JOURNAL OF URBAN ECONOMICS 40, 13]37 Ž1996.

ARTICLE NO. 0021

Urbanization and Economic Development: A Bias

toward Large Cities? 1
RONALD L. MOOMAW

Oklahoma State Uni¨ ersity, Stillwater, Oklahoma 74078

AND

ALI M. SHATTER

Faculty of Commerce and Economics, Sana’a Uni¨ ersity, Sana’a, Republic of Yemen

Received February 16, 1994; revised January 18, 1995

We find that a nation’s urban population percentage increases with GDP per
capita; industrialization; export orientation; and possibly, foreign assistance. It
decreases with the importance of agriculture. Industrialization and agricultural
importance have the same implications for the concentration of urban population
in cities with 100,000 q population as for the urban percentage. Greater export
orientation reduces such concentration. Finally, GDP per capita, population, and
export orientation reduce primacy. Political factors, such as whether a country’s
largest city is also its capital, affect primacy. Our results do not seem to imply that
developing-country urbanization today differs fundamentally from urbanization in
the past. Q 1996 Academic Press, Inc.

INTRODUCTION
Urbanization is closely linked to economic development. As economies
develop, relative and absolute changes in demand increase the relative and
absolute importance of the manufacturing and service sectors. These
sectors are much less land intensive than the agricultural sector, and they
allow easier substitution of nonland for land inputs. Thus, firms in these
sectors can concentrate in urban areas in spite of urban land’s high prices
ŽGraves and Sexton w8x, Mills w11x, Mills and Becker w12x, and Mills and
Hamilton w13x.. Internal economies of scale, economies of urbanization
1
We presented an early version of this paper at the 1992 meetings of the Regional Science
Association in Chicago. It was one of several papers presented by former students in honor of
Edwin Mills. We thank Pat Norton for arranging the sessions; we also thank Charles Becker,
Jan Brueckner, Mike Edgmand, Edwin Mills, Andreas Savvides, and two anonymous referees
for their helpful suggestions. They, of course, are not responsible for any remaining errors.

13
0094-1190r96 $18.00
Copyright Q 1996 by Academic Press, Inc.
All rights of reproduction in any form reserved.
14 MOOMAW AND SHATTER

and localization, and positive transportation costs increase the urban

advantage for both production and consumption ŽGraves and Sexton w8x
and Mills and Becker w12x.. As markets grow, the division and specializa-
tion of labor increases. This places a premium on physical proximity to
reduce transportation costs and often increases the importance of face-to-
face contact to reduce communication costs ŽEvans w7x and Stigler w20x..
These ideas underlie formal models of urbanization, such as Brueckner’s
w4x Mills-type model. Brueckner’s model determines a country’s urban
population or, more precisely, the population of the largest city in a
country. Equilibrium requires that individuals be indifferent between ur-
ban and rural location, which usually requires higher income in the urban
area because of urban land’s higher price. This approach emphasizes he
increasing productivity and therefore the increasing attractiveness of urban
areas as economies develop; it considers the urbanization that accompa-
nies development as natural and desirable.
Although they do not dispute this analysis for developed countries,
economists such as Bairoch w2x and Todaro w21x do not think that urbaniza-
tion in developing countries today is benign or in response to positive
economic features of urban areas. Instead, they believe it is excessive.
Bairoch, for instance, argues that urbanization in developing countries is
much greater than their level of development would suggest; in short, he
regards it as excessive. He argues that this ‘‘excessive’’ urbanization follows
rapid population growth caused by, among other factors, declines in
mortality rates; rapid population growth in turn leads to rural crowding
and stimulates rural to urban migration. In addition, he argues that
artificially high urban wages pull a disproportionate part of the population
to urban areas. Bairoch invokes a duality between the productive modern
urban sector, which uses modern technology, and the backward agricul-
tural sector, which uses traditional technology, to partially explain artifi-
cially high urban wages. Todaro takes a similar position concerning wages;
he differs in his greater emphasis on economically inefficient migration
caused by legally and socially determined minimum wage rates, govern-
ment subsidies of industrialization and urbanization, and a general bias
toward the largest city.
In a study that we build on, Mills and Becker w12x regress the percentage
of a country’s population in urban areas Žthe percent urban. on the
percentage of the labor force in agriculture, GNP per capita, and other
variables. As expected, they found a positive relationship between the
percent urban and GNP per capita and a negative one between percent
urban and the agricultural share. Because our focus is on the concentra-
tion of population in large cities, we estimate a similar percent-urban
equation. Our intent, however, is to use it as a benchmark in examining
urban concentration.
 URBANIZATION AND ECONOMIC DEVELOPMENT 15

One survey ŽHamer and Lin w9x. of the empirical urbanization literature
cites only two studies of the relationship between city size Žurban concen-
tration. and economic development. Rosen and Resnick’s study w18x uses
the exponent calculated from a Pareto distribution of city sizes as a
dependent variable; it finds that countries with smaller populations and
with lower per capita incomes have more concentrated population centers.
The other study, Henderson w10x, uses a Herfindahl index to measure
urban concentration; it finds that urban concentration decreases with the
total urban population, the importance of agriculture, the ratio of manu-
facturing employment to service employment, a federalized system of
government, and literacy.
Wheaton and Shishido w24x also measure urban concentration with a
Herfindahl index; using a nonlinear model, they relate urban concentra-
tion to GNP per capita. They interpret their results as showing that
optimal city population increases with development ŽGNP per capita. up to
a middle level of development and subsequently falls. Wheaton and
Shishido’s implied optimal city size is much larger than that found in most
developing countries. Presumably, they would not claim that developing-
country urbanization is excessive.
A recent paper by Ades and Glaeser w1x emphasizes the role of govern-
ment and politics in determining the population of the largest city in a
country. In their model, government uses spatial tax policy to survive while
exploiting the population. The model assumes that favorable treatment of
the largest city enhances the government’s survival. It implies that dictator-
ship increases the desired population of the largest city and that political
instability increases it, especially for democracies. In their empirical model
of the determinants of the size of a nation’s largest city, Ades and Glaeser
use measures of dictatorship and of political instability and also such
variables as a capital-city dummy, nonurbanized population, land area, the
share of the labor force outside agriculture, and the share of trade in
GDP. One of these variables, the trade variable, has a statistically signifi-
cant coefficient with the expected negative sign. The other variables have
statistically significant coefficients with the expected positive signs. An-
other variable, GDP per capita, has a positive but insignificant coefficient.
In an equation that they do not discuss in detail, Ades and Glaeser find
that both population share of the largest city and urbanization outside the
main city have negative and significant effects on growth of GDP per
capita. Moomaw and Shatter w16x also find that the population share of the
largest city retards growth; a crucial difference, however, is that they find
that concentration of the urban population in large cities, ceteris paribus,
enhances growth. Thus, Moomaw and Shatter’s results do not completely
16 MOOMAW AND SHATTER

