Toward A General Theory of Hierarchy: Books, Bureaucrats, Basketball Tournaments,

and the Administrative Structure of the Nation-State

Author(s): Thomas H. Hammond
Source: Journal of Public Administration Research and Theory: J-PART , Jan., 1993,
Vol. 3, No. 1 (Jan., 1993), pp. 120-145
Published by: Oxford University Press on behalf of the Public Management Research
Association

Stable URL: https://www.jstor.org/stable/1181571

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

and Oxford University Press are collaborating with JSTOR to digitize, preserve and extend
access to Journal of Public Administration Research and Theory: J-PART

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward A General Theory Of
Hierarchy: Books, Bureaucrats,
Basketball Tournaments,
And The
Administrative Structure Of The
Nation-State'

Thomas H. Hammond
Michigan State University

ABSTRACT

Information processing and policymaking, in any kind of

tution, involve the making of comparisons. For information
ing, the comparisons are among bits of data. For policymakin
comparisons are among alternative options. Many of our inst
tions are hierarchies, and this essay examines, for a wide ran
institutions, how hierarchies structure the making of the com
sons.

It has long been understood that social institutions enable

'The author thanks Rick Albrecht,
Bob Bates, Mike Bratton, Kenneth
John Meier, Paul Thomas, Nick Van
de Walle, and B. Dan Wood for help-
ful comments on an earlier draft of
this paper, which was presented at the
annual meeting of the Midwest Politi-
cal Science Association in Chicago in
1990.
people to make choices and engage in activities that otherwise
would be impossible. It has also been understood that the
rules by which these institutions are governed are not neutral
in their impact: Changing the rules can be expected to change
the choices and activities. While early students of politics,
economics, and public administration in the United States were
undoubtedly motivated by these understandings, their ap-
proach to institutional studies in the 1920s and 1930s was long
on description and short on analysis: Institutional rules were
described and cataloged, but relatively little effort was made to
examine the nature of their impact.

J-PART, 3(1993):1:120-145 The behavioral revolution of the 1940s and 1950s was
directed against this overemphasis on description and catalog-
ing of institutional rules. More important in the newer view

120/Journal of Public Administration Research and Theory

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms

2In his recent autobiography, Mod-
els of My Life (1991, 270), Simon
reflected on his 1946 essay:
"Urwick never quite forgave me
this attack on his life work, but
Gulick was quite friendly in later
years. Presumably he made al-
lowance for the hubris of a young
man. Hubris, arrogance, or what-
ever, that article secured my in-
stant and permanent visibility in
public administration." For a
defense of Gulick (1937) against
Simon's attack see Hammond
(1 990).
3In my view, two of the most pro-
minent neoinstitutional approach-
es to the study of hierarchy-trans-
action cost analysis (Williamson
1975, 1985) and principal-agent
theory--have greater limits to their
applicability, especially in public
administration, than is sometimes
recognized. There is not space
here to summarize my doubts; see
Hammond (forthcoming) for a
discussion.
140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
Toward a General Theory of Hierarchy
were the attitudes and behavior of the individuals in the
institutions; inspiration for a science of politics was to be
found in the burgeoning fields of sociology (e.g., Lazarsfeld,
Berelson, and Gaudet 1944) and psychology (e.g., Simon 1944,
16; 1947, ch.5). The attitudes, norms, and decisionmaking
practices of individuals were to be described in detail and their
impact on the individuals' political behavior examined. With
this focus on the individuals, less attention could be paid to
the institutions which provided motive, meaning, and structure
to their behavior. Indeed, the earlier generation of studies in
public administration which had attempted, however imper-
fectly, to elucidate the impact of institutional rules (e.g., Gulick
and Urwick 1937) were subjected to withering and almost
scornful criticism (Simon 1946). In this fashion the behavioral
revolution swept away what we now call the "old institutional-
ism."'
The 1970s and 1980s brought about a revival of interest in
the institutional rules governing electoral, legislative, and
administrative systems. The main enterprise of the old institu-
tionalism--the description of institutional rules--is still seen as
essential. The "new institutionalism" goes beyond its predeces-
sor by exploring the impact of institutional rules on institution-
al outcomes (Ostrom 1986). So substantial is the literature in
the new institutionalism that it has already warranted review
essays in several different subfields; (see Moe 1984; Bendor
1988; Krehbiel 1988; Mueller 1989; and Dunleavy 1991 for a
sample). Despite this burgeoning literature, there exists no
general theory that illuminates the workings of a wide range
of social institutions. Thus, much remains to be done to char-
acterize the impact of institutional rules on institutional out-
comes.
In recent years it has been argued that "markets" and
"hierarchies" are two of the most fundamental ways of organiz-
ing transactions among individuals. The development of some
understanding of the properties of markets as social institu-
tions has been, of course, a major achievement in the discipline
of economics. In comparison, the study of hierarchy has been
neglected?
In this essay I offer some thoughts on how the institution-
al rules defining a hierarchy affect the nature of the policy-
making that occurs with the hierarchy. My argument is that a
hierarchy's rules affect outcomes because they affect the nature
of the comparisons made of information and options during the
policymaking process. An understanding of the nature of
121 /1-PART, January 1993
This content downloaded from
 Toward a General Theory of Hierarchy

these comparisons is central to an understanding of hierarchies

as institutions and may contribute more generally to the devel-
opment of a theory of institutions. This paper illustrates the
argument about comparisons in hierarchies in two ways. First,
it discusses the impact of hierarchical rules on tasks in which
information processing--the collection, classification, and under-
standing of information--is the central activity. Library cata-
logs, budget formats, and intelligence agencies are examined
from this perspective. Next, the paper discusses the impact of
hierarchical rules on tasks in which the making of choices --
listing options and deciding which is the best course of ac-
tion -- is the central activity. Basketball tournaments, bureau-
cratic structures, budget formats, legislatures, electoral systems,
and administrative systems for the nation-state are all exam-
ined from this perspective. The paper ends by discussing
what research is necessary to further advance this institutional-
ist enterprise.

CATEGORIES AND COMPARISONS IN HIERARCHIES

The people in an institution must routinely engage in two

kinds of activities: They must collect and process information
so as to learn what problems they need to address, and they
must decide what to do about these problems. Their insti-
tution's rules will affect these activities in two ways.

First, problem recognition by the decisionmakers in an

institution does not occur automatically; they must perceive
that there is a problem requiring attention. Moreover, the
problem that is perceived can usually be perceived in several
ways. The perception and definition of a problem depend on
the collection and assessment of some body of data. For most
institutions no single decisionmaker can possibly give due
consideration to every piece of relevant data. My argument is
that different institutional rules can create different sets of
information for the decisionmakers; these different sets of
information can in turn be expected to lead decisionmakers to
different definitions of the problem, or even to lack of recogni-
tion that there is a problem at all.

Second, whatever problem is perceived and defined,

addressing it requires consideration of possible options and a
choice from among them. The act of choice by the decision-
maker involves a comparison of the options that are available.
My argument is that institutional rules can be expected to
constrain the choice set available to a decisionmaker. Since
different sets of institutional rules will constrain the choice set
in different ways, the decisionmaker's comparisons and choices
will be biased in different ways.

