Juflufiiflflmvilzhs.‘ 33;.

(an!

.3 Wm}...

. 113)
ark} i
x l.

.. i . .
3:. I. . .i
.r.§,....r§9rmnwuu.
his .1 Funfinmflnx.
“mum
.... .. .
M. r. $5

.12.
3.. I!
:1 a.

a...
4 3.5.3.. u.

I). .
him”.
5.

5.4% .

“hit-l
‘13:}:
9;}... v.

.10.!..}

V31»

3!.
.Efit‘xbut.
9 . ..
ifi:....2x:l' ..
in“ in“,
~ I“ V

‘2‘.

513‘. l":-
- \
‘33" 1‘.

i
M

 

. .'.1.9V 1.
.(taliq i 4
e?.y\l ‘
t» 111?}; , ‘ ‘
W!
I! a-

 

vi....

4 ill»):

 

 

 

 

 

 

 

 

H

MICHIGAN STATE UflflfifiSITY LIBRARI 8

<3; i 111111111111111 1111111

3 1293 01421 5002 f r

11

This is to certify that the

dissertation entitled

The Politics and Process of International
Trade: Domestic and International Factors
Affecting Trade Policy Matters

presented by

Sherry Bennett Quinones

has been accepted towards fulfillment
of the requirements for

Ph.D. degreein Political Science

mam/4

Major professor

 

 

Dr. Gretchen Hower

November 17, 1995
Date

 

MS U i: an Affirmative Action/Equal Opportunity Institution 0-12771

 

d—l'fl

 

 

LIBRARY
Mlchlgan State
Unlverslty

 

 

 

PLACE ll RETURN BOX to remove We checkout from your record.

TO AVOID FINES return on or before «to due.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU leAn Nfirmetlve Action/E“ Opportunlty lnetttuion

1

ABSTRACT

THE POLITICS AND PROCESS OF INTERNATIONAL TRADE:
DOMESTIC AND INTERNATIONAL FACTORS
AFFECTING TRADE POLICY MATTERS

By

Sherry Bennett Quifiones

This research consists of three separate studies, which share a common
substantive topic on international trade. Thus, although each of the models presented
share a substantive focus, they vary considerably in theoretical orientation, scope, and
method. As such, they comprise distinct chapters. The first chapter, Congressional
Responsiveness to Constituency Demands: The Political Economy of the North
American Free Trade Agreement, is a study that examines domestic political
processes influencing trade policy in the United States. Specifically a theoretical
argument is developed that contends protectionist (import-competing) and anti-
protectionist (export-competing) interests will demand responsiveness from their
representatives on trade policy matters. To test the theoretical model, the vote
outcome associated with NAFTA is analyzed. Using political and economic data
collected from 104th congressional districts, a set of propositions are tested with

ordinary least squares and probit estimations. The overall results obtained from the

analysis of the model, suggest that the economic effects of the vote tend to dominate
over traditional political issues.

The second chapter, Modeling the Dynamics of Power, Trade and Conflict:
Fluctuations in the Economy, Alternating Distributions of Power and the Prospects
for Peace, proposes a novel theoretical and empirical framework for analyzing long
cycles. The theoretcial model integrates the analysis of economic and political cycles
into a unified causal framework capable of explaining the onset of systemic conflict
in the international system. Using political and economic data from 1850-1976, the
empirical analysis draws upon new theory on contegration in econometrics to test the
interrleationship between systemic conflict, the distribution of power, and trade.
Integration tests find that each of the phenomena are independently characterized by
stochastic trends, while cointegration tests indicate that these series share a common
trend over time. This subsequently results in another of tests, which demonstrate that
conflict in the system responds alone to changes in the distribution of power, trade
and economic waves.

The third chapter, Regime Structure, Leadership Uncertainty and the
Maintenance of Cooperation: The Gains in Modeling International Economic
Regimes as Organizational Teams, seeks to formalize a model of an international
trade regime by conceiving the structure as an organizational team. The formal model
takes into account an N -person environment, which incorporates Bayesian updating

among the actors, to test for the prospects of cooperation within a regime under

conditions of leadership uncertainty. A simulation is conducted to test several
propositions. The results of the analysis suggest that as the probability of confronting
a strong leader rises, the payoffs from defection decrease among states comprising the
regime. However, as the probability of confronting a weak leader capable of
punishing rises, the payoffs from defection increase among the states comprising the
regime. The implications from these results are drawn out, while suggestions for

future research are delineated.

ACKNOWLEDGMENTS

It is a happy day that I sit here in front of my computer typing up my final
remarks and thanks to my numerous friends, colleagues and family, whose support
have made this accomplishment mean something. First and foremost, I would like to
extend my heart felt thanks to my committee members, Gretchen Hower, David
Rohde, Scott Gates and Jim Granato. Each of them have contributed immensely, and
in very different ways, to my training and development as a student. They have made
this last hurdle a seemingly, painless process...well, maybe just a little painful. In
addition, Renée Smith has been an important influence in my life over the last six
years. She is originally responsible for steering me down this path as an
undergraduate in the winter quarter of 1989. Her honesty, generosity, and integrity
make her unusual and a great friend. Her determinism, ambition, intensity, and high
standards are contagious. Also, a big thanks to Cliff Morgan, Ric Stoll and Rick
Wilson for their encouragement and support during the past year. Also Chris
Anderson, Jeff Bowers, Kathleen O’Connor, and Dan Ward were immensely
supportive.

I am also grateful to my numerous friends that made my years in East Lansing

some of the most memorable experiences of my life. Kevin, Mel, Kate (and now Ty)

are great friends and were the best neighbors. I already miss the warm fires and
engaging conversation. Big thanks to Kim Eby & Jose Cortina, two truly interesting
people, who I treasure knowing always. Also, Jean Phillips & Stan Gulley for being
kind and generous friends and aiding in my endeavors to consume great food and
wine. In addition, I would like to thank my friends and colleagues in the political
science department at Michigan State, particularly Shelly Arsneault (and Tony), Bob
Orrne, Phil Alderfer, Misha Barnes, Jeff Coupe, Kathleen Dowley, John Davis, Eric
Duchesne, Brian J aniskee, and Sara McLaughlin. Each of them made my days in
Kedzie hall fun. Also, a big thanks to Gretchen and Paul for softball and poker. Also,
a big thanks to Keith Hattrup, Sandy & Ron Landis, Jeff Schneider, Joanne Spear
Sorra, and Steve Kowslowski for great memories, too numerous to mention. I’ll
treasure my friendships with all of them forever.