support the Ades and Glaeser w1, p. 14x conclusion that ‘‘Large cities
generate rent-seeking and instability, not long term economic growth.’’
This paper provides evidence regarding the determinants of three as-
pects of urbanization. First, it estimates the determinants of the urban
percentage with a Mills]Becker-type model as a benchmark against which
the determinants of metropolitan concentration, defined as the percentage
of the urban population in cities of 100,000 q population, are compared. It
also estimates the determinants of urban primacy defined as the share of
the urbanized population in the largest city.
This study extends the existing literature in several ways. First, depend-
ing upon the availability of the urbanization data, it uses 68, 74, or 90
countries for the years 1960, 1970, and 1980. Thus, it includes more
countries than other studies ŽHenderson w10x, Rosen and Resnick w18x,
Wheaton and Shishido w24x. that focus on urban concentration and uses 3
years of cross-sectional data. Ades and Glaeser w1x use more countries than
our study Ž85 compared to 74. and use data from more years Žfour years
compared to three.. They, however, average the data over time; thus their
study is pure cross-sectional, whereas we use a panel data set. We use
ordinary least squares ŽOLS. to estimate equations that include dummy
variables for regions of the world and for time; such estimates provide
information regarding ‘‘overurbanization’’ and possibly about changing
temporal patterns of urbanization. We also use panel estimation tech-
niques in an attempt to extract more information from the data. We
present estimates based on the within estimator}a dummy variable for
each country. Because we have only three cross-sections in the panel, we
believe that the pooled OLS estimates with regional and time dummies
yield the preferred results.
This study examines urbanization and development across countries. By
distinguishing between the urban percentage, metropolitan concentration,
and primacy, it allows a variable, say export orientation, to have different
effects on different aspects of urbanization. It finds, for instance, that the
relative importance of exports increases the urban percentage but reduces
both metropolitan concentration and primacy. These important results
provide empirical support for recent theories of urbanization Že.g.,
Elizondo and Krugman w6x.. The study also shows that the importance of
agriculture and industry affects the urban percentage and metropolitan
concentration in expected ways, but that it does not affect primacy.
Perhaps this means that the urbanization and metropolitan concentration
are more responsive to economic forces than is primacy, implying that
primacy may be more responsive to political forces. In short, the paper
studies the determinants of various aspects of urbanization, whereas most
papers study only one aspect, usually some element of primacy or of the
urban percentage.
 URBANIZATION AND ECONOMIC DEVELOPMENT 17

The paper next discusses general hypotheses regarding urbanization and

the variables used to represent them. Subsequently, it presents the esti-
mates of the determinants of the urban percentage, metropolitan concen-
tration, and urban primacy.
AN OVERVIEW OF URBANIZATION
This section discusses the ideas behind and variables used in the
equations estimated in the next two sections. A fundamental idea is that
economic development increases market size, which in turn results in
increased division and specialization of labor. Other important ideas relate
to urban economic theory, changing demand patterns with development,
and development indicators other than per capita production. GDP per
capita and the shares of agriculture and industry in the aggregate economy
are the overall development indicators. Other variables considered are
literacy, commodity exports as a share of GDP, and foreign assistance as a
share of GDP.
As economies develop, the division and specialization of labor increase
in response to the size of the market ŽEvans w7x and Stigler w20x.. Increased
reliance on external sources for inputs and for the distribution of outputs
replaces reliance on the subsistence farm, the village market, and the
company town. Greater specialization and trade ensues, placing a premium
on proximity. Economizing on transportation and communication leads to
urbanization. Thus, economic development, holding constant the distribu-
tion of economic activity across agriculture, industry, and services, may
lead to increased urbanization}a greater share of economic activity in
cities.
Changes in demand patterns that accompany development increase the
importance of services and manufactured goods Žsee, e.g., Graves and
Sexton w8x and Mills and Becker w12x.. With the shift toward services and
manufacturing goods, cost advantages due to agglomeration economies
favor cities. The ‘‘primary agglomeration benefits of a city are captured by
manufacturing industries, less so in services Žand of course agriculture .’’
ŽGraves and Sexton w8, p. 161x.. Thus, demand and cost changes associated
with increases in per capita income may also lead to greater industrializa-
tion and urbanization.
In short, economic development may lead to greater urbanization for
two reasons. First, the increased division of labor associated with larger
markets makes economizing on communication and transportation costs
more important. This in turn implies an increasing advantage for urban
location. Second, a likely shift in economic structure away from agriculture
Žoften associated with development. may lead to greater urbanization.
Changes in economic structure and the level of economic development can
be independent. Thus, an empirical model should include variables to test
18 MOOMAW AND SHATTER

the independent effects of economic development Žmeasured by per capita

production. and sectoral structure Žmeasured by shares of economic activ-
ity in agriculture, industry, and services..
An increased literacy rate also accompanies economic development.
Does it also have an independent effect on urbanization? A presumption
exists that an increased level of education leads to greater urbanization. If
this is a direct, positive, independent determinant of urbanization, it might
imply that education alters preferences to make urban areas more attrac-
tive. Henderson w10x, however, argues that the literacy rate reflects the
technological sophistication of the economy. Therefore, he expects a
positive but indirect association of the literacy rate with urbanization
because literacy is a proxy for the type of technology that leads firms to
urban areas.
Mills and Song w14x suggest that the unexpectedly high level of urbaniza-
tion in South Korea might be due to the openness of the Korean economy.
Openness might lead to urbanization for two reasons. First, it may increase
the importance of transportation nodes, which are present mostly in urban
areas. Second, it may increase demand for marketing, financing, and
communication}face-to-face contact}making an urban location more
important. Elizondo and Krugman w6x, on the other hand, expect increased
trade to decrease the importance of large cities and more particularly the
largest city. They argue that in a small, closed economy, backward and
forward linkages combined with economies of scale cause production to be
concentrated in a single city. Economies of scale dictate that most indus-
trial products will be produced at a single location, and backward and
forward linkages to customers and suppliers dictate production at the
same, single location. Therefore, excessive urban size or primacy develops.
The process is reversed with greater openness because industrial firms buy
and sell in international markets, releasing them from the same, single
location. Therefore, openness reduces the size of large cities.
Because foreign aid may be allocated in response to political as well as
economic considerations, we use it to test for a political bias toward
urbanization. If developing countries wish to promote urbanization, for-
eign aid provides a potential instrument.
These considerations lead to the use of the following variables in the
regression models of the next section. The models take the percent urban
as a log linear function of: Ži. GDP per capita ŽGDPrCAP., Žii. the
percentage of the labor force in agriculture ŽAGRrLAB., Žiii. the percent-
age of the labor force in industry ŽINDrLAB., Živ. the literacy rate for
those 15 years and older ŽLITR., Žv. commodity exports as a proportion of
GDP ŽEXPrGDP., and Žvi. foreign assistance as a proportion of GDP
ŽASSISTrGDP..
 URBANIZATION AND ECONOMIC DEVELOPMENT 19

