I will demonstrate in this chapter that Model 11E is valuable in
thinking about regional economic development.
4 Were the farm to double both the amount of labor and land it uses, the ratio of labor to land used
by the farm (n) would remain constant, the amount of output per unit land (q) would also remain
the same, and therefore the total output of the farm would double.
5 The elasticity of substitution is the percentage change in ratio of land to labor when the ratio of
factor prices is increased by 1%.
11.2 Model 11A: Factor Substitution with One Crop and in the Absence of Shipping Cost 303
Table 11.1 Model 11A: farms produce wheat with substitutability between land and labor when
unit shipping cost is zero
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): a—Intercept of labor
productivity equation for wheat; N—Population of consumers at customer point; Ra —Rent bid for
square kilometer of land by alternative use everywhere; w—Wage rate; α—Intercept in individual
linear inverse demand curve; β—Slope in individual linear inverse demand curve; γ —Exponent
in labor productivity equations. Outcomes (endogenous): m—Labor used per unit land produc-
ing wheat; P—Market equilibrium price for wheat; Q—Aggregate quantity of wheat in market
equilibrium; q—Wheat produced per unit land.
revenue spent on labor inputs. This limits the exponent: 0 < γ < 1.6 Within the
interval from 0 to 1, the larger is γ , the more intensively the farm operates.
6 It is implausible to assume γ < 0 since this would mean gross output decreases as we increase
the amount of labor used on a square kilometer of land. It is also implausible to assume γ = 0
304 11 The City and Its Hinterland
The farm’s profit per square kilometer of land is the amount by which revenue
derived from the area exceeds the cost of labor and the opportunity cost of land: see
(11.1.2). The farm is assumed to hire the amount of labor that maximizes this profit:
see (11.1.3). Substituting this amount of labor, m, into (11.1.2) gives us the max-
imized profit (11.1.4). With competition in the wheat market, farms keep entering
the market until excess profit is driven to zero; this happens when price (P) falls to
OQ a Quantity demanded
OA 4.50 Intercept of demand curve (a = 4.50)
OA 4.50 EQ aO Consumer benefit
OC Intercept of supply curve (11.1.6)
A6.00 OCE aQaO Producer cost
A4.50
A2.20
Ea
C D
A0.85
0
B0.85 B2.20 Qa B4.50 B6.00 B7.64
Quantity
Fig. 11.1 Model 11A: equilibrium in the market for wheat.
Notes: α = 4.5, β = 12, a = 1, γ = 0.80, N = 1,000, w = 1, and Ra = 0.65.
Equilibrium price is 1.51 and equilibrium quantity is 249. The horizontal axis is scaled from 0 to
700; the vertical axis is scaled from 0 to 8
since the farm then could produce the same output with any amount of labor on a unit of land;
given the cost of labor (w), the profit-maximizing farmer would therefore choose zero labor and
incur only land rent and shipping cost in production of wheat for sale at the market. Finally, the
farm cannot have γ > 1 because this implies increasing returns which means the firm would
maximize profit by shrinking the amount of land used to near zero, thereby making m very large.
11.2 Model 11A: Factor Substitution with One Crop and in the Absence of Shipping Cost 305
the level given in (11.1.5). Because I have simplified the model by ignoring inputs to
production other than labor and land, there is no place here for normal profit arising
from unpriced inputs of the farmer as entrepreneur; I therefore assume normal profit
is zero. Further, that farms keep entering wheat production until (11.1.5) is reestab-
lished implies that the supply curve for wheat is a horizontal line (11.1.7) that cuts
the Y-axis at (11.1.6).
We are now ready to solve for market equilibrium price and quantity. The famil-
iar demand curve (11.1.8) can be laid on top of a supply curve that is horizontal
here. Equilibrium price is therefore given by (11.1.5) and equilibrium quantity by
(11.1.8). The total amounts of land and labor used for wheat production are then
given by (11.1.9) and (11.1.10), respectively. As in earlier models, we are then
able to calculate producer cost (11.1.11), consumer benefit (11.1.12), and net social
welfare (11.1.13).
I illustrate these ideas in Fig. 11.1. There, the supply curve is the horizontal line
CD. Suppose the demand curve is A4.50 B4.50 . The intersection point Ea gives the
equilibrium quantity of wheat (OQa ) and price (OC). In that diagram, I also show,
using faint lines, the demand curve shifted by various amounts to the left or right.
In all cases, because the supply curve is horizontal, the equilibrium price stays the
same; only the quantity adjusts.
What are the comparative statics here? There are two stories to be told: see
Table 11.2.
• One story is where the given (parameter value or exogenous variable) alters the
position of the supply curve: this includes a, γ , Ra , and w. The supply curve
remains horizontal but can shift either up or down. Anything (specifically an
increase in Ra or w or a decrease in a) that causes the supply curve to shift
306 11 The City and Its Hinterland
up causes the equilibrium to run up the demand curve: i.e., P increases and Q
decreases.7
• The other is where the parameter value shapes the position of the demand curve:
this includes N, α„ and β. If α is increased as noted above, the demand curve
shifts to the right, and the equilibrium travels to the right along the horizontal
supply curve: P stays the same and Q increases. This kind of parallel shift in the
demand curve is illustrated by various demand lines (Av Bv ) in Fig. 11.1. If β is
increased or N decreased, the demand curve sweeps clockwise around the point
(0, α), and the equilibrium point travels to the left along the horizontal supply
curve: P stays the same and Q decreases.
What does the regional economy look like? In column [1] of Table 11.3, I sum-
marize the regional economy corresponding to equilibrium point Ea in Fig. 11.1.
Here, 20% of all farm receipts are paid out to landlords in rents and the remaining
80% to workers as wages (since γ = 0.80 in this example). This is the first model
in the book that has allowed us to separate out the wage bill in the regional economy.
Now, let me introduce shipping costs into this regional economy. I retain the
same production function for wheat—now labeled (11.4.1)—as in Model 11A. See
Table 11.4. Given that the farm now incurs shipping costs as well, its profit per
square kilometer of land is given by (11.4.2).8 I envisage the farm at any location
choosing labor intensity (m) to maximize profit per square kilometer (r[x]) for wheat
at any given distance: see (11.4.3). Substituting this optimized labor intensity back
into (11.4.2), I show this profit for wheat, now its bid rent for land, decreasing with
distance x from the market because of shipping cost: see (11.4.4). There is a radius
X beyond which profit cannot possibly be positive regardless of how much (or lit-
tle) labor is employed in production: see (11.4.5). As farms compete for sites that
would otherwise be more profitable, they drive market rent at a given site up to the
bid rent so that farmers everywhere earn only normal profits. How does this differ
from Model 11A? In Model 11A, farmers enter the market until price is driven down
to the point where excess profit is everywhere equal to Ra ; in Model 11B, price is
driven down until excess profit at the boundary (X) is equal to Ra .
Table 11.4 Model 11B: farms producing wheat on a rectangular plane with substitutability
between land and labor in presence of shipping cost
We can now derive details about the market for wheat in equilibrium. I use
numerical integration (specifically, the midpoint rule version of a Riemann sum9 )
to solve for the total quantities of wheat produced (Q) and labor employed (M),
and total expenditure on shipping (S): see (11.4.6), (11.4.7), and (11.4.8). From
these, we can then calculate consumer benefit, producer cost, and net social welfare:
see (11.4.11), (11.4.9), and (11.4.12). The equations in Table 11.4 do not permit
an explicit algebraic solution for price. I use numerical methods to solve P. As in
Chapter 10, I use quadratic programming to find the P that maximizes net social wel-
fare in (11.4.12) subject to the constraints imposed by (11.4.1) through (11.4.11). As
we did in Fig. 11.1, consider again the case where α = 4.5, β = 12, N = 1,000,
a = 1, γ = 0.80, w = 1, and Ra = 0.65 and now let s = 0.05. The solution to
this example of Model 11B is P = 1.77, Q = 227, X = 5.17, M = 293, L = 84,
and R[0] = 1.43: see column [2] of Table 11.3. The empirical example of Model
10B that I used to generate Fig. 10.3 gave a lower P and higher Q compared to 11A
at the same values of α, β, and N. Model 11B generates a smaller radius X and a
higher central rent R[0]: in Model 10B, these were 9.00 and 0.45, respectively: see
Fig. 11.3. These two differences reflect that (1) Ra was zero in Model 10B and (2)
the lack of substitution in Model 10B denies the opportunity to seek a less costly
mix of land and labor in producing wheat at any given distance.
What does the supply curve for wheat look like? As in Chapter 10, I trace out a
supply curve by systematically varying the intercept (α) of the demand curve. See
the sample curve CEb Db in Fig. 11.2, but keep in mind that the exact position and
shape of this curve will depend on the value we assume for s. The supply curve in
Model A (the faint line CD now reproduced in Fig. 11.2) has the same vertical inter-
cept as CEb Db : after all, if only a few farms are shipping, they will be all near the
customer point, and hence shipping cost will be negligible. In general, the larger the
P the faster the CEb Db rises, reflecting the fact that the area available for production
increases with the square of the radius.
What happens to land rent in this model? In Fig. 11.3, I sketch land rent (see
curve Kb Lb ) by location corresponding to the case of point Eb in Fig. 11.2. For
comparison purposes, I also show land rents (curve Gb Hb in Fig. 11.3) from Model
10B. The opportunity cost of land (Ra ) keeps rents higher everywhere in Model 11B
compared to Model 10B. Further, by incorporating substitutability of labor and land,
farms in Model 11A are able to bid rents up even more when nearer the market. The
steeper rent gradient in Model 11B is what enables the decrease in maximum radius
(X) in this model relative to Model 10B.
What about the comparative statics here? There are basically three stories: see
Table 11.5.
• One story is where the given (parameter value or exogenous variable shapes
the position of the supply curve: this includes a, γ , Ra , w, and s. The supply
curve—now upward sloped—can shift either up or down. Anything (specifically
an increase in Ra , s, or w or a decrease in a) that causes the supply curve to shift
up causes the equilibrium to run up the demand curve: i.e., P increases and Q
decreases.10 When P is increased, the bid rent curve shifts up.
A4.50
A2.20
Eb Db
Pb
C D
A0.85
0
B0.85 B2.20Qb B4.50 B6.00 B7.60
Quantity
Fig. 11.2 Model 11B: supply curve for wheat.
Notes: a = 1, N = 1,000, s = 0.05, w = 1, β = 12, Ra = 0.65, and γ = 0.80.
Equilibrium price is 1.77 and equilibrium quantity is 227. The horizontal axis is scaled from 0 to
700; vertical axis from 0 to 8
• The second story is where the parameter value shapes the position of the demand
curve: this includes N, α, and β. If α is increased as noted above, the demand
curve shifts to the right and the equilibrium travels to the right along the supply
curve: P and Q increase. When P is increased, the bid rent curve shifts up. If β is
increased or N decreased, the demand curve sweeps clockwise around the point
(0, α), and the equilibrium point runs down the supply curve: P and Q decrease.
When P is decreased, the bid rent curve shifts down.
• The third story is about two parameters—γ and s—that affect the shape, not just
the position, of the bid rent curve. When s is larger, the bid rent curve becomes
steeper. When γ is larger, the farm is better able to substitute labor for land, and
this too causes the bid rent curve to become steeper.
11.3 Model 11B: Factor Substitution with One Crop and in Presence of Shipping Cost 311
Fig. 11.3 Model 11B: bid Model 11B: Market for land
rents and market rent as a GbHb Bid rent line for land in Model 10B (for comparison purposes)
function of distance from the Kb KbLb Bid rent curve for land by wheat farmers
Bid rent for land adacent to market (R[0])
OK b
market where both crops are ORa Opportunity cost of land (R a)
OX Outer radius (X) enclosing farms producing wheat for this market
sold.
Notes: a = 1, N = 1,000,
Rent
s = 0.05, w = 1 , α = 4.5,
β = 12, and γ = 0.80. In
Model 11B, R[0] = 1.43
and R[X] = 0.65. The
horizontal axis is scaled from
0 to 10; the vertical axis is
scaled from 0 to 1.6 Ra Lb
Gb
0
X Hb
Distance from market
Outcome
Given P Q X R[x]
at x= 0 X
[1] [2] [3] [4] [5]
a − + − + 0
N 0 + + + 0
Ra + − − + +
s + − − + 0
w + − + − 0
α + + + + 0
β 0 − − − 0
γ − + − + 0
Notes: See also Table 11.4. Comparative statics numerically estimated from the
base case: α = 4.5, β = 12, s = 0.05, a = 1, γ = 0.8, w = 1,
N = 1,000, Ra = 0.65; +, Effect on outcome of change in given is positive; −,
Effect on outcome of change in given is negative; 0, Change in given has no effect
on outcome.
What happens to the regional economy? See column [2] of Table 11.3. Compared
to Model 11A, the regional economy here produces less wheat in total, uses less land
in total (in part also because of increased labor intensity), yet incurs only slightly
lower rents in total (because of the Ricardian rents), and a substantial shipping cost.
312 11 The City and Its Hinterland
The equilibrium price for wheat is therefore higher; aggregate consumer expenditure
on wheat is also higher. Aggregate farm disbursements for labor and land are slightly
lower in 11B.
So far, we have looked at a market for wheat. Suppose, however, consumers at the
city also purchase a second crop, say corn. Let us imagine two concentric rings
around the city; an inner ring of width d1 occupied by producers of crop 1 and
an outer ring of width d2 occupied by producers of crop 2. X1 and X2 are derived
from these: see (11.6.1) in Table 11.6. Cobb–Douglas production functions for the
two crops are given in (11.6.2); note that the scalar differs (a1 vs. a2 ) but that the
exponent (γ ) is assumed, for simplicity of exposition, to be the same for the two
crops. The profits per square kilometer of both kinds of farms are given in (11.6.3).
Using the condition that the profit-maximizing farm employs labor on each square
kilometer of land at distance x up to the level where the marginal value product
of labor is equal to its wage, we get (11.6.4). Substituting these expressions for
labor intensity (m1 [x] and m2 [x]) back into the profit equations (11.6.3), we get the
optimized profit for each type of farm at each possible site. Under the assumption
of free entry by crop 2 farms, we can find the P2 that makes the excess profit (its bid
rent) of a crop 2 farmer at distance X2 just equal to the alternative land rent (Ra ): see
(11.6.5). Under the assumption of free entry by crop 1 farms, we can then find the
P1 that makes the bid rent (excess profit) of a crop 1 farm just equal to the bid rent
(excess profit) of a crop 2 farm at distance X1 : see (11.6.6). Note here a necessary
condition for crop 1 to outbid crop 2 in the inner ring and be the lower bidder in ring
2 is that (11.6.7) be satisfied.
I leave aside for the moment the question of whether P1 or P2 are high enough
to permit any production of that crop. Ignoring for the moment the possibility of
such a corner solution, we are now ready to model supply and market equilib-
rium using numerical integration (midpoint rule version of a Riemann sum11 ) for
(11.6.8), (11.6.9), and (11.6.10). Total production levels for the two crops are given
in (11.6.8), total labor inputs in (11.6.9), and shipping costs in (11.6.10). That allows
us to calculate Producer Cost in each market (11.6.11). Given the quantities of crop
supplied, we can calculate the market clearing prices, V1 and V2 : see (11.6.12). This
then allows us to calculate Consumer Benefit for each crop (11.6.13) and then aggre-
gate Net Social Welfare (11.6.14). As in Model 10C, we can then solve a quadratic
program to find the d1 and d2 that maximize (11.6.14) subject to the constraints
in (11.6.1) through (11.6.13). The consequence of this maximization is that we get
market clearing prices in the two markets: P1 = V1 and P2 = V2 .
Table 11.6 Model 11C: Beckmann’s version of Thünen regional economy with 2 crops
Radii
X1 = d1 and X2 = d1 + d2 (11.6.1)
Production of crop per square kilometer of land used for that crop
γ γ
q1 = a1 m1 and q2 = a2 m2 where 0 < γ < 1 (11.6.2)
Profit per square kilometer of land for farm producing that crop
r1 [x] = (P1 − s1 x)q1 − wm1 and r2 [x] = (P2 − s2 x)q2 − wm2 (11.6.3)
Profit maximizing labor per square kilometer of land for farm producing crop k
m1 [x] = (γ a1 (P1 − s1 x)/w)(1/(1 − γ )) m2 [x] = (γ a2 (P2 − s2 x)/w)(1/(1 − γ )) (11.6.4)
P2 required to make crop 2 no longer profitable for farm beyond distance X2
P2 = s2 X2 + (Ra / (1 − γ ))(1 − γ ) /(a2 (γ /w)γ ) (11.6.5)
P1 required to make crop 1 equally profitable as crop 2 for farm at distance X1
P1 = (a2 /a1 )P2 + ((a1 s1 − a2 s2 )/a1 )X1 (11.6.6)
Condition for crop 1 to be produced in inner ring, crop 2 in outer ring
a1 s1 > a2 s2 (11.6.7)
Quantity of crops produced
X X
Q1 = 2π 0 1 xq1 [x]dx and Q2 = 2π X12 xq2 [x]dx (11.6.8)
Producer cost
PC1 = wM1 + S1 + Ra π X12 and PC2 = wM2 + S2 + Ra π (X22 − X12 ) (11.6.11)
Market clearing price
V1 = α1 − β1 Q1 /N and V2 = α2 − β2 Q2 /N (11.6.12)
Consumer benefit
CB1 = (1/2)(V1 + α1 )Q1 and CB2 = (1/2)(V2 + α2 )Q2 (11.6.13)
Net social welfare
SW = CB1 + CB2 − S1 − S2 (11.6.14)
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): ai —Intercept of labor
productivity equation for crop i; N—Population of consumers at customer point; Ra —Rent bid
for square kilometer of land by alternative use everywhere; si —Unit shipping rate for crop i;
w—Wage rate; x—Distance; α i —Intercept in individual linear inverse demand curve for crop i;
β i —Slope in individual linear inverse demand curve for crop i; γ —Exponent in labor productivity
equations. Outcomes (endogenous): di —Width of ring in which crop i produced; mi —Labor used
per unit land producing crop i; Mi —Total labor in production of crop i; Pi —Market equilibrium
price for crop i; Qi —Aggregate quantity of crop i in market equilibrium; qi —Crop i produced per
unit land; Ri [x]—Profit per square kilometer from crop i; Si —Total shipping cost; Vi —Price at
which given quantity clears the market; Xi —Outer radius of area supplying crop 1.
314 11 The City and Its Hinterland
To illustrate, consider the case of two crops that are otherwise similar except
that one is more costly to ship: a1 = a2 = 1, γ = 0.80, w = 1, Ra = 0.65,
N = 1,000, α1 = α2 = 4.5, β1 = β2 = 12, s1 = 0.05, and s2 = 0.035. If two
rings are present, crop 1 is therefore produced in the inside ring. I present details of
the numerical solution in column [3] of Table 11.3. Note there that even though the
demand and supply parameters are otherwise the same for the two crops, the higher
s1 alone causes P1 to be higher than P2 . Less of crop 1 is demanded compared to
crop 2. Because crop 1 is produced on the inside, where rents are higher, farms
there compensate by using labor more intensively. In Fig. 11.4, I show the schedule
of labor intensity (curve ABC) by distance. In Fig. 11.5, I show the schedule of
market rents (as curve ABC) where the bid rents by crop 1 farms form AB while the
bid rents by crop 2 farms form BC.
As in Model 10C, we can now draw a quasi supply curve for crop 1: quasi again
in the sense that we cannot hold the price of the other crop (P2 ) constant. As previ-
ously, I trace this supply curve by systematically varying α 1 ; Cc Ec Dc in Fig. 11.6 is
the locus of equilibrium prices and quantities so obtained. The market equilibrium
for the numerical example in the preceding paragraph is labeled Ec . For compari-
son purposes, I show the supply curve for the corresponding one-crop Model 11B:
F B
G C
0
E D
Distance from demand point
F B
G C
0
E D
Distance from demand point
CEb Db from Fig. 11.2. Equilibrium shifts from Eb to Ec when we introduce the sec-
ond crop. Competition for land at the margin of crop 1 production (i.e., near X1 )
pushes up the cost of the marginal unit of crop 1 supplied and thereby increases
Ricardian rent for all sites still closer to the customer point. My interpretation of
Cc Ec Dc in comparison to CEb Db is as follows. The introduction of crop 2 pushes
the equilibrium point in market 1 back up its demand curve; for any given quantity
of crop 1, the price at which that quantity is forthcoming has gone up. Overall, the
effect of a second crop is to twist the supply curve so that Cc Ec Dc looks closer to a
horizontal supply (the kind of supply curve consistent with a perfectly competitive
market) than does CEb Db .
Notice here that a corner solution wherein crop 1 is no longer produced occurs
when α 1 is no larger than the amount OCc in Fig. 11.6. Similarly, we can imagine
an α 1 so large that crop 2 is no longer able to compete for land anywhere. At that
level of α 1 , Cc Ec Dc converges to CEb Db ; this would happen somewhere to the right
of the graph area shown in Fig. 11.6.
What about comparative statics in Model 11C? See Table 11.7.
316 11 The City and Its Hinterland
A6.00
Model 11C Equilibrium in market for crop 1 of 2
A vBv Demand curve (at a1 = v)
CcEcDc Supply Curve for wheat
CEbDb Supply curve from Model 11B
CcPcEcCc Producer surplus
OA4.50EcQcO Consumer benefit
OCcEcQcO Producer cost
OQc Equilibrium quantity
A4.50 OPc Equilibrium price
Pc A4.50EcPc Consumer surplus
Price
A2.20
Pc Ec Dc
Pb Db
Cc Eb
C D
A0.85
0
B0.85 B2.20 Qb B4.50 B6.00
Qc Quantity
Outcome
a1 − − + − + + − − + + 0
a2 − + − + + + − − + − 0
N + + + + + + + + + + 0
Ra− − + − + + + − + − + +
s1 − + − − + + − − + + 0
s2 − + − + + + − + − − 0
w + + − − − − + + − − 0
α1 + + + + + − + + − − 0
α2 − + − + + + + + + + 0
β1 − − − − − + − − + + 0
β2 + − + − − − − − − − 0
γ − − + + + = − − + + 0
Notes: See also Table 11.6. Comparative statics numerically estimated from the base case: α1 =
α2 = 4.5, β1 = β2 = 12, s1 = 0.05, s2 = 0.035, a1 = a2 = 1, γ = 0.8, w = 1,
N = 1,000, Ra = 0.65; +, Effect on outcome of change in given is positive; −, Effect on
outcome of change in given is negative; 0, Change in given has no effect on outcome.
crops are now more costly to produce, so the price of each crop rises, the
quantity demanded drops, and farms use less labor and land in total.
αι When α ι is increased, the price each consumer is willing to pay for a given
quantity of crop i goes up. The price of crop i rises; so too does the amount
demanded of the crop and of the inputs (labor and land) used to produce
it. The resulting competition in the land market means that the price of the
other crop rises but that quantity demanded (and input usage) declines.
βι When β ι is increased, the price each consumer is willing to pay drops faster
the more quantity of crop i farms supply. The effect here is the opposite of an
increase in α i . The price of crop i drops: so too does the amount demanded of
the crop and of the inputs (labor and land) used to produce it. The resulting
decrease in competition in the land market means that the price of the other
crop drops and that quantity demanded (and input usage) both increase for
that crop.
γ When γ is increased, crop production becomes more dependent on labor
input relative to land input. As farmers use less land and more labor to
produce their crops, land rents decline. For the parameter values used to cal-
culate the comparative statics here, each crop becomes less costly to produce,
its price drops, and quantity demanded increases.12
12 One can envisage here a counter-scenario in which the wage rate is high enough that the farms
find it more costly (not less costly) to become more labor intensive in production of the two crops.
In such a case, the price of each crop would rise, and the quantity demanded would decrease.
318 11 The City and Its Hinterland
I end this section with a final comment on the effect of including the substi-
tutability of labor and land. The effect of incorporating substitutability is to increase
the impact of proximity to the customer point on land rent. Put differently, substi-
tutability makes locational advantage more important to the extent that producers
can make more use of ubiquitous inputs to offset the higher land rents in closer
proximity to the city.
The three models above introduce the idea that substitutability of land and labor
is important in location theory. However, to advance Thünen’s analysis of the dis-
tribution of income and to better understand the linkages between commodity and
factor prices, we need to think more about outcomes in the labor market. In Models
11A through 11C, I assumed that the wage is fixed: i.e., labor is a ubiquitous input.
However, Thünen had in mind the idea that the markets for labor, land, capital,
and commodities had to be jointly in equilibrium. To incorporate a labor market
into location theory, I use an approach derived from Samuelson (1983).13 To set
the stage for this model, let us once again first look at a non-spatial version of the
model—one wherein shipping costs are assumed to be zero everywhere.
Samuelson envisages a competitive regional economy in which there is produc-
tion of only two commodities.14 Labor is used to produce both commodities. Every
unit of labor is also a consumer who purchases these commodities for consump-
tion. In Models 11A through 11C, I used M to refer to the total amount of labor
employed on farms. Earlier in the book, I used N to refer to the total number of
consumers at a place. In the remainder of this book, I look at each person as both
labor and consumer. I will therefore use N to refer to counts of either workers or
consumers.
For ease of exposition, Samuelson envisages a barter economy in which he treats
wheat as numéraire. Farms produce wheat that they then use to pay their wage bill
and land rent. Farm workers use some of their wages in wheat to pay for soap and
for shipping. At the city, soap producers pay their workers in wheat too; which is
the revenue they gain from selling soap to farm workers. Soap workers then use part
of their wheat income to purchase soap.
One commodity, that I call soap, is produced only at the customer point (which I
characterize as the city). I assume that soap is produced with labor only, at constant
returns to scale and by a perfectly competitive industry (efficient firms). In other
words, every soap worker produces the same amount (l) of soap. This is a fixed
coefficients technology. I also assume that fixed costs can be ignored. Since soap
production takes up no land, Samuelson treats the city as a place: ignored here is
the idea that city labor might also require land for housing, roads, or other uses. As
did Beckmann, Samuelson ignores capital as a factor of production. See (11.8.1) in
Table 11.8.
Soap production
v2 = lN2 (11.8.1)
Wheat yield per square kilometer of land
q1 = a(N1 /L)γ where 0 < γ < 1 (11.8.2)
Wheat production net of rent
v1 = L(q1 − Ra ) (11.8.3)
Land input that maximizes net wheat production
L = ((1 − γ )a/Ra )1/γ N1 (11.8.4)
Maximized net wheat
v1 = γ a1/γ ((1 − γ )/Ra )(1 − γ )/γ N1 (11.8.5)
Labor
N = N 1 + N2 (11.8.6)
Production possibility frontier
v1 = PlN − Pv2 (11.8.7)
Walrasian implicit exchange rate (units of wheat per unit soap)
P = γ a1/γ ((1 − γ )/Ra )(1 − γ )/γ /l (11.8.8)
Utility-maximizing allocation of labor
N2 = nN N1 = (1 − n)N (11.8.9)
Wage (in wheat) for farm labor in competitive labor market
w1 = γ a1/γ ((1 − γ )/Ra )(1 − γ )/γ 1 (11.8.10)
Wage (in soap) for soap labor in competitive labor
market
w2 = l (11.8.11)
Per capita consumption and utility
q11 = (1 − n)l/P q12 = nl U = q111− n qn12 (11.8.12)
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): a—Scalar in wheat
production; l—Marginal product of labor in soap production; N—Total labor in region; Ra —
Opportunity cost of land; ν—Preference for soap; γ —Exponent in wheat yield. Outcomes
(endogenous): q1 —Wheat yield per square kilometer of land; L—Total amount of land used in
wheat production; N1 —Regional labor allocated to wheat production; N2 —Regional labor allo-
cated to soap production; P—Walrasian exchange rate in commodity market (units of wheat per
unit soap); q11 —Amount of wheat consumed per laborer; q12 —Amount of soap consumed per
laborer; U—Utility level of laborer; v1 —Amount of wheat produced in region net of land rent;
v2 —Amount of soap produced in region; w1 —Wage rate (in units of wheat) for wheat labor;
w2 —Wage rate (in units of soap) for soap labor.
320 11 The City and Its Hinterland
The other commodity, wheat, is produced only on farms outside the city, using
only labor and land inputs. It is possible to incorporate two or more agricultural
commodities here; I consider just one for the sake of exposition. Once again, capital
inputs are ignored. Farms are also assumed to be efficient firms. Wheat produc-
tion is assumed to be a perfectly competitive industry. Yield (wheat output per unit
land) is a decreasing returns to scale function of labor intensity (labor per unit land),
N1 : see the Cobb–Douglas production function in (11.8.2). Assume wheat produc-
tion does not require capital or any other input (other than land and labor). Fixed
costs are ignored. Note also that (11.8.2) implies constant returns to scale in wheat
production overall; doubling the amount of land and labor leads to a doubling of
wheat production. Wheat production net of land rent, v1 , is now given by (11.8.3).
See the example in Fig. 11.7 where the most profitable intensity is OF, the wheat out-
put per unit land is OG, the total wage paid per square kilometer of land is OH, and
the excess profit (inclusive of land rent) is HG. Given we allocate N1 units of labor
in total to wheat production, the amount of land used for wheat production is given
by (11.8.4). The maximized level of net wheat production is then given by (11.8.5).
Note the implication that Ra must be strictly greater than zero here. If Ra were zero
CD Line parallel to OB
GH Wheat yield net of wage bill at labor intensity OF
OA Wheat yield as a function of labor intensity
OB Wage bill per square kilometer of land farmed as function of labor intensity
OF Labor intensity that maximizes profit
OG Wheat yield at labor intensity OF
OG Wage bill at labor intensity OF
A
D
B
G
E
H I
0
F Labor intensity
Fig. 11.7 Model 11D: wheat yield and profit in the Samuelson model.
Notes: ν = 0.5, l = 4, a = 2, γ = 0.8, N = 500, and Ra = 0.5. The profit-maximizing
solution is a labor intensity of 1.84 and a wheat yield per square kilometer of 3.25. The horizontal
axis is scaled from 0.0 to 4; vertical axis from 0 to 8
11.5 Model 11D: Non-spatial Version of Samuelson’s Model of a Thünen Economy 321
as was assumed by Thünen, there would be no limit to the amount of land used in
wheat production. Indeed, this is why I introduced Ra at the start of this chapter.
Because of the constant returns to scale in wheat production, wheat revenue is
exhausted in factor payments for land and labor; put differently, no excess profits
are earned by the farm.
Assume land rent is paid to absentee landlords and therefore does not affect the
demand for soap in the region. What happens to factor income in this regional
economy? The only other factor of production is wheat labor. The labor market
(which encompasses both city and farm areas) is assumed competitive.15 Wheat
labor receives all the rest of the income generated. Soap production is assumed to be
constant returns to scale and to require only one input (labor). Soap workers are paid
their marginal product, l, in soap (and this exhausts soap revenue); in labor market
equilibrium, wheat workers are each paid the equivalent amount in wheat, lP.
We are now ready to draw the production possibility frontier for this regional
economy. The total amount of labor in the region is fixed: see (11.8.6). Substituting
soap production (11.8.1) and maximized net wheat (11.8.5) into (11.8.6) gives the
region’s production possibility frontier, (11.8.7). This production possibility frontier
takes the form of a straight line with a negative slope (the negative of the implicit
exchange rate). See the production possibility frontier AB in Fig. 11.8. To under-
stand the linearity, imagine initially all labor in the region is in soap production.
A
AB Production possibility frontier
CD Highest indifference curve reachable on AB
OF Quantity of wheat produced net of rent
C OG Quantity of soap produced
F E
0
G B
Soap production (Q2)
Fig. 11.8 Model 11D: production possibility frontier in a non-spatial Samuelson model.
Notes: ν = 0.5, l = 4, a = 2, γ = 0.8, N = 500, and Ra = 0.5. At point E, v1 = 378 and
v2 = 1,000. Horizontal axis is scaled from 0 to 2,500; vertical axis from 0 to 800
15 Among others, Curry (1985a) considers the operation of geographical labor markets in the
presence of imperfect information.
322 11 The City and Its Hinterland
The total amount of soap produced is lN, and the net amount of wheat produced is
zero. Now, suppose 1 worker shifts from soap to wheat production. The amount of
soap production declines by l units; from (11.8.5), the amount of wheat produced
(β − 1)/β
net of rent rises by β(1 − β)(1 − β)/β a1/β Ra units. If we move a second unit
of labor from soap to wheat, we observe the same reduction in soap and the same
gain in net wheat production. The reason for this linearity is that both soap and net
wheat production have constant returns to scale.
Where along this production possibility frontier does our region consume? In
effect, each worker has the potential to earn l units of soap (by being a soap worker).
Everyone in the model, whether wheat or soap worker, has the same log-linear util-
ity function based on the consumption of two indispensable commodities: soap and
wheat. See (11.8.9). As in Chapter 9, this kind of utility function means no one
can live in autarky; wheat workers and soap workers must each trade with the other
to consume both of these indispensable commodities. Utility maximization implies
that each worker, as a consumer, will spend a fraction ν on soap (thereby consum-
ing Nνl units of soap in total) and a fraction 1 − ν on wheat (thereby consuming
N(1 − ν)lP units of wheat in total). In Fig. 11.8, this is shown as combination E on
the indifference curve CED.
This is a general equilibrium model.16 Up until Chapter 9 in this book, I had
assumed at least one of either the demand curve or supply curve is exogenous.
Samuelson’s version therefore is more ambitious because, like Chapter 9, it seeks
to answer the question of why and how exchange, and thereby a regional economy,
exists. There are three markets in this model: a market for labor, a market for land,
and a market for the exchange of soap for wheat. In this model, there is a demand
for labor and a demand for land. However, these are not exogenous to the model.
Rather, as so nicely illustrated in Fig. 11.8, they are determined within the model as
an outcome of the interaction between, on the one hand, the preferences for wheat
vs. soap and, on the other hand, the tradeoff between production of wheat vs. soap.
What about comparative statics here? How do the market outcomes change when
one of the givens is changed? See Table 11.9.
a When a is increased, the farm produces more wheat from a given level of
land and labor inputs. Because wheat is now more plentiful, the exchange
rate increases (more wheat is needed to purchase a unit of soap). Farms use
more land in wheat production. The split of labor between soap and wheat
remains unchanged, but utility rises because consumers have more wheat
available (though not more soap).
l When l is increased, soap workers each produce more output. Because soap
is now more plentiful, the exchange rate falls (less wheat needed to pur-
chase a unit of soap). The split of labor between soap and wheat remains
unchanged. Utility rises because consumers have more soap available
(though not more wheat).
Outcome
Given N1 N2 L v1 v2 w1 w2 P U
[1] [2] [3] [4] [5] [6] [7] [8] [9]
a 0 0 + + 0 0 + + +
l 0 0 0 0 + 0 0 − +
N + + + + + 0 0 0 0
Ra 0 0 − − 0 0 − − −
γ 0 0 − − 0 0 − − −
ν − + + − + 0 + 0 +
Notes: See also Table 11.8; +, Effect on outcome of change in given is positive; −, Effect on
outcome of change in given is negative; 0, Change in given has no effect on outcome.
N When N is increased, there are more workers available to produce both com-
modities. Production of both commodities is increased proportionally, and
the exchange rate remains the same. However, per capita consumption and
therefore utility remain the same.
Ra When Ra is increased, farms must pay more for each unit of land rented. This
makes wheat relatively more costly, pushing down the exchange rate, rais-
ing labor intensity, decreasing the amount of wheat consumed, and lowering
utility.
γ When γ is increased, wheat production becomes more reliant on labor rel-
ative to land. The amount of land used in wheat production drops. The
exchange rate declines (less wheat needed to purchase a unit of soap).
ν When ν is increased, consumers have a stronger preference for soap over
wheat. Labor shifts from wheat into soap production, the amount of land
used in wheat production decreases, and net wheat production declines.
What does the regional economy look like? See the example in Table 11.10.
Here, soap firms produce 1,000 units of soap; farms produce 473 units of wheat
(equivalent to 1,250 units of soap at the going exchange rate of 0.3783 units of
wheat per unit of soap). The labor force is split equally between soap and wheat
production (since I have assumed ν = 0.50) and the total wage bill in city and farm
is the same (when measured in units of soap). There are no shipping costs in Model
11D, and free entry ensures that excess profit is zero for both farms and soap firms.