Lastly, Terry, Brett and Reba made this accomplishment mean something
special. Their wisdom, trust, respect, and unconditional love give me encouragement
everyday. I thank Eupi and Rosin for their love and respect, and most of all for
Mickey. Mickey is my best friend. My accomplishments are in part a reflection of
his unyielding love, support, respect, and admiration. Our life together has been a fun

experience and I look forward to what the future holds!

vi

TABLE OF CONTENTS

Page
LIST OF TABLES ................................................................................................ ix
LIST OF FIGURES ............................................................................................... xi

CHAPTER (1) CONGRESSIONAL RESPONSIVENESS TO CONSTITUENCY
DEMANDS: THE POLITICAL ECONOMY OF THE
NORTH AMERICAN FREE TRADE AGREEMENT

Introduction ........................................................................................................... 1 .
The Politics Surrounding NAFI‘A ........................................................................ S
A Generalizable Theory of Congressional Responsiveness .................................. 10
Constituency Preference Demands .................................................................... 11
Congressional Responsiveness and the Determinants of Preference Supply... 17
A Model of Congressional Responsiveness .......................................................... 24
Data ................................................................................................................... 24
Constituency Demand Variables ................................................................ 24
Congressional Supply Variables ................................................................ 26
Method ............................................................................................................... 29
Empirical Results ............................................................................................... 30
Conclusion ............................................................................................................. 37

CHAPTER (2) MODELING THE DYNAMICS OF POWER, TRADE
AND CONFLICT: FLUCTUATIONS IN THE ECONOMY,
ALTERNATING DISTRIBUTIONS OF POWER

AND THE PROSPECTS FOR PEACE

Introduction ............................................................................................................ 41
Long Cycles in Review .......................................................................................... 51
Explaining the Dynamics of Systemic Conflict ..................................................... 61
The International Economy ............................................................................... 62
International Trade ............................................................................................ 66
The Distribution of Power ................................................................................. 70
Towards a General Co-Evolving Systemic Equilibrium Hypothesis .................... 73
A Model of Co-Evolving Cycles ........................................................................... 79
Properties of Integrated Series .......................................................................... 80

vii

Diagnosing Integration ...................................................................................... 84

Diagnosing Cointegration ................................................................................. 92
The Nature of Causality in Cointegrated Systems ................................................ 99
The Role of Exogeneity in Modeling ................................................................ 99
Defining Exogeneity ......................................................................................... 104
Exogeneity and Integrated Processes .................................................................... 106
A Test of Exogeneity ........................................................................................ 112
Specifying an Error Correction Model ................................................................. 118
The Bugle-Granger Two-Step Method ............................................................. 120
Conclusion ............................................................................................................. 1 23

CHAPTER (3) REGIME STRUCTURE, LEADERSHIP UNCERTAINTY
AND THE MAINTENANCE OF COOPERATION:

THE GAINS IN MODELING INTERNATIONAL ECONOMIC REGIMES
AS ORGANIZATIONAL TEAMS

Introduction ........................................................................................................... 129
International Regimes as Organizational Structures ............................................. 132
Economic Policy Regimes ................................................................................ 132
Organizational Teams ....................................................................................... 135
Leadership ......................................................................................................... 137
Toward a General Formal Model .......................................................................... 139
The Maintenance of Cooperation Under Leadership Uncertainty ........................ 149
Bayesian Updating ............................................................................................ 149
A Simulation ..................................................................................................... 152
Conclusion ............................................................................................................. 160
APPENDIX A: Serial Correlation Tests ............................................................... 163
APPENDIX B: Formal Model Proofs .................................................................. 165
LIST OF REFERENCES ...................................................................................... 168

viii

 

T3

T2

T2

LIST OF TABLES

Page
Table 1.1 Variable Descriptions ............................................................................ 25
Table 1.2 Descriptive Statistics ............................................................................. 31
Table 1.3 Ordinary Least Squares Model .............................................................. 32
Table 1.4 Probit Model .......................................................................................... 32
Table 2.1 Variable Descriptions ............................................................................ 65
Table 2.2 Properties of Stationary and Integrated Processes ................................ 81
Table 2.3 Descriptive Statistics ............................................................................. 89
Table 2.4 Dickey-Fuller Tests for Nonstationarity ................................................ 90
Table 2.5 MacKinnon’s Critical Values ................................................................ 91
Table 2.6 Cointegration Regression and Evidence of Cointegration .................... 95
Table 2.7 Johansen Cointegration Tests ................................................................ 98
Table 2.8 Systemic Conflict Granger Causality Tests .......................................... 113
Table 2.9 Diagnosing Exogeneity: Restricted VEC Test ...................................... 117
Table 2.10 Error Correction Model Estimates ...................................................... 122
Table 3.1 Nl Payoff When a1=2 and c=1 ..................................... 142
Table 3.2 N 1 Payoff When 6:99 and y=.8 .......................................................... 155
Table 3.3 Nl Payoff When 6:99 and y=.5 .......................................................... 155

ix

Table 3.4 N, Payoff When 6:99 and y=.2 .......................................................... 155