EMPIRICAL DETERMINANTS OF THE URBAN

PERCENTAGE
Economic development, as measured by income per capita, has a strong
association with urbanization, as measured by the urban percentage. For
example, a simple regression of the log of the urban percentage on the log
of GDP per capita for 90 countries for 1980 yields a coefficient of 0.49 and
an R 2 of 0.71. In this section, we test for multiple determinants of the
urban percentage using log-linear models and a panel data set that
includes 90 countries and three years Ž1960, 1970, and 1980..2
The first equation in Table 1 shows pure cross-section results using 1980
data.3 In this multiple regression the coefficient of GDPrCAP is 0.32,
compared to 0.49 in the simple regression. The coefficients of AGRrLAB
and INDrLAB have the expected signs and are significant. The level of
economic development and the sectoral distribution of production exert
significant, independent effects on urbanization.
The coefficients of AGRrLAB and INDrLAB support Sexton and
Graves’s hypothesis regarding the relative attraction that urban areas have
for industry, agriculture, and services. An increase in INDrLAB, holding
AGRrLAB constant, is at the expense of the residual sector, which we call
services. Therefore, the positive and significant coefficient of INDrLAB is
consistent with the Graves and Sexton conjecture; it implies that an
increase in industry’s share at the expense of the residual}services}in-
creases urbanization. Similarly, the negative and significant coefficient of
AGRrLAB implies that a substitution of agriculture for services reduces
urbanization.
EXPrGDP has a positive effect on the urban percentage. Other things
equal, more open economies are more urbanized.
The other significant coefficient in Eq. Ž1. is for ASSISTrGDP, foreign
assistance as a proportion of GDP.4 The positive coefficient suggests that
foreign assistance has promoted urbanization, independent of its effect on
development or industrialization. It may suggest a political bias toward
urbanization. Such a bias has long been suspected. In the early 1970s an
OECD report said that foreign assistance ‘‘was instrument in favouring the
towns’’ w17, p. 204x.. To the best of our knowledge, however, our study is

2
See the Appendix for details about the data.
3
Estimates using only 1980 data are reported for two reasons. One, data for ASSISTrGDP
were unavailable for 1960; because we were particularly interested in this variable we
estimated equations using the available data. Two, we want to compare the pooled results
with the results for 1980, a year for which the data may be more accurate than those for
earlier years.
4
ASSISTrGDP is foreign assistance received relative to GDP. Countries that do not
receive foreign assistance, including donor countries, are assigned a value close to zero rather
than zero, so that logarithms can be taken.
20 MOOMAW AND SHATTER

TABLE 1
Determinants of the Urban Percentage a

Variable Equation Ž1. Equation Ž2. Equation Ž3. Equation Ž4. Equation Ž5.

GDPrCAP a
0.32*** 0.31*** 0.35*** 0.19*** y0.08
Ž4.27. Ž4.11. Ž6.18. Ž3.48. Žy1.20.
ASSISTrGDP 0.01**
Ž2.57.
DASSIST 0.26** 0.24*** y0.12
Ž2.15. Ž2.68. Žy1.16.
AGRrLAB y0.14*** y0.13** y0.17*** y0.12*** y0.07*
Žy2.66. Žy2.49. Žy4.48. Žy3.64. Žy1.82.
INDrLAB 0.13*** 0.13** 0.23*** 0.19*** 0.12***
Ž2.61. Ž2.51. Ž5.92. Ž5.19. Ž3.10.
LITR 0.11 0.13 0.10** 0.27*** 0.16***
Ž1.37. Ž1.57. Ž2.51. Ž5.80. Ž4.31.
EXPrGDP 0.10** 0.09** 0.06* 0.08** 0.01
Ž2.53. Ž2.36. Ž1.91. Ž2.54. Ž0.34.
D70 y0.76* y1.04** y0.66
Žy1.73. Žy2.34. Žy1.49.
D80 0.02 y0.01 y0.39**
Ž0.19. Žy0.06. Žy1.96.
CONSTANT 2.40*** 2.37*** 2.49*** 1.92***
Ž6.05. Ž5.87. Ž11.62. Ž7.88.
DNAMER y0.01
Žy0.10.
DCAMER 0.29**
Ž2.44.
DSAMER 0.51***
Ž4.38.
DMEAST 0.48***
Ž4.20.
DSASIA y0.09
Žy0.57.
DSEASIA y0.11
Žy0.91.
DCENAF 0.35**
Ž2.08.
DWAF 0.35**
Ž2.45.
DSAF y0.21
Žy1.48.
DOCEANIA 0.33*
Ž1.81.
R2 0.76 0.76 0.74 0.81 0.99
Observation 90 90 270 270 270
Year 1980 1980 1960]1980 1960]1980 1960]1980
a
The number in parentheses is the t statistic. The number above it is the coefficient.
*Significance at 90%.
**Significance at 95%.
***Significance at 99%.
 URBANIZATION AND ECONOMIC DEVELOPMENT 21

the first to find a statistical association between foreign assistance and

urbanization.
To provide additional evidence for the relationship between urbaniza-
tion and development, we pooled data for 1960, 1970, and 1980. Data are
available for all three years for all variables except ASSISTrGDP. Infor-
mation is available, however, that allows the construction of a dummy
variable, DASSIST, which takes the value of one if a country is a recipient
of foreign assistance. Compared to the dummy variable, the continuous
variable, ASSISTrGDP, has the advantage of measuring the extent of
foreign assistance. To decide if ASSISTrGDP and DASSIST measure the
same thing, we run otherwise identical regressions with one variable, then
the other. Equation Ž2. contains the same variables and data as Eq. Ž1.,
except that DASSIST replaces ASSISTrGDP. The coefficient of DASSIST
is significant and positive in the second regression, just as the coefficient of
ASSISTrGDP is in the first. Just as importantly, the other coefficients
remain essentially unchanged.5
Equation Ž3. reports regressions computed with pooled data for 1960,
1970, and 1980. Dummy variables for 1970 and 1980}D70 and D80}cap-
ture time effects. The coefficients reported in Eq. Ž3. are similar to those
in Eqs. Ž1. and Ž2.. For example, the coefficient of GDPrCAP only
changes from 0.31 or 0.32 to 0.35. Exceptions are the larger coefficient of
INDrLAB and the smaller coefficient of EXPrGDP in Eq. Ž3.. Pooling
the data increases the precision of the estimates as seen in a comparison
across the equations. For instance, the coefficient of the literacy rate
attains increased significance. In Eq. Ž3. all coefficients are significant
except for that of D80. The results conform with expectations regarding
the effects of economic development on urbanization. In addition, the
coefficient on the foreign assistance variable supports the assertion that
political factors in developing countries may lead to a bias toward urban-
ization.
The coefficients of the dummy variables for 1970 and 1980 in Eq. Ž3.
and subsequent regressions imply that the urban percentage function
either was lower or unchanged vis-a-vis` 1960. In particular, there is no
evidence of an upward drift of the function, a drift feared by some.
Equation Ž4. augments Eq. Ž3. by adding dummy variables for separate
geographic regions of the world to capture regional effects. Several notice-
able changes emerge. First, the coefficient of GDPrCAP is smaller.
Second, the coefficient of LITR is substantially larger. Third, the coeffi-
cient of DASSIST, the foreign assistance variable, becomes insignificant.
ŽIn a regression not reported here using 1970 and 1980 data and replacing