Soap firms pay no land rent (since they don’t use land as an input to production);
farms pay 20% of their output on land rent (since 1 − γ = 0.20).
Model 11D
No shipping Model 11E
cost With shipping cost
s1 = s2 = 0 s1 = s2 = 0.01 s1 = s2 = 0.02
[1] [2] [3]
Firm disbursements
Total wage bill
Wheat producers 354 354 352
Soap producers 1,000 979 962
Total land rent
Wheat producers 89 88 88
Soap producers 0 0 0
Land used
Wheat producers 136 125 114
Soap producers 0 0 0
Land rent (wheat per
square kilometer)
At x = 0 0.65 0.83 1.05
At x = X – 0.65 0.65
Wage rate for wheat labor
(in kind)
At x = 0 1.42 1.33 1.26
At x = X – 1.42 1.42
Labor employed
Wheat producers 250 255 259
Soap producers 250 245 240
Exchange rate (units of
wheat per unit soap)
At x=0 0.35 0.33 0.31
At x=X – 0.38 0.40
Firm production
Wheat 443 442 440
Soap 1,000 979 962
Utility 1.19 1.15 1.12
Note: – indicates Model 11D has no outer radius. Parameter values in common to all 3 columns
above: ν = 0.5, l = 4, a = 2, γ = 0.8, N = 500, and Ra = 0.65. Calculations by author.
away from the city, the effective exchange rate—P[x] at Euclidean distancex away
from the city—is different because of the cost of shipping. Samuelson here thinks
of an iceberg model where a proportion of the amount hauled evaporates, spoils, or
is fed to the horse pulling the wagon on the way to and from the market.17 Here,
once again, unit shipping cost is envisaged strictly as resource use. Samuelson uses
Exchange ratio
P[x] = e − s0 x − s1 x P[0] (11.11.1)
Wheat yield
q1 [x] = a(N1 [x])γ (11.11.2)
Excess profit
q1 [x] − w1 [x]N1 [x] (11.11.3)
Labor
X
N = N2 + 2π 0 xN1 [x]dx (11.11.4)
Soap wage
w2 = l (11.11.5)
Wheat wage
w1 [x] = γ a(N1 [x])γ − 1 (11.11.6)
Bid rent
R1 [x] = q1 [x] − w1 [x]N1 [x] (11.11.7)
Radius of production
R1 [X] = Ra (11.11.8)
Utility-maximization for soap worker
q22 = vw2 q21 = (1 − v)w2 P[0] (11.11.9)
U2 = qn22 q121− n (11.11.10)
Utility-maximization for wheat worker at distance x
q12 [x] = nw1 [x]/P[X] q11 [x] = (1 − v)w1 [x] (11.11.11)
U1 [x] = (q12 [x])n (q11 [x])1 − n (11.11.12)
Equilibrium in market for exchange of wheat and
soap
X
(1 − v)lN2 = 2π 0 N1 [x]q12 [x]dx (11.11.13)
Notes: Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitation of shipping cost. Givens (parameter or exogenous): a—Scalar in wheat
production; l—Marginal product of labor in soap production; N—Total labor in region; Ra —
Opportunity cost of land; s1 —Unit shipping rate for wheat (iceberg loss); s2 —Unit shipping rate
for soap (iceberg loss); ν—Preference for soap; γ —Exponent in wheat yield. Outcomes (endoge-
nous): q1 —Wheat yield per square kilometer of land; L—Total amount of land used in wheat
production; N1 —Regional labor allocated to wheat production; N2 —Regional labor allocated
to soap production; P—Walrasian exchange rate in commodity market (units of wheat per unit
soap); q11 —Amount of wheat consumed per laborer; q12 —Amount of soap consumed per laborer;
U—Utility level of laborer; v1 —Amount of wheat produced in region net of land rent; v2 —Amount
of soap produced in region; w1 —Wage rate (in units of wheat) for wheat labor; w2 —Wage rate (in
units of soap) for soap labor.
an iceberg model here, as opposed to the shipping rate in earlier chapters because
he does not want to include a transport sector with its attendant factor incomes
in his general equilibrium model. Following the idea from Chapter 3 with respect
to a perfectly competitive industry that the price difference between two places
should not exceed the unit shipping cost involved, Samuelson assumes (11.11.1):
See Table 11.11.
326 11 The City and Its Hinterland
The existence of shipping costs has several implications in the model. First, con-
sumers at different places each face a different exchange rate of soap for wheat and
can therefore be expected to shift their consumption of the two commodities accord-
ingly. Second, the wage rate for a farm worker therefore also has to vary by location
as a result; if it did not, the utility of a worker might be higher in one location com-
pared to another and this would contravene the assumption of equilibrium in the
labor market. Third, the bid rent of farms will vary with location because profit is
now affected by proximity to the city. Fourth, the farm can be expected to substitute
between land and labor inputs as the prices of these inputs vary with location.
In labor market equilibrium, all workers (N in total) are allocated to soap produc-
tion in the city, or wheat production at each distance x from the city: see (11.11.4).
Because the market for labor is competitive, each soap worker is paid his or her
marginal product (l): see (11.11.5) in Table 11.11. The wage here is paid in soap
and therefore all soap made is paid out as wages; there is no profit in soap produc-
tion. Each wheat worker is also paid his or her marginal product: see (11.11.6). This
wage is paid in wheat.
In a market for land that is also competitive, landlords rent land to the highest
bidder. In competition, tenant farmers continue to enter the market, raising their bid
rents for a site until market rent absorbs excess profit: see (11.11.7). Samuelson
assumes land is available in unlimited quantity: presumably there is an alternative
use for land that offers the same bid rent (Ra ) for every location.18 Farms that pro-
duce wheat to exchange for soap in the city must be able to pay a rent at least equal
to Ra : that defines the radius of wheat production (X): see (11.11.8). Tenant farmers
bid themselves into a spatial equilibrium, just as they did in Chapter 10. Just as in
Model 11D, the farms have no excess profits in equilibrium; free entry ensures that
their entire production of wheat is expended in payments for labor and for land.
Now, consider the market for the exchange of wheat for soap. It too is compet-
itive. As in Chapter 9, trade occurs between individuals (in this case, farmers and
city residents) who each start off with only one commodity, thereby a utility of zero,
and must trade to achieve a positive utility. An exchange rate—the amount of wheat
paid for one unit of soap—gets established at the city: P[0]. This is the exchange
ratio confronting soap workers, and they make utility-maximizing choices based on
this: see (11.11.9) and (11.11.10). For a wheat worker at distance x, the effective
exchange rate is P[x] and the utility-maximizing choices are shown in (11.11.11)
and (11.11.12). The total amount of labor (N) is exogenous (fixed): see (11.11.4).
However, labor is allocated endogenously between city production and rural pro-
duction. Put differently, labor is thought to seek out the sector (soap or wheat) and,
in the case of wheat, location (i.e., distance from city) that maximizes utility. We are
now ready to solve this model. For soap workers, there is an amount of soap they
are willing to give up at the exchange rate, P[0]. For wheat workers at distance x,
18 Curiously, this is missing from Samuelson’s article. Samuelson appears to suggest that the alter-
native rent is zero. However, as is explained later in this chapter, the solution of the model requires
Ra be larger than zero.
11.6 Model 11E: Spatial Version of Samuelson Model 327
there is an amount of soap they are willing to purchase at the exchange rate P[x].
What then is the exchange rate P[0]—which in turn translates into P[x] at each dis-
tance x—that clears the market: i.e., equalizes the amount of soap that soap workers
are willing to give up with the amount wheat workers are willing to purchase. See
(11.11.13).
In Chapter 10, we saw that it is difficult to solve algebraically the agricultural
location problem on a rectangular plane. Model 11E is even more difficult to solve
because it also incorporates equilibrium in the labor market. To solve the model, I
have found it helpful to use an algorithm to solve market radius (X) nested within
an iterative process to solve the exchange rate at the city (P[0]). The steps are as
follows.
Although Model 11E does not permit explicit (algebraic) solution,19 one can
readily comprehend the patterns that arise in spatial equilibrium as derived from
numerical solutions. See Fig. 11.9.
P[x] Exchange rate. In spatial equilibrium, the exchange rate (the amount of
wheat given up to purchase one unit of soap) increases with distance from
the city because of the cost of shipment: see curve AB in Fig. 11.9. Put
differently, a wheat worker further away from the city must give up more
19 I
do not consider uniqueness of the equilibrium solution to this problem. However, Nerlove and
Sadka (1991) prove uniqueness for a version of the Samuelson model that is similar to Model 11E.
328 11 The City and Its Hinterland
Fig. 11.9 Model 11E: Model 11E: Outcomes by distance from city
equilibrium outcomes in a AB Exchange rate (wheat per unit soap) P[x]
CD Land rent (in wheat) R[x]
spatial version of the EF Wage paid (in wheat) to farm laborer w1[x]
Samuelson model. GH Wheat yield per square kilometer of land
IJ Soap consumed in ratio to farm wage q2[x]/w1[x]
Notes: ν = 0.5, l = 4, LD Opportunity cost of land Ra
a = 2, γ = 0.8, s2 = 0.01, OC Land rent adjacent to city R[0]
OL Outer radius of wheat production X
s1 = 0.01, N = 500, and
Ra = 0.65 . Horizontal axis
is scaled from 0 to 7; left G F
vertical from 0 to 6; right E
Axis for CD, GH, and IJ
vertical from 0 to 2
I
J
B
A
C
K D
0
Distance from city L
wheat per unit of soap to cover the cost of shipping the wheat in from the
farm and soap back to the farm.
q1 [x] Wheat yield. Wheat yield (per square kilometer) falls steadily with distance
from the city in spatial equilibrium: see curve GH. This should not be sur-
prising. After all, as noted above, the wheat wage increases with distance
from the city. Therefore, farms further away from the city would want to use
costly labor more sparingly. Because farming is less labor intensive there,
wheat yield is lower.
R[x] Market rent. Excess profit, dissipated as land rent, falls steadily with
increased distance from the city in spatial equilibrium: see curve CD. This
is also not surprising. After all, in the absence of any factor substitution, we
expect market rent to fall with distance to the market because of shipping
costs: that was a lesson learned in Chapter 10. As seen in Model 11B, the
farm’s ability to substitute between land and labor allows it to switch to a
lower labor intensity at more remote locations offsetting in part the impact
of higher shipping costs.
w1 [x] Wheat wage. That the exchange rate increases with distance from the city
implies that the nominal wage paid to a wheat worker must also increase
with distance in a spatial equilibrium: see curve EF. If wage did not increase
11.6 Model 11E: Spatial Version of Samuelson Model 329
with distance, workers there would not be able to achieve the same level of
utility as wheat workers closer to the city. Wheat workers reach the same
level of utility in part through a higher wage that permits them to offset the
effect of higher shipping costs. In part also, wheat workers achieve the same
level through their ability to substitute between soap and wheat. At locations
where remoteness from the market makes soap more costly to consume,
wheat workers substitute between consumption of soap and consumption of
wheat.
What happens in this model if shipping rates were to become very low? In
Chapter 10, we saw that the Lee–Averous model predicts that the supply curve for
the farm commodity becomes horizontal: i.e., suppliers will provide any quantity at
a price equal to the (constant) marginal cost of production. Corresponding to this,
there is no outer radius to production of the farm commodity. Thus, although the
production area is unbounded, only a finite amount of land will be used for pro-
duction. In this chapter, a similar result arises in Model 11D. Model 11E differs
from the Lee–Averous model; agricultural yield is not fixed. Nonetheless, Model
11E is similar in that as shipping cost goes to zero, the location of farms becomes
indeterminate. Farms are free to locate, unbounded by shipping costs.
What about comparative statics here?20 How does a spatial equilibrium change
when one of the givens is changed. See Table 11.12.
Table 11.12 Model 11E: spatial version: comparative statics of an increase in exogenous
variable (assumings1 > s2 )
Outcome
Given N1 N2 Lt v1 v2 P[0] U
[1] [2] [3] [4] [5] [6] [7]
a + − + + − + +
l 0 0 + + + − +
N + + + + + − −
Ra − + − − + − −
s1 + − − − − − −
s2 + − − − − − −
γ − + − − + − −
ν − + − − + − −
Note: See also Table 11.11; +, Effect on outcome of change in given is positive; −, Effect
on outcome of change in given is negative; 0, Change in given has no effect on outcome.
a When a is increased, each farm produces more wheat from the same amounts
of land and labor. With wheat more plentiful, soap becomes relatively more
costly: i.e., P[0] increases. Utility rises everywhere. With enhanced productiv-
ity, farms are able to supply the market from further away: i.e., X increases. In
the regional economy, less soap is demanded at the higher exchange rate, and
the amount of labor allocated to wheat production increases.
l When l is increased, each soap worker produces more. With soap now more
plentiful, wheat becomes relatively more costly: i.e., P[0] drops. Utility rises
everywhere. The income and substitution effects of the price change for wheat
cancel each other out, so the demand for wheat overall and the allocation of
labor in the regional economy remain unchanged
N When N is increased, there are more workers available to produce both com-
modities. Given that the income elasticity of demand21 for each product is 1.0,
we would expect demand for each product to rise proportionally. However,
to get sufficient additional wheat, the area under cultivation, radius X, must
increase, and nonzero shipping cost means that wheat becomes relatively more
costly: i.e., P[0] drops. Labor allocated to wheat production increases as a
proportion of total regional labor, N.
Ra When Ra is increased, farms must pay more for each unit of land rented.
This makes wheat more costly to produce. Wheat becomes relatively more
expensive (P[0] drops), and the area under cultivation (radius X) drops. In the
regional economy, more labor is now allocated to soap production.
s1 When s1 is increased, it costs more to ship wheat to soap workers. Less wheat
reaches the market and the exchange rate, P[0], therefore drops. The demand
for wheat decreases, and the radius of production (X) drops. In the regional
economy, this is compensated in part by a shift in labor from soap to wheat
production.22 In spatial equilibrium, utility is now everywhere lower.
s2 When s2 is increased, it costs more to ship soap to wheat workers. Put differ-
ently, a wheat worker at distance x pays only a smaller amount of wheat on net
(i.e., after shipping cost) for each unit of soap purchased. The exchange rate at
the city, P[0], drops. At remote distances, farm workers are no longer able to
achieve the same utility: hence, X drops. In spatial equilibrium, utility is now
everywhere lower. The amount of labor in soap production rises on net. The
principal spatial effect is that the exchange rate now increases more rapidly
with distance from the city.
γ When γ is increased, wheat production becomes more reliant on labor relative
to land. Access to land is less important in the regional economy, and the radius
(X) of wheat production therefore shrinks. This in turn makes wheat relatively
less costly to produce, and thus P[0] drops.
ν When ν is increased, consumers have a stronger preference for soap compared
to wheat. In the regional economy, labor is reallocated from wheat to soap
production. The area under cultivation (radius X) decreases. Because wheat is
now marginally less costly to supply to the city, soap becomes relatively less
expensive: i.e., P[0] declines.
21 For a given good, income elasticity of demand (i) is the percentage change in quantity of the
good demanded (q) given a one percent change in the income (y) of the consumer: i = (y/q)dq/dy.
22 Equivalently, Nerlove and Sadka (1991, p. 97) conclude that a drop in transport costs is sufficient
to induce labor to shift from rural to urban areas.
11.7 Final Comments 331
What about the regional economy in this model? In column [2] of Table 11.10,
I show the regional economy for an example comparable to that in Model 11D
except that s1 = s2 = 0.01. In column [3], I show a second example where
s1 = s2 = 0.02. Contrasting these with the outcomes for Model 11D in column
[1] gives us an idea of the role played by shipping cost.
First, consider the market for labor. Suppose we were able to decrease shipping
costs, sweeping from column [3] to column [1] of Table 11.10, the wage (paid
in wheat) to a wheat worker adjacent to the city increases. In contrast, the wage
(paid in soap) to a city worker remains fixed regardless of shipping cost. As we
decrease shipping costs, labor is redeployed from wheat production to soap produc-
tion. Finally, as we decrease shipping costs, the wage bills paid increase: slightly for
farms and more for soap firms.
Second, consider the market for land. As we decrease shipping costs, the rent
(paid in wheat) per unit of land adjacent to the city decreases. As we decrease ship-
ping costs, the total amount of land used in wheat production increases. Finally, as
we decrease shipping costs, the total land rents paid increase slightly.
Finally, consider the market for the two commodities: wheat and soap. As we
decrease shipping costs, soap becomes more valuable: it costs more wheat to pur-
chase one unit of soap. Total wheat production decreases and total soap production
rises. In so doing, utility goes up.
I explain these effects as follows. When shipping costs are high, the regional
economy has to devote considerable labor to production of commodities that require
land. Commodities that do not require land must therefore be shipped at great cost
to remote consumers. The regional economy has a focus on wheat production and
consumption. When shipping costs are decreased, the regional economy condenses
around the city, the price of the city good falls, and consumers everywhere purchase
relatively more soap and less wheat and benefit overall from the reduced cost of
shipping.
From the perspective of this book, it is both a strength and a shortcoming of
Model 11E that it is a general equilibrium model. The strength is that we have
now been able to link, for a simple regional economy, the prices of land, labor, and
commodities. We can now see how a change in any of the givens in this model
translates into changes in the equilibrium outcomes in all three sets of markets.
This is a powerful analytical tool. The weakness is that, in a sense, I began this
book with a much simpler question: how does the location of one firm affect the
prices paid or received by other firms in the vicinity. Unfortunately, we can’t look
at such questions directly with a general equilibrium model since all prices are now
endogenous. However, there is an important linkage here; the competition that leads
firms to choose a location is pushing firms toward the kind of equilibrium that Model
11E describes.
the assumptions that underlie Model 11A through 11E. Many assumptions are in
common to all these models: see the list in panel (a) of Table 11.13. The models
differ in that (1) Models 11A through 11C assume given demand curves for agri-
cultural products in the city and an infinitely elastic supply of labor to farms at a
given wage, (2) Models 11D and 11E assume a fixed amount of labor in the region
to be allocated between city and farm production, (3) Models 11A and 11D assume
shipping costs are zero, (4) Models 11B and 11E assume that farms produce only
one crop, and (5) Model 11C assumes farms produce two crops.
What about Walrasian equilibrium and localization?
• Model 11A has no shipping costs. It models three markets: wheat, labor, and land.
The supply of labor and the supply of land are each infinitely elastic at the going
wage (w) or rent (Ra ), respectively. The amounts of labor and land used by a farm
are determined in the market for wheat. There is no Walrasian simultaneity here.
Prices are not jointly determined here: instead, Ra , w, and other givens determine
P (the price of wheat).
• Model 11D also has no shipping costs. It too identifies three markets: wheat/soap,
labor, and land. The supply of land is infinitely elastic at the going rent (Ra ). What
11.7 Final Comments 333
makes it different from 11A is that (1) the total supply of labor in the region is
fixed and is allocated to either city or farm production and (2) firms in the city
produce soap that is exchanged for wheat. There is Walrasian simultaneity here.
The exchange rate of wheat for soap (P) and the wage (w) are jointly determined
by Ra and other givens.
• Models 11B and 11C incorporate shipping costs. Model 11B has three markets:
wheat, labor, and land. Model 11C adds a fourth market (crop 2). The supply of
labor is infinitely elastic at the going wage (w). Land may also be obtained but
now subject to differentials in shipping cost and land rent. There is Walrasian
simultaneity here. The locational rent (the amount by which rent exceeds Ra at a
given location) and the price of wheat and the price of crop 2 in Model 11C are
now jointly determined by Ra , w, and other givens.
• Model 11E also incorporates shipping costs. Land may also be obtained, subject
to differentials in shipping cost and land rent. There is Walrasian simultaneity
here. The locational rent, the price of wheat, and the wage rate are now jointly
determined by Ra and other givens.
What about the regional economy? Model 11E has brought us a long way toward
a comprehensive depiction of the regional economy. From Table 11.10, we can mea-
sure regional output and regional income. We are also able to divvy up regional
income into a wage bill and a rent bill. In Model 11E, there is no profit to either soap
or wheat production. The only profit is the advantage that arises to farms nearer the
city, and this is dissipated in locational rents.
In terms of the objectives of this book, this chapter is important for the ideas
raised in the last two paragraphs. At the same time, there remain some troubling
issues. One is that capital has no role to play in the models in this chapter; we
cannot use these models to look at how firm receipts are distributed to labor, land,
and capital.
Second, I am troubled by curve EF in Fig. 11.9: the wage paid to wheat labor.
As drawn, EF rises steadily as we consider farms further away from the city. The
wage is highest at the outer edge of wheat production. I recognize that this is a
nominal wage only. Since utility is the same for laborers throughout the region, soap
is becoming more costly with distance from the city at an even faster rate than the
wage. However, what bothers me is that wage reaches a precipice at the outer radius.
Thünen envisaged it as a frontier wage. The outer boundary of wheat production is
akin to a “big bang”: wage rises ever faster as we move away from the factory, but
a step beyond that boundary there is no further demand for labor. Why the abrupt
change?
Third, the models in this chapter are concerned solely with the consequences
of perfect competition. In this regard, this chapter is similar to Chapters 4 and 5.
Throughout this book however, I have emphasized the implications of geography
and shipping cost for the existence of local monopolies. There is no monopoly in the
market for either soap or wheat in this chapter. Indeed, the only monopoly element
evident is the locational rent generated at sites near the city. At this stage, we need
a model in which local monopolies can arise.
Chapter 12
Local Production and Consumption
Substitutability, Saturation, and the Regional Economy
(Thünen–Miron Problem)
A factory, using labor and other inputs, produces a commodity for sale to farms
that replaces local production. Because of transportation costs (including shipping
costs and commuting cost), the factory purchases inputs (including labor) and sells
its product within a market area. In so doing, the factory leads to a rearrangement
of local production and consumption. As a monopolist, the firm sets price know-
ing this affects the locations of consumers as well as their demands. In Model
12A, the behavior of the farm in autarky is examined. In Model 12B, I consider
a farm within the market area of the factory. In Model 12C, I introduce a monopo-
list with constant marginal cost. In Model 12D, the monopolist, also a monopsonist
in its market for labor, sets an f.o.b. wage to attract labor to work in the factory.
In Models 12B through 12D, I explore the significance of having a saturated mar-
ket area wherein farm demand for the commodity is price inelastic. The models in
this chapter build on Chapters 2 and 11 in that they introduce substitution between
a locally produced good and the factory good. This allows us to better understand
the role of market saturation and its impact on decisions by the monopolist regard-
ing wage and price as the regional economy grows. Unlike Chapter 11, this chapter
allows us to see how prices and localization get jointly determined in a regional
economy when the firm is able to exploit the monopoly advantage created by a
geography.
To me, Model 11E is intriguing because it links—for the first time in this book—
prices in the commodity, land, and labor markets. At the same time, there are
important respects in which it is also wanting. Model 11E assumes soap production
occurs solely in the city. Why is this? After all, this might seem surprising given
that model assumes constant returns to scale in soap production. If no economies of
scale, why not have each farm produce its own soap as well as wheat? That way,
the farm could save on shipping costs. Presumably the answer is, although the city
has constant returns to scale in soap, it is not possible to replicate that low unit cost
1 This is the same argument made in Nerlove and Sadka (1991, p. 100) that I quoted in Chapter 1.
12.2 Local Production in the Farm Economy 337
2 In that sense, I treat the farm as equivalent to the region in Model 11E: a fixed amount of labor.
An alternative model is developed in Curry (1984) wherein it is assumed that farms may each have
a given, but different, amount of labor. Such differences introduce the possibility of trade among
farms—an outcome that I ignore here by assuming all farms have the same amount of labor.
3 I ignore here problems that arise when decision-making unit (the farm here) is composed of more
than one individual. Where individuals differ in their preferences, it is no longer necessarily true
that their preferences can be well-ordered in the sense required in Models 12A or 12B.
4 The algebraically inclined reader might ask whether it is possible that z might be negative; that is,
the farm sells soap rather than purchases it. I rule out that possibility here: implicitly I assume that
338 12 Local Production and Consumption
Utility of farm
u = qν2 q11 − ν 0 < ν < 1 (12.1.1)
Notes: Rationale for localization (see Appendix A): Z3—Implicit unit cost advantage at some
locales; Z8—Limitations of shipping cost. Givens (parameter or exogenous): b—Scalar in wheat
production; c—Marginal product of labor in on-farm soap production; h—Total labor available to
the farm; r—Marginal cost of commuting; s—Unit shipping rate; x—Distance from factory; β—
Exponent of labor in wheat production; γ —Exponent in land in wheat production; ν—Preference
for soap. Outcomes (endogenous): h1 —Labor allocated by farm to wheat production; h2 —Labor
allocated by farm to soap production; h3 —Labor allocated by farm to factory work; L—Land
used by farm in wheat production; P[x]—Effective price of factory soap at x; Pb —F.o.b. price
of soap (exogenous in Model 12B); Q1 —Amount of wheat produced by farm; q1 —Consumption
of wheat by farm; q2 —Consumption of soap by farm; R[x]—Market rent per unit land at x; u—
Utility of farm; ν 1 —Production of wheat by farm net of land rent; ν 2 —Production of soap by
farm; wb —Factory wage (Model 12D only); —Amount of soap purchased from factory.
In allocating labor among activities, the farm finds that each additional unit of
labor allocated to wheat production earns a smaller additional return: a consequence
of the diminishing returns assumed in (12.1.2). The implication of this argument, in
reverse, is that the farm would never allow h1 to drop to zero; for a sufficiently small,
but nonzero, h1 , we can make the marginal productivity of labor in wheat production
as high as is needed by altering the amount of land used in production. Therefore, at
least a small h1 is always more productive than labor allocated to soap production on
the farm. In contrast, each additional unit of labor allocated to farm soap production
earns the same marginal return, c: see (12.1.4). Since I treat wheat as numéraire
here, the marginal value product of labor in soap production is cP[x]. The farm at
distance x therefore allocates labor to wheat production until the marginal product
there falls to cP[x].
whatever price at which the factory sells its soap, it is less costly than producing soap on the farm.
If this were not true, why have a factory at all?
12.3 Model 12A: Farm in Autarky 339
Autarky
z∗ = 0 (12.2.1)
R∗ = Ra (12.2.2)
h∗3 = 0 (12.2.3)
Land use (from maximizing farm’s net production of wheat)
β / (1 − γ )
L = (γ b / Ra )1 / (1 − γ ) h1 (12.2.4)
β / (1 − γ )
v1 = (1 − γ )b1 / (1 − γ ) (γ / Ra )γ / (1 − γ ) h1 (12.2.5)
v2 = ch − cb − 1 / β γ γ /β (1 − γ ) − (1 − γ ) / β Rγ / β v(1 − γ ) / β (12.2.6)
a 1
Notes: See also Table 12.1. Rationale for localization (see Appendix A): NA—No localization for
farm in autarky. Givens (parameter or exogenous): b—Scalar in wheat production; c—Marginal
product of labor in on-farm soap production; h—Total labor available to the farm; Ra —Opportunity
cost of land; s—Unit shipping rate; β—Exponent of labor in wheat production; γ —Exponent in
land in wheat production; ν—Preference for soap. Outcomes (endogenous): h1 ∗ —Labor allocated
by farm to wheat production; h2 ∗ —Labor allocated by farm to soap production; h3 ∗ —Labor allo-
cated by farm to factory work; k1 —A constant; L∗ —Land used by farm in wheat production;
p∗—Effective price of soap (units of wheat per unit soap); Pb —F.o.b. price set by factory; Q1 ∗ —
Amount of wheat produced by farm; q1 ∗ —Consumption of wheat by farm; q2 ∗ —Consumption of
soap by farm; R∗ —Rent paid per unit land; u∗ —Utility of farm; v1 ∗ —Production of wheat by farm
net of land rent: also amount of wheat consumed by farm; v2 ∗ —Production of soap by farm: also
amount of soap consumed by farm; X∗ —Distance from factory beyond which farm is in autarky;
z∗ —Amount of soap farm purchases from factory.
340 12 Local Production and Consumption
First, consider the use of land. Given an amount of labor in wheat production,
there is a quantity of land input that maximizes production of wheat net of land rent
in (12.1.3); the partial solution for L (“partial” because we have not yet solved for
the amount of labor to be employed in wheat production) is shown in (12.2.4) and
for v1 in (12.2.5). Because of diminishing returns in wheat production (β < 1 − γ ),
the farm’s net production of wheat rises less than proportionally with the allocation
of labor (h1 ). By substituting (12.2.4), and (12.1.4) into (12.1.6), we get the farm’s
production possibility frontier in autarky, which is shown in (12.2.6). It is convex
with respect to the origin in (v1 , v2 ) space: see curve AB in Fig. 12.1.
Fig. 12.1 Model 12A: Model 12A: Production and consumption in autarky
production and consumption: AB Farm’s production possibility frontier
CD Indifference curve
farm in autarky. E Utility-maximizing production and consumption combination
Notes: ν = 0.4, b = 1.5, HI
OF
Implicit price ratio
Wheat produced (net of rent); also wheat consumed
β = 0.6, γ = 0.05, H OG Soap produced; also soap consumed
A
c = 0.001, h = 4,000,
C
Ra = 100. Here, q1 = 116.6
and q2 = 2.05. Outcomes:
h1 ∗ = 1,946, h2 ∗ = 2,054,
L∗ = 0.06, Q1 ∗ = 122.73,
F E
v1 ∗ = 116.6, ν2 ∗ = 2.0,
Wheat (net of land rent)
u∗ = 23.18, p∗ = 37.84.
Horizontal axis scaled from 0
to 6; vertical from 0 to 250 D
G Soap B I
Now, consider the allocation of labor. How does the farm allocate labor in autarky
in order to reach point E in Fig. 12.1? Log-linear utility functions instance homoth-
etic preferences; as its income is varied, the farm varies its consumption of each of
the two commodities proportionally. Put differently, the log-linear utility function
implies that each commodity has an income elasticity of demand of 1. The farm’s
income corresponds to the total amount of labor (h) at its disposal. In autarky, it
turns this labor into (1) soap at constant returns to scale or (2) wheat at diminish-
ing returns to scale. In the absence of diminishing returns, the farm would therefore
allocate its labor to wheat and soap production in the proportions ν and 1 – ν, respec-
tively. However, with diminishing returns to wheat production, a utility-maximizing
farm instead allocates a proportion, k1 , of h (its income here) for wheat with the
remainder going to farm soap production. See (12.2.7) and (12.2.8). The propor-
tion k1 varies directly with the returns to scale in wheat production (β and γ ) and
inversely with ν. The closer to constant returns to scale, the more attractive it is
to put labor in wheat production; the upper limit on k1 —as we approach constant
returns to scale in wheat production—is 1 – ν.
12.3 Model 12A: Farm in Autarky 341
By setting h∗1 and h∗2 in this way, and then setting L∗ according to (12.2.4), the
farm then gets net productions, ν1∗ and ν2∗ , that maximize utility. This happens at the
point on the farm’s highest indifference curve: the one just tangent to the production
possibility frontier. See indifference curve CD in Fig. 12.1. We can solve for all the
other variables by back substitution starting from (12.2.7), (12.2.8), and (12.2.9).
The level of utility, u∗ , that corresponds to curve CD is given in (12.2.13). The farm
then produces the combination (ν1∗ , ν2∗ ) at point E along AB—given by (12.2.12)
and (12.2.14), respectively—and then consumes those same quantities: q∗1 = ν1∗
and q∗2 = ν2∗ . These amounts are OF and OG, respectively in Fig. 12.1.
Even though the farm in autarky is not exchanging wheat for soap with the fac-
tory, it still behaves as though there were an exchange rate (i.e., a ratio of prices) of
wheat for soap. The negative of the slope of the line passing through E, and tangent
to both curve AB and curve CD, is the implicit exchange rate, p∗ : see (12.2.15). The
implicit exchange rate—which corresponds to line HI in Fig. 12.1—shows us the
tradeoff at the margin (point E in Fig. 12.1): how much wheat (the numéraire good)
the farm would have to give up to get one more unit of soap. From this, we can now
calculate how far the farm must be from the factory to make the purchase of factory
soap unattractive.5 At the boundary of the factory’s market area, the effective price
of soap, (12.1.5), must just be equal to the implicit exchange rate. This implies that
the maximum distance a farm can be from the factory is given by (12.2.16). Beyond
X∗ , the farm is in autarky in that z∗ = 0.
What about comparative statics in Model 12A? See Table 12.3.
b When b is increased, the farm can produce more wheat from the same com-
bination of labor and land. Now relatively scarce, soap becomes more costly.
The amount of land used for wheat production increases as does net output.
Overall, the farm becomes better off.
b 0 0 + + + + 0 +
c 0 0 – 0 0 0 + +
h + + – + + + + +
Ra 0 0 – – – – 0 –
β + – ? ? ? ? – ?
γ + – ? ? ? ? – ?
ν – + + – – – + ?
5 For the moment, I leave aside complications that arise when the farm is part of the factory’s labor
market area. These are discussed below in Model 12D.
342 12 Local Production and Consumption
c When c is increased, the farm can produce more soap from the same amount
of labor. Now being relatively scarce, wheat becomes more costly. More
soap is produced in total. Overall, the farm becomes better off.
h When h is increased, the farm can produce more wheat and more soap. The
farm allocates more labor to each. As land now becomes relatively scarce,
wheat becomes more valuable. The amount of land used for wheat produc-
tion increases as does net output. More soap is produced in total. Overall,
the farm becomes better off.
Ra When Ra is increased, land becomes a relatively more costly input. As land
is now relatively more expensive, wheat becomes more costly. The amount
of land used for wheat production declines as less wheat is produced on net.
β When β is increased, labor becomes more important in the production of
wheat. The farm reallocates labor from wheat to soap production.
γ When γ is increased, land becomes more important in the production of
wheat. The farm reallocates labor from wheat to soap production.
ν When ν is increased, consumers have a stronger preference for soap com-
pared to wheat. The farm reallocates labor from wheat to soap production.
Soap becomes more valuable, and its output increases. The amount of land
used for wheat production declines as less wheat is produced on net. Being
more preferred, soap is now more valuable.
In the models presented in Chapter 11, I had assumed constant returns to scale
in wheat production. There, we saw (in Fig. 11.8 for example) a linear production
possibility frontier. In contrast, Model 12A assumes diminishing returns to scale,
and this generates a production possibility frontier in Fig. 12.1 that is nonlinear.6 Is
the assumption of diminishing returns important in Chapter 12? Specifically, what
happens to Model 12A as we let β + γ approach 1: i.e., constant returns to scale?
The answer is that Curve AB in Fig. 12.1 gets pulled into the shape of a straight
line, k1 converges to 1 – ν, and curve AB and the implicit price ratio line (curve
HI in Fig. 12.1) converge. However, even in the absence of diminishing returns, it
is still possible to find a combination of wheat and soap production that leaves the
farm best off in autarky. The significance of diminishing returns becomes clearer
when we consider—as we will do in Model 12B—farms inside the market area of
the soap factory.
What about the regional economy here? In Table 12.4, I summarize the economic
characteristics of a utility-maximizing farm in autarky based on the same numerical
example used to generate Fig. 12.1.7 Using p∗ to convert amounts of soap into equiv-
alent amounts of wheat, the farm in autarky has a total output equivalent to about
200 units of wheat. As part of its output, about 123 units of wheat are produced
(using 1,946 units of labor and 0.06 units of land) of which γ (here 5%) is spent on
6 More specifically, curve AB in Fig. 12.1 is concave with respect to the origin.
7 To solve Model 12A, I first use (12.2.13), (12.2.6), and (12.2.9) to find the labor allocated by
the farm to wheat (h∗1 ) and soap (h∗2 ), and (12.2.16) to find the implicit price ratio (p∗ ). Next, I
use (12.2.11) to solve for the land rented (L∗ ). Finally, I find total wheat output (Q∗1 ), net wheat
produced (ν1∗ ), own production of soap (ν2∗ ), and utility (u∗ ).