122/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General Theory of Hierarchy

While it is possible to present a formal definition of hier-

archy, for my purposes here a simple, intuitive definition will
suffice: A hierarchy is a classification of objects, events, or
activities into sets of nested categories, that is, into categories
within categories within categories. If we begin with some
basic set of elements (e.g., a collection of objects, events, or
activities), we can classify these elements into two or more
subsets; the elements in each subset can be classified further
into two or more sub-subsets; and the elements in each of
these sub-subsets can be classified further into two or more
sub-sub-subsets. This process of classification can proceed
until we want to go no further or until each element from the
original set is in its own unique sub-sub-sub...set and no
further subdivisions are possible. In this fashion a hierarchy is
created. Whether we are focusing on processing information
or on making choices, these hierarchical rules will shape pro-
foundly the character of the comparisons that lie at the heart of
each process. That is, the hierarchy's categories shape the hier-
archy's comparisons.

CATEGORIES, COMPARISONS, AND THE ORGANIZATION

OF LEARNING

The particular configuration of a hierarchical institution

greatly influences what happens to information as it is collect-
ed, classified, and made available to decisionmakers. If two
institutions are identical in every respect (e.g., they have the
same tasks, the same personnel, and even have access to the
same raw data) but the two institutions' hierarchies differ, the
institutions may classify the data differently, and thus the top
level decisionmakers in each may learn different things from
the information. Indeed, once it is collected and classified by
an institution the information usually comes to have its own
unique hierarchical structure, and what can be most easily
learned from one structure of information may be different
from what can be most easily learned from another. Let us
consider three illustrations of this argument.

Library Catalogs and the Hierarchical Organization

of Knowledge

Few readers, I suspect, have ever thought about the impli-

cations of how books are organized in a library, beyond per-
haps puzzling over why some book was classified in one
subject matter category rather than in another; the book may
have seemed "out of place." Yet different cataloging systems
such as the Dewey Decimal and the Library of Congress sys-
tems may have a substantial impact on what we are able to
learn.

123/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms

'It should be apparent that there is
an essentially infinite number of
different criteria by which books
can be seen as "similar to" each
other. This is what accounts for
the occasional odd location of a
book in the library: The book
cataloger perceived some feature
of the book, which we did not
perceive, that made it appear
similar to some other set of books.
140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
Toward a General Theory of Hierarchy
The most elemental act involved in cataloging a library
book is to assert that "this book is like that book"; all the books
judged to be similar by some criteria are given a similar cata-
log number and placed close to each other on the shelves.
What results is a hierarchy: The books are organized in terms
of a complex set of nested classification categories.4 If we are
looking for books relevant to a current research project, how
the books are organized in a library will affect what books we
find and thus what we learn. There are three reasons this is
so.
First, if we have limited time or energy, the number of
books we gather for our project can be affected by the catalog-
ing system. If our topic is recognized as a distinct subject
matter category by the cataloging system, we may be able to
collect essentially all the relevant books in a short time from
the same section of the library. If our subject is not recognized
as a distinct category by the cataloging system, we may have
to spend the day chasing all over the library system to collect
the desired books. This problem is particularly acute for those
who are physically disabled and unable to move around easily.
Thus how much we are able to learn about our subject might
be substantially affected by the cataloging scheme the library
uses.
Second, when we have limited time to devote to our
project, the cataloging system can be expected to affect what
we learn serendipitously while trekking through the library.
When they search for a particular book, most academics also
spend considerable time browsing along the shelves in the
vicinity of the target book. Just by browsing in this fashion we
have all stumbled across some important book that we did not
know existed. What is interesting here is that the books which
are physically close to our target book--and thus within "brow-
sing range"-will be different under different cataloging sys-
tems; our attention will be drawn to a different set of books in
a Library of Congress library, for example, than in a Dewey
Decimal library. Thus what we learn serendipitously from our
browsing will be different under the two different systems.
Third, when we know a library has some item but we
cannot recall the title or author's name and we cannot recall
where it is !ocated in the library, it may be easier to retrieve
the item via one cataloging system than another. For example,
when I am working to define a research topic for a student, I
sometimes realize that I have a photocopy of a relevant article
but cannot remember the author's name. If my personal
collection of photocopies were organized alphabetically by
authors' names with no further subdivisions, I might have to
124/1-PART, January 1993
This content downloaded from

'The following discussion draws
on Hammond and Knott (1980,
13-15).
140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
Toward a General Theory of Hierarchy
look through the entire collection to find the article. Indeed,
when I have had to find something in one of my collection's
categories that has a large number of items, the search has
sometimes taken ten or twenty minutes. When an article has
been classified in a part of my collection which has a more
detailed set of nested categories, it has taken much less time to
find. This is because I can narrow the search by working my
way through the nested category labels, eliminating big chunks
of my collection from consideration, before I have to start
ruffling through a set of articles alphabetically. This means
that the search time required to find an item is affected by the
cataloging system.
In sum, library catalogs organize knowledge in hierar-
chies, and since different cataloging systems do this in differ-
ent ways, they make different bodies of knowledge accessible
to us. Since we revise our current understanding of a subject
by comparing it with new material, what we learn and how
much we learn will be a function of the library's hierarchical
cataloging system.
Concealing and Revealing Information Via Budget Formats5
If the budget of a large organization is to be comprehend-
ed by some decisionmaker, there has to be some aggregation
of expenditures. Otherwise the decisionmaker will be inundat-
ed by such a mass of detail that it cannot be digested. The
common solution is to give the budget a hierarchical structure
of nested categories of expenditures. Of course, agency expen-
ditures can be categorized in a variety of ways; for example,
the budget format can be based on "line items" (for objects
such as pencils, salaries, etc.), on "functions," on "programs,"
on geographical regions, and so forth. Whatever format is
adopted, what the decisionmaker learns about the budget will
be structured by the budget's major categories of expenditures.
Consider an agency whose work is done in local geo-
graphic districts. The agency's budget could be defined in two
ways. One way would be to have the major budget categories
correspond to the agency's geographic regions: Expenditures
are proposed for district 1, for district 2, and so forth. Within
each major geographic category might be subcategories for
various functional expenditures; for each district there would
be subcategories for administration, purchasing, training,
maintenance, and so forth. Alternatively, the major categories
might correspond to functional categories: Expenditures are
proposed for administration, purchasing, training, mainte-
nance, fieldwork, and so forth. Within each of these major
functional categories would be subcategories for the expendi-
125/J-PART, January 1993
This content downloaded from