5
The same conclusion is obtained from other estimates not reported here. They are
available in Shatter w19x or on request.
22 MOOMAW AND SHATTER

DASSIST with the continuous foreign assistance variable, ASSISTrGDP,

the same three changes emerge..
The geographic region dummies are DNAMER ŽNorth America.,
DCAMER ŽCentral America and Caribbean., DSAMER ŽSouth America.,
DMEAST ŽMiddle East and North Africa., DSASIA ŽSouth Asia., DSE-
ASIA ŽSoutheast Asia., DCENAF ŽCentral Africa., DWAF ŽWest Africa.,
DSAF ŽSouthern Africa., DEUR ŽWestern Europe., and DOCEANIA
ŽAustralia and New Zealand.. Western Europe is the omitted category.
ŽSee the Appendix for details.. The dummy variables for Central and
South America, Central and West Africa, and the Middle East have
significant, positive coefficients. Countries in these regions have urban
percentages that are greater than those predicted by the determinants of
urbanization in Europe. This result is consistent with Bairoch’s contention
that Third World countries have urbanized without developing. But, as we
shall argue below, such a result is also consistent with other explanations.
Next we report the results for Eq. Ž5., the specification with
country fixed effects. This equation is estimated using the within, or
dummy variable, estimator.6 Three variables retain significant coefficients
}AGRrLAB, INDrLAB, and LITR}thus economic structure and edu-
cation have robust effects. Although the coefficients of GDPrCAP and
EXPrCAP are not significant in this model, the earlier estimates with
significant coefficients for all these variables are preferred because of the
lack of sufficient time-series data for each country.
The results for the urban percentage model in Table 1 are generally
consistent with expectations and with previous research. GDP per capita,
the share of economic activity in agriculture, services and industry, and
literacy have the expected effects on the urban percentage. We also find
Žwe believe for the first time. a positive effect of export orientation on the
urban percentage. Furthermore, some estimates support the conjecture
that foreign aid may be biased toward urban areas. Finally, the fixed
effects for some regions that include developing countries have positive
coefficients. This says that urbanization is greater in these regions}per-
haps because of political factors}than would be predicted based on
European data; this may be Bairoch’s ‘‘urbanization without development.’’
With these results as a benchmark, the paper now turns to the main
question: Does economic development bias urbanization toward large
cities?

6
Readers of an earlier draft encouraged us to exploit our panel data by including country
fixed effects. In doing so, we omitted the foreign assistance variable because it also is a fixed
effect.
 URBANIZATION AND ECONOMIC DEVELOPMENT 23

DETERMINANTS OF METROPOLITAN CONCENTRATION

Two measures of a bias toward large cities are relevant. One is
metropolitan concentration, which we define as the share of a country’s
urbanized population in cities of greater than 100,000 population. The
other is primacy, which we define as the largest city’s share of a country’s
urbanized population.7 First we deal with metropolitan concentration, then
primacy.
Metropolitan concentration Ž MrU . is equivalent to the metropolitan
share of the national population Ž MrN . divided by the urban share of the
national population ŽUrN ., where M, U, and N are metropolitan, urban,
and national population, respectively. The log of metropolitan concentra-
tion, therefore, is equivalent to the log of the metropolitan percentage
minus the log of the urban percentage. We assume that the metropolitan
percentage depends on the same variables as the urban percentage, so we
use the specification of Table 1 for the metropolitan concentration equa-
tion.8 If an explanatory variable has a positive coefficient, we say it
‘‘biases’’ urbanization toward larger cities.9
This section considers whether metropolitan concentration increases
with economic development. Before discussing the multiple regressions, we
note that the simple regression of the log of metropolitan concentration on
the log of GDPrCAP yields a coefficient of 0.07 with a t statistic of 2.06;
R 2 is 0.06. This suggests that economic development measured by produc-

7
We decided not to use the exponent of the Pareto distribution ŽRosen and Resnick w18x.
and the Herfindahl index ŽHenderson w10x. because we wanted to distinguish between
population concentration in large cities and population concentration in the primate city. The
former measures do not allow for this distinction. Furthermore, we consider the concentra-
tion of the urbanized population in large cities and in the primate city because we want to
test whether economic development biases population toward large cities and the primate city
relative to smaller urban places. To do this, we have to accept the different definitions of
urban places used by different countries.
8
In estimates reported in Shatter w19x, all of the variables significant in, for instance, the
pooled urban percentage model are significant in the pooled metropolitan percentage model,
except for the time dummy for 1970. Although the coefficients differ quantitatively,
the pertinent qualitative differences are three: Ži. the coefficient of EXPrGDP is negative in
the metropolitan percentage model; Žii. the time dummy for 1970 is not significant in the
metropolitan percentage model; and Žiii. the time dummy for 1980 is negative and significant.
The fact that the 1970 metropolitan data were projected by Davis and Golden w5x may explain
why the 1970 coefficient in the metropolitan percentage is not significant, whereas the same
coefficient for the urban percentage equation is significant. The different source for the 1980
metropolitan data might explain why the 1980 coefficient in the metropolitan equation is
significant compared to the corresponding coefficient in the urban equation.
9
Suppose lnŽ MrN . s a q b ln ŽGDPrCAP. and lnŽUrN . s a9 q b9 ln ŽGDPrCAP.. Then,
lnŽ MrN . y lnŽUrN . s lnŽ MrU . s Ž a y a9. q Ž b y b9.ln ŽGDPrCAP.. The coefficient in
this regression gives the differential effect of the independent variable on the metropolitan
percentage relative to the urbanization percentage.
24 MOOMAW AND SHATTER