12.3 Model 12A: Farm in Autarky 343
renting land. The remaining 95% is the amount of wheat retained as income by the
farm. The farm uses its remaining labor to produce soap. Its total income—(1) net
wheat production plus (2) soap production valued in wheat—is 194 and it splits this
with share ν spent on soap consumption and share 1 – ν spent on wheat consump-
tion. The expression for k1 in (12.2.13) guarantees that these amounts consumed
correspond exactly to the net amounts produced, hence autarky.
Because here I have assumed diminishing returns to scale in wheat production, I
can derive one further result about the regional economy. From the wheat production
function, we know that the farm will allocate a proportion γ of its output of wheat to
land rents. If it were purchasing labor in a competitive market, it would also allocate
a proportion β to its wage bill; that would leave a proportion 1 − β − γ which is
the profit of the farm enterprise. Of course, the farm here is not purchasing its labor
from a market; the amount of labor on the farm is given. Nonetheless, we can divvy
farm income from wheat production into an implicit wage bill and an implicit profit
proportionally using β and 1 − β − γ . I show these amounts in Table 12.4. In the
case of soap production, there are constant returns to scale, so the implicit wage bill
would equal farm income from soap production; the implicit profit associated with
soap production is zero.
I add one final comment in light of Thünen’s concern with the distribution of
income. In Model 12A, rent (Ra ) is given. Effectively, Ra and the other givens of
this model determine the outcomes of the model, including farm consumption and
utility. If Ra were to increase, farm utility would decline. So, to rephrase Thünen’s
question, what determines whether Ra is low (farms pay low rents and have a high
344 12 Local Production and Consumption
utility) or Ra is high (farms pay a high rent and have a low utility)? As with earlier
models in this book, Model 12A is silent on this question; it tells us how much land
the farm will consume and hence how much is spent on the rent bill. However, it
says nothing about how Ra , the rent per unit land, is determined.
Now, consider the farm within the output market area of the soap factory. I leave
aside until Model 12D the possibility of the farm being in the factory’s labor mar-
ket. Within the factory’s output market area, the farm exchanges wheat for factory
soap. In autarky, the farm produced for its own needs: in Fig. 12.1, E is the both the
combination of output levels and the combination of consumption levels that maxi-
mize utility. Once inside the market area of the firm, the combination of wheat and
soap that a farm produces (say V2 in Fig. 12.2) may now differ from the combina-
tion that it consumes (say Q2 in Fig. 12.2). The farm then trades away wheat to get
the combination of wheat and soap that maximizes its utility.
Let us assume here a vast population of farms, many of them in autarky, and
wherein the market area for the soap factory is relatively small. In that way, we
can think of the supply of farms, given free entry, as infinitely elastic at the level
of utility (u∗ ) achievable in autarky. Given the f.o.b. price set by the soap factory
and the unit shipping rate, farms freely relocate until the level of utility achievable
anywhere in the market area is just equal to the utility achievable in autarky. Where
an advantage might otherwise exist at a particular location, I assume that farms bid
up the rent on land locally until that utility advantage disappears. In other words,
spatial equilibrium means that rent rises everywhere within the factory’s market
area by just enough to leave the farm there at the same level of utility as in autarky.
The implication of this is that almost every measure of interest—the amounts of
wheat and soap produced by the farm, the amounts of wheat and soap consumed,
the allocation of farm labor, the amount of land used in wheat production, and the
market rent for land will now vary by location within the factory’s market area.8
Because of that variation, I refer to the utility-maximizing solution for each variable
as a schedule by location (a gradient): e.g., R[x].
To begin, consider a farm just inside the boundary: i.e., where x is only slightly
smaller than X∗ . There are two differences here compared to the farm in autarky. For
one, the effective price of factory soap now affects the farm’s consumption bundle:
the mix of soap and wheat that maximizes its utility. For another, the lower price for
factory soap encourages the farm to reallocate labor from soap to wheat production.
Put differently, the farm finds it attractive to purchase factory soap, substitute some
farm-made soap in favor of factory soap, shift some labor from soap production
8 Another implication is that we are using utility as though it were ratio or interval-scaled rather
than ordinal.
12.4 Model 12B: Farm Purchasing Soap from the Factory 345
F E
F2 Q2
F1 Q1
F0 Q0
Wheat (net of land rent)
0
Soap G G2 G1 G0 B
to wheat production and thereby increase the amount of land that it uses to pro-
duce wheat. The extent of substitution in both production and consumption depends
on (1) the effective price of factory soap, (2) the extent of diminishing returns to
wheat production, and (3) the relative preference for soap as opposed to wheat. If,
for example, the farm had constant returns to scale in wheat production, then all
farms within the factory’s market area would give up all soap production and spe-
cialize entirely in wheat production. In the presence of diminishing returns, the farm
may well find it advantageous to continue producing soap on the farm even when it
purchases soap from the factory.
Now, consider farms still closer to the factory. Starting from X∗ and moving
closer to the factory, we observe the following. Because soap is now relatively
cheaper, farms closer to the factory consume relatively more of it. Nonetheless,
an efficient farm will produce somewhere along its production possibility frontier.
As we consider farms still closer to the factory, the effective price of factory soap
drops further and the farm is inclined to give up still more soap production despite
the diminishing returns to wheat production. In fact, we may reach a distance (I
call it X1 ) at which the farm has entirely given up its own soap production: there is
346 12 Local Production and Consumption
Table 12.5 Model 12B: farm in X1 area (purchasing from factory, but still producing some soap
of its own)
Farm’s objective
β
Max U = (bh1 Lγ − RL − Pz)1 − ν (c(h − h1 ) + z)ν (12.5.1)
Land input (from ∂U /∂L = 0)
L[x] = (γ b / R[x])1 / (1 − γ ) h1 [x]β / (1 − γ ) (12.5.2)
Output
Q[x] = b1 / (1 − γ ) h1 [x]β / .(1 − γ )) (γ / R[x])γ / (1 − γ ) (12.5.3)
V1 [x] = (1 − γ )Q[x] (12.5.4)
Soap purchases (from ∂U / ∂z = 0 )
P[x](z[x] + c(h − h1 [x])) = ν(V1 [x] + P[x]c(h − h1 [x])) (12.5.5)
Utility (after back substitution to eliminate L and z)
U[x] = (ν ν (1 − ν)1 − ν / P[x]ν ) ((1 − γ )b1 / (1 − γ )
h1 [x]β / (1 − γ ) (γ / R[x])γ / (1 − γ ) + P[x]c(h − h1 [x])) (12.5.6)
Labor input (from ∂U / ∂h1 = 0)
h1 [x] = b1 / (1 − β − γ ) (γ / R[x])γ / (1 − β − γ ) (β / (cP[x]))(1 − γ ) / (1 − β − γ ) (12.5.7)
Utility (after back substitution to eliminate h1 )
U[x] = (ν ν (1 − ν)1 − ν / P[x]ν ) (P[x]ch + (1 − b − γ )b1 / (1 − β − γ )
(γ / R[x])γ / (1 − β − γ ) (β / (cP[x]))β / (1 − β − γ ) ) (12.5.8)
Land rent that preserves U[x] = u∗
everywhere
R[x] = γ ((1 − β − γ )ν ν (1 − ν)1 − ν / (u∗ P[x]ν
− P[x]chν ν (1 − ν)1 − ν ))(1 − β − γ )/γ b1 / γ (β / (cP[x]))β / γ (12.5.9)
Optimized labor input (after back substitution to eliminate R)
h1 [x] = (β / (cP[x]))(u∗ P[x]ν − P[x]chν ν (1 − ν)1 − ν ) /
((1 − β − γ )ν ν (1 − ν)1 − ν ) (12.5.10)
Effective price at which h1 [x] = h
P[x] = ((βu∗ ) / (chν ν (1 − ν)1 − ν ))1 / (1 − ν) (12.5.11)
Inner radius of X1 area (if X2 area exists) under f.o.b. pricing
X1 = (((βu∗ ) / (chν ν (1 − ν)1 − ν ))1 / (1 − ν) − Pb ) / s (12.5.11)
Notes: See also Table 12.1. Rationale for localization (see Appendix A): Z2—Implicit unit cost
advantage at some locales; Z8—Limitations of shipping cost. Givens (parameter or exogenous):
b—Scalar in wheat production; c—Marginal product of labor in on-farm soap production; h—Total
labor available to the farm; Pb —F.o.b. price for soap set by factory; Ra —Opportunity cost of land;
s—Unit shipping rate; x—Distance from factory; β—Exponent of labor in wheat production; γ —
Exponent in land in wheat production; ν—Preference for soap. Outcomes (endogenous): g1 [x]—A
constant at x; h1 [x]—Farm labor in wheat production at x; h2 [x]—Farm labor in soap production
at x; h3 [x]—Farm labor in factory work at x; k2 —A constant; L[x]—Land used by farm in wheat
production at x; P[x]—Effective price of factory soap at x; q1 [x]—Consumption of wheat on farm
at x; Q1 [x]—Production of wheat on farm at x; q2 [x]—Consumption of soap on farm at x; R[x]—
Bid rent per unit land at x; u[x]—Utility of farm at x; v1 [x]—Production of wheat (net of rent) on
farm at x; v2 [x]—Production of soap on farm at x; X∗—Radius beyond which farm is in autarky;
X1 —Radius within which farm produces only wheat; z[x]—Quantity of factory soap purchased by
farm at x.
12.4 Model 12B: Farm Purchasing Soap from the Factory 347
Table 12.6 Model 12B: farm in X2 area (producing only wheat; all soap purchased from factory)
Farm’s objective
β
Max U = (bh1 Lγ − RL − Pz)1 − ν (c(h − h1 ) + z)ν (12.6.1)
Land input (from ∂U / ∂L = 0)
L[x] = (γ b / R[x])1/(1 − γ ) hβ / (1 − γ ) (12.6.2)
Output
Q[x] = b1 / (1 − γ hβ / (1 − γ )) (γ / R[x])γ / (1 − γ ) (12.6.3)
V1 [x] = (1 − γ )Q[x] (12.6.4)
Soap purchases (from ∂U / ∂z = 0)
P[x]z[x] = νV1 [x] (12.6.5)
Utility (after back substitution to eliminate L and z)
U[x] = (ν ν (1 − ν)1 − ν / P[x]ν )(1 − γ )b1/(1 − γ ) hβ / (1 − γ ) (γ / R[x])γ / (1 − γ ) (12.6.6)
Land rent that preserves U[x] = u∗
everywhere
R[x] = g((1 − γ )ν ν (1 − ν)1 − ν / (u∗ P[x]ν ))(1 − γ ) / γ b1 / γ hβ / γ (12.6.7)
Figure 12.1 was useful to us in thinking about the farm in autarky; a simi-
lar diagram is helpful in thinking also about a farm within the factory’s market
area. Consider a farm’s production possibility frontier: curve AB from Fig. 12.1
now reproduced in Fig. 12.2. The quantity OB is the amount of soap (ν2 − ch)
produced when all farm labor is used to produce soap alone. OB is not affected
by land rent since soap production is assumed to require no land. Put differ-
ently, if we consider a site for the farm closer to the factory where rent is
higher, point B on the curve will not move. The quantity OA is the amount of
wheat (ν1 = b1/(1 − γ ) (yγ /(1 − γ ) − y1/(1 − γ ) )R − γ /(1 − γ ) hβ/(1 − γ ) ) produced when
all farm labor is used to produce wheat alone. OA is net of land rent; OA will be
smaller where land rent is higher. Put differently, as we consider farm sites within
the factory’s market area, rent per square kilometer will be higher the closer the farm
348 12 Local Production and Consumption
is to the market in a way that offsets the advantage of being able to purchase soap
for a lower unit shipping cost. The effect of competition among farms—in push-
ing up rents in the market for land—is to twist the production possibility frontier
counterclockwise about point B on Fig. 12.2. The resulting production possibility
frontiers, e.g., AB, A2 B, A1 B, and A0 B in Fig. 12.2, must be such as to leave farms
at the respective sites indifferent among them.
Consider a location at distance x somewhere in the X1 market area. Since the rent
here is higher than Ra , the production possibility frontier twists counterclockwise
around point B on Fig. 12.2. Suppose that it forms curve A2 B. The exchange rate is
the effective price: Pb + sx. Imagine a straight line with a slope of − (Pb + sx);
a so-called terms of trade line. To maximize utility for a farm, the terms of trade
line must be tangent both to (1) the production possibility frontier (A2 B) at the
production combination (point V2 in Fig. 12.2) and (2) the indifference curve CD
at the consumption combination (point Q2 in Fig. 12.2). Put differently, if we put
a terms of trade line at its point of tangency to the indifference curve, the rent per
square kilometer must be such as to make the production possibility frontier reach
the terms of trade line at just one point: either as a tangent (as in at point V2 on
A2 B) or as a corner solution (as at A1 on A1 B or at A0 on A0 B). In each case,
the terms of trade line guarantees that (1) the amount of wheat given up is just the
amount needed to purchase the factory soap and (2) no other pair of production and
consumption combinations allows the farm to reach a higher level of utility. As an
example, suppose the farm at x = 0 in Fig. 12.2 produces 0A0 units of wheat net
of rent and no soap (i.e., is in the X2 area), trades F0 A0 units of wheat (including
shipping cost) to get F0 Q0 units of soap, and thereby consumes 0F0 units of wheat
and 0G0 units of soap. In general, as we consider farms closer to the market, the
locus of consumption combinations is traced out by the part of the indifference curve
below and to the right of point E. Similarly, the locus of net production combinations
is traced out initially by the curve EV2 A1 as we pass through the X1 market area
and then curve A1 A0 as we pass through the X2 area.
To solve Model 12B for a farm in the X1 market area, I first find the amount
of land used by a farm at distance x from the factory. The partial solution for
L—“partial” again because we have not yet solved for the amount of labor to be
employed in wheat production—is shown in (12.5.2) and for v1 in (12.5.4). Next,
I consider the farm’s allocation of labor. For the farm at distance x, each unit of
labor allocated to soap production has a marginal value product of cP[x]. The farm
therefore allocates labor to wheat production—with its diminishing returns—until
the marginal value product there drops to cP[x]; any remaining labor is allocated
to soap production. In terms of wheat production, the labor allocation is (12.5.7).
Finally, I consider the amount of factory soap purchased by the farm. Because I
have assumed a log-linear utility function, the farm allocates shares ν and 1 − ν
of its total net income, ν1 + P[x]c(h − h1 ), to consumption of wheat and soap,
respectively: see (12.5.5). I can then determine the utility of a farm at distance x:
see (12.5.8). Under the assumption of spatial equilibrium, I can then reverse this to
derive an expression for R[x] at each distance from the factory that gives the same
level of utility. See (12.5.9).
12.4 Model 12B: Farm Purchasing Soap from the Factory 349
9 This may seem surprising to readers who are algebraically inclined. After all, there is a potentially
important difference between the rent gradients in the X2 and X1 areas: i.e., between (12.6.7) and
(12.5.9). In (12.6.7), the denominator of R[x] is u∗ P[x]ν. In (12.5.9), the denominator includes
the factor u∗ P[x]ν − P[x]ch ν ν (1 − ν)1 − ν . Is it possible in (12.5.9), unlike in (12.6.7), for the
350 12 Local Production and Consumption
D B
F C
0
E Distance from factory G
What happens to the intensity of land use in wheat production (h1 /L)? Because
it is more costly to rent land nearer the factory, we might expect that the farm finds
it profitable to use land more intensively there. The production function (12.1.2)
allows for substitution between land and labor in production of wheat. As noted
above, we see the rent for a unit of land higher for a farm nearer the factory com-
pared to a farm near the boundary X∗ . Therefore, we might expect substitution. This
is exactly what happens in the X1 and X2 market as illustrated by the schedules
for h1 /L in Fig. 12.4. In the X1 area, h1 /L rises as we move closer to the factory:
the farm has labor being freed up from soap production that it can now apply to
wheat production. In the X2 area, h1 /L continues to rise. What is perhaps surpris-
ing, again here, is the way in which h1 /L rises consistently as we cross over from
the X1 to the X2 market areas. In the X2 area, farms here do not have any on-farm
denominator factor to go to zero? If so, it would create a discontinuity in R[x] that would make it
different from (12.5.9). The answer here is that, the denominator factor in (12.5.9) always drops as
we move further away from the factory: albeit more slowly than does the denominator in (12.6.7).
Put differently, farms that substitute (i.e., where h1 [x] < h) are able to outbid a farm that opts not to
substitute (i.e., h1 [x] = h) everywhere in the X1 area. Nonetheless, rents decline with increasing
distance everywhere in the X1 area, and the market boundary is reached at R[X ∗ ] = Ra .
12.4 Model 12B: Farm Purchasing Soap from the Factory 351
Fig. 12.4 Model 12B: labor Model 12B; Labor intensity and locaton
β = 0.6, γ = 0.05, OD
OE
Labor intensity at x = X1
Width of the X2 market area
c = 0.001, h = 4,000, OF Labor intensity at x = X*
Ra = 100, Pb = 20,
s = 0.10. Outcomes are
X ∗ = 178.43, X1 = 66.80,
h1 /L = 31,710 at x = X ∗ ;
h1 /L = 222,180 at
x = X1· h1 /L = 2,228,206
at x = 0. Horizontal axis
scaled from 0 to 200; vertical
from 0 to 2,500,000
D B
F C
0
E G
Distance from factory
soap production from which to find additional labor. The argument is the same as
in Fig. 12.3. As we approach the boundary X1 heading for the factory, the farm is
setting its own soap production ever closer to zero so that at the boundary (X1 ), there
is no discontinuity in the rate at which h1 /L changes.
Finally, what happens to the density of farms (or, its reciprocal, the land area
occupied by a farm) across the factory’s market area? Fig. 12.5 illustrates an out-
come possible here. Again, consider first a farm at X∗ and then examine farms that
are closer to the factory. Just inside the X1 area, as we move closer to the factory,
lot size L initially increases. This might seem strange because we argued above that
labor intensity, h1 /L, was rising with the increase in land rent as we move toward
the factory. However, because h1 increases quickly as the farm gets out of soap
production just inside X∗ , the farm is best off there to increase L. Once we move
from further into the X1 area and then into the X2 area, we see the more conven-
tional result, that lot size shrinks as we move toward the factory because of higher
rent. This is not surprising since the possibility of further substitution among uses
of labor is decreasing (in the X1 area) or zero (in the X2 area) as we consider farms
still closer to the factory.
What about comparative statics here? See Table 12.7. A change in given typically
affects the boundaries of the X1 and X2 areas: i.e., X1 and X∗ . It may also affect the
value of an outcome at x = 0, x = X1 , or x = X ∗ .
b When b is increased, the farm produces more wheat from the same combina-
tion of labor and land. Now relatively scarce, soap becomes more valuable.
The outer boundary (X∗ ) and inner boundary (X1 ) both increase. The amount
of land used by a farm for wheat production increases as does net output. Land
rents rise but only in the X2 area.
352 12 Local Production and Consumption
Fig. 12.5 Model 12B: land Model 12B: land area and location
Ra = 100, Pb = 20, C
D
s = 0.10. Outcomes are
X ∗ = 178.43, X1 = 66.80,
L[X ∗ ] = 0.0614;
F B
A
0
E Distance from factory G
c When c is increased, the farm produces more soap from the same amount
of labor. Factory soap is now less competitive. The outer boundary (X∗ )
and inner boundary (X1 ) both decrease. More soap is produced on-farm in
total. The attractiveness of proximity to the factory is reduced, so land rent
declines.
h When h is increased, the farm produces more wheat and more soap. The outer
boundary (X∗ ) and inner boundary (X1 ) both decrease. The farm allocates
more labor to each activity. The amount of land used for wheat production
increases as does net output. The attractiveness of proximity to the factory is
reduced, so land rent declines.
Pb When Pb is increased, soap purchases become more costly to the farm. The
outer boundary (X∗ ) and inner boundary (X1 ) both decrease.
Ra When Ra is increased, land becomes a relatively more costly input. The outer
boundary (X∗ ) and inner boundary (X1 ) both decrease. As land is now rela-
tively more expensive, wheat becomes more costly. The amount of land used
for wheat production declines and less wheat is produced on net. Land rents
rise everywhere in the factory’s market area.
s When s is increased, soap purchases become more costly to the remote farm.
The outer boundary (X∗ ) and inner boundary (X1 ) both decrease. Rent at both
distances remains unchanged.
β When β is increased, labor is more important in the production of wheat.
Farms reallocate labor into wheat production because of its improved yield.
12.4
0 X1 X∗ 0 X1 X∗ 0 X1 X∗ 0 X1 X∗ 0 X1 X∗ 0 X1 X∗ X1 X∗
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20]
b . . 0 − + + + + + + + + . . 0 + 0 . + +
c . . 0 + 0 0 + 0 0 + 0 0 . . + − 0 . − −
h + + + + + + + + + + + + . . + − 0 . − −
Pb . . 0 + 0 0 + 0 0 + 0 0 . . 0 − 0 . − −
Ra . . 0 − − − − − − − − − . . 0 + + + − −
s . . 0 0 0 0 0 0 0 0 0 0 . . 0 0 0 0 − −
Model 12B: Farm Purchasing Soap from the Factory
β . . + − + + + + + + + + . . − + − . + +
γ . . + + + + − − − − − − . . 0 − − . − −
ν 0 0 − − − − − − − − − − . . + + + . − +
Note: Effects estimated numerically by increasing each given in turn compared to values for givens shown in Table 12.8; +, Effect on outcome of change
in given is positive; –, Effect on outcome of change in given is negative; 0, Change in given has no effect on outcome; ., Not applicable.
353
354 12 Local Production and Consumption
Now relatively scarce, soap becomes more valuable. The outer boundary (X∗ )
and inner boundary (X1 ) both increase.
γ When γ is increased, land is more important in the production of wheat. As
the farm is less able to substitute labor for land, the outer boundary (X∗ )
and inner boundary (X1 ) both decrease. The farm produces less wheat. The
attractiveness of proximity to the factory is reduced, so land rent declines.
ν When ν is increased, consumers have a stronger preference for soap compared
to wheat. As soap is now more valuable, farms are willing to purchase factory
soap from further away: the outer boundary (X∗ ) increases. At the same time,
they are reallocating farm labor in favor of soap production, so farm soap
production increases, and the inner boundary (X1 ) decreases. Less land is used
for wheat production, less wheat is produced both gross and net of rent. Land
rent rises because proximity to the factory is now more valuable.
What about the regional economy here? Panel (b) of Table 12.8 summarizes
examples of a typical farm within the factory’s market area. This example uses the
same numerical values for the givens as we did for Model 12A in Table 12.4: plus
values for the two new givens: Pb and s. The summary values in panel (b) include
farms in both the X1 and X2 market areas. To help us better understand these sum-
mary data, I present in panel (a) a breakdown of key measures by location within
the factory’s market area. Because of the assumptions in common to Models 12A
and 12B, there are similarities in outcome between farms in autarky and farms in
the factory’s trading area evidenced by a comparison of Tables 12.4 and 12.8. By
assumption, the utility of a farm is the same in both. Both models also assume that
the farm (1) splits its income 60–40 between wheat and soap consumption and (2)
expends 5% of its wheat production on land rent. What then are the differences?
Compared to the farm in autarky, the average farm within the factory’s market area
produces more wheat (161 vs. 123) and less soap (0.54 vs. 2.05), consumes less
wheat (103 vs. 116) and more soap (2.51 vs. 2.05), occupies less land (0.03 vs. 0.06),
pays a higher rent bill (8 vs. 6), and puts more labor into wheat production (3,460
vs. 1,946). In addition, the two models differ in terms of leakages from the regional
economy. In Model 12A, the only leakage is the rent bill; the farm in autarky—
from Table 12.4—pays 6 units of wheat to an absentee landlord. In Table 12.8,
the typical farm within the factory’s market area pays 8 in rent plus another 11 on
shipping.
factory commodity, Model 12B, allows for economic life outside the factory’s mar-
ket area. Although Model 12B does have farm incomes that rise steadily out to the
boundary of the factory’s market area as in Model 11E, Model 12B envisages a
transition from a farm just inside the factory’s market area to a farm in autarky.
Let me digress for a moment on a related thought. As a student in Economics,
I remember professors explaining the notion of a compensated demand curve. The
idea was simple enough. Since price varies (by definition) along an ordinary demand
curve, the real income (well-being) of a consumer is also varying. Imagine instead
a demand curve in which, at every price, the income of the consumer is adjusted
to leave them equally well off at that price compared to another price. As a stu-
dent, I thought this an interesting theoretical concept that illustrated the important
distinction between income and price substitution effects but without a counter-
part in the real world. However, let us take a sample of farms at various distances
and record data on the quantity of factory soap purchased and the effective price
paid. We observe farms further from the factory paying a higher effective price.
A graph of this schedule of data as a curve constitutes a compensated demand
curve. At every distance from the factory, land rent adjusts just enough to ensure
that the farm achieves the same level of utility. In spatial equilibrium, what we
observe across the landscape is therefore a compensated demand curve for factory
soap.
Finally, note that the firm in 12D is a bilateral monopolist. It faces a downward-
sloped demand curve for its output and an upward-sloped supply curve for its input.
Where a monopolist engages in mill pricing of its output and/or labor input,
it quickly becomes interested in the price elasticity of demand of any consumer
or the price elasticity of its labor supply. These are important in determining the
most profitable price and wage. In Models 12C and 12D, I will show that these two
elasticities vary with distance from the firm. This makes the aggregate elasticities
facing the firm dependent on the nature and extent of its labor supply and output
demand markets: i.e., on the localization of farms. Let me put the same argument
differently. Implicit in Model 11E is the idea that a wage, wb , is set to attract the
farm labor needed to produce soap in the city. This is an idea explored further in
Model 12D. The implication of this is that the firm faces an upward-sloped supply
curve for labor; it can get more workers in its labor supply area only if it pays a
higher wage. Congestion arises because, to attract still more labor, the firm must
pay more to (1) attract workers from further away and (2) induce farms that already
supply labor to provide still more. Put differently, congestion here is a lower price
elasticity of labor supply.
In Models 12C and 12D, I continue to assume a rectangular plane wherein dis-
tance can be measured using a straight line. This plane is taken to be boundless; we
never have to worry about edge effects. Land on this plane is everywhere identical
in terms of yield. Suppose that this land is initially occupied by identical farms as
in Model 12A, each producing and consuming in autarky. There is no other use for
this land. I continue to assume that the supply of farms is infinitely elastic at the
autarky utility level, u∗ ; if anything were to happen that might cause utility to rise
even momentarily, the concept of free entry implies new farms would flood into the
region, pushing up land rents, until utility drops back to u∗ .
Using comparative statics, Models 12C and 12D also allow us to think about
how a spatial economy might change over time as a result of capital investment or
improvements in technology. This is an underdeveloped area of competitive loca-
tion theory. There have been efforts made, for example, to estimate the effects
of transportation investments—by reducing the unit cost of shipping—on spatial
patterns of agricultural production.10 In Models 12C and 12D, I focus on how the
firm might respond to an improvement in productivity. I do this as an exercise in
comparative statics. I make the assumption that the farm’s ability to produce wheat
and soap remain the same over time but that the factory produces soap more and
more efficiently (imagining here that productivity improves as time goes on). In
part, this might be because the factory is accumulating capital that it reinvests in the
production process. In part, it might also be because the firm’s ability to apply or
innovate technology is improving. However, consistent with my comparative statics
10 Day and Tinney (1969) extend a Thünen model by introducing the need for working capital
which the farmer can accumulate over time. Also in regard to a Thünen model. Okabe and Kume
(1983) employ a kind of cobweb cycle model to study how market prices for crops in the current
season affect the farm’s decision of what crop to plant the following year. Fujita and Krugman
(1995) look at how to model the growth and form of urban and rural areas jointly over time.
358 12 Local Production and Consumption
approach, I make the assumption that the firm cannot or does not anticipate these
changes in productivity.
Demand for
factory soap (if no X2 area)
Q = 2π x∈X1 (xz[x]/L[x])dx (12.9.5)
Notes: For inner boundary to X1 area when h1 [X1 ] = h, see (12.5.11). See Table 12.5 and Table
12.6. Rationale for localization (see Appendix A): Z2—Implicit unit cost advantage at some
locales; Z8—Limitations of shipping cost. Givens (parameter or exogenous): C—Marginal cost
of factory soap production; Co —Fixed cost of factory soap production; x—Distance from fac-
tory. Outcomes (endogenous): L[x]—Land used per farm in wheat production at x; Pb —F.o.b.
price for soap set by factory; P[x]—For factory, price for soap at farm at x that maximizes profit;
Q—Factory output of soap; Z—Profit of factory; z[x]—Quantity of factory soap purchased by a
farm at x.
12.7 Model 12C: The Factory as Monopolist Using f.o.b. Pricing 359
12D. Implicitly, Model 12C assumes all inputs to factory soap production, including
labor, have horizontal supply curves. In effect, the factory can purchase as much or
as little of each input, and the unit cost will always stay the same. Again, this is as
in Chapter 2.
As in Model 12B, I allow for the possibility of concentric ring areas around the
factory. As in Model 12B, an X2 area where farms allocate labor only to wheat
production and an outer X1 area where farms allocate labor to wheat and farm soap
production are possible. The X2 area may not exist in all cases; I include it here for
comprehensiveness. Demand for factory soap is therefore given by either (12.9.4)
or (12.9.5) depending on the presence of an X2 area. The amount demanded by any
one farm is given by (12.9.1) or (12.9.2); see the kinked curve ABC in Fig. 12.6
of which segments AB and BC are the demand schedules for factory soap by a
farm in the X1 area and X2 area, respectively. Note that (12.9.1) and (12.9.2) are
compensated demand curves; they already incorporate the adjustment in land rent
that makes consumption of factory soap at each effective price give the same level
of utility.
Now consider the introduction of the soap factory into this landscape. For the
soap factory to be profitable, several conditions must be met. First, in this barter
economy, the factory’s f.o.b. price (Pb ) must be below p∗ ; otherwise it would not
D B
Effective price for soap to farm
0
E Soap purchased (z)
Fig. 12.6 Models 12B and 12C: demand by farm (z) for factory soap.
Notes: ν = 0.2, b = 900, β = 0.2, γ = 0.6, c = 0.16, h = 4,000, Ra = 9,600, s = 0.02,
C0 = 0, and C = 0.03. Outcomes are p∗ = 0.2906 X ∗ = 9.64 in 12B and 13.03 in 12C.
Horizontal axis scaled from 0 to 600; vertical from 0 to 0.35
360 12 Local Production and Consumption
get any customers. Second, the factory’s average cost of production (C + C0 /Q)
must be sufficiently low to make production profitable.
Assume the following hypothetical situation without a factory. Each farm is in
autarky. The period of time is a year. The preference—budget share—for soap (ν)
is 0.20. The scale parameter for wheat production (b) is 900. The returns on labor
and land in wheat production (β and γ ) are 0.2 and 0.6, respectively. The marginal
productivity of labor in farm soap production (c) is 0.16 blocks of soap per person-
hour of labor. The unit shipping rate (s) is 0.02 per unit of wheat transported 1 km.
The total amount of labor available to the farm is 4,000 person-hours per year, and
the annual opportunity cost of land (Ra ) is 9,600 m3 of wheat per square kilometer
for the year. In this case, we can calculate the outcome for the farm in autarky from
Table 12.2. From v, we know that the farm would like to allocate its 4,000 person-
hours annually into 3,200 person-hours for wheat production and 800 person-hours
for farm soap production. However, because of diminishing marginal productivity
in wheat production, the farm instead allocates only 2,667 person-hours to wheat
production (h∗1 ) and the rest (1,333 person-hours) to soap (h∗2 ). This allows the farm
to produce 213 blocks of soap (v∗2 ); in autarky, this is also the amount of soap con-
sumed (q∗2 ) by the farm. The farm rents 0.0388 km2 of land and is thereby able
to produce 620 units of wheat annually (Q∗1 ). The land rent bill (Ra L∗ ) totals 372
units of wheat, leaving a net output (v∗1 ) of at most 248 units of wheat; in autarky,
this is also the amount of wheat consumed by the farm. The utility level achieved
by the farm (u∗ ) is 241. For farms in autarky, the implicit exchange rate (p∗ ) is
0.2906. At the utility maximum, the farm is giving up 0.2906 m3 in potential wheat
consumption for the last block of soap consumed.
Assume now that a soap factory is placed somewhere on this rectangular plane.
This factory occupies no space. This factory sells (exchanges) its product to farms
(in return for wheat). Shipments arise because of the flows of wheat sold to the fac-
tory and soap purchased there. These movements incur shipping costs, but I assume
these movements require no geographic space. In other words, even in the presence
of the factory, all land on the rectangular plane is used by farms and only for wheat
production.
I assume initially here that the factory sells soap at an f.o.b. price (exchange rate),
Pb . Then, I consider the case where a monopolist can improve profit using discrim-
inatory pricing. Suppose the factory has C = 0.27, which is modestly smaller
than the autarky-implicit exchange rate (p∗ ) in Table 12.4. Suppose we also assume
C0 = 0 so that we make soap production in the factory much like soap production
on the farm. Both are constant returns to scale; since C < p∗ here, factory produc-
tion is more efficient. In this case, what Pb will earn the factory its maximum profit?
More directly, how do we solve here for the exchange rate (Pb ) that maximizes fac-
tory profit? Our problem here is that neither equation (12.9.4) nor equation (12.9.5)
can be reduced algebraically to gives an explicit solution for Pb . We must therefore
rely on numerical approximations to the integral expressions in them (once again, I
use the midpoint rule version of a Riemann sum11 ). We also have to use numerical
methods to find the solution; one approach is to iterate to find the Pb that makes the
profit earned by the factory as large as possible
Doing this, I find the profit-maximizing factory will set an f.o.b. exchange rate
(Pb ) of 0.2752 units of wheat per unit of soap. Appropriately, this is below the
implicit exchange rate (0.2906) for farms in autarky. Under profit maximization, the
radius of the factory’s output market is a mere 0.77 km and, there is no X2 area here.
The total amount demanded from the factory is 881 units of soap.
In terms of comparative statics, what would happen were C in fact smaller? In
panel (a) of Table 12.10, I show the profit-maximizing exchange rate for a factory
with f.o.b. pricing for values of C ranging from 0.27 down to 0.03. In each case,
befitting a comparative statics analysis, I assume the firm treats C as a fixed param-
eter. Not surprisingly, when C is lower, so too is the profit-maximizing Pb . Where it
is more efficient (i.e., has a smaller C), the firm sets a lower f.o.b. price; as it does so,
radius, output, and profit all increase. If the profit-maximizing price becomes low
enough, it becomes possible for farms sufficiently close to the factory to specialize
entirely in wheat production: hence the emergence of an X2 area.
As we know from a first course in Economics, the price set by a monopolist
depends on the price elasticity of demand for soap; the more elastic the demand,
Table 12.10 Model 12C: factory behavior as function of marginal cost (C)
the greater the revenue associated with a lower price. Here, elasticity varies from
one consumer to the next for two principal reasons. One is that because the factory
prices f.o.b. consumers pay the freight and face a higher effective price. As far as
the factory is concerned, consumers closer to the factory appear to have demands
that are more sensitive to f.o.b. price than do consumers who are further away. In
Table 12.10 we see that as C is lowered, the radius of the market (X∗ ) increases
which implies that the factory has more customers further away; hence the aggregate
demand facing the factory becomes less price elastic from the factory’s perspective.
The second is that to the extent that consumers are in the X2 area, their demand
will be less sensitive to price because they no longer substitute farm-made soap for
factory-made. In Table 12.10, I show that as C drops, eventually Pb becomes low
enough for an X2 area to emerge and that as C drops further, the demand from the
X2 area becomes an ever-more important share of its total demand. This too implies
that, as C drops still further, the factory’s demand will become ever less price elastic.
Panel (a) of Table 12.10 also shows “Additional rent”: i.e., the aggregate rent
paid by farms over and above (π X ∗2 )Ra . Because we have assumed that landlords
are absentee, the factory sees rent paid as a leakage from the local economy in that
this takes away from the wheat being offered in exchange for soap. After all, in the
absence of the factory, additional rent everywhere would be zero. At C = 0.03, the
factory’s profit and the additional rent are 97 and 73,000 units of wheat, respectively.