6If the decisionmaker knows a
priori what questions he would
like to ask, he may be able to
force the agency to prepare a
version of the budget that better
suits these kinds of questions.
However, the decisionmaker
would somehow have to know
more at the outset about what
kinds of questions he should ask.
140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
Toward a General Theory of Hierarchy
tures which take place within each geographic district; there
would be a subcategory for administration in district 1, for
administration in district 2, and so forth.
The political importance of the budget format can be seen
by looking at the format from the viewpoint of the agency's
own director. Assume that the director has the authority to
determine the agency's budget format. Assume also that the
director suspects that if higher-level decisionmakers knew
about workload disparities among the geographic districts,
they might well close some district office. Assume finally that
the director does not want any district office closed. The
question is, What budget format would be politically most
convenient for the agency director to adopt?
The first budget format, based on geographic categories,
might lead the decisionmakers to ask questions about the
relative workloads of the district offices, (e.g., Should district
l's budget be increased and district 2's budget cut because of
changes in the workloads of the districts?). The decisionmaker
would be less likely to raise a question about what particular
functions this money would be used for. The agency director
would then have to supply the requested information, and
elimination of the vulnerable, low workload district might
follow. For this reason the director might want to adopt the
second budget format, based on functional categories, because
it would steer the decisionmaker away from sensitive ques-
tions involving comparisons of district offices. This format
would lead the decisionmaker to ask whether money should
be shifted back and forth across functional categories (e.g., Is
administration soaking up too much money? Do we need to
shift some of this money into maintenance?). It would be less
likely that the decisionmaker would raise the question: For
which district will this additional maintenance money be used?
In sum, dealing with highly aggregated budget categories
economizes on a decisionmaker's time and energy, but only at
the cost of having the nature of the decisionmaker's questions,
and thus what is learned, substantially structured by the
budget categories themselves.6
National Security and the Organization of
Intelligence Agencies
The director of an intelligence agency cannot pay attention
personally to all the problems that confront the organization.
He has limited time and energy, and he probably does not
have the specialized knowledge that would allow him even to
recognize the existence of particular kinds of problems. Hence,
126/J-PART, January 1993
This content downloaded from

7While the background setting for
this example is from Allison
(1971), the specifics are adapted
from an example in Hammond
(1986, 399401); see also Bendor
and Hammond (1992).
140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
Toward a General Theory of Hierarchy
he must rely on other people, especially his subordinates, to
keep him informed about the problems his organization faces.
It is often the agency's bottom level field officials who
initially collect raw data indicating a problem that needs the
director's attention. In even a modest-sized organization,
many field officials will be scanning the environment and
sending messages upward about potential problems. If the
director tries to read all these messages he will be over-
whelmed by their volume. Hence middle-level managers in
the agency will normally collate and aggregate the field offi-
cials' reports into summary documents which then are for-
warded to the director. So what the director sees is not the
raw data that the field officials see but a highly aggregated or
condensed interpretation by the middle managers of what the
field officials see. The director will base his choices about
what problems to attend to on these interpretations. March
and Simon (1958, 165) refer to this process as "uncertainty
absorption," and they note that "through the process of uncer-
tainty absorption, the recipient of a communication is severely
limited in his ability to judge its correctness. Although there
may be various tests of apparent validity, internal consistency,
and consistency with other communications, the recipient
must, by and large, repose his confidence in the editing pro-
cess that has taken place, and, if he accepts the communication
at all, accept it pretty much as it stands."
My argument is that the way in which subordinates are
clustered together into offices, sections, and divisions, and how
responsibilities for assessing particular kinds of data are as-
signed to these organizational units, can affect what kinds of
inferences are drawn from the data. That is, it can affect just
how uncertainty absorption takes place. To illustrate this
point, consider the following example about how the organiza-
tional structure might plausibly have affected the drawing of
inferences from raw data in the Cuban missile crisis of 1962.7
During the months that led up to the October crisis, the
Central Intelligence Agency (CIA) had primary responsibility
for ascertaining what, if anything, the Soviets were planning in
Cuba. The CIA gathered information from a variety of sourc-
es, such as from emigres who had recently left Cuba and from
intelligence sources inside Cuba and the Soviet Union. The
information could be classified into two categories: informa-
tion about diplomatic activities of the Soviets and the Cubans
and information about military activities of the Soviets and the
Cubans. Assume there were two basic bits of data in each
127/J-PART, January 1993
This content downloaded from
 Toward a General Theory of Hierarchy

category. The CIA could detect increased diplomatic activity

between the Soviets and the Cubans (call this bit x) or only
routine diplomatic activity (bit w). Similarly, the CIA could
detect increased military activity involving the Soviets and
Cubans (bit z) or only routine military activity (bit y).

For simplicity, assume the CIA had just four field agents:
Agents D1 and D2 had responsibility for detecting changes in
Soviet diplomatic activities regarding Cuba, while agents Ml
and M2 had responsibility for detecting changes in Soviet
military activities regarding Cuba. Assume agent D1 was
located in the American embassy in Moscow and had devel-
oped a broad network of contacts in the diplomatic community
there. Agent D2 resided in Havana and had developed a
network of contacts in that diplomatic community. Agent Ml
was located in the American embassy in Moscow and had
developed a network of contacts in the military attache com-
munity there, while agent M2 resided in Havana and had
developed a network of contacts in that military attache com-
munity.

These four specialists could be grouped in two ways, by

subject matter specialty (diplomatic versus military informa-
tion) or by geographic residence (Moscow versus Havana).
Whatever the grouping, each pair of specialists reported to one
of the two middle-level managers back at CIA headquarters.
Assume these two managers had different thresholds for
alarm, Manager 1 having a higher threshold than Manager 2;
he will raise an alarm only if both bits of information he re-
ceives indicate increased activity by the Soviets and Cuba.
Manager 2, on the other hand, will raise an alarm even if only
one of the two bits of information he receives indicates in-
creased activity. Assume also that the director of the CIA,
mindful of other demands on the president's time, is as cau-
tious as Manager 1; he will raise an alarm only if he receives
reports about increased activity from both managers.

Exhibit 1 shows two ways in which these four agents can

be grouped under the managers. In the functional structure
the diplomatic specialists, D1 and D2, report to Manager 1
(who becomes, in effect, a specialist in processing diplomatic
information), while the military specialists, M1 and M2, report
to Manager 2 (who becomes a specialist in processing military
information).

128/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General Theory of Hierarchy

Exhibit 1

How Hierarchy Affects Information Processing and the Drawing of Inferences

Functional Structure Geographic Structure

"Routine "Increased
Activity" Activity"
l l

"Routine "Increased "Increased "Increased

Activity" Activity" Activity" Activity"

1 2 1 2

"Increased "Routine "Increased "Increased "Increased "Increased "Routine "Increased

Diplomatic Diplomatic Military Military Diplomatic Military Diplomatic Military
Activity" Activity" Activity" Activity" Activity" Activity" Activity" Activity"

Di D2 Ml M2 D, M1 D2 M2
IlI I I I I I I
x w z z x z w z
i Raw Data - l l Raw Data -

Diplomatic Infonnation Manager l's Decision Rules

w = No increase in diplomatic activity If he sees x and w, he reports

(Routine) "Routine Activity"

x = Increased diplomatic activity If he sees x and z, he reports

(Increased) "Increased Activity'

Military Infonnation Manager 2's Decision Rules

y = No increase in military activity If he sees z and z, he reports

(Routine) "Increased Activity"

z = Increased military activity If he sees w and z, he reports

(Increased) "Increased Activity'

Director's Decision Rules

If he sees "Routine Activity" and "Increased Activity,"

he concludes "Routine Activity" and reports "Routine Activity" to the president.