Table 3.5 N, Payoff When 6:75 and y=.8 .......................................................... 156
Table 3.6 N, Payoff When 6:75 and y=.5 .......................................................... 156
Table 3.7 N, Payoff When 6:75 and y=.2 .......................................................... 156
Table 3.8 N, Payoff When 6:50 and y=.8 .......................................................... 157
Table 3.9 N, Payoff When 6:50 and y=.5 .......................................................... 157
Table 3.10 N, Payoff When 6:50 and y=.2 ........................................................ 157
Table 3.11 N, Payoff When 6:25 and y=.8 ........................................................ 158
Table 3.12 N, Payoff When 6:25 and y=.5 ........................................................ 158
Table 3.13 N, Payoff When 6:25 and y=.2 ........................................................ 158
Table 3.14 N, Payoff When 0:01 and y=.8 ........................................................ 159
Table 3.15 N, Payoff When 6:01 and y=.5 ........................................................ 159
Table 3.16 N, Payoff When 6:01 and y=.2 ........................................................ 159

LIST OF FIGURES

Page
Figure 2.1 Relationship Between Systemic Conflict and Inflation ....................... 64
Figure 2.2 Relationship Between Trade and Systemic Conflict ............................ 69

Figure 2.3 Relationship Between Distribution of Power and Systemic Conflict. 72
Figure 2.4 The Co-Evolving Systems Theory ....................................................... 75

Figure 3.1 N-Person Cooperative Dilemma .......................................................... 141

xi

n5

CI

CHAPTER 1

CONGRESSIONAL RESPONSIVENESS TO CONSTITUENCY DEMANDS:
THE POLITICAL ECONOMY OF
THE NORTH AMERICAN FREE TRADE AGREEMENT

Introduction

The large international trade deficits the United States accrued during the
1980's have heightened concerns about the future of domestic industries and job
security. The loss of manufacturing jobs, heightened competition from import
producers, and lack of access to foreign markets have left numerous economists
and policy makers debating the merits of free trade.1 Subsequently, adversaries
and proponents of free trade alike have purposefully sought to influence the
composition of trade policy. These pressures have subsequently been associated
with an increase in the number of trade bills introduced in the House within the
last couple of decades. Some opponents of free trade have sought specific

protection for particular domestic industrial sectors.2 However, other forms of

 

lSee Lawrence & Schultze (1990) and Destler (1992).

2The Textile Bill of 1985 and the Fair Practices in Automotive Products Act passed in 1982 are two examples
of domestic content legislation seeking specific protection for domestic industries enduring some form of
economic stress under free trade.

2

legislation have recently sought to expand the breadth of trade in the form of
bilateral and multi-lateral free trading agreements.

Presidential power to initiate free trade agreements (FT A) was initially
granted by Congress in the Tariff and Trade Act of 1984. Subsequently a bilateral
free trade agreement was initiated with Israel in 1985.3 This pact was to set the
stage for future free trade accords and was passed with no opposition in the
House.4 In 1988 another bilateral FT A was signed between the United States and
Canada. The passage of this agreement proved more controversial than its
predecessor. Because Canada is the United State's largest trading partner, this
agreement was more likely to affect more import and exporting industries than the
agreement with Israel. As such, more special interest groups campaigned against
it.5

However, the vote on the North American Free Trade Agreement (NAFTA)

last year afforded a true test of congressional support for free trade, in part,

 

3The House approved the agreement by a 422-0 vote May 7, 1985. Under the agreement Israel was to drop over
one half of existing restraints on US. imports, with the rest to be phased out by 1995. In addition, the United
States would immediately lift 80% of its barriers on Israeli goods (Congressional Quarterly Almanac, 1985,
p.260).

‘Initially the president's power to negotiate a free trade agreement was granted by Congress only in the case of
Israel. Congressional concerns over granting the president free reign in drafting similar trade agreements with
other nations, were manifested in a senate report. This report made clear that the free trade agreement with
Israel was a special case and that future agreements would have to be negotiated on a nation-by-nation basis
(see Congressional Quarterly Almanac, 1984, p.171).

’Nevertheless, the House approved the free trade agreement by a 366.40 vote August 9, 1988 (Congressional
Quarterly Almanac, 1988, p.86-H). The pact was designed to phase out tariffs between the United States and
Canada over a ten-year period (the United States tariffs on Canadian goods averaged 4%, while Canadian duties
on US. goods averaged 10%). In addition it contained some restrictions on cross-border investments and trade
services. Moreover, it guaranteed U.S. access to Canadian oil, gas, and uranium (see Congressional Quarterly
Almanac, 1988, pp.222-23).

3

because it was drafted at a time in which strong protectionistic forces were
sweeping the nation at levels not known since the passage of the Smoot-Hawley
tariff in 1930.‘5 NAFTA is by far the most comprehensive free trade agreement
passed by Congress to date. Similar to the other bilateral trade agreements, it
sought to eliminate all tariffs on goods produced and sold in North America. In
addition, it included numerous provisions concerning the maintenance of free and
fair trade on such things as investments, intellectual property rights, and services.
However, the passage of NAFTA made numerous labor and manufacturing groups
vulnerable. More often than not, the agreement was publicly equated with the loss
of labor and manufacturing jobs.7 Although, economists and trade experts alike
forecast long-run benefits associated with the free trade agreement (i.e. stimulated
growth resulting in a net plus of jobs), concerns over potential joblessness
resonated in the media.8 The crux of the issue really concerned who would lose in
the short-run. Few economists denied the inevitable period of short-term

adjustment that would take place. Specifically, short-term adjustments would

 

“The House passed NAFTA with a 234-200 vote. A strong majority (75%) of the Republicans voted in support
of the agreement (132-43), while a majority of the Democrats (60%) voted against the agreement (102-156)
(Congressional Quarterly Weekly Reports, November 20, 1993, p. 3224).

7An extensive debate in economics exists over the effects of trade on employment. The purpose of this paper
is not to address this debate per se, but rather, in some sense, to examine the politics evolving from it. See
Hufbauer et a1. (1993) for a discussion of the impact of NAFTA on employment. Generally, the empirical
effects of trade on employment remain inconclusive. See Tyson et a]. (1988) and Deardoff & Stern (1979) for
a general discussion of labor's stake in international trade matters.

1US. Trade representative Mickey Kantor, along with numerous economists, defended the employment benefits
associated with the agreement. Specifically, he claimed that it was a better deal for US. workers than existing
status quo policies. The creation of over 200,000 jobs was projected within the first two years after NAFT A's
passage. Potential gains for US. industry would come in various forms over a wide spectrum of sectors: new
car sales, beef exports, financial services (Congressional Quarterly Weekly Reports, September 18, 1993).