Presumably, if they were not paying this added rent, farms would have spent a share
of this on factory soap.
At the same time, note what happens to the price elasticity of demand over the
range of C shown in Table 12.10. When C is large, there is no X2 area; all farm-
ers are in the X1 area where there is substitution between farm and factory soap.
As we make C smaller, an X2 area begins to emerge. When C = 0.03, demand is
dominated by customers from the X2 area where there is no possibility of substitu-
tion. Add to that the idea that the consumer at X∗ is increasingly further away as C
becomes smaller, and the overall price elasticity of demand is decreasing.
Now, let me put the ideas in the last two paragraphs together. Despite earning
larger profits, the factory owner for whom C drops has reason to be unhappy. First,
the owner sees the leakage of profit in the form of higher rents paid by farms to
absentee landlords. Second, f.o.b. pricing is proving to be an increasingly blunt tool
for the job at hand. To maximize profit, the factory owner would want to decrease
prices in the part of the market where demand is more price elastic. However, under
f.o.b. pricing, the factory raises or lowers effective price by the same amount at
every location.
outlines Model 12C in the case of the factory pursuing discriminatory pricing.
Table 12.5 and Table 12.6 still describe outcomes as long as we remember that P[x]
is now the delivered price rather than effective price (f.o.b. price plus unit shipping
cost). From those two tables, I can derive the farm’s demand for factory soap. See
(12.9.1) and (12.9.2). The factory now sets a discriminatory price (P) for farms at
each distance x so as to maximize overall profit. See (12.9.7).
Because of mathematical complexity, I once again rely here on numerical meth-
ods to solve Model 12C for the price-discriminating monopolist. In panel (b) of
Table 12.10, I show discriminatory pricing solutions for the same values of C exam-
ined above in respect to a factory pricing f.o.b. There are some obvious differences
between f.o.b. and discriminatory pricing schemes. First, not unexpectedly, profit
is higher when the firm uses discriminatory instead of f.o.b. pricing: e.g., 13,000
units of wheat vs. 97,000 at C = 0.03. Second, the price-discriminating factory
has a larger market radius than does the f.o.b. pricer: e.g., 13.0 km vs. 9.6 km at
C = 0.03. Third, as we move down Table 12.10, for the values of C listed, an X2
area first appears with f.o.b. pricing (at C = 0.18) and with discriminatory pricing
only at C = 0.13. Further, at low values of C, the boundary radius X1 for discrim-
inatory pricing is not the same as for f.o.b. pricing. However, these are differences
that the reader might well have expected. There are two other differences that I think
might be more surprising.
First, in the case of discriminatory pricing, the factory sets the same delivered
price across its X2 area even while it sets a delivered price that increases with dis-
tance across the X1 area. See Fig. 12.7 where I compare the schedule of effective
prices under f.o.b. pricing (curve AJB there) with the schedule of delivered prices
under discriminatory pricing (curve CDE). Earlier in Fig. 8.2, I showed another
example where delivered prices (under discriminatory pricing) increase less quickly
with distance than do effective prices (under f.o.b. pricing). Still earlier, in Chapter
2, I had labeled this behavior by the monopolist “partial freight absorption”. Partial
freight absorption is still evident here in the X1 market area: compare the slopes
of curves JB and DE in Fig. 12.7. What is novel about Fig. 12.7 is the horizontal
segment (curve CD) of the schedule of delivered prices in Model 12C that spans
the X2 market area there. To explain why this happens, I go back to the equation
(12.9.2) for the farm’s demand for factory soap in the X2 area. This is not the linear
demand curve that I had used earlier as in Chapter 2. Instead, it is a negative expo-
nential demand curve for which the price elasticity (ignoring the negative sign) is
the exponent on price: namely 1 − ν. Since 0 < ν < 1, the price elasticity must
lie between 0 and 1. Put differently, the farm’s demand for soap in the X2 area is
price inelastic. As such, it is most profitable for the firm to set the delivered price as
high as possible there (consistent with being in the X2 area). The factory therefore
sets delivered price everywhere in its X2 market area just low enough to keep farms
from producing soap of their own: see (12.9.8).12
12 Thisis the same kind of argument that I made in Model 7D about price setting by a repair
contractor.
364 12 Local Production and Consumption
B E
J
C
D
Effective price (P[x])
0
I F G H
Distance from factory (x)
Fig. 12.7 Schedule of price by location in Models 12B and 12C compared.
Notes: ν = 0.2, b = 900, β = 0.2, γ = 0.6, c = 0.16, h = 4,000, Ra = 9,600, s = 0.02,
C0 = 0, and C = 0.03. Outcomes are Pb = 0.0978, X ∗ = 9.64 in 12B and 13.03 in 12C.
Horizontal axis scaled from 0 to 14; vertical from 0 to 0.4
Second, from the perspective of the factory, there is less leakage of rent with dis-
criminatory pricing. For example, at C = 0.03, total additional rent paid by farms
(rent paid by farms over and above (πX ∗2 )Ra ) is only 44,000 units of wheat: com-
pared to 73,000 under f.o.b. pricing). Put differently, the factory is better able to
benefit from the gains to be made from its efficient production. Much of the reduc-
tion in additional land rent comes in the X2 market area. See Fig. 12.8 which again
illustrates the case of C = 0.03. There, the faint curve (ABC) is the schedule of
rents under f.o.b. pricing while the kinked solid curve (DEF) is the schedule of rents
under discriminatory pricing. Remember here that farms are in spatial equilibrium
and that I have assumed they achieve the same level of utility under either pricing
scheme. What the consumer saves in land rent in one pricing scheme over the other
is offset by a higher price for soap.
Now, let me put the ideas in the last two paragraphs together. In my view, the
factory finds itself constrained by f.o.b. pricing in the sense of a leakage of potential
12.8 Model 12C: The Factory Using Discriminatory Pricing 365
B E
D
K F
C
0
G H I J
Distance from factory (x)
Fig. 12.8 Schedule of land rent (R[x]) by location in Models 12B and 12C compared.
Notes: ν = 0.2, b = 900, β = 0.2, γ = 0.6, c = 0.16, h = 4,000, Ra = 9,600, s = 0.02,
C0 = 0, and C = 0.03. Outcomes are Pb = 0.0978, X ∗ = 9.64 in 12B and 13.03 in 12C.
Horizontal axis scaled from 0 to 14; vertical from 9,000 to 11,500
profit in the form of additional land rent (i.e., over and above Ra ). As the efficiency
of the factory improves (i.e., C drops), f.o.b. pricing means that the factory lowers
its effective price for everyone: even though customers in the adjacent X2 market
area have a relatively inelastic demand. Using price discrimination, the firm would
drop its price only for the more remote (and more elastic) customers still in the X1
market area.
Some students, particularly those from outside Economics, may see price dis-
crimination as something undesirable, reprehensible, or even illegal. They think that
the pricing of commodities should somehow be more closely tied to the cost of man-
ufacturing and distribution. I think that many economists would argue that, in fact,
it can be difficult to discern whether firms are indeed using price discrimination.
Suppose, for example, that our soap manufacturer announces that he will sell his
product at a uniform price out to distance X1 and that beyond that the firm will pass
along only a fraction of any additional unit shipping cost to the consumer. This can
be cast in a way that alludes to (1) an efficient regional infrastructure for shipping so
that the difference in the marginal cost of serving a customer here rather than there
366 12 Local Production and Consumption
within the region is negligible and (2) the cooperative spirit of the firm in helping
out remote customers. These assertions may or may not be true; what is clear is that
the pricing scheme of the firm here is indistinguishable from price discrimination.
This brings us to an important idea in this book. I began this book by arguing
that geography is important to economic thought. However, we have now come to a
point where, at least within a radius X1 , the factory maximizes its profit by a pricing
scheme (uniform pricing) in which geography appears not to matter. In fact, to avoid
the appearance that it is indeed engaged in price discrimination, the firm has the
incentive to say things like “an efficient infrastructure for shipping services means
that we can serve a large market area without differentiating price by location”. At
the same time, just such a pricing policy—viewed as price discrimination—allows
the factory to extract profit that would otherwise accrue to landlords. And, such
profit arises specifically because of the role of unit shipping costs.
Two caveats are in order here. First, is it always the case that the factory will
have an X2 area around it? No. As argued above, the existence of an X2 area is
dependent on the extent of diminishing returns to scale in wheat production. Only
if wheat production is sufficiently close to constant returns to scale will we see the
emergence of an X2 area. Second, is it always the case that farms everywhere in the
X2 area will have a price-inelastic demand for soap? Again no. We get this result
in Model 12C because we have assumed a log-linear utility function for each farm
combined with a competitive market for land. What is clear, however, is that the
absence of farm soap production in the X2 area makes farms less able to substitute
when the price of factory soap is reduced and that this makes demand less price
elastic.
13 Note that my treatment of commuting cost here is smooth; for each additional unit of labor
allocated to factory work, the cost of commuting rises proportionally. In reality, commuting cost
12.9 Model 12D: The Factory as Bilateral Monopolist 367
production rises and to factory work declines. The critical distance from the factory
is X2 : h3 > 0 and h2 = 0 within radius X2 ;h3 = 0 and h2 > 0 outside it. I label
the geographic area inside radius X2 as the X3 area (the firm’s labor market area).
There are two distinct ways of thinking about the factory here. One is that the
factory is essentially equivalent to a cooperative at which farms gather to produce
their needed soap more efficiently. In this case, the output area served by the fac-
tory and the area from which it draws its labor would be the same. The other is to
think that the factory uses labor from a small area to produce a good sold over a
larger area. The latter is more akin to the concept of a manufacturing plant used by
Marshall. Therefore, in what follows, I assume X2 < X ∗ : i.e., the factory’s labor
market area is smaller than its output market area.
Model 12D examines how a bilateral monopolist with labor as its sole input
adjusts wage and price (exchange rate) so as to maximize profit. Whether such
adjustments are upward or downward will depend on the elasticities of demand
for the factory’s output and of its supply of labor. In the non-spatial spatial, a usual
tendency would be to assume fixed elasticities so that successive increments of eco-
nomic growth or technical change would consistently give similar kinds of wage
and price changes. Does the same result hold in a world incorporating geographic
space? If not, why not? I return to this question shortly.
Model 12D is laid out in Table 12.11. In contrast to Model 12C, assume the
soap factory has only one variable input, labor. The production function (12.11.1)
assumes that there are returns to scale: diminishing if 0 < δ < 1, constant if δ = 1,
or increasing if δ > 1. In contrast, soap production on the farm, as we saw in
Chapter 12, had constant returns to scale. At the same time, output is determined by
a scalar k0 that I assume is fixed in any one time period but that changes over the
years; however, assume the firm is unaware of this and treats k0 as fixed for each
time period. Assume here that the factory is a relatively efficient producer of soap
compared to farms. The demand for factory soap is given by (12.11.3) which totals
the demands in the X1, X2, and X3 areas. As in Model 12C, the X2 area may or
may not exist; once again, I include it here for comprehensiveness. The supply of
factory labor is given in (12.11.5). It hires all workers at the same wage, wb .14 The
net profit of the firm is given by (12.11.2). Assume the factory has a fixed cost, C0 ,
each time period. This is its only cost other than the wage bill.
Now consider the introduction of a soap factory into this landscape. For the soap
factory to be profitable, several conditions must be met. First, in this barter economy,
the factory’s f.o.b. exchange rate (Pb ) must be below p∗ ; otherwise it would not get
any customers. Second, its wage (wb ) must be high enough to attract the required
workers. In the X1 area, a farm that produces soap of its own, and the marginal
value product of labor is cP[x]; therefore, to have labor attracted to factory work
is usually more lumpy than this. You might choose for example to work 8 hours a day, rather
than 7, without incurring any added commuting cost. The models in this chapter do not take such
considerations into account.
14 This too may seem strange. After all, why wouldn’t the factory use discriminatory pricing in its
labor market. I use a constant wage here once again to simplify calculation.
368 12 Local Production and Consumption
Factory’ profit
Z = P b Q − C0 − w b N (12.11.2)
Demand for
factory soap (assuming X2 area)
Q = 2π x∈X3 (xz[x]/L[x])dx + (12.11.3)
2π x∈X2 (xz[x]/L[x])dx + 2π x∈X1 (xz[x]/L[x])dx
Demand for
factory soap (assuming no
X2 area)
Q = 2π x∈X3 (xz[x]/L[x])dx + 2π x∈X1 (xz[x]/L[x])dx (12.11.4)
Notes: In the absence of X2 area, outer boundary (X2 ) of X3 area occurs where the marginal
return on factory work equals the marginal return on on-farm soap production, i.e., wb -rX2 =
c(Pb + sX2 ). In the presence of X2 area, outer boundary (X2 ) of X3 area occurs where the
marginal return on factory work equals the marginal return on wheat production, i.e., wb -rX2 =
(βk2 /(ch(1-γ )))(Pb +tX2 )ν . See Table 12.5, Table 12.6, and Table 12.12. Rationale for localization
(see Appendix A): Z1—Presence of fixed cost for factory; Z2—Implicit unit cost advantage at
some locales; Z8—Limitations of shipping cost. Givens (parameter or exogenous): Co —Fixed
cost of factory soap production; ko —Scale of production; δ—Returns to scale on labor in factory
soap production. Outcomes (endogenous): h3 [x]—Schedule of labor allocated by farm to factory
work; L[x]—Schedule of and used per farm in wheat production; N—Total supply of labor to
factory; Pb —F.o.b. price for soap set by factory; Q—Total quantity of factory soap demanded;
wb —Wage rate paid by factory; Z—Factory s profit; z[x]—Schedule of quantity of factory soap
purchased by a farm.
implies wb > cP[0]. In fact, wb must be even higher than this if there is an X2 area
present. Third, the factory’s scale of production (k0 ) must be sufficiently large to
make production profitable.
How do farms behave in this model? Farms that fall into the X1 or X2 areas will
continue to have the behavior described in Tables 12.5 and 12.6. What is new here
are the farms in the labor market area (X3). Here, farms allocate no labor to on-farm
soap production: see (12.12.1) and (12.12.2) in Table 12.12. For a given allocation of
labor to wheat production, we can then find the optimal size of farm (12.12.3) and
the gross and net outputs, (12.12.4) and (12.12.5). For a given allocation of labor
to wheat production, we can determine z from the familiar budget share equations
(12.12.6) and (12.12.7) and reexpress utility as a function of h1 [x] in (12.12.8).
Finally, we can then find the h1 [x] that maximizes utility in (12.12.9), reexpress
utility in (12.12.10), and then invert it to get an expression for R[x] at each distance
(12.12.11) that makes U[x] = u∗ everywhere.
Let us now try a simple example. Suppose the factory has δ = 1.0, that is,
constant returns to scale, just like farm soap production. Suppose we also assume
C0 = 0 so that we make soap production in the factory using labor only, much
like soap production on the farm. In this case, what k0 at the minimum is required
for the factory to earn a profit? The answer to that question is fairly easy; k0 would
12.9 Model 12D: The Factory as Bilateral Monopolist 369
Table 12.12 Model 12D: farm in X3 area (farm engaged only in wheat production and factory
work)
Notes: Rationale for localization (see Appendix A): Z1—Presence of fixed cost for factory; Z2—
Implicit unit cost advantage at some locales; Z8—Limitations of shipping cost. Givens (parameter
or exogenous): b—Scalar in wheat production; c—Marginal product of labor in on-farm soap
production; h—Total labor available to the farm; Ra —Opportunity cost of land; s—Unit shipping
rate; β—Exponent of labor in wheat production; γ —Exponent in land in wheat production; ν—
Preference for soap. Outcomes (endogenous): h1 [x]—Farm labor in wheat production at x; h2 [x]—
Farm labor in soap production at x; h2 [x]—Farm labor in factory work at x; k2 —A constant (see
Table 12.5); L[x]—Land used per farm in wheat production at x; q1 [x]—Consumption of wheat
on farm at x; Q1 [x]—Production of wheat on farm at x; q2 [x]—Consumption of soap on farm at
x; R[x]—Bid rent per unit land at x; u[x]—Utility at x; v1 [x]—Production of wheat (net of rent)
on farm at x; v2 [x]—Production of soap on farm at x; z[x]—Quantity of factory soap purchased
by farm at x.
have to be larger than c. Since we have assumed c = 0.16 above, let us therefore
assume k0 = 0.18 as an illustrative case: put differently, assume factory workers
are modestly more productive than are farms in producing soap per unit of labor
allocated to this activity.
How do we solve here for the exchange rate and wage (Pb , wb ) combination that
maximizes factory profit. Our problem here is twofold. First, equations (12.11.3)
and (12.11.5) cannot be reduced algebraically. We must therefore rely on numer-
ical approximations (midpoint rule version of a Riemann sum15 ) to the integral
expressions in them. The second problem is that we have to use numerical meth-
ods to find the solution. One approach is to scan values of Pb ; for each chosen Pb ,
we can then iterate to find the wb that provides sufficient labor to meet the demand
for soap forthcoming at that price and thereby calculate the profit earned by the
factory.
Doing this, I find the profit-maximizing factory will set an f.o.b. exchange rate
(Pb ) of 0.2800 m3 of wheat per block of soap. See Table 12.13. As expected, this
is below the implicit exchange rate (0.2906) for farms in autarky. To sustain the
demand this generates, the factory will offer a wage (wb ) of 0.0482 per person-hour.
For a farm in autarky, the marginal value product of labor is cp∗ which is 0.0465 per
person-hour; the factory is, as is necessary, offering a relatively better wage. Under
profit maximization, the radius of the factory’s output market is a mere 0.53 km;
the radius of its labor market is 0.26 km; and, there is no X2 area here. From the
small sizes of the X1 and X3 areas, the transport rates, s and r, must be relatively
high. The total amount demanded from the factory is 1,330 blocks of soap. For the
factory, total revenue is 372, total cost is 356, and net profit is 16.
Table 12.13 Model 12D: factory behavior under f.o.b. pricing as function of scalar (k0 )
Radiuses Output
Scalar Price Wage Labor Profit X2 Add’l
k0 Pb wb X2 X1 X∗ N All X3 X2 X1 Z area rent
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]
Finally, let us look at the price elasticity of demand facing the factory here.
Suppose the factory were to reduce its f.o.b. exchange rate by 10%: from 0.2800
to 0.2520. To meet the increased demand, the factory would then have to set
wb = 0.0544. The total amount demanded from the factory would then be 29,351
blocks of soap. For the factory, total revenue would now be 7,396, total cost would
be 8,876, and net profit would drop to –1,479. The price elasticity of demand here
is ((8,876 − 1,330)/1,330)/((0.2520 − 0.2800)/0.2800) = −210.7. At first glance,
this may seem surprising. After all, if the price elasticity is so large (in absolute
value), why wouldn’t the factory further lower its price? However, the trick here
is to be mindful also of what is happening to cost. In our example, cost is simply
the wage bill. To meet the increase in demand when it drops price, the factory has
to raise its wage to attract sufficient labor. Total cost is now 8,876: up from 356
previously. Put differently, cost rises almost 24-fold, whereas revenue goes up only
19-fold; that is why profit drops when Pb is set 10% below 0.2800.
Model 12D allows us to think about economic development. Important here is
the role of k0 . As k0 is made larger, the firm is able to produce more output from a
12.9 Model 12D: The Factory as Bilateral Monopolist 371
given quantity of labor. In practice, k0 might increase because the firm is using more
capital to produce its output or because the technology of the firm has improved.16
In what follows, I treat changes in k0 as comparative statics. I do not ask what the
firm might do if it knew that its k0 was going to increase over time. Instead, I treat
ko as exogenous and simply look at the setting of Pb and wb .
What happens when k0 is larger? Table 12.13 shows the profit-maximizing com-
binations of exchange rate and wage as k0 is varied systematically from 0.18 to 0.80.
Not surprisingly, as k0 is increased, the profit-maximizing Pb decreases. Perhaps
more surprising is that wb at first increases and then decreases as k0 is increased.
What causes Pb to decline so quickly at first and then almost level out? See Fig. 12.9.
What does this have to do with the first up then down trajectory for wb ?
Fig. 12.9 Model 12D: f.o.b. Model 12D: F.o.b. price and wage set by factory
AB Wage set by factory to maximize profit
price and wage setting by a CD F.o.b. price set by firm to maximize profit
profit-maximizing factory.
Notes: ν = 0.2, b = 900,
β = 0.2, γ = 0.6, C
c = 0.16, h = 4,000,
Ra = 9,600, s = 0.02, B
C0 = 0, and C = 0.03.
A
D
0
Scale of factory production (k0)
How a factory sets price depends on both the price elasticity of demand for soap
and elasticity of supply of labor. Here, demand elasticity varies from one consumer
to the next for two principal reasons. As we saw in Model 12C, because the factory
prices f.o.b., consumers pay freight, and hence consumer demand appears to be
more sensitive to f.o.b. price for consumers closer to the factory than for consumers
further away. In Table 12.13 we see that as k0 rises, the radius of the market (X∗ )
16 An extension to the models considered in this chapter would treat capital and technology, by
both firm and farms, as endogenous. See, for example, Zhang (1993).
372 12 Local Production and Consumption
increases which implies that we have more customers further away from the factory,
and the aggregate demand facing the factory will become less price elastic from the
factory’s perspective. The second reason is that to the extent that consumers are in
the X2 or X3 area their demand will be less sensitive to price because they no longer
have the ability to substitute farm-made soap for factory-made.
In Table 12.13, I show what happens to Pb and wb as k0 rises from 0.18 to 0.80.
First, when k0 is greater than about 0.4, Pb becomes low enough for an X3 area to
emerge; as k0 rises further, the demand from the X2 and X3 areas becomes an ever-
more important share of its total demand. This implies that above k0 , about 0.4, the
factory’s demand becomes less price elastic as k0 rises. Second, when k0 is smaller
than about 0.4, the effect is opposite. There, as k0 is increased, the factory finds that
demand in the X1 area (more elastic demand) is growing faster than demand in the
X2 area (less elastic demand).
The final column of Table 12.13 shows “Additional rent”: i.e., the aggregate rent
paid by farms over and above (π X∗2 )Ra . Because we have assumed that landlords
are absentee, the factory sees rent paid as a leakage from the local economy in that
this takes away from the wheat being offered by farmers in exchange for soap. The
total additional rent is negligible for k0 below about 0.20. However, by k0 = 0.80,
the total additional rent is 43,000 units of wheat. As in Model 12C, additional rent is
a substantial relative to the total profit earned by the factory. Once again here, there
is an incentive for the factory to use discriminatory pricing to capture some of the
additional rent.
What about the comparative statics of Model 12D? Suppose a factory finds itself
in the situation described in the bottom row of Table 12.13 and correspondingly
sets its f.o.b. price at 0.132 and its wage a 0.058. Now, suppose this factory were
to increase either its Pb or wb by 10%. What effects would such changes have on
outcomes in its market area. In Fig. 12.10, I show the effect of each change on
the wheat labor gradient. Under profit maximization, h1 [x] is given by ABCD in
Fig. 12.10: kink B happens at the boundary between the X3 and X2 areas and Kink
C happens at the boundary between X2 and X1. If wb is increased by 10%, the
gradient shifts to EFCD. The shift in kink from B to F is because of the expansion
of the X3 area, since factory work is now more attractive. Within the X3 area, farms
everywhere reduce the amount of labor, h1 [x], devoted to wheat labor in order to
supply more factory labor. Beyond the new X3 area boundary, the higher factory
wage has no effect on the wheat labor gradient. If Pb is increased by 10%, the
gradient shifts to GHIJ. The shift in kinks from B to H and C to I and in X∗ from D
to J is because all three areas (X3, X2, and X1) shrink because factory soap is now
less attractive. In the X3 area, farms reallocate labor in favor of wheat production; in
the X1 area, farms put more labor into farm soap production to offset factory soap
that is now less attractive. A similar story can be told about the impacts of a change
in wb or Pb on the rents (see Fig. 12.11), land use (see Fig. 12.12), and net wheat
production (see Fig. 12.13).
What about the regional economy in Model 12D. Continuing with the same
numerical example, let me start with column [1] of Table 12.14 where ko is just
0.20. The factory’s market area is small: just 1.1 km in radius or about 100 farms.
The factory earns a profit here of just 0.1. The average farm pays a rent of 374.
12.10 Final Comments About This Chapter 373
H F I C
B
Wheat labor (h1[x ] )
J D
G
A
E
0
Distance from factory (x )
Moving to the right in Table 12.14 gives basic information about the factory and
the farms in its market area. In addition to transportation costs, there are two leak-
ages from the regional economy here that depend on ko : one is factory profit and
the other is the additional (scarcity) rents paid to landlords. Across these results, the
scarcity rents in total (after multiplying the per farm additional rent by the number
of farms) is on the order of 3/4 of the landlord profit. The two leakages become
more important the larger is ko . Taken together as a proportion of aggregate wheat
production in the factory’s market area, they rise from 0.2% of wheat production
when ko = 0.18 to 3.0% when ko = 0.80.
In this chapter, the principal model has been Model 12D. I included Models
12A through 12C to help readers better understand aspects of Model 12D. In
Table 12.15, I summarize the assumptions that underlie Model 12A through 12D.
Many assumptions are in common to all these models: see panel (a) of Table 12.15.
The models differ in that (i) a farm is assumed to be within the factory’s market
374 12 Local Production and Consumption
Fig. 12.11 Model 12D: Model 12D: Comparative statics of rent (R)
comparative statics of land ABC Rent per unit land by a farm at profit-maximizing solution
DBC Rent per unit land by a farm if factory wage (wb) increased by 10%
rent. EF Rent per unit land by a farm if factory f.o.b. price (Pb) increased by 10%
Notes: ν = 0.2, b = 900, D
B
0
Distance from factory (x) F C
Fig. 12.12 Model 12D: Model 12D: Comparative statics of land (L)
comparative statics of land ABCD Amount of land rented by a farm at profit-maximizing solution
EFCD Amount of land rented by a farm if factory wage (wb) increased by 10%
use. GHIJ Amount of land rented by a farm if factory f.o.b. price (Pb) increased by 10%
Notes: ν = 0.2, b = 900,
I C
β = 0.2, γ = 0.6, H
Land used by farm (L [x ])
c = 0.16, h = 4,000, B F
Ra = 9,600, s = 0.02, J D
C0 = 0, k0 = 0.80,
δ = 1.0, r = 0.01,
Pb = 0.1445, and
wb = 0.0585. Horizontal
axis scaled from 0 to 9;
vertical axis scaled from 0.00
to 0.06. Viewed from left to
right, kinks occur at
boundaries of X3, X2, and X1 G
A
areas, respectively
E
0
Distance from factory (x)
area in Models 12B through 12D, (ii) the monopolist’s f.o.b. price gets set endoge-
nously in Model 12C, and (iii) the monopsonist gets to set both price and wage in
Models 12D.
What about Walrasian equilibrium and localization?
• Model 12A has no shipping costs. It models one exchange albeit implicitly: the
implicit exchange of wheat for soap. The supply of land is infinitely elastic at
12.10 Final Comments About This Chapter 375
I C
H
B F
J D
G
A
E
0
Distance from factory [x]
the rent (Ra ). The amounts of labor and land used by a farm in autarky in wheat
production are implicitly traded off against the alternative use of labor in soap
production. There is no Walrasian simultaneity here. Prices are not jointly deter-
mined here: instead, Ra and other givens determine p∗ (the implicit exchange
rate).
• Model 12B has shipping costs. It models exchange in three markets: wheat/soap,
labor (implicitly), and land. Overall, the supply of land is infinitely elastic at the
opportunity cost (Ra ); however, shipping costs now mean that, in Models 12D
through 12D, there will be rent differentials by location that preserve a utility
equilibrium. What makes 12B different from 12A is that farms in the market area
now exchange some wheat for factory soap. There is still no Walrasian simul-
taneity here. The rent gradient (R[x]) is now determined by the factory’s f.o.b.
price (Pb ), the opportunity cost of land (Ra ), and other givens.
• Model 12C endogenizes f.o.b. price (Pb ). Model 12C has three markets: wheat,
labor (implicitly), and land. There is now a Walrasian simultaneity here. The
locational rent (the amount by which rent exceeds Ra at a given location) and
factory’s f.o.b. price (Pb ) are now jointly determined by Ra and other givens.
• Model 12D endogenizes f.o.b. price (Pb ) and wage (wb ). Model 12D has three
markets—wheat, labor, and land—all now explicit. Land may also be obtained,
376 12 Local Production and Consumption
subject to differentials in shipping cost and land rent. There is Walrasian simul-
taneity here. The locational rent, the factory’s f.o.b. price (Pb ), and the factory
wage (wb ) are now jointly determined by Ra and other givens.
What about the regional economy? Model 12D is the final step in this book
toward a comprehensive depiction of the regional economy. We can measure
regional output and regional income. We are also able to divvy up regional income
into a wage bill, a rent bill, and profits. In Model 12D, there is the possibility of
a profit to factory soap production. For a farm, the only profit is the advantage
that arises when it locates nearer the factory, and this is dissipated in locational
rents.
Who gains and who loses from improvement in productivity at the factory? The
factory gains. So do the absentee landlords. However, farms are no better off after
the improvement in factory productivity. The reason here is our assumption of free
entry. In this chapter, the utility of a farm never changes. If the price of factory soap
drops, the size of the market and the intensity of land use shift until utility is same
as before. In effect, we are assuming an infinitely elastic supply of farms at a given
level of utility.
12.10 Final Comments About This Chapter 377
There is an interesting relationship here between land rent and factory behavior
in Model 12D. As k0 increases, the factory becomes more productive, and this trans-
lates directly into higher profit for the factory. However, it also translates into higher
land rents as farms bid to be closer to the factory.
What about impact of transportation costs? Model 12D assumes both commuting
and shipping costs. What would happen if both s and r were to approach zero? If
s is zero, farms everywhere face the same effective price for factory soap. If r is
zero, farms everywhere have a choice between two activities with constant returns
to scale; factory labor and on-farm soap production. It would engage only in the
activity with the higher marginal value product of labor. Land rent everywhere
would be Ra , and no farm would produce soap itself. Each farm would purchase
the same amount of factory soap (z), and the factory wage would be set so as to
draw the appropriate amount of labor from the farm. The factory would then have an
infinitely elastic supply of labor. In such a case, the factory becomes like a coopera-
tive where workers gather to produce more efficiently than they could, each alone, at
the farm.
378 12 Local Production and Consumption
In developing Model 12D, I seek an argument here applicable to any local pro-
duction: not just on farms. Put differently, I use the farm as a metaphor for the local
economy in general. To me, farm and factory are key elements in a story about the
effects of globalization on the space economy. I view a farm as a local production
unit; it might consist for example of the farms in a district along with ancillary local
services and commodities: e.g., construction, education, finance, health, insurance,
real estate, religion, and retail. In turn, I view the soap factory as any globalizing
economic activity that is seeking to establish itself in local economies by replacing
a commodity or service that had been locally produced with a commodity or ser-
vice now produced elsewhere. In the models of this chapter, I want to explore how
the introduction of an efficient large-scale manufacturer reshapes prices locally and
thereby the local economy within its reach. At the same time, there are limits to
this analogy. First, if farm is indeed a community or region it is not clear how one
might determine a utility function for such a unit. I am referring here to the well-
known problem that divergent individual preferences can mean to the existence of a
social welfare function for the community as a whole; a problem sometimes labeled
Arrow’s Impossibility Theorem: see Arrow (1965). Second, if the farm is an urban
area, it presumably includes people who do no farming at all. What is the nature of
land consumption in that case?
In this book, several location models give us insights into the organization of
the firm and market. What are these insights? How do they build on the work
of Coase and others on the economics of organization?
Are these concepts in fact descriptions of the same thing? If not, when, how,
and why are the differences among them important in the models in this book?
12.11 The Connecting Topics 381
The models in this book envisage that cost savings originate in different ways.
In Chapter 2, the firm has a fixed cost that can be spread over whatever quantity the
firm produces. The firm trades that cost off against the savings from the elimination
of shipping costs. In Chapter 3, the firm weights shipping cost against local varia-
tions in unit production costs in deciding how to allocate the limited capacities of its
factories. Similarly, in Chapters 4 and 5, arbitrageurs take advantage of lower cost
advantages locally where unit shipping cost is not prohibitive. In Chapter 6, the firm
finds local cost advantages only because of the shipping costs of inputs or products
that are not ubiquitous. Chapter 7 (risk reduction by outsourcing) and Chapter 9
(risk reduction by market formation) explore the urbanization economies made pos-
sible by application of an insurance principle. In Chapter 10, the only local cost
advantage is a smaller shipping cost. While these origins are different, the effects
that they have on the location of firms are similar; where price of a product is given,
firms are attracted to sites with low costs of production and shipment. At the same
time, the significance of these different origins lies in what they tell us about model
predictions.
The early chapters in this book tell us nothing about the simultaneity of prices; at
best, they tell us something about how other prices affect the price at which a firm
or industry sells. In Chapters 2, 4, and 5, the only prices that are jointly determined
are the prices of the commodity at different geographic places. Chapters 3 and 6
do not solve for market prices. However, the use of shadow prices (particularly,
the opportunity costs of using the limited production capacities of its factories) in
Chapter 3 has parallels with market prices. Chapter 7 explores the link between the
price for a firm’s good and the price it is willing to pay for outsourcing. Chapter 8
looks at the definition of market areas and the setting of price by a firm in response
to the prices of competitors within its range or trade area. Chapter 9 explores price
setting across markets of various sizes based on risk aversion.
However, the later chapters are different. Chapter 10 introduces a simultaneity
between commodity price(s) and land rent. The Samuelson models in Chapter 11
extend this simultaneity to also include wages. Chapter 12 extends this still further
by allowing us to think about the simultaneous determination of profits, land rents,
and commodity prices in a regional economy. In these models, it is the sharing of
factor markets—the fact that firms producing one commodity have to compete with
other firms for labor, land, and capital—that ensures a Walrasian equilibrium across
markets.
382 12 Local Production and Consumption
What does each of our models tell us about the impacts of shipments on the
regional economy?
The early chapters of this book do not tell us much about the regional economy as
a whole. Chapters 2 through 7 each (1) raise the possibility that we might usefully
think of a regional economy as an endogenously determined set of places linked
by shipments or other flows and (2) make it possible to think about autarky as a
state where customers either supply for their own needs or are supplied only by
adjacent factories but (3) provide no way of measuring factor incomes. Models in
this chapters at best give us an idea of producer surplus and consumer surplus for
one industry in the region.
The later chapters are more helpful. Although Chapter 9 has the potential for us
to sketch out a regional economy, the model starts with farm outputs given. It is only
starting in Chapter 10 that we are able to begin thinking about farm product, land
rents, and farm income. In Chapter 11, our regional income and product accounts
can be extended to include wages. Finally, in Chapter 12, we are also able to incor-
porate profits and show a breakdown of regional income into wages, land rents, and
profits.
The diverse links between geographical space and a monopoly that is at least
local in extent make it difficult to know when or how to apply basic economic
ideas about the consequences of competition. How, and to what extent, do the
models in this book allow us to do this?
Over the course of the book, we dip back and forth between monopolistic and
competitive perspectives. Chapter 2 introduced the idea of a monopolist engaged in
discriminatory pricing. To a monopolist, the difference in price between two cus-
tomers may never be as large as the difference in unit shipping cost between them.
Chapter 8 helps us think about the nature of a monopoly when there are competitors
within the firm’s range or trade area. Chapter 3, being silent on price, tells us little
about how a monopolist behaves in a spatial economy. The same is true of Chapters
6 and 7 in terms of the price of the commodity sold by the firm. However, Chapter
7 also gives us ideas about the monopolist contractor providing inputs to firms.
Chapters 4 and 5 introduced the idea that, in competitive equilibrium, the difference
in price between any two geographic places would not exceed the unit shipping cost
12.11 The Connecting Topics 383
between them. Chapter 9 is at odds with this topic because of a confounding of pro-
ducers and consumers. Chapters 10 and 11 do not get us much further along on this
topic because they assume competitive markets. It is not until Chapter 12 that we
are able to reincorporate the impact of a local monopolist (the soap factory).