If he sees "Increased Activity" and "Increased Activity,"

he concludes "Increased Activity" and reports "Increased Activity" to the president.

1 29/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General Theory of Hierarchy

In the geographical structure the two specialists in Moscow, D1

and M1, report to Manager I (who becomes, in effect, a special-
ist in processing information related to overall Soviet activities
in Moscow), while the two specialists in Havana report to
Manager 2 (who becomes, in effect, a specialist in processing
information related to overall Soviet activities in Havana).
However they are grouped, though, assume that the field
agents in each structure see the same bit of data: D, sees x
(increased diplomatic activity) in either case, D2 sees w (no
increased diplomatic activity), M1 sees z (increased military
activity), and M2 likewise sees z (increased military activity).

In the functional structure, Manager 1 receives reports of

x and w from D1 and D2 respectively. With conflicting mes-
sages about whether Soviet diplomatic activity has increased,
he concludes that nothing out of the ordinary is going on. As
a result, he sends a "Routine Activity" message to the director.
Manager 2 receives reports of z and z from M1 and M2 respec-
tively. These reports constitute a clear signal of increased
Soviet military activity, and so he sends a message of "Increased
Activity" to the director. The director is now receiving conflict-
ing messages--one says "Routine Activity" and one says "In-
creased Activity"--and so he concludes that nothing special is
afoot. As a result, he sends a "Routine Activity" message to the
president. In the geographical structure, on the other hand,
the same messages and decision rules would lead the director
to send a message of "Increased Activity" to the president.

Neither Allison (1971) nor more recent studies of the

Cuban missile crisis say much about the actual procedures
81t is interesting to note that one governing the aggregation and assessment of information in
of the actions taken by William
the months preceding the crisis; hence the reconstruction here
Casey on assuming control of the
CIA early in the Reagan adminis-
is necessarily fictitious. Allison did note that the CIA provid-
tration was to change the agency's ed a considerable amount of information about possible Soviet
structure from a functional form missile deployments, but the policymakers judged this infor-
to a geographic form. As mation to be ambiguous. It was not until construction of the
Taubman (1983, 35) notes, "Instead
missile sites was well advanced that the evidence was judged
of a system where subjects were
divided by discipline, with experts clear enough to warrant sending U-2s into Cuban airspace to
on the Soviet economy, for in- take photographs. Allison suggests that bureaucratic in-fight-
stance, separated from experts on ing among the State Department, the CIA, and the Air Force
Soviet politics, he restructured the delayed the first flight for several days. Another possible
operation along geographical
cause rests on the organizational structure of the CIA; a differ-
lines, putting together all the
specialists on a given country or ent structure might have led to a considerably earlier inference
region." It would be an interest- about what the Soviets were doing.8
ing matter to determine whether
this reorganization had significant
As empirical support for the plausibility of a structural
consequences for the character of
the intelligence produced.
hypothesis, it is noted that one of the factors that contributed

130/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General Theory of Hierarchy

to the successful Japanese attack on Pearl Harbor in December

1941 was that ambiguous information collected by various
intelligence units was processed by a structure that failed to
allow appropriate comparisons to be made (Wohlstetter 1962).
As a consequence, the U.S. military in Hawaii was led to
downgrade the possibility of a Japanese naval air strike.

In sum, we see that even when an organization's field

agents are seeing exactly the same raw data about the organi-
zation's environment and so are sending identical reports
upward, different kinds of comparisons among different sets of
information can be expected to take place in different struc-
tures; the different things that are thereby learned can have
important political consequences.

CATEGORIES, COMPARISONS, AND THE ORGANIZATION

OF CHOICE

It is obvious that whatever individuals learn via a hierar-

chy can affect the choices they make. But the hierarchy has
own independent influence on what choices are made. It does
this by shaping the set of options from which a decisionmak
can choose; changing the structure can change the available s
of options, thereby producing a different choice. What the
decisionmaker compares with what else is thus a function of
the nature of the hierarchy. This point can also be illustrate
by a wide range of examples.

Which Team Is the Best?: The NCAA Basketball Tourna-

ment as a Hierarchy

The players, coaches, and spectators involved in various

kinds of sports often want to know which teams are the best in
the game. To find out which is best, every team should be
required to play every other team in a round robin tourna-
ment. In this way, each team's strength can be assessed
against that of every other team. With even a modest number
of teams the required number of games would be prohibitive.
For the men's basketball tournament of the National Collegiate
Athletic Association (NCAA), for example, sixty-four teams
currently are selected to play, and sixty-four teams taken two
at a time means that 2016 games would have to be played. To
avoid this problem the NCAA holds the tournament under
single elimination rules--a single loss eliminates a team from
the tournament-and this single elimination tournament is
organized in a hierarchical format; in fact, when the who-
plays-whom listings are announced by the NCAA, newspapers
routinely show the pairings as a set of tree-like diagrams.
What is significant here, as every college basketball fan knows,

131 /J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General Theory of Hierarchy

is that these who-must-play-whom grouping decisions have a

substantial impact on the tournament outcome.

The sixty-four teams in the NCAA tournament are the

winners of the thirty basketball conferences around the country
(e.g., the Big Ten, the PAC Ten, the Big Eight) plus thirty-four
additional teams that are either independents (e.g., Notre
Dame) or conference members that did not happen to win
their conference championships. Once a team is admitted into
the tournament, the NCAA groups the teams into four "re-
gions"--East, Midwest, Southeast, West--with sixteen teams in
each region. These sixteen teams start the tournament by
playing a set of eight games; the eight winners are then paired
against each other in four games; the four winners are paired
against each other in two games; these two winners are then
paired against each other to determine the regional winner.
The winners of the four regional tournaments become the Final
Four. These four teams are then paired in the first round of
the finals (e.g., East versus West and Midwest versus South-
east), and the winners of the first round of the finals play for
the national championship. The winner of the final game is
the national champion, and the loser comes in second.

How teams are grouped into the regions, and whom they
have to play in each region, can affect which teams make it to
the Final Four. In fact, the teams in the Final Four are not
necessarily the four best teams overall. To provide a hypothet-
ical example, it would be possible to assign what are consid-
ered the thirty-two best teams (perhaps as measured by the
final pretournament Associated Press national rankings) to two
of the regions, and the thirty-two worst teams to the other two
regions. If the teams in the Final Four were grouped in such a
way that the winners of the two regions with the thirty-two
best teams play each other and the winners of the two regions
with the thirty-two worst teams play each other, it would
guarantee that the final game, for the overall championship, be
played between the team which is the best of the best and the
team which is the best of the worst. The loser of this final
game would then be considered the second best team in the
country, at least according to the tournament result. But it
should be apparent that the best of the worst team, which
ranked only thirty-third in the country but came in second in
the tournament, might well have been beaten by as many as
half of the other teams in the tournament.