tion per capita is associated with a greater concentration of urban popula-

tion in metropolitan areas}a bias toward big cities. The relationship,
however, is much weaker than the association between the urban percent-
age and development.
Does the bias toward large cities persist in the multiple regressions? The
results in Table 2 suggest that the percentages of the labor force in
agriculture, industry, and services have significant effects on metropolitan
concentration. In particular, an increase in AGRrLAB, holding INDrLAB
constant, pulls population to rural areas and smaller urban areas. This
result conforms with Henderson’s w10x work. An increase in INDrLAB, at
the expense of services Žholding AGRrLAB constant., increases
metropolitan concentration in contrast to Henderson’s result that an
increase in the manufacturing]services ratio reduces urban concentration.
In general, these results are consistent with the position that metropolitan
concentration is determined by the interaction of demand and cost. Note,
however, that the result for AGRrLAB holds for the country-fixed-effects
equation, Eq. Ž10., but that the result for INDrLAB does not.
An increase in the relative importance of exports reduces metropolitan
concentration in all of the equations. The result is consistent with
Elizondo and Krugman’s w6x arguments regarding the lesser importance of
large cities in more open economies. It is also an empirical result consis-
tent with, but stronger than, that of Ades and Glaeser w1x for a similar
variable. Results for this variable from the two sets of equations show that
a greater export orientation Ži. increases the urban percentage and Žii.
reduces metropolitan concentration. Trade increases urbanization, but its
effect is biased toward small rather than large cities.
The variable for foreign assistance has a positive coefficient, which is
significant at a 0.10 level, in Eqs. Ž6. and Ž8.. Equation Ž6. uses DASSIST
and 1980 data, whereas Eq. Ž8. uses ASSISTrGDP and pools data for 1970
and 1980. This would imply that foreign assistance has a bias toward large
cities relative to all urban places. But the coefficient of DASSIST is not
significant in Ža. Eq. Ž7. with pooled 1970 and 1980 data, Žb. an equation
with pooled data for the three years Žresults not reported., or Žc. Eq. Ž9.
with pooled data and regional fixed effects.10 The variable is omitted from
Eq. Ž10., the equation with country fixed effects.
The development indicator, GDPrCAP, and the literacy rate, LITR, do
not have significant coefficients in any of the equations in Table 2.
Economic development per se is neutral with respect to metropolitan
concentration.

10
In Shatter w19x the continuous variable and the dummy variable for foreign assistance are
used in equations with regional fixed effects. In these equations, the foreign assistance is
never significant.
 URBANIZATION AND ECONOMIC DEVELOPMENT 25

TABLE 2
Determinants of Metropolitan Concentration a

Variable Equation Ž6. Equation Ž7. Equation Ž8. Equation Ž9. Equation Ž10.

GDPrCAP a
0.05 y0.01 0.07 y0.03 y0.00
Ž0.64. Žy0.26. Ž1.19. Žy0.53. Žy0.01.
ASSISTrGDP 0.01*
Ž1.67.
DASSIST 0.20* y0.02 y0.04
Ž1.71. Žy0.28. Žy0.50.
AGRrLAB y0.10* y0.07** y0.07* y0.10*** y0.13***
Žy1.94. Žy2.37. Žy1.92. Žy3.33. Žy3.14.
INDrLAB 0.14*** 0.08*** 0.07** 0.12*** y0.03
Ž2.73. Ž2.65. Ž2.17. Ž4.06. Žy0.50.
LITR y0.08 y0.01 y0.04 y0.07 y0.00
Žy0.91. Žy0.05. Žy0.67. Žy1.39. Žy0.24.
EXPrGDP y0.08* y0.10*** y0.09*** y0.05** y0.14***
Žy1.81. Žy3.78. Žy2.68. Žy1.80. Žy3.37.
D70 0.10** 0.10** y0.35**
Ž2.30. Ž2.35. Žy2.68.
D80 y0.48*** y0.56*** y0.24 0.02***
Žy3.48. Žy3.08. Žy1.55. Ž19.90.
CONSTANT 6.81*** 7.31*** 7.31*** 7.00***
Ž15.24. Ž23.21. Ž17.57. Ž18.36.
DNAMER 0.21**
Ž2.02.
DCAMER 0.12
Ž1.33.
DSAMER 0.21**
Ž2.45.
DMEAST 0.21**
Ž2.46.
DSASIA 0.20
Ž1.43.
DSEASIA 0.42***
Ž4.49.
DCENAF y0.01
Žy0.08.
DWAF 0.01
Ž0.08.
DSAF 0.30***
Ž2.82.
DOCEANIA 0.23*
Ž1.94.
R2 0.18 0.17 0.14 0.28 0.99
Observations 68 204 136 204 204
Year 1980 1960]1980 1970]1980 1960]1980 1960]1980
a
The number in parentheses is the t statistic. The number above it is the coefficient.
*Significance at 90%.
**Significance at 95%.
***Significance at 99%.
26 MOOMAW AND SHATTER

The coefficients for the dummy variables for 1970 and 1980 in Table 2
are usually significant but of opposite sign. Furthermore, in Eq. Ž10. the
coefficient of D70 changes from positive in the earlier equation to negative
and that of D80 changes in the opposite direction. The results for the time
dummies are suspect because the metropolitan data for 1970 were pro-
jected ŽDavis and Golden w5x. and the metropolitan data for 1980 were
from a different source ŽUnited Nations w22x. than the 1960 and 1970 data.
Equation Ž9. includes fixed effects for geographic regions. Three of the
six regions with positive coefficients for the dummy variables}South
America, the Middle East, and Southern Africa}contain many developing
countries. This could be taken as support for Bairoch’s concept of urban-
ization without development, particularly considering that two of these
regions were ‘‘overurbanized’’ in the urban-percentage results. But four
other regions with developing countries do not show such an effect.
Three regions}Oceania, North America, and Southeast Asia}consist-
ing entirely or partially of more-developed countries also have positive
fixed effects. The fixed effects for these regions might have a technological
explanation. Much of the original development of large cities in North
America and Oceania occurred under different technology than did those
of Europe. Southeast Asia, with its much more recent economic develop-
ment, also has much urbanization under different conditions than other
regions. Europe, on the other hand, has had its urban systems in place for
a long time. Perhaps what seems to be urbanization without development
is really appropriate urbanization under modern technological conditions.
We tentatively conclude that metropolitan concentration does not de-
pend on economic development per se, although the results for agriculture
and industry suggest a dependence on economic structure. To the extent
that economic development is intrinsically associated with reductions in
agriculture and increases in industry, however, it does lead to metropolitan
concentration. Conversely, countries in which exports are more important
have lower levels of metropolitan concentration than they otherwise would.
Finally, there is an indication that ‘‘newer’’ regions have greater metropoli-
tan concentration, an effect perhaps due to technology.
DETERMINANTS OF URBAN PRIMACY
The specification of the urban primacy equation differs from that of
previous models. We add three variables based on suggestions in the
literature. Mills and Hamilton w13x state that ‘‘As a rule, large countries
tend to be less primate than small countries . . . and high-income countries
tend to be less primate than low-income countries’’ Žp. 411.. To control the
size effect, which may also control the arbitrariness of national boundaries,
we add the log of the national population, POP, as an independent
variable. Henderson w10x includes and finds significant a dummy variable
 URBANIZATION AND ECONOMIC DEVELOPMENT 27

for a unitary relative to a federal system of national government in his

equation for urban deconcentration. Mills and Hamilton w13x summarize
the rationale:
In federal systems, states or pro¨ inces ha¨ e some constitutional autonomy and can
raise their own taxes to pro¨ ide the local go¨ ernment ser¨ ices . . . needed to make
urban areas grow. In centralized go¨ ernments, state or pro¨ incial and local go¨ ern-
ments recei¨ e their spending instructions, as well as much of their re¨ enues, from the
central go¨ ernment. National go¨ ernments in such countries in¨ ariably pour money
into the national capital . . . . Žp. 412.