Geography
In some models in this book, the modeling of geography was relatively unimpor-
tant. In Chapters 2 through 5 and 7, for example, it is possible to generate almost
all results using unit shipping costs rather than unit shipping rates; in that sense, the
results of the model depend only on the cost of shipment, not on distance directly,
and therefore the models are independent of how distance is measured.
Chapter 6 is the first point in the book where the way that distances get mea-
sured substantially affects model outcomes. However, even here, the conditions for
localization depend only on map angles and not on map distances. The models in
Chapters 8, 9, 10, 11, and 12 all rest on the idea that distance is measured as the crow
flies and that shipping cost is proportional to distance. Conclusions in the second
half of the book are more sensitive to the way in which distances are measured.
Appendix A
Assumptions and Rationale for Localization
the same commodity and is a price taker in that market. This industry is also
characterized by free entry. Firms keep entering market until excess profit for
marginal new firm entering industry is driven to zero. Relocation of firms is
therefore assumed to be costless.
Each market clears: an exchange rate is established such that for each com-
modity or crop the quantity demanded at that rate is equal to the quantity that
suppliers have the incentive to supply. There is sufficient production capacity
overall to meet overall demand at the going price.
Commodities are perishable and cannot be carried over from one time period
to the next. No one carries over money (e.g., through savings or debt) from one
period to the next.
Unless otherwise assumed, there is no uncertainty.
A2 Barter market. There is a barter market in the exchange of two commodities.
A3 Punctiform landscape. A simplifying assumption in which each producer and
customer for a commodity can be thought to be a point—occupying no space—
on the landscape. Regardless of the number of customers at one point and
the number of producers at another, the distance between the two points—and
hence unit shipping cost—is invariant.
A4 Rectangular plane. Location of any point can be described by Cartesian coor-
dinates. Travel is possible in any direction on this plane; distances are therefore
Euclidean (as the crow flies). Where appropriate, land is available throughout
the region and has the same yield when used in the same way.
A5 Linear landscape. Economic activity occurs only as points along a line in
geographic space: i.e., a one-dimensional spatial economy.
A6 Ribbon landscape. A regional landscape in which farms are limited to a strip
of land of unit (say 1 km) width. There is a cost of shipping along the length
of the ribbon; any cost of shipping across the width of the ribbon is ignored
(assumed zero).
A7 Transportation network. Region modeled as (1) set of vertexes; (2) some pairs
of vertexes are joined by a transportation link; (3) a route—combination of
links—joins every possible pair of vertexes; (4) every customer place is a ver-
tex; (5) every input supply place is a vertex; (6) factory may be sited at any
vertex or along any link joining two vertexes.
C1 Fixed demand locations. The firm faces fixed (price-insensitive) demands for
its products from customers at one or more locations.
C2 Fixed local customers. The number of consumers in the local market is fixed.
C3 Fixed remote customers. The number of consumers in each remote market is
fixed.
C4 Identical customers. Consumers are identical in terms of preferences and
incomes. However, effective price may vary from among customers.
C5 Identical linear demand. For each industry being modeled, each customer has
the same linear inverse demand curve.
C6 Identical individuals.
C7 Maximize same utility function. Customers maximize the same log-linear
utility function defined over the consumption of two goods.
C8 Other demand curve. Each customer has the same individual inverse demand
curve, but this is not linear as in Assumption C5.
C9 Customers uniformly distributed. The density of customers (customers per
square kilometer) is the same in every part of the geographic region being
modeled.
C10 Two kinds of customers. All customers are either of kind A or kind B. All
customers of kind A are identical. All customers of kind B are identical.
J1 Zero opportunity cost (rent) for land. As in a frontier setting, no other use for
land other than those specified in model.
J2 Nonzero opportunity cost (rent) for land.
J3 Farm occupies fixed amount of land.
J4 Amount of land used endogenous.
J5 Land is input to production.
J6 Competitive market for land. All participants are price-takers. Land allocated
to highest bidder.
K1 Fixed capital requirements. Firm has fixed capital requirements. In this book,
capital refers to the plant and equipment required by the firm to produce a
commodity. Capital is durable; it is typically used to produce goods repeatedly
over many periods of time. In principle, capital is a physical measure; however,
we typically assume for simplicity that it can be represented in diverse ways as
a money value; e.g., the purchase cost, the resale value, the replacement value,
or the stream of future profits that can be earned from it. At the same time,
we need be mindful that in competitive markets for capital goods, the money
value associated with a unit of capital is itself a price determined by supply
and demand factors.
Z1 Presence of fixed cost. With larger output, firm spreads the fixed cost and
lowers its unit costs.
Z2 Implicit unit cost advantage at some locales. Unit (variable) production costs
and/or unit shipping costs are lower at some locales than others. There are
many possible reasons for such differences: e.g., differences in effective prices
for inputs (e.g., labor, land, capital, or materials), the firm’s division of labor
across establishments, or local regulation.
Z3 Implicit unit price advantage at some locales.
Z4 Risk spreading and insurance.
Z5 Capacity constraints and congestion.
Z6 Differences among consumers.
Z7 Variations in goods.
Z8 Limitations of shipping cost or travel delay.
Z9 Shipping cost on non-ubiquitous inputs.
Appendix B
Glossary
393
394 Appendix B
Capacity The amount of output that can be produced by a given factory over a
stated period of time. In a simple case, we imagine that the marginal cost of produc-
tion is fixed per unit for any quantity up to capacity (marginal cost of production is a
horizontal line up to capacity). To allow for the concept that no quantity greater than
capacity is possible, we typically assume the marginal cost curve then becomes ver-
tical. In other words, no matter how much the firm might want to increase quantity
above capacity, it simply cannot produce a quantity in excess of capacity.
Cartesian coordinates Representation of a point in two-dimensional geographic
space by a pair of distance coordinates: (x, y). An example of such a planar repre-
sentation is given by Universal Transverse Mercator (UTM) coordinates which take
the form of easting and northing coordinates. Cartesian coordinates are an approx-
imation to location in that they ignore both elevation and the fact that the earth is
a sphere. The advantage of Cartesian coordinates is that distance calculations are
simplified compared to spherical coordinates. . Euclidean distance
Cobb–Douglas production function A firm using two inputs to produce an
output is said to have a Cobb–Douglas production function if the maximum level
of output, Q, obtainable from q1 units of input 1 and q2 units of input 2 can be
g
expressed as Q = aqb1 q2 where a, b, and g are parameters. This firm has returns
to scale that are constant if b + g = 1, decreasing If b + g < 1, and increasing
if b + g > 1. Assuming competitive markets for inputs and output and constant or
diminishing returns to scale, a profit-maximizing firm will spend the proportion b
of its revenue on purchases of input 1 and g on purchases of input 2, the cross price
elasticity will be zero, and the firm’s expansion path will be linear.
Comparative cost analysis A technique for finding the least cost location for
a firm by enumerating possible locations and then calculating the unit cost of
production at each.
Comparative statics A comparison of outcomes (endogenous values) predicted
by a model when a given (exogenous variable or parameter) is changed by a small
amount. Some models describe market equilibrium; here comparative statics details
the changes in equilibrium when a given is changed by a small amount. In other
cases, models describe optimal outcomes; here comparative statics details changes
in optimal outcome when a given is changed.
Competitive location theory Area of scholarship that looks at how competition
among firms and households leads to geographic patterns in market equilibrium
(otherwise known as spatial equilibrium).
Complementary slackness theorem In a Linear Programming solution, each
inequality constraint has both a slack (or surplus) and a shadow price. The
Complementary Slackness Theorem says that at least one of the slack (or surplus)
and shadow price must be zero for every constraint.
Congestion A consequence arising in an economic landscape whereby the firm
finds it relatively more costly or less profitable to marginally increase the quantity of
Appendix B 395
a commodity that it supplies. Congestion may arise, for example, because of (1) lim-
itations in the firm’s ability to manage a larger output, (2) limitations at the factory
or in the supply chain (supplier, warehouse, mode transfer station, or transportation
network), or (3) a deterioration in the firm’s profit from the response of competi-
tors. Usually, congestion is measured over the short run: i.e., before the firm has the
opportunity to adjust its investment in land, plant, and equipment.
Conjectural variation In game theory, a player’s perception (assumption) about
the reaction of another player to the first player’s choice of action.
Consumer benefit For a market, the amount that consumers might have been
willing to pay rather than go without the commodity; alternatively, the area under
the demand curve to the left of the quantity demanded.
Consumer surplus The area below the demand curve and above the market price
to the left of market quantity. Put differently, Consumer Surplus is the amount (in
dollars) that consumers would have been willing to pay over and above the market
price rather than go without the product.
Core theory As this book is focused on microeconomic applications, I take this
to include neoclassical theory of consumer demand, theory of the firm, and welfare
economics.
Corner solution A condition in which the solution to an optimization problem
Y = Max x [f(x)] occurs at a limiting value of x.
Cournot Antoine Augustin Cournot (born 1801), a French economist and math-
ematician, published the original version of his Recherches sur les principes
mathématiques de la théorie des richesses in 1883.
Cross-price elasticity For any two commodities, labeled 1 and 2, respec-
tively, cross price elasticity (c) is the percent change in quantity of commodity 1
demanded (q1 ) given a one percent change in the price of commodity 2(p2 ):c =
(p2 /q1 )(dq1 /dp2 ).
Delaunay triangulation See Thiessen polygon
Delivered price Firm sets price for commodity delivered to customer; consumer
does not pay a separate shipping charge. Where every consumer pays the same
price, the firm engages in uniform pricing. Where each consumer pays potentially a
different price, the profit-maximizing firm engages in discriminatory pricing.
Demand cone A term characterizing the tendency for individual quantity
demanded to fall off with increasing distance from a supplier pricing f.o.b. because
of the rise in effective price.
Diminishing marginal utility The idea that utility increases at a decreasing rate
as a person consumes more of the commodity. In the twentieth century, economists
in general moved away the notion of diminishing marginal utility of commodities
396 Appendix B
Efficient firm An efficient firm (1) incurs the least possible cost in achieving its
desired output and (2) seeks the maximum revenue possible from that output. It
adopts an organizational structure than enables it to be efficient. It has a production
function which shows, for each combination of inputs, the maximum possible output
that can be produced with those inputs. The firm also has a cost function which
shows, for each level of output (Q), the minimum possible cost of achieving that
Q. Finally, the firm knows the demand for its product and exploits that information
along with the knowledge of its cost and production functions to maximize its own
profit. See also “firm.”
Endogenous In a model, an endogenous value is an outcome; a value predicted
by the model based on other (endogenous) values. Endogenous variables may have
a stochastic component.
Establishment As commonly used in Censuses, the whole or part of a firm that
is based, or carries on business, at a particular physical site (i.e., a contiguous set of
facilities). Each firm can be partitioned into one or more establishments.
Euclidean distance In two-dimensional space, the straight-line distance √ between
two points with Cartesian coordinates (x1 , y1 ) and (x2 , y2 ) is given by ((x1 −x2 )2
+ (y1 − y2 )2 ). See shortest path
Excess demand In a market at a given price P, excess demand is the amount if
any by which local demand exceeds local supply. Local here excludes demand or
supply by arbitrageurs.
Excess profit Any profit in excess of normal profit, that is, profit in excess of the
return attributable to an unpriced factor of production such as entrepreneurial skill
or owner equity. See also Ricardian rent and monopoly excess profit.
Excess supply In a market at a given price P, excess supply is the amount if any
by which local supply exceeds local demand. Local here excludes demand or supply
by arbitrageurs.
Exclusion Theorem In the Weber–Launhardt model, where input places
and market places lie along a straight line, no location can be more efficient
than one of these places. As a result of this theorem, we may use compar-
ative cost analysis at each place to find the best location. See also Hakimi
Theorem.
Exogenous In a model, an exogenous value is a given; a value used in the model to
predict other (endogenous) values. Usually, exogenous values include parameters,
independent (or predictor) variables, and lagged or nearby values of endogenous
variables. Usually, exogenous variables are not thought to be stochastic.
Expansion path See linear expansion path.
Expected value The expected value, E[X], for a discrete stochastic variable X is
defined as E[X] = x xP[X = x] where x is a realization of X (i.e., a value that X
398 Appendix B
can take on), P[X = x] is the probability of that realization, and the summation is
over all possible realizations.
Expendable An attribute of consumer demand for a product such that there
is a price above which the consumer demands none of it. The opposite of an
expendable commodity is an indispensable commodity: i.e., a commodity, which
must be consumed in some quantity, however small, even when its price is
high.
Externality A consequence on one economic actor (e.g., a firm) arising from the
behavior of another that is not priced. Also known as a spillover.
Fetter Frank A. Fetter (born 1863), an American economist, authored “The
economic law of market areas” published in the Quarterly Journal of Economics
in 1924.
Fiat money economy Any national economy in which the medium of exchange is
a paper currency issued on behalf of the national government that does not obligate
it to convert that currency into another store of value (such as gold) on demand.
I assume here that the paper money supply is maintained by an authority (central
bank) to ensure that the currency is a good store of value (or, equivalently, that the
level of inflation is low).
Firm A group of persons engaged—that is, organized by contract or fiat rather
than market price mechanisms—in the production of a commodity (be it a com-
modity or service) for the purpose of earning a profit. In a competitive economy,
profit arises because the firm incurs lower cost than is possible relying on market
price mechanisms alone.
Fixed coefficients technology See Leontief technology.
Fixed cost A cost incurred by the firm for a period of operation that does not
depend on the quantity of output produced.
F.o.b. price Abbreviation for “free on board.” Firm sets price at factory, ware-
house, or store; customer pays freight from that place. Also known as mill pricing.
According to Marshall (1907, p. 325), the label “f.o.b.” arises from the practice
of merchants to quote a price for their commodity on board a vessel in port, each
purchaser then incurring any shipping cost from there.
Free entry A market condition in which nothing prevents new firms from entering
an industry other than considerations of profitability. If demand shifts so as to cause
a rise in market price, new firms enter the industry until the price is driven down
to the point where potential new firms find it is no longer profitable to get into that
industry. In competitive location theory, the concept of free entry is often associated
with the work of Lösch on imperfect competition in space. However, free entry is
also integral to perfect competition in general and to the supply curve over the long
run in particular.
Appendix B 399
Freight In this book, freight refers to the cost of shipping a quantity of commodity
from one place to another. For the most part, this means fuel, labor, maintenance,
and other costs associated with the shipment as though the firm was undertaking its
own shipping activity. Elsewhere, the book refers to shipping rates which presum-
ably are the prices charged by a firm for the service of picking up a shipment at one
place and delivering it to another. In this book, we do not distinguish between costs
and prices associated with shipping.
General equilibrium model A representation of an economy in which supply
and demand curves are endogenous.
Global net social welfare Net Social Welfare (See definition below) summed
over all regions and net of shipping costs.
Hakimi Theorem In the Weber–Launhardt model, on a transportation network
defined as consisting of M vertexes (each vertex at least one of a customer location,
a supplier location, or a place where two or more network segments intersect) with
accompanying network segments, no location can be more efficient than one or more
of the vertexes. As a result of this theorem, we may use comparative cost analysis
at each vertex to find the best location. See also Exclusion Theorem.
Half-freight A market outcome in which the prices at which a monopolist sells
the same commodity in two markets differ by half the difference in shipping costs
of shipping to the two markets. In general, this arises when consumers in the two
markets are identical and have linear demand curves.
Hitchcock Frank Lauren Hitchcock (born 1875), an American mathemati-
cian, published “The distribution of a product from several sources to numerous
localities” in Journal of Mathematics and Physics 20 (1941), pp. 224–230.
Hitchcock–Koopmans problem A Linear Program wherein the firm seeks to
minimize the sum of the costs of production at a set of factories, subject to capac-
ity constraints (one for each factory), and the shipments from these factories to
meet the demand requirements of customers at each distinct place. Also known as
the Transportation Linear Program, it was originally solved using a Stepping Stone
Algorithm. In general, linear programs are solved using the Simplex Algorithm.
Among others who also contributed in important ways to the Transportation Linear
Program and its solution were Boldyre, Dantzig, Ford and Fulkerson, Kantorovich,
Robinson, and Tolstoi.
Home market See Local Market.
Homotheticity See Linear Expansion Path.
Hoover Edgar Malone Hoover (born 1907), an American economist, published
Location Theory and the Shoe Leather Industries in 1937.
Hotelling Harold Hotelling (born 1895), an American economist and statisti-
cian, published a seminal article on locational competition entitled “Stability in
Competition” in the Economic Journal in 1929. See also Samuelson (1960).
400 Appendix B
Local market A local market is a set of agents (suppliers and demanders) engaged
in the sale and purchase of a commodity wherein market price—not a single price
set in a global market—varies from one local market to the next because of some
impediment (characterized by a unit shipping cost between markets). Where a
mechanism links participants so that market price varies systematically from local
market to local market (i.e., local markets are not in autarky), local markets can
be termed submarkets. A domestic market or home market each instance a local
market.
Local supply At a place or submarket, the supply of a product by firms from
local production. Local supply does not include supply offered by anyone in their
capacity as an arbitrageur importing for resale.
Localization economies Localization economies are reductions in unit production
cost that arise when several firms in the same industry locate in close proximity. For
some reason, having other firms in the same business in close proximity allows your
firm to be more efficient.
Log-linear utility function A log-linear utility function in two commodities takes
the form ln [U] = aln[q1 ] + (1 − a) ln [q2 ] or equivalently U = qa1 q1−a
2 where q1
is the quantity of commodity 1 and q2 is the quantity of commodity 2. This utility
function has the desirable properties that marginal utility is positive for each com-
modity and that there is diminishing marginal utility. At the same time, a log-linear
utility function has the special properties that each commodity is indispensable, that
the proportion of the consumer budget spent on each commodity is fixed (a for com-
modity 1, 1 – a for commodity 2), and that the cross- price elasticity of demand is
zero (in other words, the demand for one commodity is independent of the price of
the other commodity). See also utility function.
Logistics The management of production, inventory, and shipments so as to enable
a firm to achieve its objectives.
Lösch August Lösch (born 1906), a German economist, worked on equilib-
rium in a spatial economy. His main contribution was Die raumliche Ordnung der
Wirtschaft, published in German in 1939. An English translation of this book, The
Economics of Location, was first published in 1954.
Löschian Equilibrium In an industry where firms produce the same commodity
And compete by choosing geographic locations, a long-run equilibrium wherein
firms no longer have the incentive to enter or leave the industry. Every firm earns
normal profit only.
Marginal cost curve A schedule showing the marginal (additional) cost incurred
by the firm (usually over the short run wherein capital invested is held constant) as
a function of the quantity to be supplied.
Marginal cost of production The increment to the firm’s total cost incurred by
production of the last unit produced.
Appendix B 403
(a − b)/N and height f [xi ] for that panel. This is known as the midpoint rule version
of a Riemann sum.
Mill pricing See f.o.b. pricing.
Minimum price The lowest price at which a profit-maximizing firm would partic-
ipate in the market. Revenue just covers the variable cost of production. Minimum
price is not sustainable over the longer run because fixed costs of production are not
covered.
Model A simplified representation of the real world whose purpose is to describe,
simulate, explain, predict, or control a phenomenon of interest. Although models
can take a variety of forms from physical to chart to mathematical, I use the term
in this book to mean only something that can be expressed mathematically. In this
situation, a model includes values that are exogenously given (typically parameters
and exogenous or lagged variables) and equations that link these to the values of
endogenous variables.
Monopolist Firm can affect the price it receives for its product by varying quantity
it supplies to the market. In popular usage, a monopolist is sometimes thought to be
the only producer of a commodity. That is not the case here. Instead, I envisage the
firm to be (1) one of a small number of firms supplying a market and/or (2) produc-
ing a commodity differentiated from, but substitutable for, commodities produced
by rival firms.
Monopoly excess profit For the marginal firm, any profit in excess of normal
profit arising from the firm’s ability to exploit a downward-sloping demand curve.
See also Ricardian rent and excess profit.
Myopic Any behavior of a firm such that it does not foresee any reaction by its
competitors to its choices: e.g., with respect to price, quality of commodities sold,
or geographic location.
Net revenue See Profit.
Net social welfare For a region, the economic well-being to society as a whole:
measured either as (1) consumer surplus plus producer surplus, or (2) consumer
benefit minus producer cost. See global net social welfare.
No money illusion In a demand model, no money illusion means that price is
relative to the units in which other money variables are measured. Put differently, if
all money variables were to double, quantity demanded would be unchanged.
Non-spatial A feature of a model wherein geography plays no role. Typically,
shipping cost and/or commuting costs are assumed zero or are otherwise ignored.
Non-tariff barrier A limitation, other than a tariff, such as a quota or other
regulation that makes importing less attractive or feasible.
Normal profit The return attributable to an unpriced factor of production such as
entrepreneurial skill or owner equity.
Appendix B 405
value of e. However, for the benefit of noneconomists, I leave the negative sign to
remind the reader that when price goes up, we expect quantity demanded to drop.
Price risk A loss (or increase in cost) arising because of an unforeseen change in
market conditions that causes price to change over the short term; price risk is asso-
ciated with price volatility. In a search-theoretic perspective. sellers hold an asset
until the price bid by a potential purchaser exceeds the vendor’s reservation price.
Here, a distinction can be drawn between price risk and liquidity risk. Liquidity risk
is the loss arising because of the delay in obtaining a bid at or above the reservation
price. In practice, it is difficult to distinguish between price risk and liquidity risk.
The approach in this book is to treat liquidity risk as simply an element of price
risk.
Price taker A condition under which a market participant (supplier or demander)
is unable to affect the price they receive or pay for a unit of the product by varying
the quantity that they supply or demand. The supplier (demander) sees the demand
(supply) for its product as horizontal: i.e., infinitely elastic at the given market price.
Primal In Linear Programming, a problem in general cast as a maximization
subject to less-than-or-equal-to and nonnegativity constraints. See Dual.
Producer cost For a market, the aggregate cost incurred by suppliers; alterna-
tively, the area under the supply curve to the left of the quantity supplied.
Producer surplus The area under the market price and above the supply curve
to the left of the market equilibrium. In common parlance, producer surplus is the
amount (in dollars) received by efficient producers over and above what is needed
to secure the participation of the marginal producer in the market.
Profit The amount by which a firm’s revenue for a period exceeds its costs inclu-
sive of a normal return on any unpriced factors such as owner equity or management
skill. Also known as net revenue or excess profit.
Punctiform A commonly used abstraction of geographic space wherein eco-
nomic activity is clustered at distinct geographic points (places). Put differently,
economic activities themselves do not use land or otherwise occupy space. In this
abstraction, shipping between activities located at the same place incurs a negligi-
ble (zero) cost; shipping between activities at two different places incurs a nonzero
cost that is invariant with respect to the number or volume of economic activities
there.
Pythagorean formula The square of the length of the hypotenuse of a right angle
triangle in two-dimensional space is the sum of the squares of the other two sides.
Quadratic programming The solving of optimization problems in which a
quadratic objective function is to be maximized or minimized subject to linear
inequality and nonnegativity constraints.
Range For an expendable commodity sold at an f.o.b. price, the distance at which
shipping cost is sufficiently high to cause demand to drop to zero.
Appendix B 407
Rational By rational, I would like to mean simply that an economic actor makes
choices consistently:). Becker (1962). A consumer is deemed to be rational if, in
choosing between alternatives, he or she exhibits preferences among these alterna-
tives that are (1) well-ordered and (2) stable over time. Here well-ordered means that
if the consumer prefers alternative A to B and also prefers B to C, then that consumer
will prefer A to C. The phrase “stable over time” is to suggest that if the consumer
preferred A to B yesterday, then other things being equal, he or she will still pre-
fer A to B today. However, economists typically impose additional constraints on
rationality, including diminishing marginal utility, limitations on separability, and
notions of expected utility. For an interesting discussion of the nature and paradox
of rational choice, see Sugden (1991).
Rectangular plane In this book, a representation of the world as though it
were simply a two-dimensional surface. Ignored here, for simplicity of exposi-
tion, are (1) spherical properties of a globe and (2) differences in elevation. On
a rectangular plane, a place can be represented simply by a pair of Cartesian
coordinates.
Reductionism An approach to understanding which focuses on an aspect of the
process under study so as to better analyze the problem. In that way, we hope to
better our understanding of the process under study: i.e., our ability to describe,
simulate, explain, predict, or control it.
Relativist A relativist sees explanations simply as bettering current understanding.
Ricardian Rent An excess profit that arises because of an asset or market situation
unique to a firm that prevents competitors from entering the market and/or earning
the same profit.
Samuelson Paul Anthony Samuelson (born 1915), an American economist
and Nobel Laureate in 1970, published “Spatial price equilibrium and linear
programming” in the American Economic Review in 1952.
Samuelson conjecture Samuelson conjectured that if trade was to be desirable,
it had to improve well-being across regions. He proposed that global net social wel-
fare (GNSW) be measured as the sum of net social welfare for all regions less the
shipping costs (treated here as a deadweight loss.
Semi-net revenue For a firm, revenue minus variable cost. In this book, variable
cost includes both production and shipment. The firm’s net revenue (profit) is semi-
net revenue minus fixed cost.
Shadow price A standard Linear Programming problem maximizes an objective
function linear in endogenous variables subject to linear constraints and nonneg-
ativity constraints on the endogenous variables. A shadow price, one for each
linear constraint, is the amount by which the value of the objective function could
be increased if only the constraint were relaxed by one unit (made one unit less
binding).
408 Appendix B
values and provides the basis for the relationships used to solve for the endogenous
variables. The other is that the theory is a set of deductions arising from a set of
assumptions. The conundrum is that it is not easy to move back and forth between
these two ways of viewing theory. It would be nice, for example, if assumptions cor-
responded to givens and deductions to relationships. However, there is no particular
reason to expect this. This makes for some interpretation of models; we cannot just
look at a mathematical structure.
Thick A market condition in which there are many buyers and sellers. From a
search-theoretic perspective on markets a seller in a thick market does not have to
wait long to get a fair price for their commodity.
Thiessen Alfred Henry Thiessen (born 1872), an American climatologist, defined
polygons around individual rainfall stations to estimate total rainfall across a
region.
Thiessen polygon A map polygon formed on a rectangular plane by constructing
perpendicular bisectors to the straight lines joining a place (typically a store) to
similar places nearby. The partitioning of a map in this way is called a Voronoi
Diagram. The map of the lines for which perpendicular bisectors are drawn is called
the Delaunay triangulation.
Thin A market condition in which there are few buyers and sellers. From a search-
theoretic perspective on markets a seller in a thin market typically must wait longer
to get a fair price for their commodity.
Thünen Johann Heinrich von Thünen (born 1783), a German economist, pub-
lished the original version of The Isolated State in 1826. An English translation
of this book, Isolated State: An English Edition of Der Isolierte Staat, was first
published in 1966.
Total cost For a firm, the sum of variable and fixed costs of production inclusive
of any unpriced resources such as entrepreneurial talent.
Trade area The geographic area covered by a market in which a firm (often, a
store) participates. The trade area is thought to be composed of market areas for
each store in the market. See market area.
Transportation Linear Program See Hitchcock–Koopmans problem.
Ubiquitous A feature of an economic landscape wherein some input (factor of
production) is available everywhere at the same price.
Uniform pricing The practice of a monopolist by which the same price is set
for different customer places even though the marginal cost of serving a customer
differs from one place to the next.
Unit shipping cost See unit transaction cost.
Unit shipping rate Unit transaction cost per kilometer shipped.
410 Appendix B
Unit transaction cost Cost incurred by buyer related to search and informa-
tion gathering, negotiation, and acquisition (including freight and transfer, storage
and inventory, agency and brokerage fees, credit, cost of insurance and other loss
risks, installation and removal, warranty and service, and taxes and tariffs) per unit
shipped (purchased) inclusive of normal profit (the profit attributable to an unpriced
factor of production such as entrepreneurial skill or owner equity). Unit transaction
cost may vary seasonally. Throughout this book, this is termed unit shipping cost.
Urbanization economies Urbanization economies are reductions in unit produc-
tion cost made possible when firms in different industries locate in close proximity.
For some reason, having a diversity of firms in other businesses nearby makes your
firm more efficient.
Utility function An ordered (i.e., ordinal) scoring of choices (bundles of com-
modities whose consumption is desirable) that reflect consumer preferences and
that are transitive and evidence diminishing marginal utility. As a ranking, a given
utility function is said to be unique up to a monotonic transformation. For example,
the utility functions f [x, y] = xb y1−b where 1 < a < 0 and g[x, y] = axb yc where
b > 0, c > 0, and b + c < 1, calculated at consumption of x units of commodity 1
and y units of commodity 2, generate the same rank ordering: i.e., g[x, y] is a mono-
tonic transformation of f [x, y]. The two utility functions above exhibit diminishing
marginal utility. Also see “Rational.”
Variable cost A cost incurred by the firm for a period of operation that varies with
the quantity of output produced.
Von Thünen See Thünen.
Voronoi Georgy Fedoseevich Voronoi (born 1868), a Russian mathematician
worked on polygonal partitioning of a two-dimensional plane. This follows on ear-
lier work by Rene Descartes (1644) and Johann Peter Gustav Lejeune Dirchlet
(1850).
Voronoi diagram See Thiessen polygon.
Walras Marie-Ésprit Léon Walras (born 1834), a French (Swiss) economist, first
published his Élements d’économie politique pure in 1874.
Weber Alfred Weber (born 1868), a German economist, published über den
Standort der Industrie in 1909. An English translation of this book, Theory of the
Location of Industries, was first published in 1929.
Weber–Launhardt map triangle In the I = 2, J = 1 version of the Weber–
Launhardt Model, the triangle formed when three map points on a rectangular
plane—each with a pair of Cartesian coordinates—are drawn on a map.
Weber–Launhardt weight triangle In the I = 2, J = 1 version of the Weber–
Launhardt Model, the “equilibrium of forces” triangle created using the weights w1 ,
w2 , and wm . See Marginal Shipping Cost.
Bibliography
Adams, R. M., Fleming, R. A., Chang, C. C., McCarl, B., & Rosenzweig, C. (1996). A reassess-
ment of the economic effects of global climate change on U.S. agriculture. Climate Change,
30(2), 147–165.
Aguirre, I., Espinosa, M. P., & Macho-Stadler, I. (1998). Strategic entry deterrence
through spatial price discrimination. Regional Science and Urban Economics, 28(2),
297–314.
Aitken, S. C., & Bjorklund, E. M. (1988). Transactional and transformational theories in behavioral
geography. Professional Geographer, 40(1), 54–64.
Aji, M. A. (1995). Intra-metropolitan productivity variations of selected manufacturing and
business service sectors: What can we learn from Los Angeles? Urban Studies, 32(7),
1081–1096.
Alao, N. (1974). An approach to intraurban location theory. Economic Geography, 50(1), 59–69.
Alba, R. D., & Logan, J. R. (1993). Minority proximity to whites in suburbs – an individual-level
analysis of segregation. American Journal of Sociology, 98(6), 1388–1427.
Alcaly, R. E. (1976). Transportation and urban land values—review of theoretical literature. Land
Economics, 52(1), 42–53.
Alchian, A. A. (1950). Uncertainty, evolution, and economic theory. Journal of Political Economy,
58(3), 211–221.
Alderighi, M. & Piga, C. A. (2008). Why should a firm choose to limit the size of its market area?
Regional Science and Urban Economics, 38(1), 191–201.
Alexandersson, G. (1961). Changes in the location pattern of the Anglo-American steel industry:
1948–1959. Economic Geography, 37(2), 95–114.
Allen, W. B. (1977). Transportation Economics—Mohring, H. Land Economics, 53(2), 257–261.
Alonso, W. (1964). Location and land use: Toward a general theory of land rent. Cambridge,
Mass.: Harvard University Press.
Alperovich, G., & Katz, E. (1983). Transport rate uncertainty and the optimal location of the firm.
Journal of Regional Science, 23(3), 389–396.
Anas, A., Arnott, R., & Small, K. A. (1998). Urban spatial structure. Journal of Economic
Literature, 36(3), 1426–1464.
Anderson, S. P., de Palma, A., & Hong, G.-S. (1992). Firm mobility and location equilibrium.
Canadian Journal of Economics, 25(1), 76–88.
Anderson, S. P., & Ginsberg, V. A. (1999). International pricing with costly consumer arbitrage.
Journal of International Economics, 7(1), 126–139.
Anderson, S. P., Goeree, J. K., & Ramer, R. (1997). Location, location, location. Journal of
Economic Theory, 77, 102–127.
Appa, G. M. (1973). The transportation problem and its variants. Operational Research Quarterly,
24(1), 79–99.
Aranya, R. (2008). Location theory in reverse? Location for global production in the IT industry
of Bangalore. Environment and Planning A, 40(2), 446–463.
411
412 Bibliography
Baumol, W. J., Panzar, J. C., & Willig, R. D. (1982). Contestable markets and the theory of industry
structure. New York: Harcourt Brace Jovanvich.
Becker, G. S. (1962). Irrational behavior and economic theory. Journal of Political Economy, 70(1),
1–13.
Beckmann, M. (1952). A continuous model of transportation. Econometrica 20(4), 643–660.
Beckmann, M. J. (1968). Location theory. New York: Random House.
Beckmann, M. J. (1971). Market share, distance, and potential. Regional and Urban Economics,
1, 3–18.
Beckmann, M. J. (1972). Thünen revisited-neoclassical land use model. Swedish Journal of
Economics, 74(1), 1–6.
Beckmann, M. J. (1999). Lectures on location theory. Berlin: Springer-Verlag.
Beckmann, M. J., & Marschak, T. (1955). An activity analysis approach to location theory. Kyklos,
8, 125–143.
Beckmann, M. J., & Thisse, J. (1986a). The location of production activities. In P. Nijkamp (Ed.),
Handbook of regional and urban economics (pp. 24–27). Amsterdam: Elsevier.
Beckmann, M. J., & Thisse, J. F. (1986b). The location of production activities. In P. Nijkamp
(Ed.), Handbook of regional and urban economics (Vol. 1, pp. 49–62). Amsterdam: Elsevier.
Beckmann, M. J., & Thisse, J. F. (1987). The location of production activities. In E. S. Mills (Ed.),
Handbook of regional and urban economics (pp. 21–95). Amsterdam: North-Holland.
Belderbos, R., & Sleuwaegen, L. (1998). Tariff jumping DFI and export substitution: Japanese
electronics firms in Europe. International Journal of Industrial Organization, 16(5), 601–638.
Belderbos, R. A. (1997). Antidumping and tariff jumping: Japanese firms’ DFI in the European
Union and the United States. Weltwirtschaftliches Archiv, 133(3), 419–457.
Belding, R. (1981). A test of the von Thünen locational model of agricultural land-use with accoun-
tancy data from the European economic community. Transactions of the Institute of British
Geographers, 6(2), 176–187.
Belleflamme, P., Picard, P., Thisse, J. F. (2000). An economic theory of regional clusters. Journal
of Urban Economics, 48(1), 158–184.
Benassi, C., & Chirco, A. (2008). An elasticity approach to equilibrium and preference concentra-
tion in the Hotelling game. Journal of Economics, 94, 125–141.
Bennett, R. J., Graham, D. J., & Bratton, W. (1999). The location and concentration of businesses in
Britain: Business clusters, business services, market coverage and local economic development.
Transactions of the Institute of British Geographers, 24(4), 393–420.
Benson, B. L. (1980). Löschian competition under alternative demand conditions. American
Economic Review, 70(5), 1098–1105.
Berliant, M., & Dunz, K. (1995). Existence of equilibrium with nonconvexities and finitely many
agents. Journal of Mathematical Economics, 24(1), 83–93.
Berliant, M., & Dunz, K. (2004). A foundation of location theory: Existence of equilibrium, the
welfare theorems, and core. Journal of Mathematical Economics, 40(5), 593–618.
Berliant, M., & ten Raa, T. (1988). A foundation of location theory: Consumer preferences and
demand. Journal of Economic Theory, 44(2), 336–353.
Berliant, M., & ten Raa, T. (1992). Corrigendum [a foundation of location theory: Consumer
preference and demand]. Journal of Economic Theory, 58(1), 112–113.