I6

[1

4

result in a loss of jobs in some industries, even though in the long run more jobs
would be created.9 If NAFTA, in the long run, would accrue benefits to American
industries, why was there such resistance to its passage?

To answer this question appropriately, a theoretical framework is deve10ped
and subsequently tested. An argument is developed that suggests constituents
vulnerable to short-term adjustment costs demand regulation of protection. By
testing a majoritarian politics hypothesis, I will show how the outcome of the vote
on NAFTA was a function of political constraints imposed by constituents
confronting potential economic risk in their districts. Generally speaking, this
research finds support for the notion that representatives will supply protection to
vulnerable interests if it is in their best "re-election" interest to do so.

Specifically, the analysis confirms several interesting results. Initially,
protectionist interests affiliated with manufacturing labor (import oriented) and
union groups do encounter greater unemployment risk, while their anti-
protectionist counterparts (export oriented) do not. Moreover, the risk of
unemployment subsequently results in protectionist interests significantly affecting
the likelihood of a representative voting against the free trade agreement, while
anti-protectionism groups tend to decrease the probability of a representative
voting against free trade. Surprisingly, the overall results obtained from the

analysis of the model, suggest that the economic affects of the vote tend to

 

9See Hufbauer et al. (1993) for a thorough discussion about the net gains associated with the passage of
NAFTA.

5

dominate over traditional political issues such as partisan affiliation, presidential
support, and a representative’s previous election support. These results suggest
support for Wallerstein’ s (1987) argument that unemployment tends to dominate
the politics associated with the demand for protection. However, initially it is
necessary to precede with a discussion about the nature of the politics surrounding
the debate over NAFTA's, to demonstrate its appropriateness as a test of the

theoretical model.

The Politics Surrounding NAFTA

The debate over NAFTA dominated the media weeks before it was to be
proposed to the House for a vote. Concerns over short-term adjustment costs were
reflected in polls taken before the NAFTA vote. Molyneux (1994) found that clear
majorities believed NAFTA would cause US. companies to relocate to Mexico
(73%) and result in lower wages for American labor. Moreover, a commanding
three-to-one majority felt the agreement would result in fewer jobs in the United
States. This fact is reinforced by an NBC News/Wall Street Journal survey, which
found by a five-to-three margin that Americans felt the treaty would lose more
jobs for the United States than it would gain. Even though conventional wisdom
suggested that NAFTA would actually result in a net gain of employment over the

long run, the alleged threat of a giant "sucking sound" of jobs going south

 

 

6
permeated the public debate.lo Opinions of Populist politicians, such as Perot,
permeated the media with anti-NAFT A rhetoric. Perot's attempts to influence the
NAFTA debate culminated into a book and subsequently a debate with Vice
President Al Gore. “

Representatives and governmental officials sensed extreme opposition to
NAFTA's passage. Most of this opposition was largely attributed to fears over
losing jobs. Labor Secretary Robert B. Reich commented,

"This opposition has little to do with the agreement and much to do

with the pervasive anxieties arising from economic changes that are

already affecting Americans" (Congressional Quarterly Weekly

Report, September 4, 1993, p.2335).

Similar sentiments were echoed by representatives. Rep. Anna G. Eshoo, a first-
term Democrat from California, was inclined to support NAFTA but admitted that
she was susceptible to pressures from labor groups to vote against it. Dennis

Hitchcock, a member of the Air Transport Employees Union from Eshoo's district,

was against NAFTA and warned Eshoo about losing support from labor if she

 

I"The reference of a "giant sucking sound" in the US. job market was made by Ross Perot in his book, Sari
Your Job, Save Our Coung. Senate Assistant Minority Leader Alan K. Simpson (Republican Wyoming)
offered an alternative hypothesis that suggested, "the (giant) sucking sound in Ross Perot's comer....is from
some extraterrestrial vehicle pulling his brains from his body" (Congressional Quarterly Weekly Report,
September 13, 1993, p.2439).

"However, ironically, evidence suggests that Perot may have actually hurt the anti-NAFT A cause and that his
performance in the debate, if anything, actually served to decrease the public’s opposition. Perot's negligible
effect on the NAFTA debate was reinforced by an ABC poll that found a 5 percentage point opposition margin
the night of the debate that grew merely to an eight percentage point gap two nights later (Molyneux, 1994).
Even in the state of Michigan, a highly touted anti-NAFT A state, Perot's effect was negligible on opposition
forces. A state poll found a 15 percentage point opposition margin the night of the debate, which decreased
to an nine percentage point gap after the debate (EPIC/MRA Report, Vol.1, No.20, December 15, 1993).

7

voted for the agreement.” Moreover, Representative Mike Kopetski, Democrat-
Oregon, had similar threats levied against him. In fact, he reported that his support
for the trade pact ended his relationship with the International Association of
Machinists and Aerospace Workers, which reportedly will no longer give him
PAC funds.l3 However, despite labor's pressure to constrain Eshoo's and
Kopetski's support for NAFTA, both voted for the agreement.

However, this is not the case for a Republican representative in South
Carolina. Republican Bob Inglis from South Carolina's 4th district, a place where
economic prosperity and international trade are supposed to be closely associated,
leaned against voting for NAFTA. Inglis was quoted as saying that he would,
"...vote against NAFTA because I fear the loss of the manufacturing base of the
country." Inglis received pressure from prominent local business people in his
district. For example, Milliken, a big textile industrialist and influential figure in
Republican politics in Inglis's district (not to mention throughout the state) was
vehemently against NAFTA's passage. It was Milliken's support that helped Inglis
secure his seat. Thus, it was no surprise to learn that Inglis later voted against the

I. I4

agreemen What seems clear, is that labor unions and manufacturing interests

can use carrots to gain influence, in addition to sticks to shape legislative

 

"In fact, labor unions in her district went so far as to pass resolutions calling for member unions not to endorse
or provide campaign cash to any members of Congress who supported the agreement (Congressional Quarterly
Weekly Reports, November 6, 1993, p. 3016).