To capture these effects on primacy we include the dummy variable

DCENTR, which takes the value of one for a unitary national government.11
We also add DCAPCTY, which takes the value of one if a country’s largest
city is also its national capital. Ades and Glaeser w1x use the latter, but not
the former, variable to capture political effects.
We removed the foreign assistance variable from the primacy equation
because its coefficient was never significant Žsee Shatter w19x.. Although
AGRrLAB and INDrLAB also had insignificant coefficients in that
preliminary study, we report equations that include these variables because
Ades and Glaeser use one of them, percent agriculture, in their estimates.
Consequently, in Table 3 we report results for four sets of equations with
the following attributes: the set that includes Ža. GDPrCAP, AGRrLAB,
and INDrLAB wEqs. Ž11. and Ž12.x; Žb. GDPrCAP and excludes
AGRrLAB and INDrLAB wEqs. Ž13. and Ž14.x; Žc. AGRrLAB and
INDrLAB and excludes GDPrCAP wEqs. Ž15. and Ž16.x; and Žd.
GDPrCAP and AGRrLAB and excludes INDrLAB wEqs. Ž17. and Ž18.x.
Only the estimates that pool the observations for different periods are
reported and discussed. Primacy is measured as the population of the
largest city as a proportion of total urban population. GDPrCAP has a
negative coefficient in the six equations in which it appears; the t ratios
range from 1.01 to 4.17. The coefficients of GDPrCAP are significant only
in the equations that exclude both AGRrLAB and INDrLAB, suggesting
troublesome collinearity among the variables. In addition, the coefficients
for POP and EXPrGDP are negative and significant in all equations in
Table 3, while the coefficients of LITR are positive and significant.
So primacy is greater in countries with lower GDPrCAP. But why? Is it
because of economic forces or some combination of cultural, political, and
social forces? We argue that economic forces are extremely important, on
both theoretical and empirical grounds. The theoretical argument ad-
vanced by Elizondo and Krugman w6x derives from Adam Smith’s famous
theorem that ‘‘the division of labor is limited by the extent of the market.’’

11
We followed Mutlu w15x in classifying governments as unitary or federal.
28 MOOMAW AND SHATTER
URBANIZATION AND ECONOMIC DEVELOPMENT 29
30 MOOMAW AND SHATTER

Indeed, Stigler Žw20, p. 139. states that

. . . geographic dispersion is a luxury that can be afforded by industries w countries?x

only after they ha¨ e grown large Ž so that e¨ en the smaller production centers can reap
the major gains of specialization. . . . . w my questionx

As relatively small production centers become large enough to gain

some advantages of specialization, transportation costs dictate geographic
dispersion of production. With the increase in the extent of the market,
relatively small centers become more important}larger}at the expense
of the largest cities. This geographic dispersion is exactly what the coeffi-
cients of GDPrCAP, POP, and EXPrGDP imply. An increase in
GDPrCAP, holding POP and the other variables constant, means that the
economy is larger, allowing the advantages of specialization to be reaped,
even if production is dispersed. An increase in population, holding
GDPrCAP constant, also implies a larger economy. The coefficients of
POP is negative and significant in all equations, implying that a larger
economy has less primacy. Similarly, the coefficients for EXPrGDP are
negative and significant in all equations. As predicted directly by Elizondo
and Krugman and indirectly by Smith and Stigler, as the ‘‘extent of
the market’’ increases with EXPrGDP, primacy declines}dispersion
increases.
Although the coefficients of GDPrCAP and POP are quite similar in
the equation with and without regional fixed effects, the coefficient of the
export variable is stronger in those with regional effects. The regional fixed
effects themselves suggest greater primacy in countries classified in Cen-
tral America, the Middle East, Southeast Asia, and Oceania. The countries
classified in South America and West Africa have positive coefficients
for the regional fixed effects that just miss significance. These results
may support an independent political or underdevelopment bias toward
primacy.
We do not completely ignore political variables. The dummy variable for
a country whose national capital is its largest city ŽDCAPCTY. has
significant coefficients, with the expected sign, in all equations. But the
dummy variable for the unitary national government, DCENTR, has
insignificant coefficients in all equations.
The results for economic structure suggest that it does not play a role in
primacy. First, the coefficient of INDrLAB is never significant. Second,
AGRrLAB has significant but positive coefficients in the equations with-
out the regional fixed effects. This result may be consistent with Bairoch’s
w2x argument that the density of agricultural population pushes people to
cities, particularly to large cities. But third, the coefficients of AGRrLAB
are not significant in the estimates with the regional fixed effects.
 URBANIZATION AND ECONOMIC DEVELOPMENT 31

Although the coefficients of LITR are smaller and less significant in the
equations with regional fixed effects, literacy rates are associated positively
with more primacy. This might be the ‘‘bright lights’’ effect differentially
attracting more educated people to the largest city. Or it might be, as
Henderson argues, that greater literacy is associated with more advanced
technologies. Thus, the more advanced technologies could be responsible
for the concentration in the largest city.
The coefficients of the dummy variables for 1970 and 1980 are consis-
tent across the eight equations. The coefficients for 1970 are not signifi-
cant, but the coefficients for 1980 are positive and significant, suggesting
that the primacy function shifted up from its 1960 level by 1980. So, if
primacy is a problem, the problem may be getting worse.
Table 4 shows the results of estimating these equations with country
fixed effects, time effects, and the various combinations of the continuous
variables. As in the previous section, the results are not as strong in these
equations as in the equations without country fixed effects. With the

TABLE 4
Determinants of Primacy: Country Fixed Effects

Variable Equation Ž19. Equation Ž20. Equation Ž21. Equation Ž22.