Bernoulli, D. (1954). Exposition of a new theory on the measurement of risk. Econometrica, 22(1),
23–36. (Originally published in Latin in 1738; translated by Louise Sommer).
Berry, B. J. L., Parr, J. B., Epstein, B. J., Ghosh, A., & Smith, R. H. T. (1988). Market centers and
retail location: Theory and applications. Englewood Cliffs, NJ: Prentice-Hall.
Bester, H., de Palma, A., Leininger, W., Thomas, J., & Thadden, E.-L. von. (1996). A non-
cooperative analysis of Hotelling’s location game. Games and Economic Behavior, 12,
165–186.
Bhagwati, J. N., & Srinivas, T. N. (1974). Reanalyzing Harris-Todaro model – policy rankings in
case of sector-specific sticky wages. American Economic Review, 64(3), 502–508.
Bigman, D., & ReVelle, C. (1978). Theory of welfare considerations in public facility location
problems. Geographical Analysis, 10(3), 229–240.
414 Bibliography
Bigman, D., & ReVelle, C. (1979). Operational approach to welfare considerations in applied
public facility location models. Environment and Planning A, 11(1), 83–95.
Birkett, M. S. (1930). The iron and steel industry since the war. Journal of the Royal Statistical
Society, 93(3), 343–397.
Bishop, R. L. (1963). Game-theoretic analyses of bargaining. Quarterly Journal of Economics,
77(4), 559–602.
Black, D., & Henderson, V. (1999). Spatial evolution of population and industry in the United
States. American Economic Review, 89(2), 321–327.
Blackorby, C., Donaldson, D., & Moloney, D. (1984). Consumers surplus and welfare change in a
simple dynamic model. Review of Economic Studies, 51(1), 171–176.
Blaug, M. (1979). The German hegemony of location theory: A puzzle in the history of economic
thought. History of Political Economy, 11(1), 21–29.
Blaug, M. (Ed.) (1992). Johann von Thünen (1783–1850), Augustin Cournot (1801–1877), Jules
Dupuit (1804–1866). Elgar Reference Collection Series. Pioneers in Economics series (Vol.
24). Aldershot, UK: Elgar.
Block, D., & DuPuis, E. M. (2001). Making the country work for the city-von Thünen’s ideas
in geography, agricultural economics and the sociology of agriculture. American Journal of
Economics and Sociology, 60(1), 79–98.
Blomqvist, A. G. (1978). Urban job creation and unemployment in LDCs – Todaro vs. Harris and
Todaro. Journal of Development Economics, 5(1), 3–18.
Bobst, B. W., & Waananen, M. V. (1968). Cost and price effects of concentration restrictions in the
plant location problem. American Journal of Agricultural Economics, 50(3), 676–686.
Bodington, C. E., & Baker, T. E. (1990). A history of mathematical programming in the petroleum
industry. Interfaces, 20(4), 117–127.
Boots, B. N. (1980). Weighting Thiessen polygons. Economic Geography, 56(3), 248–259.
Borts, G. H., & Stein, J. L. (1964). Economic growth in a free market. New York: Columbia
University Press.
Bos, H. C. (1965). Spatial dispersion of economic activity. Rotterdam: Rotterdam University Press.
Boventer, E. von. (1961). The relationship between transportation costs and location rent in
transportation problems. Journal of Regional Science, 3(2), 27–40.
Bowles, S., & Gintis, H. (1988). Contested exchange: Political economy and modern economic
theory. American Economic Review, 78(2), 145–150.
Bowles, S., & Gintis, H. (2000). Walrasian economics in retrospect. Quarterly Journal of
Economics, 115(4), 1411–1439.
Bowring, J. (1986). Competition in a dual economy. Princeton, NJ: Princeton University Press.
Boyd, R. (1983a). Lumber transport and the Jones act: A multicommodity spatial equilibrium. Bell
Journal of Economics, 14, 202–212.
Boyd, R. (1983b). Costs and distortions of U.S. regulations policy in intercoastal lumber shipping.
Journal of Policy Modeling, 5(2), 293–307.
Boyd, R., Doroodian, K., & Abdul-Latif, S. (1993). The effects of tariff removals on the North
American lumber trade. Canadian Journal of Agricultural Economics, 41(3), 311–328.
Boyer, M., Laffont, J.-J., Mahene, P., & Moreaux, M. (1995). Sequential location equilibria under
incomplete information. Economic Theory, 6(2), 323–350.
Bradfield, M. (1971). Note on location and theory of production. Journal of Regional Science,
11(2), 263–266.
Braid, R. M. (1991). 2-dimensional Bertrand competition – block metric, Euclidean metric, and
waves of entry. Journal of Regional Science, 31(1), 35–48.
Braid, R. M. (2008). Spatial price discrimination and the locations of firms with different product
selections or product varieties. Economic Letters, 98, 342–347.
Brand, C. J. (1937). The fertilizer industry. Annals of the American Academy of Political and Social
Science, 193, 22–33.
Brander, J. A., & Spencer, B. J. (1987). Foreign direct investment with unemployment and
endogenous taxes and tariffs. Journal of International Economics, 22(3/4), 257–279.
Bibliography 415
Breedlove, L. B. (1932). The ice industry: Its economics and future. Journal of Land and Public
Utility Economics, 8(4), 404–417.
Bressler, R. G., & King, R. A. (1970). Markets, prices and interregional trade. New York: John
Wiley and Sons.
Brown, S. E. (1957). An index of relative market area strength. Annals of the Association of
American Geographers, 47(2), 155–155.
Brueckner, J. K., & Zenou, Y. (1999). Harris-Todaro models with a land market. Regional Science
and Urban Economics, 29(3), 317–339.
Buckley, P. J., & Casson, M. (1991). The future of the multinational enterprise (2nd ed.). London:
Macmillan.
Bukenya, J. O., & Labys, W. C. (2005). Price convergence on world commodity markets: Fact or
fiction? International Regional Science Review, 28(3), 302–329.
Bullard, C., & Engelbrecht-Wiggans, R. (1988). Intelligent data compression in a coal model.
Operations Research, 36(4), 521–531.
Bunge, W. (1966). Theoretical geography. Lund: C. W.K. Gleerup Publishers (published for Royal
University of Lund: Department of Geography).
Buongiorno, J., & Uusivuori, J. (1992). The law of one price in the trade of forest com-
modities: Co-integration tests for U.S. exports of pulp and paper. Forest Science, 38(3),
539–553.
Burns, M. E. (1973). Concept and measure of consumers surplus. American Economic Review,
63(3), 335–344.
Byers, J. A. (1996). Correct calculation of Dirichlet polygon areas. Journal of Animal Ecology,
65(4), 528–529.
Cai, D. H. (2008). An economic growth model with endogenous labor shift under a dual economy.
Applied Mathematics Letters, 21(8), 774–779.
Caldas, M., Walker, R., Perz, S., Arima, E., Aldrich, S., & Simmons, C. (2007). Theorizing land
cover and land use change: The peasant economy of Amazonian deforestation. Annals of the
Association of American Geographers, 97(1), 86–110.
Calvo, G. A. (1978). Urban unemployment and wage determination in LDCs – trade unions in
Harris-Todaro model. International Economic Review, 19(1), 65–81.
Camacho, A., & Persky, J. (1990). Variability and location. Environment and Planning A, 22(1),
69–77.
Capozza, D. R., & Attaran, K. (1976). Pricing in urban areas under free entry. Journal of Regional
Science, 16(2), 167–182.
Capozza, D. R., & Van Order, R. (1978). A generalized model of spatial competition. American
Economic Review, 68, 896–908.
Carlson, A. S., & Gow, C. B. (1936). Scrap iron and steel industry. Economic Geography, 12(2),
175–184.
Carrothers, G. A. P. (1956). An historical review of the gravity and potential models of human
interaction. Journal of the American Institute of Planners, 22(1), 94–102.
Casella, A. (2001). The role of market size in the formation of jurisdictions. Review of Economic
Studies, 68(1), 83–108.
Caves, R. E. (2007). Multinational enterprise and economic analysis (3rd ed.). Cambridge:
Cambridge University Press.
Chamberlin, E. H. (1962). The theory of monopolistic competition: A reorientation of the theory of
value (8th ed.). Cambridge, MA: Harvard University Press.
Charnes, A., & Cooper, W. W. (1954). The stepping stone method of explaining linear program-
ming calculations in linear programming problems. Management Science, 1(1), 49–69.
Charnes, A., & Cooper, W. W. (1957). Management models and industrial applications of linear
programming. Management Science, 4(1), 38–91.
Charnes, A., Cooper, W. W., & Henderson, A. (1953). An introduction to linear programming.
New York: John Wiley and Sons.
Chaudhuri, S. (2008). Wage inequality in a dual economy and international mobility of factors: Do
factor intensities always matter? Economic Modelling, 25(6), 1155–1164.
416 Bibliography
Chipman, J. S., & Moore, J. C. (1980). Compensating variation, consumers surplus, and welfare.
American Economic Review, 70(5), 933–949.
Chisholm, M. (1959). Economies of scale in road goods transport? Off-farm milk collection in
England and Wales. Oxford Economic Papers New Series, 11(3), 282–290.
Chisholm, M. (1961). Agricultural production, location, and rent. Oxford Economic Papers New
Series, 13(3), 342–359.
Chisholm, M. (1970). Geography and economics (2nd ed.). London: G. Bell & Sons.
Chisholm, M. (1979a). Rural settlement and land use: An essay in location (3rd ed.). London:
Hutchinson (First edition published in 1962: second edition in 1968).
Chisholm, M. (1979b). Von Thünen anticipated. Area, 11(1), 37–39.
Cho, C. J. (2001). Amenities and urban residential structure: An amenity-embedded model of
residential choice. Papers in Regional Science, 80(4), 483–498.
Clark, C. (1967). Von Thünen’s isolated state. Oxford Economic Papers, 19(3), 370–377.
Clark, J. B. (1887). The limits of competition. Political Science Quarterly, 2(1), 45–61.
Clark, W. A. V. (1992). Comparing cross-sectional and longitudinal analyses of residential mobility
and migration. Environment and Planning A, 24(9), 1291–1302.
Coase, R. H. (1937). The nature of the firm. Economica, 4(16), 386–405.
Coelho, P. R. P., & McClure, J. E. (2005). Theory versus application: Does complexity crowd out
evidence? Southern Economic Journal, 71(3), 556–565.
Coelho, P. R. P., & McClure, J. E. (2008). The market for lemmas: Evidence that complex models
rarely operate in our world. Economic Journal Watch, 5(1), 78–90.
Combes, P.-P., & Duranton, G. (2006). Labor pooling, labor poaching, and spatial clustering.
Regional Science and Urban Economics, 36(1), 1–28.
Combes, P.-P., Mayer, T., & Thisse, J. F. (2008). Economic geography: The integration of regions
and nations. Princeton, NJ: Princeton University Press.
Connor, J. M. (1981). Food product proliferation – a market structure analysis. American Journal
of Agricultural Economics, 63(4), 607–617.
Cook, G. A. S., Pandit, N. R., Beaverstock, J. V., Taylor, P. J., & Pain, K. (2007). The role of
location in knowledge creation and diffusion: Evidence of centripetal and centrifugal forces
in the city of London financial services agglomeration. Environment and Planning A, 39(6),
1325–1345.
Cooley, C. H. (1894). The Theory of Transportation. Publications of the American Economic
Association, 9(3), 13–148.
Couclelis, H., & Golledge, R. (1983). Analytic research, positivism, and behavioral geography.
Annals of the Association of American Geographers, 73(3), 331–339.
Coughlin, C. C., & Segev, E. (2000). Location determinants of new foreign-owned manufacturing
plants. Journal of Regional Science, 40(2), 323–351
Cournot, A. A. (1960). Researches into the mathematical principles of the theory of wealth.
(Translated by N. T. Bacon with an essay on Cournot and Mathematical Economics and
a Bibliography of Mathematical Economics by I. Fisher). New York: Augustus M. Kelley.
(originally published in French in 1838)
Cox, K. R., & Golledge, R. G. (Eds.). (1969). Behavioral problems in geography: A symposium.
Evanston, IL: Northwestern University Department of Geography.
Cox, K. R., & Golledge, R. G. (Eds.). (1981). Behavioral problems in geography revisited. New
York: Methuen.
Cromley, R. G. (1982). The von Thünen model and environmental uncertainty. Annals of the
Association of American Geographers, 72(3), 404–410.
Cuddington, J. T. (1993). Further results on the macroeconomic effects of AIDS – the dualistic,
labor-surplus economy. World Bank Economic Review, 7(3), 403–417.
Cunha, C. B., & Mutarelli, F. (2007). A spreadsheet-based optimization model for the integrated
problem of producing and distributing a major weekly newsmagazine. European Journal of
Operational Research, 176, 925–940.
Currie, J. M., Murphy, J. A., & Schmitz, A. (1971). The concept of economic surplus and its use
in economic analysis. Economic Journal, 81(324), 741–799.
Bibliography 417
Curry, L. (1964). The random spatial economy: An exploration in settlement theory. Annals of the
Association of American Geographers, 54(1), 138–146.
Curry, L. (1967). Quantitative geography, 1967. Canadian Geographer, 11(4), 265–279.
Curry, L. (1972). Division of labor from geographical competition. Annals of the Association of
American Geographers, 71(2), 133–165. (Reprinted in Curry (1998, 131–164)).
Curry, L. (1976a). Fluctuations in random spatial economy and control of inflation. Geographical
Analysis, 8(4), 339–353. (Reprinted in Curry (1998, 375–390)).
Curry, L. (1976b). Location theoretic style and urban policy. Economic Geography, 52(1), 11–23.
(Reprinted in Curry (1998, 345–358)).
Curry, L. (1978). Demand in spatial economy 2: Homo stochasticus. Geographical Analysis, 10(4),
309–344. (Reprinted in Curry (1998, 31–68)).
Curry, L. (1984a). Inefficiency of spatial prices using the thermodynamic formalism. Environment
and Planning A, 16(1), 5–16. (Reprinted in Curry (1998, 287–298)).
Curry, L. (1984b). A trade localization operator: The conservative case. Canadian Geographer,
28(4), 373–376.
Curry, L. (1985a). Inefficiencies in the geographical operations of labor markets. Regional Studies,
19(3), 203–215. (Reprinted in Curry (1998, 359–374)).
Curry, L. (1985b). Factor returns and geography. Geografiska Annaler B, 67(2), 69–81. (Reprinted
in Curry (1998, 97–112)).
Curry, L. (1985c). The village economy as a hypercycle. Geographical Analysis, 17(2), 156–166.
(Reprinted in Curry (1998, 425–438)).
Curry, L. (1986). Formalized behavior in agents markets. Tijdschrift voor Economische en Sociale
Geografie, 77(3), 169–177. (Reprinted in Curry (1998, 69–78)).
Curry, L. (1989). Spatial trade and factor markets. Economic Geography, 65(4), 271–279.
(Reprinted in Curry (1998, 331–344)).
Curry, L. (1998). The random spatial economy and its evolution. Aldershott UK: Ashgate.
Daitoh, I. (2008). Environmental protection and trade liberalization in a small open dual economy.
Review of Development Economics, 12(4), 728–736.
Daly, M. T., & Webber, M. J. (1973) Growth of firms within city. Urban Studies, 10(3), 303–317.
Damania, D. (1994). The scope for pure profits in a standard location model. Journal of Regional
Science, 34(1), 27–38.
Dana, J. D. Jr., & Spier, K. E. (2001). Revenue sharing and vertical control in the video rental
industry. Journal of Industrial Economics, 49(4), 223–245.
Daniel, T. E., & Goldberg, H. M. (1981). Dynamic equilibrium energy modeling: The Canadian
BALANCE model. Operations Research, 29(5), 829–852.
Dantzig, G. B. (1951a). Application of the simplex method to a transportation problem. In
T. C. Koopmans (Ed.), Activity analysis of production and allocation (pp. 222–259). New York:
John Wiley and Sons.
Dantzig, G. B. (1951b). Maximization of a linear function of variables subject to linear inequalities,
In T. C. Koopmans (Ed.), Activity analysis of production and allocation (pp. 339–348). New
York: John Wiley and Sons.
Dantzig, G. B. (1963). Linear programming and extensions. Princeton: Princeton University
Press.
Dantzig, G. B. (1991). Linear programming. In J. K. Lenstra, et al. (Eds.), History of mathematical
programming: A collection of personal reminiscences. Amsterdam: CWI. 19–31. (Reprinted in
2002 in Operations Research, 50(1), 42–47).
d’Aspremont, C., Gabszewicz, J. J., & Thisse, J. F. (1979). On Hotelling’s “Stability in
Competition”. Econometrica, 47(5), 1145–1150.
Davies, O. (1975). Examination of properties associated with isolated producer in 2-dimensional
spatial markets. Journal of Regional Science, 15(1), 47–52.
Davis, D. R. (1998). The home market, trade, and industrial structure. American Economic Review,
88(5), 1264–1276.
Davis, D. R., & Weinstein, D. E. (1999). Economic geography and regional production structure:
An empirical investigation. European Economic Review, 43(2), 379–407.
418 Bibliography
Davis, D. R., & Weinstein, D. E. (2003). Market access, economic geography, and comparative
advantage: An empirical test. Journal of International Economics, 59(1), 1–23.
Davis, P. (2006). Spatial competition in retail markets: Movie theaters. RAND Journal of
Economics, 37(4), 964–982.
Day, R. H. (1961). Recursive programming and supply prediction. In E. O. Heady, et al. (Eds.),
Agricultural supply functions – estimating techniques and interpretation (pp. 203–227). Ames:
Iowa State University Press.
Day, R. H., & Kennedy, P. E. (1970). Dynamic location model of production. Journal of Regional
Science, 10(2), 191–197.
Day, R. H., & Tinney, E. H. (1969). Dynamic von Thünen model. Geographical Analysis, 1(2),
137–151.
De Bernardi, M. (Ed.). (1933). Jules Dupuit: De L’Utilité et de sa Mesure. Torino: La Reforma
Sociale.
De Felice, F. (1972). An application of optimum location theory. Rivista Internazionale di Scienze
Economiche e Commerciali, 19(2), 168–173.
De Fraja, G., & Manenti, F. M. (2003). How long is a piece of cable? Equilibrium determination of
local telephone areas. Journal of Economics – Zeitschrift fur National Okonomie, 79(1), 61–85.
de Frutos, M. A., Hamoudi, H., & Jarque, X. (2002). Spatial competition with concave transport
costs. Regional Science and Urban Economics, 32, 531–540.
De Geer, S. (1927). The American manufacturing belt. Geografiska Annaler, 9, 233–359.
de Palma, A., Lindsey, R., von Hohenbalken, B., & West, D. S. (1994). Spatial price and vari-
ety competition in an urban retail market: A nested logit analysis. International Journal of
Industrial Organization, 12, 331–357.
de Smith, M. J. (1981). Optimum location theory – generalizations of some network problems and
some heuristic solutions. Journal of Regional Science, 21(4), 491–505.
De Vany, A., & Walls, W. D. (1996). The law of one price in a network: Arbitrage and price
dynamics in natural gas city gate markets. Journal of Regional Science, 36(4), 555–570.
De Vroey, M. (2007). Did the market-clearing postulate pre-exist new classical economics? The
case of Marshallian theory. The Manchester School of Economic and Social Studies, 75(3),
328–348.
Dean, R. D., & Carroll, T. M. (1977). Plant location under uncertainty. Land Economics, 53(4),
423–444.
Dean, R. D., Leahy, W. H., & McKee, D. L. (Eds.). (1970). Spatial economic theory. New York:
The Free Press.
DeCanio, S. J. (1984). Delivered pricing and multiple basing point equilibria: A reevaluation.
Quarterly Journal of Economics, 99(2), 329–349.
Demeza, D. (1983). The simple welfare economics of monopolistic competition. Journal of
Economic Studies, 10(1), 60–62.
Demeza, D., & von Ungersternberg, T. (1982). Monopoly, product diversity, and welfare. Regional
Science and Urban Economics, 12(3), 313–324.
Dempsey, B. W. (1960). The wage frontier: The economic organization of free agents. Chicago:
Loyola University Press. (Includes an English translation of Part II (Section 1) of Thünen’s The
Isolated State: See pp. 187–367)
Deneckere, R., Marvel, H. P., & Peck, J. (1996). Demand uncertainty, inventories, and resale price
maintenance, Quarterly Journal of Economics, 111(3), 885–913.
Dennis, K. (1982a). Economic theory and the problem of translation. Journal of Economic Issues,
16(3), 691–712.
Dennis, K. (1982b). Economic theory and the problem of translation 2. Journal of Economic Issues,
16(4), 1039–1062.
Dennis, K. (2002). Nominalising the numeric: An alternative to mathematical reduction in
economics. Cambridge Journal of Economics, 26(1), 63–80.
Deutsch, A. (1965). Joint commodities and basing point pricing – the case of caustic soda and
liquid chlorine. Canadian Journal of Economics and Political Science, 31(3), 397–401.
Bibliography 419
Fetter, F. A. (1924). The economic law of market areas. Quarterly Journal of Economics, 38,
520–529.
Fields, G. S. (1979). Welfare economic approach to growth and distribution in the dual economy.
Quarterly Journal of Economics, 93(3), 325–353.
Fik, T. J. (1988). Spatial competition and price reporting in retail food markets. Economic
Geography, 64(4), 29–44.
Fik, T. J., & Mulligan, G. F. (1991). Spatial price competition – a network approach. Geographical
Analysis, 23(1), 79–89.
Fisher, I., (1925). Mathematical investigations in the theory of value and prices. New Haven: Yale
University Press.
Fjeldsted, B. L., & South, J. B. (1979). A note on the multiregional multi-industry forecasting
model. Journal of Regional Science, 19(4), 483–491.
Flaam, S. D., & Godal, O. (2008). Market clearing and price formation. Journal of Economic
Dynamics and Control, 32(3), 956–977.
Flahaut, B., Laurent, M. A., & Thomas, I. (2002). Locating a community recycling center within a
residential area: A Belgian case study. 1. Professional Geographer, 54(1), 67–82.
Florian, M., & Los, M. (1982). A new look at static spatial price equilibrium models. Regional
Science and Urban Economics, 12(4), 579–597.
Ford, L. R., & Fulkerson, D. R. (1956). Solving the transportation problem. Management Science,
3(1), 24–32.
Fotheringham, A. S. (1979). Polarized growth within a multi growth center environment – case
study of the United States 1920–1970. Environment and Planning A, 11(2), 193–208.
Fotheringham, A. S. (1988). Market share analysis techniques: A review and illustration of current
US practice. In N. Wrigley (Ed.), Store choice, store location, and market analysis (pp. 120–
159). London: Routledge.
Fox, K. A. (1953). A spatial equilibrium model of the livestock-feed economy in the United States.
Econometrica, 21(4), 547–566.
Fox, K. A., & Taeuber, R. C. (1955). Spatial equilibrium models of the livestock-feed economy.
American Economic Review, 45(4), 584–608.
Francis, R. L., McGinnis, L. F., & White, J. A. (1983). Locational analysis. European Journal of
Operational Research, 12(3), 220–252.
Friedman, J. W., & Thisse, J. F. (1993). Partial collusion fosters minimum product differentiation.
RAND Journal of Economics, 24(4), 631–645.
Friedman, M. (1953). Essays in positive economics. Chicago: University of Chicago Press, 3–43.
Friesz, T. L., Harker, P. T., & Tobin, R. L. (1984). Alternative algorithms for the general network
spatial price equilibrium problem. Journal of Regional Science, 24(4), 475–507.
Fujita, M., & Krugman, P. (1995). When is the economy monocentric? von Thünen and Chamberlin
unified. Regional Science and Urban Economics, 25, 505–528.
Fujita, M., Krugman, P., & Venables, A. J. (1999). The spatial economy. Cambridge,
Massachusetts: MIT Press.
Fujita, M., & Thisse, J. F. (2009). New economic geography: An appraisal on the occasion of Paul
Krugman’s 2008 nobel prize in economic sciences. Regional Science and Urban Economics,
39(1), 109–119.
Furlong, W. J., & Slotsve, G. A. (1983). Will that be pickup or delivery? An alternative spatial
pricing policy. Bell Journal of Economics, 14, 271–274.
Furtran, W. H., Nagy, J. G., & Storey, G. G. (1979). The impact on the Canadian rapeseed industry
from changes in transport and tariff rates. American Journal of Agricultural Economics, 61(2),
238–248.
Gabriel, S. A., Vikas, S., & Ribar, D. M. (2000). Measuring the influence of Canadian carbon sta-
bilization programs on natural gas exports to the United States via a ‘bottom-up’ intertemporal
spatial price equilibrium model. Energy Economics, 22(5), 497–525.
Gabszewicz, J. J., & Thisse, J. F. (1986a). Spatial competition and the location of firms. In
R. Arnott (Ed.), Location theory (pp. 1–72). Chur: Harwood Academic Publishers.
422 Bibliography
Gabszewicz, J. J., & Thisse, J. F. (1986b). On the nature of competition with differentiated
products. Economic Journal, 96(381), 160–172.
Gabszewicz, J. J., & Thisse, J. F. (1992). Location. In R. J. Aumann, & S. Hart (Eds.), Handbook
of game theory with economic applications (pp. 281–304). Amsterdam: North-Holland.
Gale, D., Kuhn, H. W., & Tucker, A. W. (1951). Linear programming and the theory of games.
In T. C. Koopmans (Ed.), Activity analysis of production and allocation (pp. 317–329). New
York: John Wiley and Sons.
Gale, S. (1972). Inexactness, fuzzy sets, and foundations of behavioral geography. Geographical
Analysis, 4(4), 337–349.
Gannon, C. A. (1972). Consumer demand, conjectural variation and location equilibria in simple
spatial duopoly. Papers of the Regional Science Association, 28, 83–107.
Gannon, C. A. (1973). Optimization of market share in spatial competition. Southern Economic
Journal, 40(1), 66–79.
Garrison, W. L. (1959a). Spatial structure of the economy. Annals of the Association of American
Geographers, 49(2), 232–239.
Garrison, W. L. (1959b). Spatial structure of the economy II. Annals of the Association of American
Geographers, 49(4), 471–482.
Garrison, W. L., & Marble, D. F. (1957). The spatial structure of agricultural activities. Annals of
the Association of American Geographers, 47(2), 137–144.
Gass, S. I. (1990). On solving the transportation problem. Journal of the Operational Research
Society, 41(4), 291–297.
Gee, J. M. A. (1985). Competitive pricing for a spatial industry. Oxford Economic Papers, 37,
466–485.
Gehrig, T. (1998). Competing markets. European Economic Review, 42(2), 277–310.
Geoghegan, J., Cortina Villar, S., Klepeis, P., Macario Mendoza, P., Ogneva-Himmelberger, Y.,
Chowdhury, R. R., et al. (2001). Modeling tropical deforestation in the southern Yucatan penin-
sular region: Comparing survey and satellite data. Agriculture, Ecosystems, and Environment,
85(1–3), 25–46.
Georgescu-Roegen, N. (1954). Choice, expectations, and measurability. Quarterly Journal of
Economics, 68(4), 503–534.
Getnet, K., Verbeke, W., & Viaene, J. (2005). Modeling spatial price transmission in the grain mar-
kets of Ethiopia with an application of ARDL approach to white teff. Agricultural Economics,
33(3), 491–502.
Ghatak, S., Mulhern, A., & Watson, J. (2008). Inter-regional migration in transition economies:
The case of Poland. Review of Development Economics, 12(1), 209–222.
Ghosh, A., & McLafferty, S. L. (1987). Location Strategies for Retail and Service Firms. Lexington
MA: Lexington Books.
Ghosh, A., & Rushton, G. (Eds.). (1987). Spatial analysis and location-allocation models.
New York: Van Nostrand Reinhold Company.
Giarratani, F., Gruver, G., & Jackson, R. (2007). Clusters, agglomeration, and economic develop-
ment potential: Empirical evidence based on the advent of slab casting by U.S. steel minimills.
Economic Development Quarterly, 21(2), 148–164.
Giersch, H. (1949). Economic union between nations and the location of industries. Review of
Economic Studies, 17(2), 87–97.
Gillen, W. J., & Guccione, A. (1993). Search and Fetter law of markets. Geographical Analysis,
25(2), 165–170.
Gilligan, T. W. (1992). Imperfect competition and basing-point pricing: Evidence from the
softwood plywood industry. American Economic Review, 82(5), 1106–1119.
Glaeser, E. L., Kallal, H. D., Scheinkman, J. A., & Shleifer, A. (1992). Growth in cities. Journal
of Political Economy, 100(6), 1126–1152.
Glazer, A., Gradstein, M., & Ranjan, P. (2003). Consumption variety and urban agglomeration.
Regional Science and Urban Economics, 33(6), 653–661.
Goettler, R., & Shachar, R. (2001). Spatial competition in the network television industry. RAND
Journal of Economics, 32(4), 624–656.
Bibliography 423
Güder, F., & Morris, J. G. (1988). Objective function approximation: An application to spatial price
equilibrium models. American Journal of Agricultural Economics, 70(2), 391–396.
Güder, F., Morris, J. G., & Yoon, S. H. (1992). Parallel and serial successive overrelax-
ation for multicommodity spatial price equilibrium problems. Transportation Science, 26(1),
48–58.
Guise, J. W. B. (1979). An expository critique of the Takayama-Judge models of interregional and
intertemporal market equilibrium. Regional Science and Urban Economics, 9(1), 83–95.
Gupta, B., Lai, F.-C., Pal, D., Sarkar, J., & Yu, C.-M. (2004). Where to locate in a circular city?
International Journal of Industrial Organization, 22, 759–782.
Gupta, B., Pal D., & Sarkar, J. (1997). Spatial Cournot competition and agglomeration in a model
of location choice. Regional Science and Urban Economics, 27, 261–282.
Haddock, D. D. (1982). Basing-point pricing: Competitive vs. collusive theories. American
Economic Review, 72(3), 289–306.
Hadley, G. (1962). Linear programming. Reading, MA: Addison-Wesley.
Hadley, G. (1964). Nonlinear and dynamic programming. Reading, MA: Addison-Wesley.
Haggett, P., Cliff, A. D., & Frey, A. (1977). Locational analysis in human geography (2nd ed.).
New York: John Wiley & Sons.
Haig, R. M. (1926a). Toward an understanding of the metropolis I Some speculations regarding
the economic basis of urban concentration. Quarterly Journal of Economics, 40(2), 179–208.
Haig, R. M. (1926b). Toward an understanding of the metropolis II The assignment of activities to
areas in urban regions. Quarterly Journal of Economics, 40(3), 402–434.
Hakimi, S. L. (1964). Optimum location of switching centers and the absolute centers and medians
of a graph. Operations Research, 12, 450–459.
Hall, R. W. (1989). Graphical interpretation of the transportation problem. Transportation Science,
23(1), 37–45.
Hamacher, H. W., & Nickel, S. (1998). Classification of location models. Location Science, 6,
229–242.
Hamilton, J. H., & Thisse, J.-F. (1992). Duopoly with spatial and quantity-dependent price
discrimination. Regional Science and Urban Economics, 22, 175–185.
Hammond, S. (1942). Location theory and the cotton industry. Journal of Economic History,
2(Supplement), 101–117.
Hamoudi, H., & Moral, M. J. (2005). Equilibrium existence in the linear model: Concave versus
convex transportation costs. Papers in Regional Science, 84(2), 201–219
Handler, G. Y., & Marchandani, P. B. (1979). Location on networks: Theory and algorithms.
Cambridge, MA: MIT Press.
Hanemann, W. M. (1991). Willingness to pay and willingness to accept – how much can they
differ? American Economic Review, 81(3), 635–647.
Hanink, D. M. (1995). Evaluation of wilderness in a spatial context. Growth and Change, 26(3),
425–441.
Hanink, D. M., & Cromley, R. G. (2005). Geographic change with trade based on comparative
advantage. Annals of the Association of American Geographers, 95(3), 511–524.
Hanson, G. H. (1996). Localization economies, vertical organization, and trade. American
Economic Review, 86(5), 1266–1278
Hanson, G. H., & Xiang, X. (2004). The home-market effect and bilateral trade patterns. American
Economic Review, 94(4), 1108–1129.
Hardie, I., Parks, P., Gottleib, P., & Wear, D. (2000). Responsiveness of rural and urban land uses
to land rent determinants in the US south. Land Economics, 76(4), 659–673.
Harker, P. T. (Ed.). (1985). Spatial price equilibrium: Advances in theory, computation and
application. Berlin: Springer-Verlag.
Harker, P. T. (1986). The spatial price equilibrium problem with path variables. Socio-Economic
Planning Sciences, 20(5), 299–310.
Harker, P. T. (1987). The core of a spatial price equilibrium game. Journal of Regional Science,
27(3), 369–389.
Bibliography 425
Harker, P. T., & Pang, J.-S. (1990). Finite-dimensional variational inequality and nonlinear
complementarity problems: A survey. Mathematical Programming, 48(2), 161–220.
Harley, C. K. (1988). Ocean freight rates and productivity, 1740–1913—The primacy of mechani-
cal invention reaffirmed. Journal of Economic History, 48(4), 851–876.
Harris, C. C. Jr. (1973). The urban economies, 1985. Lexington: D.C. Heath and Company.
Harris, C. C. Jr. (1980). New developments and extensions of the multiregional, multi-industry
forecasting model. Journal of Regional Science, 20(2), 159–171.
Harris, C. C. Jr., & Hopkins, F. E. (1972). Locational analysis: An interregional econometric model
of agriculture, mining, manufacturing, and services. Lexington, MA: Lexington Books.
Harris, C. C. Jr., & Nadji, M. (1987). Regional economic modeling in theory and in practice.
Lanham, Md. and London: University Press of America.
Harris, J. R., & Todaro, M. (1970). Migration, unemployment, and development: A two-sector
analysis. American Economic Review, 60(1), 126–142.
Harris, R. G. (1977). Economies of traffic density in rail freight industry. Bell Journal of
Economics, 8(2), 556–564.
Harsanyi, J. C. (1956). Approaches to the bargaining problem before and after the theory of
games: A critical discussion of Zeuthen’s, Hick’s, and Nash’s theories. Econometrica, 24(2),
144–157.
Harsanyi, J. C. (1965). Bargaining and conflict situations in the light of a new approach to game
theory. American Economic Review, 55(1/2), 447–457.
Harsanyi, J. C. (1966). A general theory of rational behavior in game situations. Econometrica,
34(3), 613–634.
Hart, A. B. (1890). The rise of American cities. Quarterly Journal of Economics, 4(2),
129–157.
Hart, O. D. (1979). Monopolistic competition in a large economy with differentiated commodities.
Review of Economic Studies, 46(1), 1–30.
Hart, O. D. (1985a). Monopolistic competition in the spirit of Chamberlin – special results.
Economic Journal, 95(380), 889–908.
Hart, O. D. (1985b). Monopolistic competition in the spirit of Chamberlin – a general model.
Review of Economic Studies, 52(4), 529–546.
Hart, P. W. E. (1980). Problems and potentialities of the behavioural approach to agricultural
location. Geografiska Annaler Series B Human Geography, 62(2), 99–107.
Hartshorne, R. (1927). Location as a factor in geography. Annals of the Association of American
Geographers, 17(2), 92–99.
Hartwick, J. M. (1973a). Lösch’s theorem on hexagonal market areas. Journal of Regional Science,
13(2), 213–221.
Hartwick, J. M. (1973b). The pricing of commodities and agricultural land in multiregional general
equilibrium. Southern Economic Journal, 39(1), 31–45.
Hartwick, J. M. (1998). Location theory, vol. 2. Regional Science and Urban Economics, 28(2),
237–242.
Harvey, D. W. (1966). Theoretical concepts and the analysis of agricultural land-use patterns in
Geography. Annals of the Association of American Geographers, 56(2), 361–374.
Hausmann, J. A. (1981). Exact consumers surplus and deadweight loss. American Economic
Review, 71(4), 662–676.
Hay, D. A. (1976). Sequential entry and entry-deterring strategies in spatial competition. Oxford
Economic Papers – New Series, 28(2), 240–257.