I3See (CQ Weekly Reports, September 18, 1993, p.2437).

l‘See (Congressional Quarterly Weekly Reports, November 6, 1993, p.301 8-19).

8

preferences. However, what determines when a legislator will choose to succumb
to the pressures from dominant local interests? Moreover, if a legislator is
susceptible to constituency pressure on trade matters, which interests is she most
likely to be sensitive too?

Both political scientists and economists alike have tried to explain why it is
rational for legislators to vote in favor or against their constituents.” In terms of
trade policy matters, three principle literatures exist that seek to explain policy
outcomes. The first is generally referred to as the pressure group hypothesis (or
more generally, capture theories).16 Explanations of this sort propose that trade
policy outcomes result from special interest groups competing among themselves
for political influence and wealth by lobbying politicians. The second variant, the
congressional dominance theory, suggests the bureaucracy in charge of regulation
is not susceptible to capture, or influence by interest groups.” Rather, Congress
as an institution, controls the bureaucracy through oversight and is largely
responsible for the distribution of benefits. Lastly, the majoritarian politics
hypothesis (or more generally the voting model approach) proposes that trade
policy outcomes are a function of political pressures imposed by constituency

interests within a district.‘8

 

lsFor example see Downs (1957), Stigler (1971) and Fiorina (1974).

mFor example see Schattschneider (1935), Stigler (1971), Peltzman (1976), Becker (1983), Findlay & Wellisz
(1986) and Magee et al. (1989).

17For example see Shepsle (1979), Shepsle & Weingast (1984) and Weingast (1981; 1984).

ll’For example see Laverge (1983), Mayer (1984), McArthur & Marks (1988).

9

Because the following research is concerned with examining local pressures
on congressional decision-making, a majoritarian politics approach to the study of
trade policy outcomes is most appropriate. As such, the theoretical argument
draws upon the latter research, by examining how constituency preference
demands for trade regulation become supplied by their representative. The
investigation of the demand for, and supply of protection, is by no means a novel
undertaking.19 However, previous studies seeking to predict policy outcomes have
not fully accounted for the sources of economic risk responsible for constituency
demands for trade regulation. Moreover most models fail to capture demands on
both sides of a free trade issue. Typically existing models focus exclusively on
special interests against free trade legislation, while not considering the role of
interests seeking to lobby on behalf of less restrictive agreements.20 The following
research controls for both protectionist and anti-protectionist interest groups.
Moreover, on the supply side, numerous studies have not fully considered the
political risk faced by representatives and their vote on trade policy matters.

Specifically, most existing research does not clearly delineate the link between

 

19For example see Ray (1981), Frey (1984), Coughlin (1985), Nelson (1987), Conybeare (1991), and Hansen
(1991). Wallerstein (1987) also discusses the importance of the demand for, and supply of, protection.
However, his model is incomplete as his aim is simply to capture the effects of unemployment on the demand
for protection. The model developed in this paper considers protection for interests not only against free trade,
but also, interests that advocate free trade. The former is considered in virtually all models seeking to explain
trade policy outcomes, while the latter is typically ignored.

2°Destler et al. (1987) are one of the most notable exceptions. They find that opposition interests are better at
getting specific protection, but not at general procedural changes. McArthur & Marks (1988) are another
exception. They control for farm employment and export employment and suggest that they will have a
negative effect on a legislator‘s vote for a protectionistic bill.

10

constituent pressures and congressional decision making on trade policy matters.
As such, I will draw upon existing congressional literature to model congressional

decision-making on NAFTA appropriately.

A Generalizable Theory of Congressional Responsiveness

A representative democracy is premised on the notion that there is a link
between constituency preferences and their representative's preferences on policy.
Foremost, representation manifests itself in the electoral connection (Mayhew
1974). To the extent that members of Congress depend upon the support of their
constituency to maintain tenure in office, there exists a strong incentive for
representatives to reflect the policy preferences of the people comprising their
district. One of the central tasks in developing this theoretical argument is to
explain the nature of preference formation. A supply and demand framework is
an intuitive way to model the preferences of constituents and congressional

responsiveness:

Demand = f (constituency preferences)
Supply = f (congressional responsiveness)

In this framework, the question becomes whether constituency preferences, in the
form of a demand for a particular policy outcome, are supplied by their respective

member of Congress, hence determining the extent of congressional

ll

responsiveness to constituency interests. Accordingly, it is important to consider
the factors which explain the development of constituency preferences for trade
legislation as well as the determinants which explain why representatives may or

may not consider such factors in their decision-making process.

Constituency Preference Demands

Individuals purposefully vote for a particular representative because they
embody similar beliefs and attitudes that correspond, at some level, to their own.
Fundamental to the notion of a congressmen's representation then, is identifying
who or what groups have particular preferences on any given piece of legislation.
Ironically, on the issue of NAFTA, indications from opinion polls suggest that a
majority of respondents polled were generally unaware or unconcerned with the

agreement.21

Undoubtedly however, there did exist a particular segment of society
for whom NAFTA became a particularly important issue. What distinguishes
these individuals for whom the agreement was important from others who had no
interest in NAFTA? Specifically, what characteristics differentiate constituents

who were concerned with NAFTA's outcome from those who were not?

Moreover, what type of preferences did these constituents have and what was their

 

2‘Newhouse and Matthews (1994) present polling data that demonstrate that a majority of the American
population never became polarized on the NAFTA issue. In fact, the vast majority of Americans never became
informed or thought the agreement would directly have an effect upon them. Specifically, they found that 6 out
of 10 Americans were not following the debate on NAFTA. Moreover, they found that a surprising 4 out of
5 Americans did not even know whether their representative was for or against NAFTA. In fact, Molyneux
(1994) only found a narrow margin of voters identifying themselves supporting or opposing the agreement.

12

effect on the voting behavior of legislators? Answers to these questions are crucial
in delineating the factors important in explaining constituency preference demands
and congressional responsiveness.