GDPrCAP a y0.05 0.02 y0.03

Žy0.77. Ž0.36. Žy0.53.
AGRrLAB y0.13 y0.98 y0.13
Žy1.61. Žy1.42. Žy1.62.
INDrLAB 0.05 0.04
Ž0.83. Ž0.61.
LITR 0.07** 0.08** 0.08** 0.08**
Ž2.02. Ž2.18. Ž2.37. Ž2.36.
EXPrGDP y0.04* y0.04* y0.04* y0.04*
Žy1.92. Žy1.91. Ž1.92. Žy1.86.
POP 0.16 0.09 0.14 0.21
Ž1.19. Ž0.85. Ž1.07. Ž1.61.
D70 y0.05 y0.02 y0.05 y0.06
Žy1.13. Žy0.65. Žy1.26. Žy1.48.
D80 y0.11 y0.06 y0.12 y0.14*
Žy1.47. Žy1.05. Žy1.56. Žy1.88.
R2 0.95 0.94 0.94 0.95
Observations 208 b 208 208 208
a
The number of parentheses in the t statistic. The number above it is the coefficient.
b
To increase the number of observations, we used an unbalanced panel for the regressions
reported in Table 3 and 4.
*Significance at 90%.
**Significance at 95%.
***Significance at 99%.
32 MOOMAW AND SHATTER

exception of the coefficient for the 1980 dummy in one of the four
equations, the only significant coefficients are for EXPrGDP and for
LITR. ŽThis specification embeds political effects, such as the effect of
DCAPCTY, in the country fixed effects..
Given the shortness of time-series data, we believe that the estimates in
Table 3 are the preferred ones. These results imply that economic influ-
ences, particularly ‘‘the extent of the market,’’ are powerful determinants
of primacy. These influences are captured by the coefficients on
EXPrGDP, GDPrCAP, and POP. Todaro’s observation that primacy is a
feature of low-income Ž‘‘developing’’. countries is supported by these
results. The results also suggest that this feature is an appropriate geo-
graphical adaption to economic conditions. Nevertheless, political factors
may be important, as suggested by the strength of the capital city variable
and perhaps by the regional fixed effects.
CONCLUSION
This study has used panel data to estimate the determinants of the
urban percentage, metropolitan concentration, and primacy. Many urban
economists see economic forces as the determinants of these different
dimensions of urbanization}in both developed and developing countries.
Other economists, including Bairoch and Todaro, believe that the urban
percentage and primacy in developing countries greatly exceed optimal
levels. Excessive urbanization results, they believe, from political forces,
inefficient migration, and dual industrial and rural sectors. Indeed, Bairoch
describes the situation as urbanization without development, meaning that
developing countries today are more urbanized than other countries were
at similar levels of development in the last century.
With respect to the percentage of the national population in urban areas
}the urban percentage}we find that countries with smaller shares of
their labor force in agriculture and larger shares in industry are more
urbanized. Countries with higher literacy rates are also more urbanized. In
addition, in all specifications save one}that with country fixed effects}we
find that urbanization increases with GDP per capita and with exports as a
proportion of GDP. When regional fixed effects are included, five regions
that contain less-developed countries have greater urbanization than ex-
pected compared with Western Europe, the base region. Three other
regions that contain less-developed countries, however, do not have signifi-
`
cantly different urban percentages vis-a-vis Europe. As Bairoch says,
urbanization is greater today, other things equal, then it was in the 19th
century. Given that the urban percentage responds as expected to develop-
ment and given the tremendous changes in communication and transporta-
tion technologies over the past century, we do not conclude, however, that
the urban percentage is in some sense too great.
 URBANIZATION AND ECONOMIC DEVELOPMENT 33

Metropolitan concentration also responds to economic forces. The most

striking result, perhaps, is the negative relationship between the export
share of GDP and metropolitan concentration. Furthermore, metropolitan
concentration decreases with agriculture’s share and increases with indus-
try’s share Žexcept in the specification with country fixed effects . of the
labor force. Literacy and production per capita do not have significant
effects. Regions consisting of both developed and less-developed countries
have positive regional effects.
Finally, with regard to primacy, there is evidence that political forces are
important. There is an apparent bias toward the capital city; furthermore,
the pattern of significant regional fixed effects suggests political influences.
Economics is also important. Primacy is greater, other things equal, the
smaller the economy, the lower GDP per capita, the smaller the share of
exports in GDP, and the lower the literacy rate. Primacy is associated with
a lower level of development, but this may be in response to economic
forces. The sectoral distribution of the labor force does not appear to
affect primacy. These results are subject to the qualification that the share
of exports in GDP and the literacy rate have the only significant coeffi-
cients in the equation with country fixed effects.
APPENDIX
We attempted to collect data for 90 countries. We started with the 121
countries listed in the ‘‘World Development Indicators’’ in the ‘‘1990
World Bank Report’’ World Bank w25x. After eliminating city-states }Hong
Kong and Singapore}and other countries for which we could not get
complete urbanization or economic data for 1960, 1970, and 1980, we
grouped them into 11 regions. The 90 countries used in the urban
percentage regressions are listed in Table A1. The table also shows the
countries used in the subsequent metropolitan concentration and urban
primacy equations.
The urban percentage and the percentage of the urban population in
the largest city}primacy}are from the ‘‘World Development Indicators’’
supplemented by ‘‘Prospects of World Urbanization, 1988’’’ w23x.
Metropolitan concentration, the percentage of the urbanized population in
cities of 100,000 or more, was computed from data in Davis and Golden w5x
and various issues of the U.N. ‘‘Demographic Yearbook’’ w22x. Population
in cities of 100,000 or more are from Davis and Golden w5x for 1960 and
1970. ŽThey projected the 1970 population data.. Population for cities of
100,000 or more for 1980 is from the ‘‘Demographic Yearbook.’’ We
identified metropolitan areas or agglomerations of 100,000 q population
and summed their populations to get metropolitan population. Unfortu-
nately, we were unable to construct this variable for the original 90
countries; it is available for only 68 countries. The urban percentages uses
34 MOOMAW AND SHATTER

TABLE A1

Metropolitan concentration Urban primacy

North America}DNAMER
Canada Yes Yes
United States Yes Yes
Central America}DCAMER
Costa Rica Yes Yes
Dominican Republic Yes Yes
El Salvador Yes Yes
Guatemala Yes Yes
Haiti Yes Yes
Honduras Yes Yes
Jamaica Yes Yes
Mexico Yes Yes
Nicaragua Yes Yes
Panama Yes Yes
Trinidad and Tobago Yes Yes
South America}DSAMER
Argentina Yes Yes
Bolivia Yes Yes
Brazil Yes Yes
Columbia Yes Yes
Chile Yes Yes
Ecuador Yes Yes
Paraguay Yes Yes
Peru Yes Yes
Uruguay Yes Yes
Uruguay Yes Yes
Venezuela No No
Middle East and North Africa}DMEAST
Algeria Yes Yes
Egypt Yes Yes
Iran Yes Yes
Iraq Yes Yes
Jordan Yes Yes
Kuwait No No
Morocco Yes Yes
N. Yemen No No
Saudi Arabia Yes Yes
Syria Yes Yes
Tunisia Yes Yes
Turkey Yes Yes
South Asia}DSASIA
Bangladesh No No
India Yes Yes
Nepal Yes No
Pakistan Yes Yes
Sri Lanka No No
 URBANIZATION AND ECONOMIC DEVELOPMENT 35