Head, K., & Ries, J. (2001). Increasing returns versus national product differentiation as an
explanation for the pattern of U.S.-Canada trade. American Economic Review, 91(4), 858–876.
Head, K., Ries, J., & Swenson, D. (1995). Agglomeration benefits and location choice – evi-
dence from Japanese manufacturing investments in the United States. Journal of International
Economics, 38(3–4), 223–247.
Heckscher, E. (1919). Utrikshandelns Verkan pa lnkomstfordelningen. Ekonomist Tidskriji, 21,
1–32. Translated as “The effect of foreign trade on the distribution of income” (Reprinted in
426 Bibliography
H. S. Ellis & L. A. Metzler (Eds.) (1949). Readings in the Theory of international Trade.
Philadelphia: Blakiston Co. Reprinted in H. Flam, & M. J. Flanders (Eds.) (1991). Heckscher-
Ohlin Trade Theory. Cambridge MA: MIT Press).
Henderson, J. M. (1955). A short-run model for the coal industry. Review of Economics and
Statistics, 37(4), 336–346.
Henderson, J. V. (1988). Urban development: Theory, fact, and illusion. New York: Oxford
University Press.
Henkel, J., Stahl, K., & Walz, U. (2000). Coalition building in a spatial economy. Journal of Urban
Economics, 47(1), 136–163.
Herbert, J. D., & Stevens, B. H. (1960). A model for the distribution of residential activity in urban
areas. Journal of Regional Science, 2(2), 21–36.
Hicks, J. R. (1956). A revision of demand theory. Oxford: Clarendon Press.
Hitchcock, F. L. (1941). The distribution of a product from several sources to numerous localities.
Journal of Mathematics and Physics, 20, 224–230.
Hoare, A. G. (1978). Industrial linkages and dual economy – case of Northern Ireland. Regional
Studies, 12(2), 167–180.
Hohenbalken, B. von, & West, D. S. (1986). Empirical tests for predatory reputation. Canadian
Journal of Economics, 19(1), 160–178.
Holmes, T. J., & Stevens, J. J. (2004). Geographic concentration and establishment size:
Analysis in an alternative economic geography model. Journal of Economic Geography, 4(3),
227–250.
Hoover, E. M. (1937). Location theory and the shoe leather industries. Cambridge, MA: Harvard
University Press.
Horn, H., & Persson, L. (2001). The equilibrium ownership of an international oligopoly. Journal
of International Economics, 53(2), 307–333.
Hornby, J. M. (1968). Investment and trade policy in dual economy. Economic Journal, 78(309),
96–107.
Horwath, R. J. (1969). Thünen’s isolated state and area around Addis-Ababa. Annals of the
Association of American Geographers, 59(2), 308–323.
Hotelling, H. (1929). Stability in competition. Economic Journal, 39, 41–57.
Houthakker, H. S. (1955). On the numerical solution of the transportation problem. Journal of the
Operations Research Society of America, 3(2), 210–214.
Hsu, C. I., & Tsai, I. J. (1999). Logistics cost, consumer demand, and retail establishment density.
Papers in Regional Science, 78(3), 243–263.
Hsu, S. K. (1997). The agroindustry: A neglected aspect of the location theory of manufacturing.
Journal of Regional Science, 37(2), 259–274.
Hsu, S. K., & Mai, C. C. (1984). Production location and random input price. Regional Science
and Urban Economics, 14(1), 45–62.
Hsu, S. K., & Tan, L. T. (1999). Agroindustry location under output price uncertainty. Annals of
Regional Science, 33(3), 289–303.
Huang, T. (2009). Hotelling competition with demand on parallel line. Economic Letters, 102(1),
155–157.
Huck, S., Müller, W., & Vriend, N. J. (2002). The east end, the west end, and King’s cross: On
clustering inn the four-player Hotelling game. Economic Inquiry, 40(2), 231–240.
Huff, D. L. (1963). A probabilistic analysis of shopping center trade areas. Land Economics, 39(1),
81–90.
Huff, D. L. (1964). Defining and estimating a trade area. Journal of Marketing, 28(1), 34–38.
Huff, D. L., & Jenks, G. F. (1968). Graphic interpretation of friction of distance in gravity models.
Annals of the Association of American Geographers, 58(4), 814–824.
Hughes, J. W., & Barbezat, D. P. (1996). Basing-point pricing and the Stahlwerksverband:
An examination of the “New Competitive School”. Journal of Economic History, 56(1),
215–22.
Hummels, D. (2007). Transportation costs and international trade in the second era of globalization.
Journal of Economic Perspectives, 21(3), 131–154.
Bibliography 427
Huntingdon, E. (1927). The quantitative phases of human geography. Scientific Monthly, 25(4),
289–305.
Huriot, J.-M., & Thisse, J.-F. (Eds.). (2000). Economics of cities. Theoretical perspectives.
Cambridge, UK: Cambridge University Press.
Hurter, A. P. Jr., & Martinich, J. S. (1989). Facility location and the theory of production. Boston:
Kluwer Academic Publishers.
Hurter, A. P. Jr., & Wendell, R. E. (1972). Location and production—a special case. Journal of
Regional Science, 12(2), 243–247.
Hurwicz, L. (1953). What has happened to the theory of games? American Economic Review,
43(2), 398–405.
Hurwicz, L. (1973). The design of mechanisms for resource allocation. American Economic
Review, 63(2), 1–30.
Hwang, H., & Mai, C. C. (2002). The tariff-jumping argument and location theory. Review of
International Economics, 10(2), 361–368.
Hwang, H., Mai, C. C., & Shieh, Y. N. (2007). Production-location decision and free entry
oligopoly: A correction. Journal of Urban Economics, 61(3), 576–679.
Hyson, C. D., & Hyson, W. P. (1950). The economic law of market areas. Quarterly Journal of
Economics, 64(2), 319–327.
Inaba, F. S., & Wallace, N. E. (1989). Spatial price competition and the demand for freight
transportation. Review of Economics and Statistics, 71(4), 614–625.
Ioannides, Y. M. (1994). Product differentiation and economic growth in a system of cities.
Regional Science and Urban Economics, 24(4), 461–484.
Ioffe, G., & Nefedova, T. (2001). Land use changes in the environs of Moscow. Area, 33(3),
273–286.
Irmen, A., & Thisse, J.-F. (1998). Competition in multi-characteristics spaces: Hotelling was
almost right. Journal of Economic Theory, 78, 76–102.
Isard, W. (1949). The general theory of location and space-economy. Quarterly Journal of
Economics, 63(4), 476–506.
Isard, W. (1951). The locational equilibrium of the firm. Quarterly Journal of Economics, 65(3),
379–399.
Isard, W. (1954). Location theory and trade theory: Short-run analysis. Quarterly Journal of
Economics, 68(2), 305–320.
Isard, W. (1956). Location and space-economy. New York: Wiley.
Isard, W., & Smith, T. (1967). Location games: With application to classic location problems.
Papers of the Regional Science Association, 19, 45–80.
Ishikawa, T., & Toda, M. (1995). An unequal spatial structure of location and price with consumer
density. Economic Record, 71(213), 167–178.
Islam, N., & Yokota, K. (2008). Lewis growth model and China’s industrialization. Asian Economic
Journal, 22(4), 359–396.
Iyer, G., & Seetharaman, P. B. (2008). Too close to be similar: Product and price competition in
retail gas markets. Quantitative Marketing Economics, 6, 205–234.
Jacobs, J. (1969). The economy of cities. New York: Vintage.
Jacobs, J. (1984). Cities and the wealth of nations: Principles of economic life. New York: Vantage.
Javorcik, B. S. (2004). Does foreign direct investment increase the productivity of domestic firms?
In search of spillovers through backward linkages. American Economic Review, 94(3) 605–627.
Jayet, H. (1997). Collective location. Annales d’Économie et de Statistique, 45, 139–146.
Jevons, H. S. (1931). The second industrial revolution. Economic Journal, 41(116), 1–18.
Johansson, B., Karlsson, C., & Stough, R. R. (Eds.). (2001). Theories of endogenous regional
growth: Lessons for regional policy. Berlin, Germany: Springer.
Johansson, B., & Quigley, J. M. (2004). Agglomeration and networks in spatial economies. Papers
in Regional Science, 83(1), 165–176.
John, F. (1948). Extremum problems with inequalities as subsidiary conditions. Studies and Essays
Presented to R Courant on his 60th Birthday, Jan 8, 1948. New York: Interscience Publishers
Inc. 187–204.
428 Bibliography
Kennedy, M. (1974). An economic model of the world oil market. Bell Journal of Economics and
Management Science, 5(2), 540–577.
Khalili, A., Mathur, V. K., & Bodenhorn, D. (1974). Location and the theory of production: A
generalization. Journal of Economic Theory, 9(4), 465–475.
Khan, M. A. (1980). The Harris-Todaro hypothesis and the Heckscher-Ohlin-Samuelson trade
model- a synthesis. Journal of International Economics, 10(4), 527–547.
Kilkenny, M., & Thisse, J.-F. (1999). Economics of location: A selective survey. Computers &
Operations Research, 26(14), 1369–1394.
Kim, S. (1995). Expansion of markets and the geographic distribution of economic activities: The
trends in U.S. regional manufacturing structure, 1860–1987. Quarterly Journal of Economics,
110(4), 881–908.
Kitsikopoulos, H. (2003). Urban demand and agrarian productivity in pre-plague England:
Reassessing the relevancy of von Thünen’s model. Agricultural History, 77(3), 482–522.
Knoblauch, V. (2002). An easy proof that a square lattice is an equilibrium for spatial competition
in the plane. Journal of Urban Economics, 51(1), 46–53.
Koo, A. Y. C. (1959). Recurrent objections to the minimax strategy. Review of Economics &
Statistics, 41(1), 36–41.
Koopmans, T. C. (1949). Optimum utilization of the transportation system. Econometrica, 17
(Supplement), 136–146. (Originally presented in 1947)
Koopmans, T. C. (Ed.). (1951). Activity analysis of production and allocation. New York: John
Wiley and Sons.
Koopmans, T. C., & Beckmann, M. (1957). Assignment problems and the location of economic
activities. Econometrica, 25, 53–76.
Koopmans, T. C., & Reiter, S. (1951). A Model of Transportation. In Koopmans, TC (ed.). Activity
Analysis of Production and Allocation. Cowles Commission Monograph 13. New York: John
Wiley and Sons. pp. 222–260.
Koropeckyj, I. S. (1967). The development of soviet location theory before the second world war:
I. Soviet Studies, 19(1), 1–28.
Kroll, Y., Levy, H., & Markowitz, H. M. (1984). Mean-variance versus direct utility maximization.
Journal of Finance, 39(1), 47–61.
Krugman, P. R. (1993a). First nature, second nature, and metropolitan location. Journal of Regional
Science, 33(2), 129–144.
Krugman, P. R. (1993b). On the relationship between trade theory and location theory. Review of
International Economics, 1(2), 110–122.
Krugman, P. R. (1995). Development, geography, and economic theory. Cambridge, USA: MIT
Press.
Krugman, P. R. (1998a). What’s new about the new economic geography? Oxford Review of
Economic Policy, 14(2), 7–17.
Krugman, P. R. (1998b). Space: The final frontier. Journal of Economic Perspectives, 12(2),
161–174.
Krumme, G. (1969). Toward a geography of enterprise. Economic Geography, 45(1), 30–40.
Krusell, P., & Smith, A. A. (1996). Rules of thumb in macroeconomic equilibrium – A quantitative
analysis. Journal of Economic Dynamics and Control, 20(4), 527–558.
Krzymowski, R., & Minneman, P. G. (1928). Graphical presentation of Thünen’s theory of
intensity. Journal of Farm Economics, 10(4), 461–482.
Krzyzanowski, W. (1927). Review of the literature of the location of industries. Journal of Political
Economy, 35(2), 278–291.
Kuhn, H. W. (1967). On a pair of dual nonlinear programs. In J. Abadie (Ed.), Nonlinear
programming (pp. 39–54). Amsterdam: North-Holland.
Kuhn, H. W., & Tucker, A. W. (1951). Nonlinear programming. In J. Neyman (Ed.), Proceedings
of the second Berkeley symposium on mathematical statistics and probability. Berkeley:
University of California Press.
Kuhn, H. W., & MacKinnon, J. G. (1975). Sandwich method for finding fixed points. Journal of
Optimization Theory and Applications, 17(3/4), 189–204.
430 Bibliography
Lentnek, B., MacPherson, A., & Phillips, D. (1992). Optimum producer-service location.
Environment and Planning A, 24, 467–479.
Leontief, W. W. (1951). The structure of the American economy, 1919–1939: An empirical
application of equilibrium analysis (2nd ed.). New York: Oxford University Press.
Lerner, A. P., & Singer, H. W. (1937). Some notes on duopoly and spatial competition. Journal of
Political Economy, 45(2), 145–196.
Levy, D. T., & Reitzes, J. D. (1993). Basing-point pricing and incomplete collusion. Journal of
Regional Science, 33(1), 27–35.
Levy, H., & Markowitz, H. M. (1979). Approximating expected utility by a function of mean and
variance. American Economic Review, 69(3), 308–317.
Lewis, G. M., & Brabec, E. (2005). Regional land pattern assessment: Development of a resource
efficiency measurement method. Landscape and Urban Planning, 72(4), 281–296.
Lewis, W. A. (1954). Development with unlimited supplies of labour. Manchester School of
Economic and Social Studies, 22, 139–192.
Lewis, W. A. (1979). Dual economy revisited. Manchester School of Economic and Social Studies,
47(3), 211–229.
Lindberg, O. (1953). An economic-geographical study of the localization of the Swedish paper
industry. Geografiska Annaler, 35(1), 28–40.
Liu, L. (1999). Labor location, conservation, and land quality: The case of West Jilin, China. Annals
of the Association of American Geographers, 89(4), 633–657.
Liu, L. (2000). Labor location and agricultural land use in Jilin, China. Professional Geographer,
52(1), 74–83.
Lo, L. (1990). A translog approach to consumer spatial behavior. Journal of Regional Science,
30(3), 393–413.
Lo, L. (1991a). Spatial structure and spatial interaction – a simulation approach. Environment and
Planning A, 23(9), 1279–1300.
Lo, L. (1991b). Substitutability, spatial structure, and spatial interaction. Geographical Analysis,
23(2), 132–146.
Lo, L. (1992). Destination interdependence and the competing destinations model. Environment
and Planning A, 24(8), 1191–1204.
Long, W. H. (1971). Demand in space: Some neglected aspects. Papers of the Regional Science
Association, 27, 45–60.
Loomes, G., & Sugden, R. (1982). Regret theory: An alternative theory of rational choice under
uncertainty. Economic Journal, 92(368), 805–824.
Lord, R. A., & Farr, W. K. (2003). Collusion and financial leverage: An analysis of the integrated
mill steel industry. Financial Management, 32(1), 127–148.
Lösch, A. (1954). The economics of location. New Haven: Yale University Press.
Louveaux, F., Thisse, J.-F., & Beguin, H. (1982). Location theory and transportation costs.
Regional Science and Urban Economics, 12(4), 529–545.
Lucas, M. T., & Chhajed, D. (2004). Applications of location analysis in agriculture: A survey.
Journal of the Operational Research Society, 55(6), 561–578.
Luna, H. P. L. (1978). Two-level national-regional planning and mathematical programming
decomposition applied to spatial price equilibrium models. Socio-economic Planning Sciences,
12(5), 251–266.
Macdonell, J. (1891). The regulation of railway rates. Economic Journal, 1(3), 490–508.
Machlup, F. (1949). The basing-point system. Philadelphia: Blakiston.
MacKinnon, J. G. (1975). An algorithm for the generalized transportation problem. Regional
Science and Urban Economics, 5(4), 445–464.
MacKinnon, J. G. (1976). A technique for the solution of spatial equilibrium models. Journal of
Regional Science, 16(3), 293–307.
Macmillan, W. D. (1992). Risk and agricultural land-use—a reformulation of the portfolio-
theoretic approach to the analysis of a von Thünen economy. Geographical Analysis, 23(2),
142–158.
432 Bibliography
Magnanti, T. L., Mirchandani, P., & Vachani, R. (1995). Modeling and solving the 2-facility
capacitated network loading problem. Operations Research, 43(1), 142–157.
Mai, C. C. (1984). Demand function and location theory of the firm under price uncertainty. Urban
Studies, 21(4), 459–464.
Mai, C. C. (1987). Demand function and location theory of the firm under price uncertainty: A
reply. Urban Studies, 24(2), 162.
Mai, C. C., & Hwang, H. (1992). Production-location decision and free entry oligopoly. Journal of
Urban Economics, 33, 257–271.
Mai, C. C., & Hwang, H. (1994). On a location theory under duopoly. Regional Science and Urban
Economics, 24(6), 773–784.
Mai, C. C., & Hwang, H. (1997). External economies of scale and optimal plant location. Pacific
Economic Review, 2(2), 115–124.
Mai, C. C., Suwanakul, S., & Yeh, C. N. (1993a). Monopsony and the production location decision
of the firm. Regional Science and Urban Economics, 23(6), 779–790.
Mai, C. C., Yeh, C. N., & Suwanakul, S. (1993b). Price uncertainty and production-location
decisions under free entry oligopoly. Journal of Regional Science, 33(4), 531–545.
Maki, U. (2004). Realism and the nature of theory: A lesson from J. H. von Thünen for economists
and geographers. Environment and Planning A, 36(10), 1719–1736.
Manne, A. S. (1967). Investment for capacity expansion: Size, location, and time phasing.
Cambridge, MA: MIT Press.
Marchionatti, R. (2004). What don’t economists know now that Marshall knew a century ago?
A note on Marshall’s sophisticated informality. Journal of Post Keynesian Economics, 26(3),
441–460.
Marcotte, P., Marquis, G., & Zubieta, L. (1992). A Newton-SQR method for spatial price
equilibrium. Transportation Science, 26(1), 36–47.
Mark, D. M., Freksa, C., Hirtle, S. C., Lloyd, R., & Tversky, B. (1999). Cognitive models
of geographical space. International Journal of Geographical Information Science, 13(8),
747–774.
Markowitz, H. M. (1952). Portfolio selection. Journal of Finance, 7(1), 77–91.
Markowitz, H. M. (1976). Markowitz revisited. Financial Analysts Journal, 32(5), 47–52.
Markusen, J. R. (1986). Explaining the volume of trade: An eclectic approach. American Economic
Review, 76, 1002–1011.
Markusen, J. R. (1995). The boundaries of multinational enterprises and the theory of international
trade. Journal of Economic Perspectives, 9(2), 169–189.
Markusen, J. R. (2002). Multinational firms and the theory of international trade. Cambridge,
USA: The MIT Press.
Marschak, J. (1950). Rational behavior, uncertain prospects, and measurable utility. Econometrica,
18(2), 111–141.
Marschak, T. (1959). Centralization and decentralization in economic organizations.
Econometrica, 27(3), 399–430.
Marshall, A. (1907). Principles of economics (5th ed.). London: The Macmillan Company.
Marshall, A. (1956). Memorials of Alfred Marshall. (A. C. Pigou (Ed.)). London: Macmillan.
Martin, R. (1999). The new ‘geographical turn’ in economics: Some critical reflections. Cambridge
Journal of Economics, 23(1), 65–91.
Martin, W., & Alston, J. M. (1997). Producer surplus without apology? Evaluating investments in
R&D. Economic Record, 73(221), 146–158.
Martins, A. P. (2008). Bilateral mobility in dualistic models. Economic Modelling, 25(3),
391–410
Massey, D. (1973). Towards a critique of industrial location theory. Antipode, 5(3), 33–39.
Mathewson, G. F., & Winter, R. A. (1983). Vertical integration by contractual restraints in spatial
markets. Journal of Business, 56(4), 497–517
Mathur, V. K. (1979). Some unresolved Issues in the location theory of the firm. Journal of Urban
Economics, 6, 299–318.
Bibliography 433
Mathur, V. K. (1982). Some unresolved issues in the location theory of the firm: Erratum. Journal
of Urban Economics, 11(3), 368–369.
Mathur, V. K. (1983). Location theory of the firm under price uncertainty. Regional Science and
Urban Economics, 13(3), 411–428.
Mathur, V. K. (1985). Location theory of the firm under price uncertainty: Some new conclusions.
Regional Science and Urban Economics, 15(4), 597–598.
Mathur, V. K. (1989). Location theory of the firm, clarification, revisions, and additions. Journal
of Urban Economics, 26(2), 189–196.
Matsumura, T., Ohkawa, T., & Shimizu, D. (2005). Partial agglomeration or dispersion in spatial
Cournot competition. Southern Economic Journal, 72(1), 224–235.
Mayer, C., & Sinai, T. (2003). Network effects, congestion externalities, and air traffic delays: Or
why not all delays are evil. American Economic Review, 93(4), 1194–1215.
McCann, P. (1993). The logistics-cost location-production problem. Journal of Regional Science,
33(4), 503–516.
McCann, P. (1995). Rethinking the economics of location and agglomeration. Urban Studies,
32(3), 563–577.
McCann, P. (1999). A note on the meaning of neo-classical location theory and its usefulness as a
basis for applied research. Papers in Regional Science, 78(3), 323–331.
McCann, P., & Mudambi, R. (1984). The location behavior of the multinational enterprise: Some
analytical issues. Growth and Change, 35(4), 491–524.
McCann, P., & Shefer, D. (2004). Location, agglomeration and infrastructure. Papers in Regional
Science, 83(1), 177–196
McCann, P., & Sheppard, S. (2003). The rise, fall and rise again of industrial location theory.
Regional Studies, 37(6), 649–663.
McCann, R. J. (1999). Exact solutions to the transportation problem on a line. Proceedings:
Mathematical, Physical, and Engineering Sciences, 455, 1341–1380.
McChesney, F. S., Shughart, W. F., & Haddock, D. D. (2004). On the internal contradictions of the
law of one price. Economic Inquiry, 42(4), 706–716.
McCool, T. (1982). Wage subsidies and distortionary taxes in a mobile capital Harris-Todaro
model. Economica, 49(193), 69–79.
McGuire, M. (1972). Private good clubs and public good clubs: Economic models of group
formation. Scandinavian Journal of Economics, 74(1), 84–99.
McLafferty, S. L., & Ghosh, A. (1986). Multipurpose shopping and the location of retail firms.
Geographical Analysis, 18, 215–226.
McLaren, J. (2000). Globalization and vertical structure. American Economic Review, 90(5),
1239–1254.
Meade, J. E. (1974). Optimal balance between economies of scale and variety of commodities –
illustrative model. Economica, 41(164), 359–367.
Meardon, S. J. (2002). On the new economic geography and the progress of geographical
economics. In S. Boehm, et al. (Eds.), Is there progress in economics? Knowledge, truth, and
the history of economic thought (pp. 217–239). Cheltenham, UK: Edward Elgar.
Megdal, S. B. (1984). A model of local demand for education. Journal of Urban Economics, 16(1),
13–30.
Miles, R. E., & Maillardet, R. J. (1982). The basic structures of Voronoi and generalized Voronoi
polygons. Journal of Applied Probability, 19, 97–111.
Miller, H. J., & Finco, M. V. (1995). Spatial search and spatial competition: A probability analysis
of basic results from the spatially-restricted theory. Annals of Regional Science, 29(1), 67–89.
Miller, S. M., & Jensen, O. W. (1978). Location and theory of production—review, summary and
critique of recent contributions. Regional Science and Urban Economics, 8(2), 117–128.
Mills, E. S., & McDonald, J. F. (Eds.). (1992). Sources of metropolitan growth. New Brunswick,
New Jersey: Center for Urban Policy Research, Rutgers University.
Mills, E. S., & Lav, M. R. (1964). A model of market areas with free entry. Journal of Political
Economy, 72(3), 278–288.
434 Bibliography
Minta, S. C. (1992). Tests of spatial and temporal interaction among animals. Ecological
Applications, 2(2), 178–188.
Mirchandani, P. B., & Francis, R. L. (Eds.). (1990). Discrete location theory. New York: Wiley.
Miron, J. R. (1975). Access, economic growth, and spatial dispersion. Journal of Regional Science,
15(2), 151–164.
Miron, J. R. (1976). City size and land rents in a closed model of urban spatial structure. In G. J.
Papageorgiou (Ed.), Mathematical land use theory (pp. 79–95). Lexington, Mass: Lexington
Books.
Miron, J. R. (1978a). Nucleated suburbanisation in a competitive market economy. Urban Studies,
15, 223–229.
Miron, J. R. (1978b). Dualistic spatial equilibrium and technical change. Journal of Regional
Science, 18(3), 337–355.
Miron, J. R. (1982). Economic equilibrium in urban land use. In L. S. Bourne (Ed.), Internal
structure of the city: Readings on urban form, growth, and policy (2nd ed.) (pp. 124–136).
Oxford: Oxford University Press.
Miron, J. R. (2002). Löschian spatial competition in an emerging retail industry. Geographical
Analysis, 34(1), 34–61.
Miron, J. R., & Lo, L. (1997). A selection bias approach to destination choice. Geographical
Analysis, 29(2), 145–168.
Miron, J. R., & Skarke, P. (1981). Non-price information and price sustainability in the Koopmans-
Beckmann problem. Journal of Regional Science, 21, 117–122.
Mirowski, P. (1984). Physics and the ‘marginalist revolution’. Cambridge Journal of Economics,
8(4), 361–379.
Moene, K. O. (1988). A reformulation of the Harris-Todaro mechanism with endogenous wages.
Economics Letters, 27(4), 387–390.
Mogus, A., Stennes, B., & van Kooten, G. C. (2006). Canada-U.S. softwood lumber trade revisited:
Substitution bias in the context of a spatial price equilibrium framework. Forest Science, 52(4),
411–421.
Montgomery, C. A., Brown, G. M., & Adams, D. M. (1994). The marginal cost of species preserva-
tion – the northern spotted owl. Journal of Environmental Economics and Management, 26(2),
111–128.
Morehouse, N. F., Strotz, R. H., & Horwitz, S. J. (1950). An electro-analog method for
investigating problems in economic dynamics: Inventory oscillations. Econometrica, 18(4),
313–328.
Morgan, F., & Bolton, R. (2002). Hexagonal economic regions solve the location problem.
American Mathematical Monthly, 109(2), 165–172.
Moses, L. N. (1958). Location and the theory of production. Quarterly Journal of Economics, 72,
259–272.
Mosteller, F., & Tukey, J. W. (1977). Data analysis and regression: A second course in statistics.
Reading, Mass.: Addison-Wesley Publishing Company, 259–262.
Motta, M. (1992). Multinational firms and the tariff-jumping argument: A game theo-
retic analysis with some unconventional conclusions. European Economic Review, 36(8),
1557–1571.
Mu, L. (2004). Polygon characterization with the multiplicatively weighted voronoi diagram.
Professional Geographer, 56(2), 223–239.
Mulligan, G. F. (1981). Lösch’s single good equilibrium. Annals of the Association of American
Geographers, 71(1), 84–94.
Mulligan, G. F. (1982). Equilibrium market area properties under 2 pricing systems. Geographical
Analysis, 14(1), 51–63.
Mulligan, G. F. (1984). Agglomeration and central place theory – a review of the literature.
International Regional Science Review, 9(1), 1–42.
Mulligan, G. F., & Reeves, R. W. (1983). The theory of the firm: Some spatial implications. Urban
Geography, 4, 156–172.
Bibliography 435
Murata, K. (1959). Areal study and location theory: Some methodological examination of eco-
nomic geography. Economic Series No. 23. Tokyo: Science Council of Japan: Division of
Economics, Commerce and Business Administration.
Murata, Y., & Thisse, J.-F. (2005). A simple model of economic geography à la Helpman-Tabuchi.
Journal of Urban Economics, 58(1), 137–155.
Murphy, F. H., & Panchanadam, V. (1997). Understanding linear programming modeling through
an examination of the early papers on model formulation. Operations Research, 45(3),
341–356.
Myrdal, G. (1957). Economic theory and underdeveloped regions. London: Duckworth.
Nachum, L., & Keeble, D. (2003). Neo-Marshallian clusters and global networks – the linkages of
media firms in Central London. Long Range Planning, 36(5), 459–480.
Neary, J. P. (1981). On the Harrs-Todaro model with intersectoral capital mobility. Economica,
48(191), 219–234.
Neary, J. P. (2001). Of hype and hyperbolas: Introducing the new economic geography. Journal of
Economic Literature, 39(2), 536–561.
Neary, J. P. (2002). Foreign direct investment and the single market. Manchester School of
Economic and Social Studies, 70(3), 291–314.
Needham, D. (1964). The rationale of basing point pricing. Southern Economic Journal, 30(4),
309–314.
Nerlove, M. L., & Sadka, E. (1991). Thünen, von model of the dual economy. Journal of
Economics, 54(2), 97–123.
Netz, J. S., & Taylor, B. A. (2002). Maximum or minimum differentiation? Location patterns of
retail outlets. Review of Economics and Statistics, 84(1), 162–175.
Neumann, J. von, & Morgenstern, O. (1947). Theory of games and economic behavior (2nd ed.).
Princeton: Princeton University Press. (1st edition published in 1944).
Neven, D., & Siotis, G. (1996). Technology sourcing and FDI in the EC: An empirical evaluation.
International Journal of Industrial Organization, 14(5), 543–560.
Nickel, S., & Puerto, J. (2005). Location theory: A unified approach. Heidelberg and New York:
Springer.
Nijkamp, P., & Paelinck, J. H. P. (1973). A solution method for neoclassical location problems.
Regional Science and Urban Economics, 3(4), 383–410.
Norback, P. J., & Persson, L. (2004). Privatization and foreign competition. Journal of
International Economics, 62(2), 409–416.
North, D. (1958). Ocean freight rates and economic development 1750–1913. Journal of Economic
History, 18(4), 537–555.
North, D. C. (1955). Location theory and regional economic growth. Journal of Political Economy,
63(3), 243–258.
Norton, W. (1979). Relevance of von Thünen theory to historical and evolutionary analysis of
agricultural land use. Journal of Agricultural Economics, 30(1), 39–47.
Nourse, H. O. (1968). Regional economics: A study in the economic structure, stability, and growth
of regions. New York: McGraw-Hill.
Ó hUallacháin, B., & Reid, N. (1991). The location and growth of business and professional
services in American metropolitan areas, 1975–1986. Annals of the Association of American
Geographers, 81(2), 254–270.
O’Brien, J. P., & Shieh, Y. N. (1989). Transportation rates, production and location in the Moses
Weber triangle-a note. Regional Science and Urban Economics, 19(1), 133–142.
Ohlin, B. (1933). Interregional and international trade. Cambridge, Mass.: Harvard University
Press.
Ohlin, B., Hesselborn, P. O., & Wijkman, P. M. (Eds.). (1977). The international allocation of
economic activity: Proceedings of a nobel symposium held at Stockholm. London: Macmillan.
Ohta, H., & Thisse, J. F. (eds.). (1993). Does Economic Space Matter? Essays in Honour of Melvin
L. Greenhut. London: St. Martins Press.
Ohta, H., Lin, Y.-S., & Naito, M. K. (2005). Spatial perfect competition: A uniform delivered
pricing model. Pacific Economic Review, 10(4), 407–420.
436 Bibliography
Okabe, A., & Kume, Y. (1983). A dynamic von Thünen model with a demand function. Journal of
Urban Economics, 14, 355–369.
O’Kelly, M., & Bryan, D. (1996). Agricultural location theory: von Thünen’s contribution to
economic geography. Progress in Human Geography, 20(4), 457–475.
O’Kelly, M. E., & Miller, H. J. (1989). A synthesis of some market area delimitation models.
Growth and Change, 20(3), 14–33.
Orden, A. (1952). Solution of systems of linear inequalities on a digital computer. Proceedings of
the Association for Computing Machinery, Pittsburgh, Pa., May 2–3, 1952, 91–95.
Osborne, M. J., & Pitchik, C. (1987). Equilibrium in Hotelling’s model of spatial competition.
Econometrica, 55(4), 911–922.
Oster, G. (1979). Factor analytic test of theory of the dual economy. Review of Economics and
Statistics, 61(1), 33–39.
Ottaviano, G., Tabuchi, T., & Thisse, J. F. (2002). Agglomeration and trade revisited. International
Economic Review, 43(2), 409–435.
Pang, J.-S. (1984). Solution of the general multicommodity spatial equilibrium problem by
variational and complementarity methods. Journal of Regional Science, 24(3), 403–414.
Papageorgiou, G. J. (1973). The impact of the environment upon the spatial distribution of
population and land values. Economic Geography, 49(3), 251–256.
Papageorgiou, G. J. (Ed.). (1976). Mathematical land use theory. Lexington, MA: Lexington
Books.
Papageorgiou, G. J. (1978). Spatial externalities I: Theory. Annals of the Association of American
Geographers, 68(4), 465–476.
Papageorgiou, G. J. (1979). Agglomeration. Regional Science and Urban Economics, 9(1),
41–59.
Papageorgiou, G. J. (1980). On sudden urban growth. Environment and Planning A, 12(9),
1035–1050.
Papageorgiou, Y. Y. (1990). The isolated city state: An economic geography of urban spatial
structure. London: Routledge.
Papageorgiou, Y. Y., & Pines, D. (1999). An essay on urban economic theory. Boston: Kluwer
Academic Publishers.
Parr, J. B. (1993). Competition, supply areas, and industrial location – an equilibrium perspective.
Annals of Regional Science, 27(3), 191–201.
Parr, J. B. (1995a). Alternative approaches to market area structure in the urban system. Urban
Studies, 32(8), 1317–1329.
Parr, J. B. (1995b). The economic law of market areas – a further discussion. Journal of Regional
Science, 35(4), 599–615.
Parr, J. B. (1997a). The law of market areas and the size distribution of urban centers. Papers in
Regional Science, 76(1), 43–68.
Parr, J. B. (1997b). The law of retail gravitation: Insights from another law. Environment and
Planning A, 29(8), 1477–1492.
Pate, J., & Loomis, J. (1997). The effect of distance on willingness to pay values: A case study of
wetlands and salmon in California. Ecological Economics, 20(3), 199–207.
Peet, J. R. (1969). Spatial expansion of commercial agriculture in the 19th century—a von Thünen
interpretation. Economic Geography, 45(4), 283–301.
Peneder, M. (2001). Entrepreneurial competition and industrial location: Investigating the
structural patterns and intangible sources of competitive performance. Cheltenham, UK:
Elgar.
Penfold, R. (2002). Risk, activity modularization, sourcing, and economic geography. Environment
and Planning A, 34(8), 1333–1353.
Perry, M., & Hui, T. B. (1998). Global manufacturing and local linkage in Singapore. Environment
and Planning A, 30(9), 1603–1624.
Phillips, D., MacPherson, A. D., & Lentnek, B. (1998). The optimum size of a producer service
firm facing uncertain demand. Environment and Planning A, 30(1), 129–141.
Bibliography 437
Phipps, M., & Langlois, A. (1997). Spatial dynamics, cellular automata, and parallel processing
computers. Environment and Planning B—Planning & Design, 24(2), 193–204.
Pinch, S., Henry, N., Jenkins, M., & Tallman, S. (2003). From ‘industrial districts’ to ‘knowl-
edge clusters’: A model of knowledge dissemination and competitive advantage in industrial
agglomerations. Journal of Economic Geography, 3(4), 373–388.
Pipkin, J. S. (1977). Some probabilistic revealed-preference models of the choice process in
recurrent urban travel. Geografiska Annaler B, 59(2), 82–94.
Pires, C. P. (2005). Constrained efficient locations under delivered pricing. Regional Science and
Urban Economics, 35, 49–56.
Pires, C. P. (2009). Location choice under delivered pricing: A reinterpretation. Annals of Regional
Science, 43, 199–213.
Pitts, T. C., & Boardman, J. P. (1998). Clustering and dispersion of duopolists in a linear market.
Geographical Analysis, 30(1), 35–44.
Plummer, P. S., Sheppard, E., & Haining, R. P. (1998). Modeling spatial price competition: Marxian
versus neoclassical approaches. Annals of the Association of American Geographers, 88(4),
575–594.
Ponsard, C. (1983). History of spatial economic theory. Berlin: Springer-Verlag.