At some general level, there were individuals who had a preference on the
outcome of NAFTA and individuals who had no preference on the issue at all. To
explain the outcome of the NAFTA vote, it is first necessary to distinguish at the
district level, which constituents had preferences (and later what those preferences
were) from those who had no preference. Arnold (1990) has developed a simple
framework which distinguishes between two different types of publics that
comprise a member's congressional district. They are the attentive and inattentive
publics. My model generalizes this idea about publics to the analysis of trade
policy. With respect to any given issue on the congressional agenda, attentive
publics are citizens who are aware of a specific issue and have a firm preference
concerning how their representative should vote. In contrast, inattentive publics
are constituents who do not have firm policy preferences on a particular issue.22

Attentive publics correspondingly have, by definition, a high salience factor
associated with a particular issue. The term "salience," represents the sensitivity
and prominence felt about a particular issue for a particular constituent or interest
group. Coinciding with the notion of saliency is what Kingdon (1989) refers to as

the intensity associated with a particular issue. Individuals who find a particular

 

22See Arnold (1990) p.64-65.

13

issue salient, do so in part, because it has an effect upon their lives. The more
directly the issue affects an individual's life, the more intense a preference
becomes. As constituents become more intense in their preferences, legislators
will generally weigh their opinions more heavily.23 Furthermore, Arnold (1990)
notes that legislators are forced to serve the interests of their attentive publics.
With respect to NAFTA, attentive publics refer to both opponents and proponents
of free trade which have strong preferences vehemently against the agreement, or
conversely, strong preferences for the agreement.

Generally, constituents who found NAFTA to be a salient issue, did so
because they felt that it would affect their employment status.24 Again, the politics
surrounding the debate on NAFTA derived in large part from the anxieties arising
from economic changes that were perceived to affect the domestic economy. Of
primary interest to numerous constituents was the risk of unemployment. Thus, in
some measure, constituency preferences for trade policy depend on the costs and
benefits associated with the outcome, or in the case of NAFTA, the perceived

outcome and its affect on employment. The economic self-interest of groups

 

23Kingdon notes that intensity among constituents is not constant. Likewise, not all issues will have the same
degree of intensity. However, in this model, NAFTA is perceived as a salient issue because of its alleged effect
on employment. Thus, constituents confronted with alleged job losses from NAFTA will have a strong intensity
associated with protecting the economic survival of their particular industry. Moreover, this would include
individuals or interests groups that would also suffer from other's job loss. That is, loss of employment has spill
over effects that affect the economic interests of groups dependent on the economic survival of other
individuals.

“This does not deny the fact that constituency preferences could have an ideological component associated with
them. However, it is argued that a majority of the constituents who found NAFTA salient, had higher levels
of intensity associated with a job dimension than with an ideological dimension.

566

Eff.

int

C01

cit

198
cffc
lEpr
lobt
Wei,
1 19‘.
filth.
find

Una
5&5},-
”Or.

l4

seeking to protect themselves has been studied in numerous settings.25 Laverge
(1983) was one of the first economists to incorporate the political process in an
empirical study of policy outcomes related to tariff policies. Extending this work
into trade policy is still a relatively new endeavor for economists.26
Unemployment is perhaps the greatest cost and most salient issue for
constituents to take into account when determining preferences for a trade policy.27
In fact, unemployment has been found to be one of the primary determinants
driving the demand for protectionist policies.28 Ironically, Wallerstein (1987) has
cited the general neglect of political scientists in modeling the threats of
unemployment in the analysis of the demand for protection. However, recently

economists seeking to explain trade policy outcomes have begun to incorporate

 

25For example see Stigler (1971) and Peltzman (1976).

“For example see Peltzman (1984), Coughlin (1985), Tosini & Tower (1987) and McArthur & Marks (1988).
Unfortunately most of these models resemble what Krehbiel (1993) refers to as "kitchen sink" models. Namely,
political variables are tossed in solely to see what explanatory power they have with little regard to what
theoretical contributions they might offer.

"Ironically, the resulting costs associated with protection are typically imposed on the general public (Ray
1981). Even though overall estimates of the costs associated with protection are low (see Feenstra (1992)), the
effect they have on driving up consumer products is indisputable. However, because consumer groups tend to
represent a large diffuse group, typically they do not mobilize. One of the most notable examples is the sugar
lobby's success in getting protection. Moreover, empirical evidence suggests that industry interests are typically
weighted more heavily than consumer interests (see for example Ray (1974), Baldwin (1976) and Caves
(1976)). In addition, consumers will more likely determine their preferences on trade policy as an employee
rather than a consumer. For example, Conybeare (1991) finds voters employed in an industry within a district
find their interests as employees outweighing their interests as consumers.

28See for example, Nowzad (1978), Anjaria et al. (1982), Wallerstein (1987), and Tyson et al. (1988).
Unemployment at the national level is believed to effect the supply of protection. There is a literature which
seeks to explain the relationship between business cycles and the tariffs (see Takacs 1981, Gallarotti 1985;
Cassing et al. 1986). Although the purpose of this paper is not to examine tariff trends over time, national
economic factors have a role in determining the supply of protection.

15

unemployment effects into their analysis.29 Likewise, the model developed in this
research considers the effects of unemployment on constituency preferences for
trade legislation. Because NAFTA would result in adjustments costs in the form
of joblessness, certain interests groups had negative expectations about NAFTA’ 3
effect on their employment if the agreement were passed into law. However, there
were also interest groups that had negative expectations about the level of
employment that would exist if NAFTA were not to pass. Specifically some jobs
remain vulnerable to free trade (such as employment associated with import-
competing industries), however, some types of employment rely on the existence
of free trade (i.e. exporting industries in general). As such, it is necessary to
control for both, protectionist forces and anti-protectionist forces.