TABLE A1}Continued

Metropolitan concentration Urban primacy

Southeast Asia}DSEASIA
Indonesia Yes Yes
Japan Yes Yes
Malaysia No Yes
Papua New Guinea No No
Philippines Yes Yes
South Korea Yes Yes
Thailand Yes Yes
Central Africa}DCENAF
Chad No No
Ethiopia Yes Yes
Mali Yes Yes
Mauritania No No
Niger No No
Somalia No No
Sudan Yes Yes
West Africa}DWAF
Angola No Yes
Benin No No
Cameroon Yes Yes
Congo No No
Cote D’Ivoire Yes Yes
Ghana Yes Yes
Nigeria Yes Yes
Senegal Yes Yes
Sierra Leone Yes Yes
Togo No No
Southern Africa}DSAF
Kenya Yes Yes
Madagascar Yes Yes
Malawi No No
Mozambique Yes Yes
South Africa Yes Yes
Tanzania No Yes
Uganda Yes Yes
Zaire No No
Zambia No No
Zimbabwe No Yes
Oceania}DOCEANIA
Australia Yes Yes
New Zealand Yes Yes
Europe}DEUR
Austria Yes Yes
Belgium No Yes
Denmark Yes Yes
Finland Yes Yes
36 MOOMAW AND SHATTER

TABLE A1}Continued

Metropolitan concentration Urban primacy

Federal Republic of Germany Yes Yes

France Yes Yes
Greece
Ireland Yes Yes
Italy Yes Yes
Netherlands Yes Yes
Norway Yes Yes
Spain Yes Yes
Sweden Yes Yes
Switzerland Yes Yes
United Kingdom Yes Yes

each country’s definition of an urban place; the size criterion for designat-
ing an urban place differs from country to country. Regional and country
fixed effects will pick up some of this variation. In collecting data for the
metropolitan population and the population of the largest city, we used, to
the extent possible, metropolitan-area designations rather than the city
proper. GDP per capita is adjusted for purchasing power; we thank Robert
Barro w3x for providing it. Labor shares, literacy rates, population, and the
export share of GDP came from the ‘‘World Development Indicators’’ of
the World Bank. Information on central or unitary form of government
came from Mutlu w15x.

REFERENCES

1. A. F. Ades and E. L. Glaeser, ‘‘Trade and Circuses: Explaining Urban Giants,’’ National
Bureau of Economic Research, Working Paper Series, Working Paper 4715 Ž1994..
2. P. Bairoch, ‘‘Cities and Economic Development,’’ Univ. of Chicago Press, Chicago Ž1988..
3. R. J. Barro, Economic growth in a cross-section of countries, Quarterly Journal of
Economics, 106, 407]443 Ž1991..
4. J. Brueckner, Analyzing third world urbanization: A model with empirical evidence,
Economic De¨ elopment and Cultural Change, 38, 587]609 Ž1990..
5. K. Davis and H. H. Golden, ‘‘World Urbanization}1950]1970. Vol. I. Basic Data for
Cities, Countries, and Regions,’’ Population Monograph Series, No. 4, revised ed.,
University of California, Berkeley Ž1969..
6. R. L. Elizondo and P. Krugman, ‘‘Trade Policy and the Third World Metropolis,’’
National Bureau of Economic Research, Working Paper Series, Working Paper 4238
Ž1992..
7. A. W. Evans, The pure theory of city size in an industrial economy, Urban Studies, 9,
49]78 Ž1972..
 URBANIZATION AND ECONOMIC DEVELOPMENT 37

8. P. E. Graves and R. L. Sexton, Overurbanization and its relation to economic growth for
less developed countries, Economic Forum, 8, 95]100, in ‘‘Urban Development in the
Third World’’ ŽP. K. Ghosh, Ed.., Greenwood Press, Westport, CT Ž1979..
9. A. M. Hamer and J. F. Linn, Urbanization in the developing world: Patterns, issues and
policies, in ‘‘Handbook of Urban and Regional Economics’’ ŽE. S. Mills, Ed.., Vol. II,
North-Holland, Amsterdam Ž1978..
10. J. V. Henderson, ‘‘Urban Development: Theory, Fact, and Illusion,’’ Oxford Univ. Press,
Oxford Ž1988..
11. E. S. Mills, ‘‘Urban Economics,’’ 2nd ed., Scott, Foresman, Glenview, IL Ž1980..
12. E. S. Mills and C. Becker, ‘‘Studies in Indian Urban Development,’’ Oxford Univ. Press,
Oxford Ž1986..
13. E. S. Mills and B. W. Hamilton, ‘‘Urban Economics,’’ 5th ed., Harper Collins, New York
Ž1994..
14. E. W. Mills and B. N. Song, ‘‘Urbanization and Urban Problems,’’ Harvard Univ. Press,
Cambridge Ž1979..
15. S. Mutlu, Urban concentration and primacy revisited: An analysis and some policy
conclusions, Economic De¨ elopment and Cultural Change, 37, 611]639 Ž1989..
16. R. Moomaw and A. M. Shatter, Urbanization as a factor in economic growth: An
empirical study, Journal of Economics, 19, 1]6 Ž1993..
17. Organization for Economic Cooperation and Development, The search for a rural-urban
balance, in ‘‘Urban Development in the third World’’ ŽP. K. Ghosh, Ed.., Greenwood
Press, Westport, CT Ž1979..
18. K. T. Rosen and M. Resnick, The size distribution of cities: An examination of the Pareto
law and primacy, Journal of Urban Economics, 8, 165]186 Ž1980..
19. A. M. Shatter, ‘‘The Relationship Between Urbanization and Economic Development:
Empirical Evidence for Developed and Less-Developed Countries,’’ Doctoral Disser-
tation, Oklahoma State University, Stillwater, OK Ž1992..
20. G. J. Stigler, The division of labor is limited by the extent of the market, Journal of
Political Economy Ž1951.; in ‘‘The Organization of Industry’’ ŽG. J. Stigler, Ed.., Irwin,
Homewood, IL Ž1968..
21. M. P. Todaro, ‘‘Urbanization in Developing Nations,’’ Center for Policy Studies, Working
Paper Series; in ‘‘Urban Development in the Third World’’ ŽP. K. Ghosh, Ed..,
Greenwood Press, Westport, CT Ž1979..
22. United Nations, ‘‘Demographic Yearbook,’’ United Nations Department of Economic
and Social Affairs, New York Žvarious years..
23. United Nations, ‘‘Prospects of World Urbanization,’’ United Nations Department of
Economic and Social Affairs, New York Ž1987..
24. W. C. Wheaton and H. Shishido, Urban concentration, agglomeration economies, and the
level of economic development, Economic De¨ elopment and Cultural Change, 30,
17]30 Ž1981..
25. World Bank, ‘‘World Development Report,’’ Oxford Univ. Press, New York Žvarious
years..