To moderate the impact of grouping, the NCAA tries to

create balanced regional tournaments: Strong and weak teams
are assigned evenly to the four regions, and within each region
the stronger teams play the weaker teams first so that the

132/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms

'This is possible, of course, be-
cause basketball is not a "transi-
tive" activity: Even if team A can
beat team B, and team B can beat
team C, it does not necessarily
follow than team A can beat team
C.
'"The following arguments are
developed at greater length in
Hammond (1986; forthcoming) the
latter paper examines information
processing, policymaking, and
implementation in the hierarchies
of business firms.
140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
Toward a General Theory of Hierarchy
stronger teams are not knocked out early. Nonetheless, there
is no guarantee that the team that ends up in second place
overall is actually second best: Any team that is beaten in a
regional tournament by the winner of the overall tournament
might conceivably have been able to beat the other three teams
in the Final Four. Indeed, it is quite possible that one of the
losing teams in a regional tournament could have beaten the
overall NCAA winner.9 The only way to eliminate completely
the impact of grouping would be to eliminate grouping itself:
For anything short of a round robin tournament there is no
way to guarantee that the two teams in the final game are
actually the two best teams in the country.
Sports tournaments provide an illuminating metaphor for
hierarchical decisionmaking since they show in graphic and
easily understood terms how the who-is-grouped-with-whom
(or what-is-grouped-with-what) assignments in a hierarchy can
affect what options remain available at later stages of the
policymaking process.
Advice and Implementation in a Bureaucratic Hierarchyl0
Hierarchy is one of the defining features of bureaucracy,
and there are two ways that the nature of the hierarchy can
affect the choices that are made via the hierarchy. First, the
agency's director must decide what to do about some problem,
and it is not unusual for subordinates to compose the menu of
options from which the director will choose. How these sub-
ordinates are organized will affect what the menu is. Second,
whatever selection the director makes, it may not be self-im-
plementing. If the subordinates must implement the director's
choice, how these subordinates are organized in the structure
will affect what policy is actually implemented.
The Advisory Process
Every director depends to some degree on subordinates
for advice about how to respond to a problem. But the direc-
tor cannot personally consider every piece of advice sent
upward for every problem. Instead, it will be the middle-level
managers who will read or listen to the advice, aggregate it in
some fashion, and send it upward. What advice the director
actually receives will consist of the bundles of advice forward-
ed by his immediate subordinates. Since the middle managers
will play the key aggregation role, what subordinates each
manager has can affect what recommendation he sends up-
ward. Thus the simple matter of who is grouped with whom
in this organizational structure, and what their official respon-
sibilities are, should be expected to influence what options the
director has available for choice.
133/J-PART, January 1993
This content downloaded from
 Toward a General Theory of Hierarchy

Consider the following example, illustrated in Exhibit 2.

There are a director, middle-level managers A and B, four field
positions (C, D, E, and F), and four field officials (1, 2, 3, and
4). Eight options are available: s, t, u, v, w, x, y, and z. The
individuals' preferences over these options (or judgments
about them) are as indicated in the exhibit. Each of the field
positions has been assigned jurisdiction over the choice be-
tween two options: Field position C has jurisdiction over the
choice from the subset {s,t), field position D has the subset
{u,v), E has the subset {w,x), and F has the subset {y,z). The
exhibit shows the officials' locations in three possible struc-
tures.

Assume that each field official recommends to his manag-

er the more preferred option within his own jurisdiction, and
assume that the manager's choice is restricted to the set of
options supplied by his immediate subordinates. In Structure
1 in Exhibit 2, for example, field official 1 in field position C
has responsibility over the choice from the set {s,t); since field
official 1 prefers s to t, he recommends s to his manager, A.
Field official 2 in position D has responsibility for the choice
from {u,v) and recommends v to manager A. Field official 3 in
position E has responsibility for the choice from {w,x) and
recommends w, while field official 4 in position F has respon-
sibility for the choice from {y,z) and recommends y. Thus
manager A receives s and v, and since he prefers s to v he
recommends s to the director. Manager B receives w and y
and, preferring y, he recommends y to the director. The
director thus receives s and y and, preferring s to y, selects s
as his choice.

Structure 2 differs from Structure 1 only in that positions

D and E, with associated jurisdictions and persons, switch
places. This minor switch completely changes the subset of
options reaching the director. Manager A now receives {s,w)
instead of {s,v); preferring w to s, he forwards w. Manager B
now receives {v,y) instead of {w,y); preferring v to y, he for-
wards v. The set of options now reaching the director thus
changes from {s,y) to {w,v), with the result that w is now the
outcome. If we hold constant the field official positions and
jurisdictions and merely switch the persons who hold these
positions, we can get a similar change in outcomes. For exam-
ple, in Structure 1 switch person 2 to field position E and
person 3 to field position D; this becomes Structure 3. The
director now receives (u,x) instead of {s,y). The final choice
accordingly switches from s to u.

134/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General Theory of Hierarchy
Exhibit 2

How Hierarchy Affects the Advisory and Implementation Processes

Field Officials' Managers' Director's

Preferences Preferences Preferences

1 2 3 4 A B

u z y x u v u

x v t u x x s
v u z s w y w

s y u w s t y
t t v y t u z
w x s z v w v

y s w t z s x
z w x v y z t

Structure 1 Structure 2 Structure 3

Outcomes: s w u

Director * * *

Advice: s y w v u x

/ \/ \/ \
Managers A B A B A B

Advice: s v w y s w v y s u x y

I \ I I \ I \I \ I \
Positions C D E F C E D F C DE F
jurisdictions: Is,t) (u,v) (w,x) (y,z) Is,t) (w,x) (u,v) (y,z) Is,t) (u,v) (w,x) (y,z)
Field Officials: 1 2 3 4 1 2 3 4 1 2 3 4

135/I-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms

"Note in addition that the juris-
dictions--a kind of organizational
"property right"-can lead to sub-
optimal choices. In Exhibit 2,
everyone prefers u to s and w, yet
Structures 1 and 2 produce s and
w respectively. This happens
because the jurisdictions grant
some official the right to eliminate
u in favor of another option
(which does not become the final
choice). See Hammond and Miller
(1985) for a discussion of the im-
pact of jursdictions on organiza-
tional policymaking.
140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
Toward a General Theory of Hierarchy
In these cases, then, changing the structure completely
alters the set of options reaching the director. In fact, one can
prove mathematically that it is impossible to design a structure
that does not affect the outcomes (see Hammond and Thomas
1989). Thus changing the structure can generally be expected to
change the outcomes.11
The Implementation Process
Even when the director has chosen some policy, this
choice must still be implemented. There can of course be a
substantial difference between the policy adopted by the direc-
tor and the policy actually implemented in the field. Some of
the difference will be due to the fact that most policies when
adopted will still entail substantial ambiguities; the director
simply will not have thought through, and specified in writing
for subordinates to read, precisely what should be done in all
possible contingencies that may arise during implementation.
So when subordinates are told to implement the director's
policy, it sometimes will not be clear to them precisely what
they are supposed to do.
If the policy will have to be implemented by a number of
interdependent subordinates, they may not see eye-to-eye on
what should be done. If agreement is needed before the policy
can be implemented, and if that agreement is elusive, they
may refer their disagreement upward to some mutual superior
for settlement. This mutual superior would then consider the
issue, make some decision, and send his decision downward
for implementation.
In some cases, this mutual superior may be the immediate
superior of the subordinates in conflict. In other cases, the
mutual superior will be several levels higher in the structure
and may in fact be the director of the overall organization. So
how subordinates are grouped in the hierarchy can be expect-
ed to affect how high in the organization conflict over imple-
mentation can be expected to rise; some structures will resolve
conflict at a low level in the hierarchy, while other structures
will route it all the way to the top. Since different structures
can mean that a given conflict will be dealt with by different
higher level managers who may have different views and
preferences, this means that the policy ultimately implemented
will itself be a function of the organizational hierarchy.
For example, consider the structures in Exhibit 2. Assume
that the director has chosen option s and that s must now be
implemented. If the individuals in field positions C and D are
136/J-PART, January 1993
This content downloaded from
 Toward a General Theory of Hierarchy