Pontes, J. P., & Parr, J. B. (2005). A note on agglomeration and the location of multinational firms.
Papers in Regional Science, 84(3), 509–517
Portugali, J. (1984). On relevance in Geography-Thünen isolated state in relation to agriculture and
political economy. Geoforum, 15(2), 201–207.
Pred, A. (1964). Toward a typology of manufacturing flows. Geographical Review, 54(1),
65–84.
Pred, A. (1965). Industrialization, initial advantage, and American metropolitan growth.
Geographical Review, 55(2), 158–185.
Pred, A. (1969a). Behaviour and location: Foundations for a geographic and dynamic location
theory. Part I. Lund Series in Geography, Series B, No. 26. Gleerup: Lund.
Pred, A. (1969b). Behaviour and location: foundations for a geographic and dynamic location
theory. Part II. Lund Series in Geography, Series B, No. 27. Gleerup: Lund.
Predohl, A. (1928). The theory of location in its relation to general economics. Journal of Political
Economy, 36, 371–390.
Prescott, E. C., & Visscher, M. (1977). Sequential location among firms with foresight. Bell Journal
of Economics, 8(2), 378–393.
Puu, T. (2003). Mathematical location and land use theory: An introduction (2nd ed.). Heidelberg
and New York: Springer.
Quigley, J. M. (1998). Urban diversity and economic growth. Journal of Economic Perspectives,
12(2), 127–138.
Quinzii, M., & Thisse J. F. (1990). On the optimality of central places. Econometrica, 58(5),
1101–1119.
Raina, J. L. (1989). Locational analysis in agricultural geography. Jaipur: Pointer Publishers.
Ranis, G., & Fei, J. C. H. (1961). A theory of economic development. American Economic Review,
51(4), 533–565.
Rautman, C. A., Reid, R. A., & Ryder, E. E. (1993). Scheduling the disposal of nuclear waste
material in a geologic repository using the transportation model. Operations Research, 41(3),
459–469.
Reggiani, A. (Ed.). (1998). Accessibility, trade, and locational behaviour. Aldershott, UK:
Ashgate.
Reilly, W. J. (1931). The law of retail gravitation. Austin, Texas: University of Texas Press.
ReVelle, C. (1986). The maximum capture or “sphere of influence” location problem: Hotelling
revisited on a network. Journal of Regional Science, 26(2), 343–358.
Ricardo, D. (1821). On the Principles of Political Economy and Taxation (3rd ed.) (P. Sraffa (Ed.)).
Cambridge, UK: Cambridge University Press (Printed in 1951).
Richardson, H. W. (1969). Regional economics: Location theory, urban structure, and regional
change. New York: Praeger.
438 Bibliography
Ritter, J. A. (1995). The transition from barter to fiat money. American Economic Review, 85(1),
134–149.
Rodriguez-Bachiller, A. (1986). Discontiguous urban growth and the new urban economics—a
review. Urban Studies, 23(2), 79–104.
Rodwin, L. (1945). Two approaches to industrial location analysis. Journal of Land & Public
Utility Economics, 21(1), 23–33.
Romer, P. M. (1986). Increasing returns and long-run growth. Journal of Political Economy, 94(5),
1002–1037.
Rossi-Hansberg, E. (2005). A spatial theory of trade. American Economic Review, 95(5),
1464–1491.
Roth, T. P. (1989). The present state of consumer theory (2nd ed.). Lanham, MD: University Press
of America.
Rothschild, R. (1987). The theory of monopolistic competition – Chamberlin, E. H. influence on
industrial organization theory over 60 years. Journal of Economic Studies, 14(1), 34–54.
Rowse, J. (1981). Solving the generalized transportation problem. Regional Science and Urban
Economics, 11(1), 57–68.
Rugman, A. M., & Verbeke, A. (2004). A perspective on regional and global strategies of
multinational enterprises. Journal of International Business Studies, 35(1), 3–18.
Rushton, G. (1971a). Postulates of central place theory and properties of central place systems.
Geographical Analysis, 3(2), 140–156.
Rushton, G. (1971b). Behavioral correlates of urban spatial structure. Economic Geography, 47(1),
49–58.
Russell, E. J. (1969). Extension of Dantzig’s algorithm to finding an initial near-optimal basis for
the transportation problem. Operations Research, 17(1), 187–191.
Russell, T. (1995). Product innovation with mass-production: Insufficient or excessive? Comment.
Japan and the World Economy, 7(4), 443–445,
Sakashita, N. (1968). Production function, demand function and location theory of the firm. Papers
in Regional Science, 20(1), 109–122.
Salop, S. C. (1979). Monopolistic competition with outside goods. Bell Journal of Economics,
10(1), 141–156.
Samuelson, P. A. (1947). Foundations of economic analysis. Cambridge, MA: Harvard University
Press.
Samuelson, P. A. (1952). Spatial price equilibrium and linear programming. American Economic
Review, 42(3), 283–303.
Samuelson, P. A. (1954). The transfer problem and transport costs, II: Analysis of effects of trade
impediments. Economic Journal, 64(254), 264–289.
Samuelson, P. A. (1960). Harold Hotelling as mathematical economist. The American Statistician,
14(3), 21–25.
Samuelson, P. A. (1983). Thünen at two hundred. Journal of Economic Literature, 21(4),
1468–1488.
Sanner, H., & Scholer, K. (1998). Spatial price discrimination in two-dimensional competitive
markets. Journal of Regional Science, 38(1), 89–107.
Sarkar, J., Gupta, B., & Pal, D. (1997). Location equilibrium for Cournot oligopoly in spatially
separated markets. Journal of Regional Science, 37(2), 195–212.
Sarker, S., Gilbert, J., & Oladi, R. (2008). Adjustment costs and Immiserizing growth in LDCs.
Review of Development Economics, 12(4), 779–791.
Sasaki, K. (Ed.). (1996). Optimum and equilibrium for regional economies: Collected papers of
Norboru Sakashita. Heidelberg and New York: Springer.
Satchi, M., & Temple, J. (2009). Labor markets and productivity in developing countries. Review
of Economic Dynamics, 12(1), 183–204.
Sauer, C. O. (1924). The survey method in geography and its objectives. Annals of the Association
of American Geographers, 14(1), 17–33.
Scarf, H. (1973). The computation of economic equilibria. New Haven: Yale University Press for
the Cowles Foundation.
Bibliography 439
Scarf, H. (1991). The origins of fixed point methods. In J. K. Lenstra, et al. (Eds.), History of math-
ematical programming: A collection of personal reminiscences (pp. 126–134). Amsterdam:
CWI.
Scherer, F. M. (1975). The economics of multiplant operation: An international comparisons study.
Harvard Economic Studies.
Schmalensee, R. (1972). Option demand and consumers surplus – valuing price changes under
uncertainty. American Economic Review, 62(5), 813–824.
Schmenner, R. W. (1982). Making business location decisions. Englewood Cliffs, NJ: Prentice Hall
Inc.
Schmitt, N. (1993). Equilibria and entry in two interdependent spatial markets. Regional Science
and Urban Economics, 23(1), 1–27.
Schneider, E. (1934). Johann Heinrich Von Thünen. Econometrica, 2(1), 1–12.
Schulz, N., & Stahl, K. (1996). Do consumers search for the highest price? Oligopoly equilib-
rium and monopoly optimum in differentiated-products markets. RAND Journal of Economics,
27(3), 542–562.
Schumpeter, J. A. (1954). History of economic analysis. New York: Oxford University Press.
Scotchmer, S., & Thisse, J. F. (1992). Space and competition – a puzzle. Annals of Regional
Science, 26(3), 269–286.
Scott, A. J. (1988). New industrial spaces: Flexible production organization and regional
development in North America and Western Europe. London: Pion.
Scott, A. J. (1998). Regions and the world economy: The coming shape of global production,
competition, and political order. Oxford: Oxford University Press.
Scott, A. J. (2000). Economic geography: The great half-century. Cambridge Journal of
Economics, 24(4), 483–504.
Scott, A. J., & Angel, D. P. (1987). The United States semiconductor industry – a locational
analysis. Environment and Planning A, 19(7), 875–912.
Seale, R. D., Seale, A. D. Jr., Leng, J. (2004). An application of business activity modeling to
regional production and national distribution of plywood. Forest Commodities Journal, 54(12),
38–44.
Seim, K. (2006). An empirical model of firm entry with endogenous product-type choices. RAND
Journal of Economics, 37(3), 619–640.
Selten, R., & Apesteguia, J. (2005). Experimentally observed imitation and cooperation in price
competition on the cycle. Games and Economic Behavior, 51, 171–192.
Serneels, S., & Lambin, E. F. (2001). Proximate causes of land-use change in Narok District,
Kenya: A spatial statistical model. Agricultural Ecosystems & Environment, 85(1–3),
65–81.
Serra, D., & ReVelle, C. (1994). Market capture by two competitors: The preemptive location
problem. Journal of Regional Science, 34(4), 549–561.
Serra, D., & ReVelle, C. (1999). Competitive location and pricing on networks. Geographical
Analysis, 31(2), 109–129.
Shaked, A., & Sutton, J. (1982). Relaxing price competition through product differentiation.
Review of Economic Studies, 49(1), 3–13.
Sharpe, W. F. (1963). A simplified model for portfolio analysis. Management Science, 9(2),
277–293.
Sharpe, W. F. (1964). Capital asset prices: A theory of market equilibrium under conditions of risk.
Journal of Finance, 19(3), 424–442.
Sheppard, E., & Curry, L. (1982). Spatial price equilibria. Geographical Analysis, 14(4),
279–304.
Sheppard, E., Haining, R. P., & Plummer, P. (1992). Spatial pricing in interdependent markets.
Journal of Regional Science, 32(1), 55–75.
Shieh, Y. N. (1999). Employment effect of wage discrimination in the Weber-Moses triangle.
Annals of Regional Science, 33(1), 15–24.
Shieh, Y. N., & Mai, C. C. (1997). Demand and location decision of a monopsonistic firm. Annals
of Regional Science, 31(3), 273–284.
440 Bibliography
Sibson, R. (1980). The Dirichlet tessellation as an aid in data analysis. Scandinavian Journal of
Statistics, 7(1), 14–20.
Siebert, H. (1969). Regional economic growth: Theory and policy. Scranton, PA: International
Textbook Company.
Simon, H. A. (1955). A behavioral model of rational choice. Quarterly Journal of Economics,
69(1), 99–118.
Simon, H. A. (1959). Theories of decision-making in economics and behavioral science. American
Economic Review, 49(3), 253–283.
Sinclair, R. (1967). Von Thünen and urban sprawl. Ekistics, 24(141), 139–143.
Siroen, J. M. (1988). La théorie de l’échange international en concurrence monopolistique: Une
comparaison des modèles. Revue Economique, 39(3), 511–544.
Skott, P., & Ryoo, S. (2008). Macroeconomic implications of financialisation. Cambridge Journal
of Economics, 32(6), 827–862.
Slade, M. E. (2005). The role of economic space in decision making. Annales d’Économie et de
Statistique, 77, 1–20,
Slichter, S. H (1932). Lines of action, adaptation, and control. American Economic Review, 22(1),
41–54.
Smart, M. (2008). Transport demand and spatial equilibria. Journal of Transport Economics and
Policy, 42(2), 323–343.
Smith, D. (1966). A theoretical framework for geographical studies of industrial location.
Economic Geography, 42(2), 95–113.
Smith, D. F., & Florida, R. (1994). Agglomeration and industrial location—an econometric
analysis of Japanese-affiliated manufacturing establishments in automotive-related industries.
Journal of Urban Economics, 36(1), 23–41
Smith, M. J. (1979). The existence, uniqueness, and stability of traffic equilibria. Transportation
Research, 13B, 295–304.
Smith, M. J. (1983). The existence and calculation of traffic equilibria. Transportation Research,
17B, 291–303.
Smith, V. L. (1963). Minimization of economic rent in spatial price equilibrium. Review of
Economic Studies, 30(1), 24–31.
Smithies, A. (1941). Optimum location in spatial competition. Journal of Political Economy, 49(3),
423–439.
Smithies, A. (1942). Aspects of the basing-point system. American Economic Review, 32(4),
705–726.
Solomon, B. D., & Pyrdol, J. J. (1986). Delineating coal market regions. Economic Geography,
62(2), 109–124.
Solow, R. M. (1956). A contribution to the theory of economic growth. Quarterly Journal of
Economics, 70(1), 65–94.
Soper, J. B., Norman, G., Greenhut, M. L., & Benson, B. L. (1991). Basing point pricing and
production concentration. Economic Journal, 101(406), 539–556.
Soubeyran, A., & Weber, S. (2002). District formation and local social capital: A (tacit)
co-opetition approach. Journal of Urban Economics, 52, 65–92.
Spengler, J. (1974). Was 1922–1972 a golden age in the history of economics? Journal of Economic
Issues, 8(3), 525–553.
Spivey, W. A. (1962). Linear programming. Science (New Series) 135(3497), 23–27.
Stahl, K., & Varaiya, P. (1978). Economics of information: Examples in location and land-use
theory. Regional Science and Urban Economics, 8, 43–56.
Steininger, K. W. (2001). International trade and transport: Spatial structure and environmental
quality in a global economy. Cheltenham, UK: Edward Elgar.
Stern, N. H. (1972a). Optimal size of market areas. Journal of Economic Theory, 4(2), 154–173.
Stern, N. H. (1972b). Optimum development in a dual economy. Review of Economic Studies,
39(118), 171–184.
Bibliography 441
Stevens, B. H. (1968). Location theory and programming models: The von Thünen case. Papers of
the Regional Science Association, 21, 19–34.
Stigler, G. J. (1937). The economics of Carl Menger. Journal of Political Economy, 45(2),
229–250.
Stigler, G. J. (1949). A theory of delivered price systems. American Economic Review, 39(6),
1143–1159.
Stigler, G. J. (1951). The division of labor is limited by the extent of the market. Journal of Political
Economy, 59(3), 185–193.
Stiglitz, J. (2000). The contributions of the economics of information to twentieth century
economics. Quarterly Journal of Economics, 115(4), 1441–1478.
Stimson, R. J. (1981). The provision and use of general practitioner serves in Adelaide, Australia
– application of tools of locational analysis and theories of provider and user spatial behavior.
Social Science & Medicine Part D – Medical Geography, 15(1D), 27–44.
Storper, M. (2000). Globalization and knowledge flows: An industrial geographer’s perspective. In
J. H. Dunning (Ed.), Regions, globalization, and the knowledge-based economy (pp. 42–62).
Oxford: Oxford University Press.
Stuart, H. W. Jr. (2004). Efficient spatial competition. Games and Economic Behavior, 49, 345–362.
Sugden, R. (1991). Rational choice: A survey of contributions from economics and philosophy.
Economic Journal, 101(407), 751–785.
Tabuchi, T. (1994). Two-stage two-dimensional spatial competition between two firms. Regional
Science and Urban Economics, 24, 207–227.
Tabuchi, T., & Thisse, J.-F. (1995). Asymmetric equilibria in spatial competition. International
Journal of Industrial Organization, 13, 213–227.
Takayama, T., & Judge, G. G. (1971). Spatial and temporal price and allocation models.
Amsterdam: North-Holland.
Takayama, T., & Labys, W. C. (1986). Spatial equilibrium analysis. In P. Nijkamp (Ed.), Handbook
of regional and urban economics (pp. 171–199). Amsterdam: Elsevier.
Tarrow, S. (1995). Bridging the quantitative-qualitative divide in political science. American
Political Science Review, 89(2), 471–475.
Teitz, M. B. (1968a). Toward a theory of urban public facility location. Papers in Regional Science,
21(1), 35–51.
Teitz, M. B. (1968b). Locational strategies for competitive systems. Journal of Regional Science,
8(2), 135–148.
Tellier, L. N. (1972). Weber problem—solution and interpretation. Geographical Analysis, 4(3),
215–233.
Thisse, J.-F., Button, K. J., & Nijkamp, P. (Eds.). (1996). Location theory (2 volumes). Brookfield,
CT: Edward Elgar.
Thisse, J.-F., & Perreur, J. (1977). Relations between point of maximum profit and point of
minimum total transportation cost — restatement. Journal of Regional Science, 17(2), 227–234.
Thisse, J.-F., & Vives, X. (1992). Basing point pricing: Competition versus collusion. Journal of
Industrial Economics, 40(3), 249–260.
Thomas, I. (2002). Transportation networks and the optimal location of human activities.
Cheltenham, UK: Edward Elgar.
Thore, S. (1986). Spatial disequilibrium. Journal of Regional Science, 26(4), 661–675.
Thrall, G. I. (1987). Land use and urban form: The consumption theory of land rent. London:
Methuen.
Thünen, J. H. von. (1826). Der Isolierte Staat in Beziehung auf Landwirtschaft und
Nationalökonomie (Part I of The Isolated State). Hamburg: Perthes. (Second edition enlarged
and improved published in 1842 (Rostock: Leopold)).
Thünen, J. H. von. (1850). Der naturgemässe Arbeitslohn und dessen Verhältnis zum Zinsfuss und
zur Landrente. (Part II (Section 1) of The Isolated State. Rostock: Leopold.
Thünen, J. H. von. (1966). Isolated state: An english edition of Der Isolierte Staat. (P. Hall, Trans.)
London: Pergamon. (This is an English translation mainly of Part 1 of The Isolated State. See
also Dempsey (1960)).
442 Bibliography
445
446 First Author Index
G Hart, A. B., 17
Gabriel, S. A., 142, 239 Hart, P. W. E., 27
Gabszewicz, J. J., 13, 203 Hartshorne, R., 14
Gale, D., 83 Hartwick, J. M., 201, 318
Gale, S., 27 Harvey, D. W., 27
Gannon, C. A., 203 Hay, D. A., 203
Garrison, W. L., 7, 14, 18, 20, 273 Head, K., 3, 15
Gass, S. I., 82 Heckscher, E., 3
Gee, J. M. A., 53 Henderson, J. M., 18, 70
Gehrig, T., 233 Henderson, J. V., 13
Georgescu-Roegen, N., 239 Henkel, J., 233
Ghatak, S., 301 Herbert, J. D., 13
Ghosh, A., 6, 13–14, 27, 202, 205 Hicks, J. R., 129
Giarratani, F., 3 Hitchcock, F. L., 13, 69–95
Giersch, H., 17 Hoare, A. G., 301
Gillen, W. J., 202 Hohenbalken, B., 62, 227
Gilligan, T. W., 15, 52, 144 Hoover, E. M., 13
Glaeser, E. L., 9 Hornby, J. M., 301
Glazer, A., 233 Horn, H., 60
Goettler, R., 203 Hotelling, H., 203
Gold, J. R., 27 Houthakker, H. S., 82
Golledge, R. G., 14, 27, 202 Hsu, S. K., 23, 43, 154, 166, 296
Gorman, W. M., 23 Huang, T., 203
Gould, P. R., 27 Huck, S., 203
Graham, R., 214 Huff, D. L., 14, 202, 227
Gras, N. S. B., 17 Hughes, J. W., 15, 52
Greenhut, M. L., 13, 15, 18, 24, 29–67, 201 Hummels, D., 296
Green, M. B., 70 Huntingdon, E., 4
Grether, E. T., 26 Huriot, J. M., 13
Griffith, D. A., 14 Hurter, A. P., 6, 166, 203
Güder, F., 126, 136 Hurwicz, L., 11, 18, 238
Guise, J. W. B., 127 Hwang, H., 62, 155, 169, 170, 219
Gupta, B., 203, 217 Hyson, C. D., 201–202
Hyson, W. P., 201–202
H I
Haddock, D. D., 15, 50, 52 Inaba, F. S., 147
Hadley, G., 18 Irmen, A., 203
Haggett, P., 27 Isard, W., 11, 13, 18, 23, 154, 157, 203
Haig, R. M., 17 Ishikawa, T., 28
Hall, R. W., 82, 220 Islam, N., 301
Hamacher, H. W., 6 Iyer, G., 203, 227
Hamilton, J. H., 203
Hammond, S., 3 J
Hamoudi, H., 203 Jacobs, J., 9
Handler, G. Y., 174 Javorcik, B. S., 9
Hanemann, W. M., 129 Jayet, H., 249
Hanson, G. H., 15, 17 Jevons, H. S., 11, 14
Harker, P. T., 126, 136 Johansson, B., 10, 13
Harley, C. K., 296 John, F., 18
Harris, C. C. Jr., 13, 17, 70 Johnson, A. S., 272
Harris, J. R., 301 Jones, D. W., 14
Harris, R. G., 296 Jones, E. D., 17
Harsanyi, J. C., 18, 238–239 Jorgenson, D. W., 301
448 First Author Index
Smart, M., 28 V
Smith, D., 157 Vaux, H. J., 142
Smithies, A., 13, 52, 203 Velupillai, K. V., 10
Smith, M. J., 6, 126 Venables, A. J., 13
Smith, R. H. T., 13 Vickrey, W. S., 22
Smith, T., 203 Vidale, M. L., 82
Smith, V. L., 135 Vining, R., 26
Solomon, B. D., 14, 202 Vollrath, D., 301
Solow, R. M., 16
Soper, J. B., 15, 52
W
Soubeyran, A., 249
Wagener, U. A., 82
Spengler, J., 16
Wagner, H. M., 83
Spivey, W. A., 18
Stahl, K., 13, 227, 233 Wagner, W. B., 14
Steininger, K. W., 13 Walker, R., 28
Stern, N. H., 13, 202, 301 Walmsley, D. J., 27
Stevens, B. H., 13 Webber, M. J., 13–14, 202–203
Stigler, G. J., 1, 10, 52, 272 Weber, S., 249
Stiglitz, J., 11 Weinschenck, G., 13
Stiglitz, J. E., 13 Wendell, R. E., 14
Stimson, R. J., 14, 27, 202 West, D. S., 62
Storper, M., 11 Wheaton, W. C., 13
Stuart, H. W., 203 Whitaker, J. K., 2
Sugden, R., 18, 238 White, A. G., 11
White, H. C., 11
T Willig, R. D., 129
Tabuchi, T., 7, 203 Wilson, M. G. A., 14
Takayama, T., 13, 18, 114, 117, 127 Wilson, R. H., 184
Tarrow, S., 21 Wingo, L. Jr., 260
Teitz, M. B., 6, 14 Wolpert, J., 27
Tellier, L. N., 166 Woodward, R. S., 166
Thisse, J., 3, 7, 11, 13, 15–16, 52, 157, 166, Wu, J. H., 126
203
Thomas, I., 6, 203 Y
Thore, S., 137 Yeh, C. N., 23
Thrall, G. I., 14
Tramel, T. E., 127
Tucker, A. W., 18, 83, 136 Z
Turk, M. H., 10 Zarembka, P., 301
Turnovsky, S. J., 129 Zenou, Y., 301
Zeuthen, F., 239
U Zhang, W. B., 14, 371
Ullman, E., 273 Zhu, T., 203
Ulph, A., 23 Ziss, S., 62
Subject Index
451
452 Subject Index
Congested/Congestion, 1–3, 8, 25, 28, 32, 44, Discriminatory pricing, 15, 22, 45, 52, 71, 202,
69–70, 72, 74, 77, 94, 97, 103–104, 205, 337, 358, 360–365, 367, 372, 382,
111, 113, 133, 138, 142, 154–155, 172, 395
184, 194–195, 197, 276, 357, 379–380, Diseconomies of scale, 3, 106
387–389, 391 Distribution of income, 7, 67, 94, 124, 130,
Constant returns to scale, 20, 166–169, 149, 176, 199, 268, 300, 318, 336, 343
301–302, 318, 320–322, 335, 337, 340, Division of labor, 1–3, 9, 25, 43, 67, 69, 95, 99,
342–343, 345, 347, 349, 360, 366, 389 124, 179, 187, 391, 391
Consumer benefit, 33–34, 46, 54, 101–102, Domestic market, 9
105, 109, 125, 129–131, 133, 140, 206, Dual, 83–88, 220, 301–303
210, 274, 276, 280, 291, 294, 303–305, Dual economy, 301–302
308–310, 312–313 Duality Theorem, 83–84
Consumer surplus, 33–34, 46, 54, 67, 98, Dynamics, 28
101–102, 105, 109, 114, 124, 126,
129–135, 137–138, 140, 206, 210, 274, E
276, 280–281, 304, 310, 316, 382 Economies of scale, 1, 3, 9, 25, 32, 70, 72–73,
Contestable market, 1, 62 99, 124, 151, 166, 176, 179, 196,
Contractor, 17, 177, 188–189, 193–199, 363, 335–336, 380, 389
382, 390 Effective price, 15, 25, 35, 38, 45, 52–53,
138, 151–153, 195, 203–207, 209, 211,
Cooperative, 231, 249–250, 252, 255, 258,
213, 215–218, 221–222, 231, 232, 264,
262, 264, 272, 366–367, 377, 380, 389
337–339, 341, 344–346, 348–349, 354,
Core theory, viii, x, xii, xiv–xv, 395
356, 359, 362–365, 385, 387, 389, 391
Corner solution, 97, 117, 163, 185–186, 189,
Efficient firm, 26, 32, 35, 69, 103, 105, 111,
191, 195, 287–288, 312, 348, 366
121, 130–131, 133, 153, 182, 318, 320
Cross-price elasticity, 225, 267, 285 Endogenous, x, xv–xvi, 11, 19, 25, 28, 41, 44,
Cumulative causation, 17 46, 54, 56, 63, 74, 78, 82, 112, 119,
139–140, 160, 167, 171–172, 176, 189,
213, 258–259, 286, 303, 322, 336, 339,
D
382, 386, 390
Deductive approach, xiii–xiv, 6
Endowment[s], 14, 234–235, 240–245,
Delaunay triangulation, 214
248–253, 259–263, 265–267, 300–301
Delivered price, 45, 48, 52, 63, 71, 87, Establishment, 11, 27, 30, 42, 53, 69, 178, 198,
209–210, 363–364, 391 202, 204, 220, 258, 378, 391
Demand cone, 207 Euclidean distance, 26, 154, 156, 159, 172, 324
Demand curve, 4, 33–43, 45–50, 54, 56, 63, Excess demand, 5, 76, 91, 110, 114–120,
71, 87, 97–98, 100–102, 105–110, 126, 135
112–116, 118, 120–122, 125–126, Excess profit, 30, 33–34, 39–40, 46–48, 63, 75,
129–136, 149, 154, 170–171, 198, 202, 78, 101, 104, 113, 129–134, 219, 269,
206–208, 210, 216–217, 219, 223–224, 274–275, 286, 294, 299, 304, 306–307,
273, 277–278, 280, 283–286, 288–291, 312, 320, 323, 328
293–294, 303–306, 308–310, 313, Excess supply, 76, 109, 114–120
315–316, 322, 332, 355–357, 359, 363, Exclusion Theorem, 158, 174
385, 387 Exogenous, 11, 28, 41–42, 44, 46, 54–56, 61,
Demand curve, free spatial, 208 63, 74, 83, 101, 107, 112, 116, 157,
[Demand/Supply] by arbitrageurs, 37, 97, 171–172, 186, 189, 191, 194, 198, 206,
109–110, 114, 120 213, 216, 224, 294, 317, 319, 323,
Diminishing marginal utility, xii, 23–24, 36, 325–327, 346, 369, 371
130, 223, 238 Expansion path, 32, 152, 166, 168–169, 253
Diminishing returns, 152, 198, 270, 301, Expected utility, xii, 233, 239
337–338, 340, 342–345, 348, 366, 379, Expected value, 239, 242–243, 245
389 Expendable, 37, 40, 48, 231–232, 262
Discipline of difference, 21 Explanation, xii–xiii, xv–xvi, 1, 3–4, 7, 21–22,
Discrete choice model, 205, 227 76, 117–118, 124
Subject Index 453
Explanatory, x–xii, xiv, xvi, 203 Information, 2, 4, 11, 15, 22, 35, 37, 44, 53, 67,
Externality, 19, 179, 250, 259, 398 147, 202, 205, 227, 233, 236, 264–265,
321, 372–373
F Installation and removal, 15, 35, 44–45, 236
Fiat money economy, 31, 69, 100, 126, 143, Instrumentalist, x, xii, xvi
153, 182, 233–234, 273, 385 Instrument variable, 82–84
Fixed coefficients technology, 318 Insurance/Insurance principle, 9, 15, 35, 44,
Fixed cost, 31, 33–34, 37, 39–40, 46, 54, 56, 177–199, 236, 240–241, 250, 258, 265,
62–63, 66–67, 72, 93, 103, 166–167, 378–379, 381, 391
171–172, 180–182, 191, 195, 206–207, Integrated market, 107–113, 117, 120, 135,
209–210, 217–218, 228, 263, 273, 318, 141
320, 358, 367–369, 377, 381, 388, 391 Inventory/Inventories, 15, 23, 33, 35, 42–44,
F.o.b. price, 15, 48, 52, 64, 154, 157, 167, 51, 53, 70, 94, 127, 152–153, 177,
171–172, 175, 201–209, 211–218, 180–193, 205, 207, 236–237
221–224, 227–228, 337–339, 344, Inverse demand curve, 33–43, 45–46, 50, 54,
346–347, 356, 358–365, 367, 368, 370, 56, 63, 100–101, 107, 110, 113, 122,
372–377, 388–389 130, 132–133, 170, 202, 206–208, 219,
F.o.b. pricing, 208–209, 211, 214, 217–218, 224, 273–274, 280, 286, 294, 303, 308,
221–223, 228, 346, 361–365, 370 313
Free entry, 17, 99, 169, 219, 299, 303, 312, Inverse supply curve, 100–101, 110, 116,
323, 326, 344, 357, 386 132–133, 171
Free spatial demand curve, see Demand curve, Investment, 3, 6, 9, 16, 21, 27–28, 31–32, 35,
free spatial 38, 44, 54–55, 57–58, 70, 74–75, 93,
Freight, 15, 21, 29, 35, 44, 47, 51–53, 116, 128, 142, 147, 179, 197, 239–240, 259,
142, 143, 145, 155–158, 165, 209, 236, 270–271, 336, 357
268, 362–363, 371 Isolation/Isolated market, 30, 93, 97, 133, 194,
Functionalist, x, xi, xii, xvi 248, 273
K
G
Kink[s]/Kinked, 76–77, 108, 111, 116–118,
General equilibrium model, 318, 322, 325, 331
125, 127, 148, 207, 216–217, 290–291,
Geographic angle[s], 163
295, 297, 349, 359, 364, 372–375
Global net social welfare, 114, 125, 127, 134,
Kuhn-Kuenne algorithm, 171
136
Gradient, 18, 60, 97, 123, 297–298, 309, 344,
349, 372, 375 L
Labor market/Market for labor, 15–16, 66,
299–302, 318–319, 321–322, 326–327,
H
331, 335, 341, 344, 367–368, 370, 380
Hakimi Theorem, 174–175
Land market/Market for land, 232, 264, 266,
Half-freight, 47, 51–52, 60, 66, 209 270, 274, 277–278, 282–283, 286–287,
Hitchcock-Koopmans problem, 69–95, 127, 311, 322, 326, 331, 348, 366, 379–380
220–221 Leontief [production] technology, 32, 152–153,
Home market, 9, 14–16, 30, 35, 40, 45, 48, 57, 156, 165–166, 167, 171, 175, 251,
60, 66–67 265, 377
Homothetic, 23, 340 Linear expansion path, 32, 166, 168, 253
Homotheticity, 253 Linear program[ming], 18, 69–70, 78, 81–86,
125, 136, 220
I Local cost advantage, 1, 3, 25, 95, 176,
Income elasticity of demand, 330, 340 380–381
Indivisibility/Indivisibilities, 1, 3, 9, 25, 31, 70, Local demand, 8, 37, 71, 76, 87, 91, 97, 102,
166, 179, 188, 198, 380 105, 107–111, 113–115, 117–118,
Industrial restructuring, 17, 193 121–123, 125, 127, 132–142, 148, 232
Industry marginal cost curve, 99, 103–104 Local economy, 2, 25, 275, 336, 356, 362, 378
454 Subject Index
Social Welfare, 25, 34, 46, 54, 101–102, 105, Unbounded, 26, 204, 267, 329
114, 125–136, 274, 276, 280–281, Uncertainty, 11, 23, 53, 62, 64, 66, 154, 179,
291–292, 294–295, 303–305, 308–309, 183, 187, 191–192, 198, 205, 226–227,
312–313, 378, 382 231, 236–240, 264, 275, 297–298
Spatial equilibrium, 7, 12, 19–20, 100, 127, Uniform pricing, 22, 52
151, 204, 231, 277, 326–329, 330, 344, Unit production cost, 3, 69, 73–76, 78, 80, 87,
348–349, 355–356, 364 90–92, 276, 381
Spatial price equilibrium, 13–14, 93, 99–100, Unit shipping cost, 9, 15–16, 21–22, 25,
113–122, 126–127, 135–136, 142–145, 28–29, 35, 43–48, 50, 52, 53–67, 69,
147–148, 205, 291–292, 300 72–77, 81, 83–87, 90–94, 97, 99–101,
Specialization, 10 107–113, 107, 132–136, 140–141, 188,
Store of value, 31, 35–36, 234 198, 202–206, 209–210, 213, 216, 218,
Submarket, 9, 25, 45, 51, 93, 97, 123, 148–149, 221–224, 231–232, 243, 245, 249–255,
202, 209 258–266, 379–383
Substitutability/Substitution/Substitutable, 5, See also Shipping cost/Unit shipping cost
15, 30, 32, 36–37, 43, 51, 86, 129, 145, Unit shipping rate, 44, 55, 83, 100, 152, 167,
151–153, 165–169, 176–177, 184, 188, 175, 206, 210, 212–213, 216, 218,
191, 201–202, 204, 207–208, 221–228, 221–224, 250, 252, 254, 270–271,
232, 240, 260, 265, 275–276, 299–300, 274–275, 277, 279–280, 283–284, 286,
302–306 292, 294, 297, 299, 306, 308, 313, 325,
Supply step function, 77 332, 338–339, 344, 346, 360, 369, 383
Surplus, 33–34, 46, 54, 67, 81, 88, 89, 98, Unit transaction cost, 231, 264
101–102, 105, 109, 114, 129–135, Unpriced, 15, 30, 32–33, 39, 101, 103, 128,
137–140, 210, 274, 280–281, 301, 304, 133, 250, 279, 305
310, 382 Urbanization [economies/economy], 9, 25,
Symmetric shipping cost, 56 154, 178–179, 198, 380–381
T Utility function, 24, 223–226, 228, 232, 238,
Takayama, 13, 18, 114, 125–149 240–242, 250, 253, 264, 322, 332, 337,
Tariff jumping, 60, 128 340, 348, 366, 378
Tariff[s], 10, 15, 35, 44–45, 60, 128, 142–146,
236
V
Taxes, 15, 35, 44–45, 236, 270
Variable cost, 32, 37, 39–40, 74, 77, 98, 273
Theory-model dissonance, 408
Thick, 236–237 Vertical integration, 11, 30, 193, 379–380
Thiessen polygon, 204, 214–215 Voronoi, 214
Thin, 236–237 Voronoi diagram, 214
Total cost, 32–34, 43, 45–46, 53–54, 70, 74,
79, 82, 84, 87–88, 131, 166–167, 175,
W
181, 184, 186, 190, 192, 210, 263, 370
Walrasian [multimarket] equilibrium/
Trade area, 211–212, 214, 232, 381–382
process/linkage/perspective, 1, 5–6,
Transaction cost, 15–16, 36, 133, 236, 264–265
24–25, 123, 233, 235, 237, 265–266,
Transfer, 3, 8, 15, 35, 44, 94, 236
268–269, 288, 297–298, 332–333,
Transportation linear program, 70
374–376, 380–381
Transportation network, 3, 8, 26, 74, 94, 133,
Warranty, 15, 35, 42, 44–45, 205, 236, 379
154, 172–175
Weber–Launhardt map triangle, 159–160, 163
U Weber-Launhardt problem, 12, 151–155
Ubiquitous, 5, 75, 111, 152, 154–155, Weber-Launhardt weight triangle, 161–162
157–158, 166–167, 171–172, 176, 178, Wellspring, 31–32, 69, 103, 179,
270, 302, 318, 381 336–337, 358