As demonstrated above, it is important to take into consideration that the
debate over free trade and its affect on job security cut two ways. If constituents
felt that their employment status was threatened by N AFTA's passage, then these
individuals were more likely to be intensely opposed to the agreement. These
constituents can generally be associated with manufacturing employment and
employment in import-competing industries threatened by foreign competition.
On the other hand if constituents thought that their employment depended on the
prevalence of free trade, then they were more likely to find themselves supporting

the agreement. These constituents can generally be associated with employment

 

29For example see Takacs (1981), Coughlin (1985), Wallerstein (1987), Magee & Young (1987), McArthur &
Marks (1988), Marks (1993).

this n
trade

folio

Prop

Pm}

Pref

16

that is dependent upon access to foreign markets and purchasing of wholesale
goods?0 Surprisingly, few existing studies capture anti-protectionist forces and
their influence upon the political process. The theoretical framework developed in
this research, takes into account the variance in constituency preferences for free
trade policy. The theoretical arguments developed above will be tested with the

following propositions:

Proposition (1): Districts composed of manufacturing interests, are more
likely to confront the risk of unemployment, and therefore
will have a lower propensity to support NAFTA.

Proposition (2): Districts composed of manufacturing interests related directly
to export markets, are less likely to confront the risk of
unemployment and therefore will have a higher propensity to
support NAFTA.

Proposition (3): Districts with a large percent of union members associated
with labor, are more likely to confront the risk of
unemployment, and therefore will have a lower propensity to
support NAFTA.

However, the negative expectations concerning NAFT A’s affect on
employment may have subsided if safety nets, or resources more generally, were

available from the federal government to offset the economic adjustments resulting

from the implementation of the free trade agreement. Specifically, the existence of

 

3°See Destler & Odell (1987) for further evidence on the sources of anti-protectionist forces. Aside from the
lack of access to foreign markets and goods, another potential threat associated with restricting free trade is the
likelihood of setting off trade wars. Restricting access to domestic markets from foreign competition has the
potential to produce retaliatory trade policies from other nation-states.

17

federal resources to small businesses, groups in need of job training, and
individuals in need of low income housing, etc. may have served to offset the
detrimental impact associated with the passage of the free trade agreement, and
hence the pressure from members to vote against NAFTA. This suggests the

following proposition:

Proposition (4): Districts with a large number of safety net resources will have
a higher propensity to support NAFTA.

Now that the fundamental theoretical components associated with
constituency preference demands have been delineated, it is necessary to turn
towards a discussion of congressional responsiveness. To elicit a complete
theoretical framework that can explain trade policy outcomes, it is vital to
incorporate how constituency preference demands are rendered in the political

process.

Congressional Responsiveness and the Determinants of Preference Supply
Although it is possible to identify the existence of constituents with strong
preferences for and against NAFTA, how do these preferences subsequently get
representation? Moreover, which constituency preferences get represented? When
considering the answers to these questions, it is necessary to analyze the

motivation behind congressional actions and decision-making. It is generally

l8

assumed that the primary goal of all members of Congress is to be re-elected.
Thus most, if not all of their activities take into account what effect a vote will
have on their bids for re-election.31 In doing so, they will identify with the
dominant industrial interests within their district. Kingdon (1989) concludes that
nearly all members of Congress defend important industries within their own
districts. As an industry becomes economically important in a district, so does the
maintenance of its economic survival. When an important industry within a
district is threatened so is the economic well-being of the district. This is the
crucial link between constituency preference demands for the regulation of trade
and congressional response. If a congressman does not heed the call for protection
by important industrial interests within a district, then the representative faces
possible retribution at the polls.32

Without question, representation tends to be rewarded in Congress. Good
service translates into a long tenure in officef"3 Kingdon (1989) and Fiorina (1974)

associate a representative's seat safety and long tenure in office with following

 

3|Although re-election is a legislator's primary goal, this by no means suggests that good policy-making is not
a concern. Members of Congress have their individual policy preferences independent of their constituents.
However, in some instances, a representative's preferences will take second priority if she finds that her
constituency's preferences are in conflict (especially if the constituency's preferences are associated with high
intensity (Kingdon 1989)).

32The notion that concentrated interests prevail in getting their preferences represented by a representative at
the expense of diffuse unorganized interests has been formally deduced in numerous settings (see for example
Downs 1957, Olson 1968, and Stigler 1971). Fiorina (1974) has posited that a member of Congress will appeal
to its largest sectional interests. In a similar vein, Fenno (1978) points out that representatives appeal primarily
to their re-election constituency.

”For example see Wilson (1986), Kemell (1977), Polsby (1968), and Thompson & Moncrief (1988).

19

their constituency's preferences on high salience votes. As such, at some basic
level, representatives have an incentive to reflect (or supply) the preferences of
their respective re-election constituency. If they repeatedly fail to supply
constituency preferences, they are most likely to face political risk. Political risk is
defined as a legislator's chance of damaging or losing support networks associated
with her re-election efforts. Specifically, political risk is associated with the loss
of economic resources and votes. When representatives choose not to vote
according to their constituents preferences then they risk the loss of support.
Again, support represents not only votes, but economic resources. One of the
biggest economic losses is campaign funds. Representatives depend on campaign
resources in their re-election efforts. Numerous individuals and groups donate
economic resources to a candidate because of the particular type of representation
she will bring to the district. Thus, contributions are to some degree, analogous to
political investments made by groups and individuals.“ This suggests the

following proposition:

Proposition (5): Representatives, which have a higher percentage of their total
campaign resources from labor political action committees,
will have a lower propensity to support NAFTA.

 

3‘For example, see Snyder (1990). He models campaign contributions as a form of investment by interest
groups in the House from 1980-1986.