assigned responsibility to implement s, and if they cannot

agree on precisely what s means, they may have to refer their
conflict upward for resolution. In Structure 1, their lowest
mutual superior is Manager A, who would have the opportu-
nity to resolve the dispute. In Structure 2, their lowest mutual
superior is the overall director, who might have to resolve the
dispute. Thus who resolves what conflicts among which field
officials will be a function of the hierarchy.12

Budget Formats and Budgetary Choices

I have already described how an agency's budget format

can influence what the decisionmaker learns. It should also be
apparent that the budget format can affect the eventual choices
as well. The reason is that the budget format constrains the
decisionmaker's options. Within budgetary limits, the choices
are to give this major expenditure category more money or to
give that major expenditure category more money. In the
example used earlier, the geographic budget format provides
the options that involve changing the amount of money given
to district 1, to district 2, and so forth. For the functional
budget format, the options that are provided involve changing
the amount of money given to various functional activities.

A top-level decisionmaker generally does have the author-

ity to make choices in ways that cut across the major budget-
ary categories. But to do this intelligently, the decisionmaker
must make himself much better informed about the meaning
of expenditures within each major category in the agency's
budget. Since it would take some of his scarce time and
energy to inform himself about these issues, the budget format
remains likely to constrain his choice set.

The Legislature as a Policymaking Hierarchy

It is commonplace to think of legislatures in the United

States as "horizontal" institutions which are governed primarily
by majority rule. A moment's reflection, though, will suggest
an alternative description: These legislatures are hierarchies
"2For each manager, the structure
since they consist not only of a floor but also of nested sets of
determines which conflicts come
to him for resolution and which committees and subcommittees. Since these committees and
do not. Because a manager learns subcommittees are granted substantial power to formulate the
about different kinds of issues in proposals to be considered at the next higher level and to
part by resolving conflicts among reject changes on matters lying within their jurisdictions, the
subordinates, it follows that the
relationships between a subcommittee and a committee, and
organizational structure will influ-
ence what he learns. Thus policy- between a committee and the floor, take on some hierarchical
making ultimately feeds back into aspects.
information processing and learn-
ing.

137/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General T)heory of Hierarchy

As in a bureaucratic hierarchy, who is assigned to the

subunits (i.e., who gains seats on which committees and sub-
committees) and what the subunits' jurisdictions are can affect
legislative decisions. For example, who is grouped with whom
on which committee can obviously affect what decisions the
committee makes: A substantial body of scholarship on the
U.S. House of Representatives has focused on the committee
assignment process (e.g., Shepsle 1978) and its implications for
policymaking. Similarly, changing a committee's or subcom-
mittee's jurisdiction can also be expected to affect outcomes.
For example, in the spatial models of committee-floor relation-
ships which assume that a committee is granted jurisdiction
over a single dimension of a multidimensional policy space
(Shepsle 1979; Krehbiel 1987), it is possible to demonstrate
(Humes forthcoming) that redefining a committee's jurisdiction
will normally lead to changes in policy outcomes.

In sum, there is a hierarchical relationship between the

chamber as a whole, its committees, and their subcommittees.
As with outcomes from bureaucratic hierarchies, legislative
outcomes are affected by the "who-is-grouped-with-whom"'
matter of committee assignments and by how each committee's
jurisdiction is defined.

Hierarchies of Representation, Gerrymandering, and

Peaceful Accommodation in Multi-Ethnic Societies

A representative system in a democracy is a simple two-

level hierarchy: Voters at one level elect representatives to a
legislature at the next level, and it is the representatives who
make public policy. If we include the legislature's leadership
positions, there are three levels to the hierarchy. For the
British Parliament, for example, the voters elect members of
Parliament (MPs), and the MPs elect the prime minister. Since
most legislatures in the United States are constitutionally
authorized to send proposals to the chief executive on a take-
it-or-leave-it basis, the chief executive can be seen as another
level. In fact, prior to ratification of the Seventeenth Amend-
ment to the Constitution in 1913, policymaking involving the
U.S. Senate had the potential for five levels: voters, who
elected state legislators, who elected U.S. senators, who in
turn elected party leaders, who had considerable influence
over what bills were sent to the president. The Supreme Court
can even be interpreted as constituting yet another level, in
that it can review statutes passed by Congress and signed into
law by the president.

138/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General Theory of Hierarchy

Every representative system is based on rules for convert-

ing citizens' votes into seats in the legislature (Rae 1971;
Taagepera and Shugart 1989). In the United States the most
common rule involves the creation of geographically based
districts, and each district is allocated one seat in the legisla-
ture; any registered voter in the district can cast a ballot for a
candidate for that district's seat in the legislature, and the
candidate with the most votes wins. A geographical district is
a particular grouping of voters who live in some geographical
space. Thus to create a representative system of this kind
requires that decisions be made about who should be grouped
with whom to constitute each district's electorate. Since there
is no agreed-upon set of rules specifying how geographic
districts should be created, the boundaries are often drawn for
political purposes. This is, of course, gerrymandering: the at-
tempt to influence what kinds of people are elected by making
politically motivated decisions about which voters should be
grouped with which other voters to form which legislative
districts.

For a simple example, assume there is a state with two

legislative districts and three kinds of voters-Democrats, Inde-
pendents, and Republicans. Assume that the Independents are
less liberal than Democrats but more liberal than Republicans.
Assume also that these three kinds of voters reside in different
geographic regions of the state, as indicated by the dashed
lines in Exhibit 3. If the boundary line between the two dis-
tricts is drawn horizontally, District 1--with all the Democrats
plus half the Independents--will elect a traditionally Democrat-
ic legislator; District 2--with all the Republicans plus the re-
maining Independents--will elect a traditionally Republican
legislator. However, if the district boundary line is drawn
vertically, each district will consist of half the Democrats, a
roughly equal number of Republicans, and half the Indepen-
dents. With the Independents being the swing voters in each
of these "vertical" districts, we might expect that a more cen-
trist pair of legislators will be elected.

We might also expect that the candidates in the vertical

districts will portray themselves differently, and use different
kinds of messages, than candidates in the party-dominated
horizontal districts. In a legislature influenced by these Inde-
pendent voters, we might expect that the kinds of issues that
the legislators stress, and the kinds of legislative programs
they adopt, will differ from the kinds of issues and programs
advanced by either traditionally Democratic legislators or
traditionally Republican legislators. Who is grouped with
whom in the districts thus has implications not only for the
legislature's composition but also for the general character of
political debate within it.

139/l-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General Theory of Hierarchy

Exhibit 3

How District Boundaries Can Influence Electorl Outcomes

Democrats live here

Independents live here

Republicans live here

"Horizontal Districts" "Vertical Districts"

Democrats Democrats Democrats

Independents - - - - - - - - - - - - - - - - - - - - - -

Independents Independents Independents

Republicans Republicans Republicans

140/1-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General Theory of Hierarchy

While one might view gerrymandering as causing essen-

tially minor, if vexing, imperfections in a representative sys-
tem, the literature on the design of constitutional rules for
governing multiethnic societies suggests that such rules can
have major consequences for the degree of peaceful accommo-
dation among the ethnic groups. For example, in his study of
the prospects for democracy and ethnic accommodation in
South Africa, Horowitz (1991, ch.5) argues that different ways
of grouping people into electoral districts can create different
kinds of incentives for political leaders when they consider
emphasizing ethnic differences in their electoral campaigns. In
effect, changing the boundaries influences what kinds of issues
can be expected to reach the national political agenda via
legislative elections. Thus creative boundary drawing may be
able to keep some inflammatory issues off the national political
agenda by making it unlikely that candidates espousing these
views can be elected.13

Drawing geographic boundaries in a political system may

affect more than just who gains office and what issues they use
to get there. Some studies of the politics of ethnicity in Africa
have argued that the very existence of ethnic self-identification
has sometimes stemmed from the creation of political bound-
aries which aggregated people into groups in ways different
from before. That is, the act of grouping induced people to
define themselves differently (see Colson 1969; Young 1976;
Ranger 1985; Horowitz 1985, 1991; Bates 1986; Kasfir 1986;
Herbst 1989, 679-80). Thus while students of comparative
politics have long argued that a country's political culture
affects the choice of constitutional rules, it may also be the case
that the constitutional rules, which in part involve the act of
grouping people in various ways, can affect not only political
outcomes, in the short run, but also the country's political
culture itself, in the long run.

Federalism and the Administrative Structure of the

Nation-State

A country's electoral rules can go only so far toward

inducing peaceful accommodation in a multiethnic society.
"There is, of course, a vast litera- This is true in part because the electoral rules determine only
ture on the workings and impact
who will occupy the government's major policymaking posi-
of different kinds of electoral
rules, but Horowitz (1991, ch.5)
tions; the electoral rules do not say who will actually imple-
argues that much of it has been ment the policies, nor do they specify the administrative sys-
insufficiently normative. Thus it tem through which this implementation will take place.
is difficult, unfortunately, to de-
rive lessons for constitutional
A national administrative system will generally consist of
engineering from it.
a modest number of large departments or ministries, each with

141/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
 Toward a General Theory of Hierarchy

jurisdiction over some particular set of issues, and each subdi-

vided into a nested set of agencies, bureaus, and offices. In the
interests of peaceful accommodation in a society, the question
is, what should this administrative system look like?

The logic of the design of a single bureaucracy's hierarchi-

cal structure, as discussed earlier, applies to the design of a
national administrative system as well. Looking at policy
implementation, for example, we might ask what kinds of
conflicts might occur in the national administrative system and
what would happen to these conflicts. If we want particular
kinds of ethnic conflicts to be worked out at lower levels of the
administrative system to minimize the number of issues that
national politicians (electoral or administrative) can seize to
inflame interethnic relations, it might be possible to design the
administrative hierarchy's jurisdictions and grouping relation-
ships so that some issues are worked out at lower administra-
tive levels. The conflicts which do rise to the top would, by
design, not be as susceptible to interethnic definitions.

Pushing this logic further, an administrative system can

be designed so that most issues have to be settled at the re-
gional or local level. For example, Horowitz (1985, ch.15)
argues that federalism--the assignment of major administrative
responsibility to subnational political units--served both to
exacerbate ethnic tensions in Nigeria and at a later date to
ameliorate them. Following independence, the First Nigerian
Republic (1960-1966) consisted of three main regions, each
corresponding to a single ethnic group. National politics
thereby became interethnic politics; one consequence of this
was the civil war involving Biafra. Starting in 1967 a much
more extensive federal system was created, and the boundaries
of the nineteen resulting administrative units were designed in
part to defuse ethnic tensions. While the Second Republic was
short-lived, Horowitz argues that the new federal system did
moderate the intensity of interethnic group conflicts, in part by
inducing some conflicts within ethnic groups. As Horowitz
(1985, 612-13) summarized his arguments about this federal
framework in Nigeria:

First, the proliferation of states dispersed some of the

conflict into more parochial forums. Second, the new states
provided arenas in which intraethnic conflict might also occur.
Third, a result of this was to enhance the position of some
political parties at the expense of others, especially in the
North, paving the way for greater interethnic cooperation in
the all-Nigerian arena. Fourth, as the new states fought to
advance their interests, a few nonethnic issues and actors were
also introduced. And fifth, the separate state bureaucracies
provided career opportunities for groups not well represented
in the federal civil service.

142/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms

"4Note, moreover, that this paper
has not exhausted the variety of
hierarchies that might be dis-
cussed in these terms. For exam-
ple, the hierarchical "elimination
by aspects" model of individual
decisionmaking by Tversky (1972)
and Tversky and Sattath (1979) is
remarkably similar to the ap-
proach to hierarchies advanced
here.
140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
Toward a General Theory of Hierarchy
The crucial issue thus involves the question of what the
basic political units of the system should be--their number,
size, boundaries, and responsibilities. The logic of hierarchical
design developed here may help us to understand these consti-
tutional and administrative design issues.
CONCLUSION
Every institution processes information so as to perceive
and define problems, and every institution's decisionmakers
choose among the available options to address these proble
The act of comparison lies at the heart of these two activitie
problem perception and definition involve the comparison of
some pieces of information with others, while choice involv
the comparison of one option with another--and the argume
in this essay has been that institutional rules creating hierar
chies have a substantial impact on the nature of these compa
sons.
The argument, however, has been presented in very
general terms. To explore this argument in greater depth will
require the development of a formal conceptual apparatus for
modeling the behavior of hierarchical institutions. This is no
small task, especially since in some of the hierarchies examined
here what flows upward is information or options (as in bud-
gets, bureaucracies, and national administrative systems) while
in other hierarchies what flows upward is teams or people (as
in basketball tournaments and electoral systems). It remains to
be seen whether the argument about comparisons in any kind
of hierarchy can be captured in one general model.'4
Even if a useful formalization of the workings of hierar-
chical institutions can be developed, this work has just begun.
To use this general approach to explain the behavior of various
kinds of hierarchical institutions will require considerable
substantive research. The understanding of several institutions
discussed here is characterized more by breadth than by depth,
and it is clear that collaboration with a wide variety of subject-
matter specialists will ultimately be necessary.
As part of the research agenda of the new institutional-
ism, understanding the properties and behavior of hierarchies
will require some demanding work. Yet only a research agen-
da such as this, aimed at developing a general theory of a
particular class of institutions, can demonstrate whether the
reach of the new institutionalism will prove commensurate
with its grasp.
143/J-PART, January 1993
This content downloaded from
 Toward a General Theory of Hierarchy
Allison, Graham T.
1971 Essence of Decision: Explaining
the Cuban Missile Crisis. Bos-
ton: Little, Brown.
Bates, Robert H.
1986 "Modernization, Ethnic Com-
petition, and the Rationality
of Politics in Contemporary
Africa." In Marion E. Doro
and Newell M. Stultz, eds.
Governing in Black Africa, rev.
ed. New York: Holmes&
Meier.
Bendor, Jonathan.
1988 "Review Article: Formal Mo-
dels of Bureaucracy." British
Journal of Political Science
18:353-95.
Bendor, Jonathan, and Hammond,
Thomas H.
1992 "Rethinking Allison's Models."
American Political Science Re-
view 86:301-22.
Colson, Elizabeth.
1969 "African Society at the Time
of the Scramble." In L. H.
Gann and Peter Duignan, eds.
Colonialism in Africa,
1870-1960, vol. 1. Cambridge:
Cambridge University Press,
27-65.
Dunleavy, Patrick.
1991 Democracy, Bureaucracy and
Public Choice: Economic Expla-
nations in Political Science.
New York: Prentice-Hall.
Gulick, Luther.
1937 "Notes on the Theory of Orga-
nization." In Luther Gulick
and Lyndall Urwick, eds.
Papers on the Science of Admin-
istration. New York: Institute
of Public Administration,
Columbia University, 1-45.
Gulick, Luther, and Lyndall Urwick,
eds.
1937 Papers on the Science of Admin-
istration. New York: Institute
of Public Administration,
Columbia University.
144/J-PART, January 1993
140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms
REFERENCES
Hammond, Thomas H.
1986 "Agenda Control, Organiza-
tional Structure, and Bureau-
cratic Politics." American
Journal of Political Science
30:379-420.
1990 "In Defense of Luther Gulick's
'Notes on the Theory of Orga-
nization."' Public Administra-
tion 68:143-73.
"Strategy, Structure, and the
Agenda of the Firm." In Rich-
ard Rumelt, Dan E. Schendel,
and David J. Teece, eds. Fun-
damental Issues in Strategic
Management: A Research Agen-
da for the 1990s. Boston:
Harvard Business School
Press. Forthcoming.
Hammond, Thomas H., and Knott,
Jack H.
1980 A Zero-Based Look at Zero-Base
Budgeting. New Brunswick,
N.J.: Transaction Books.
Hammond, Thomas H., and Miller,
Gary J.
1985 "A Social Choice Perspective
on Authority and Expertise in
Bureaucracy." American Jour-
nal of Political Science
29:611-38.
Hammond, Thomas H., and Thom-
as, Paul A.
1989 "The Impossibility of a Neu-
tral Hierarchy." Journal of
Law, Economics, and Organiza-
tion 5: 155-84.
Herbst, Jeffrey.
1989 "The Creation and Mainte-
nance of National Boundaries
in Africa." International Orga-
nization 43:673-92.
Horowitz, Donald L.
1985 Ethnic Groups in Conflict.
Berkeley: University of Cali-
fornia Press.
1991 A Democratic South Africa?:
Constitutional Engineering in a
Divided Society. Berkeley:
University of California Press.
This content downloaded from
Humes, Brian D.
"Majority Rule Outcomes and
the Choice of Germaneness
Rules." Public Choice. Forth-
coming.
Kasfir, Nelson.
1986 "Explaining Ethnic Political
Participation." In Marion E.
Doro and Newell M. Stultz,
eds. Coverning in Black Africa,
re. ed. New York: Holmes &
Meier, 5-16.
Krehbiel, Keith.
1987 "Sophisticated Committees
and Structure-Induced Equi-
libria in Congress." In
Mathew D. McCubbins and
Terry Sullivan, eds. Congress:
Structure and Policy. New
York: Cambridge University
Press, 376-402.
1988 "Spatial Models of Legislative
Choice." Legislative Studies
Quarterly 13:259-319.
Lazarsfeld, Paul F.; Berelson, Ber-
nard R.; and Gaudet, Hazel.
1944 The People's Choice. New
York: Duell, Sloan, and Pierce.
March, James G., and Simon, Her-
bert A.
1958 Organizations. New York:
Wiley.
Moe, Terry M.
1984 "The New Economics of Orga-
nization." American Journal of
Political Science 28:739-77.
Mueller, Dennis C.
1989 Public Choice II. Cambridge,
Eng.: Cambridge University
Press.
Ostrom, Elinor.
1986 "An Agenda for the Study of
Institutions." Public Choice
48:13-25.
Rae, Douglas
1971 The Political Consequences of
Electoral Laws, rev. ed. New
Haven, Conn.: Yale Universi-
ty Press.
 Toward a General Theory of Hierarchy

Ranger, Terrence Taagepera, Rein, and Shugart, Williamson, Oliver E.

1985 The Invention of Tribalism in Matthew Soberg. 1975 Markets and Hierarchies. New
Zimbabwe. Gweru [Zimba- 1989 Seats and Votes:The Effects and York: Free Press.
bwel: Mambo Press. Determinants of Electoral Sys- 1985 The Economic Institutions of
tems. New Haven, Conn.: Capitalism. New York: Free
Shepsle, Kenneth A. Yale University Press. Press.
1978 The Giant Jigsaw Puzzle: Dem-
ocratic Committee Assignments Taubman, Philip. Wohlstetter, Roberta.
in the Modern House. Chicago: 1983 "Casey and His C.I.A. on the 1962 Pearl Harbor: Warning and
University of Chicago Press. Rebound." New York Times Decision. Stanford, Calif.:
1979 "Institutional Arrangements Magazine, 16 Jan. Stanford University Press.
and Equilibrium in Multidi-
mensional Voting Models." Tversky, Amos. Young, Crawford.
American Journal of Political 1972 "Elimination by Aspects: A 1976 The Politics of Cultural Plural-
Science 23:27-59. Theory of Choice." Psychologi- ism. Madison: University of
cal Review 79:281-99. Wisconsin Press.
Simon, Herbert A.
1944 "Decision-Making and Ad- Tversky, Amos, and Sattath,
ministrative Organization." Shmuel.
Public Administration Review 1979 "Preference Trees." Psychologi-
4:16-30. cal Review 86: 542-73.
1946 "The Proverbs of Administra-
tion." Public Administration
Review 6:53-67.
1947 Administrative Behavior: A
Study in the Decision-Making
Process in Administrative Orga-
nization. New York: Free
Press.
1991 Models of My Life. New York:
Basic Books.

145/J-PART, January 1993

This content downloaded from

140.122.127.100 on Sun, 14 Aug 2022 01:58:05 UTC
All use subject to https://about.jstor.org/terms