20

Related to loss of support, is the level of vulnerability that characterizes a
legislator in his district. Vulnerability within a congressional district is associated
with the strength of support received from a representative's district. It is
postulated that the more vulnerable legislators become, the less likely they are to
risk upsetting their district’s interests. Freshmen are most likely to feel vulnerable
as they have little experience in their positions and have little power to influence
outcomes in Congress. Legislators elected to office by narrow margins of support,
have a higher risk of retribution from constituents by failing to vote according to
their preferences.” Surprisingly, the link between electoral margins and trade
policy outcomes is not clear and has not been systematically explored.36
Nevertheless, the link between vulnerability and risk is made more clear when
considering an attentive constituency's preference on a salient issue such as
NAFTA. Concern over the NAFTA vote was voiced by several Republican
legislators from marginal districts. In fact, it was the plight of several vulnerable
congressmen that inspired GOP leaders to demand that the White House and its
allies in Congress work harder to win over more support from Democrats from
competitive districts where NAFTA was unpopular.37 This suggests the following

proposition:

 

3’I'he argument suggests that legislators elected by a narrow margin, begin their service in a district that is not
completely supportive. Thus, a legislator is most likely to face a tough competitive bid for re-election that could
be exacerbated by an electorate not happy with its representation.

3"See Arnold (1990, p.45) for a discussion of this issue.

37See Congressional Quarterly Weekly Reports, September 18, 1993, p.2439.

21

Proposition (6): The percentage of vote support a representative received in
the last election, should be positively related to the propensity
of the member to support NAFTA. Specifically,
representatives from districts in which they received a large
percentage of the vote, will have a higher propensity to
support NAFTA, while representatives from districts in which
they received a small percentage of the vote, will have a lower
propensity to support NAFTA.

Also it is important to take into consideration the effect that tenure in office has on

a representative’s likelihood of voting in support of NAFTA. That is, members

that are new to Congress and have comparatively little experience in contrast to

incumbents, will be most susceptible to feel pressure to vote against NAFTA. As

such, the following proposition will be tested:

Proposition (7): Freshmen representatives will have a lower propensity to
support NAFTA.

In addition to district pressures imposed on a representative's preferences,
party organizational and executive constraints persist as well. With respect to the
former, opposition to NAFTA was strongest among the Democratic membership.
In fact, if it were not for the Republican support of NAFTA, the agreement
certainly would not have passed. Numerous studies seeking to explain trade policy
outcomes have controlled for a representative's party affiliation. However,
empirical results are mixed. Typically, such models justify the inclusion of party

as a control variable with merely ad hoc generalizations about Democrats having a

22

natural disposition against free-trade. However, this generalization is not
appropriate given Clinton's overwhelming support for the agreement.38 On the
contrary, given the historical institutional bias of the executive office, in support of
free trade principals, Clinton’s support for NAFTA may not be so surprising. To
the degree that Clinton is a Democrat, his support for NAFTA may seem unusual,
however, it is not unusual because he is the executive. The executive branch has
historically supported free trade, more so than the legislative branch, which
traditionally has been considered more protectionistic (Baldwin, 1984, Cline
1989). This makes sense to the degree that the President has more diffuse interests
to take into account when determining trade policy, more so than representatives.

Representatives generally tend to find themselves more subservient to local
pressures than the executive because of strong, concentrated interests within their
districts. In fact, contrary to the preferences of their executive, some Democratic
leaders, such as the Democratic Whip, David Bonior, were ardently campaigning
against the agreement. On the surface it would seem that the executive would
have some influence on a legislator's vote, especially from members in districts
that supported the president in the last election. However, this may not be the case
given the institutional bias of the executive to support free trade and the

contrasting protectionist bias in Congress. In trade policy matters, a president’s

 

38This makes sense to the degree that the President has more diffuse interests (i.e. a more dispersed
constituency) to take into account when determining trade policy, more so than representatives (Baldwin, 1984,
Cline 1989).

 

23

influence within a district may have no effect in stimulating free trade preferences.
Moreover, due to the fact that Clinton is a Democrat, and Democrats since the 40's
have been the party associated with supporting protectionism, his influence within

a district may serve to decrease the propensity for a representative to support free

trade. Thus, the following propositions will be tested:

Proposition (8): Representatives from districts which supported Clinton, will
have a lower propensity to support NAFTA.

Proposition (9): Democratic representatives will have a lower propensity to
support NAFTA.

Members of Congress are also undeniably bounded by their own ideological
predispositions. Numerous empirical models seeking to predict trade policy
suggest that the ideological predisposition of a representative can help explain a
vote outcome.39 The argument suggests that liberal members of Congress have a
higher propensity to support restrictions on free trade, and conversely,
conservative representatives will be more likely to vote against restrictive trade
policies. The empirical findings from these studies however are inconclusive due
to poor measures of ideology."0 However, typically it is plausible to assume some

consistency in the ideological predispositions of representatives, such that,

 

3"See Peltzman (1984), Kalt & Zupan (1990).

”Generally, most of the studies use ADA ratings. The problem associated with these measures have been noted
(see Jackson & Kingdon 1990, 1993; Krehbiel 1993).

 

24

members who typically vote liberally, or conservatively, on economic issues, will
tend to exhibit similar ideological dispositions on trade policy matters. As such,

the following proposition will be tested:

Proposition (10): Representatives with liberal rating scores on economic issues
will have a propensity to support NAFTA.

A Model of Congressional Responsiveness

Qua

To test the theory about constituency preference demands and congressional
responsiveness, data were collected for all 435 Congressional districts of the 103rd
Congress. Table 1.1 below lists the variables used in the analysis, describes how
they were measured and their respective sources.

Constituenchemand Variables. Numerous reports and opinion polls
about NAFTA indicated that many concerns arising from the agreement's passage
were related to its effect on labor and more specifically manufacturing
employment. Because these employment sectors have witnessed a decrease in the
growth of jobs available, districts which have a large percentage of these groups,
will tend to have a constituency that has negative impressions about NAFTA’s

affect on future employment.

25

 

 

.833. u—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

;unu 8.. 3280“ 32 3.5.5. 5 853k 3 302.: .255an vooou 55:3 3.5:. :02 353E
.ngnou
c2530— .? .695... 2. 3 no; nevus—u age-803.com
53 ca 3933. .32 auto—cc. E heist TV .3 036: «we 3. E 20> uo nuanuobm .5330 .3 E 80> .o Eugen GULF—.0:
.oE 5:33 fink—homo
235.5 .aaxooicoéuEOmdsnnouiBInA—cc .083 5.3
.2: 303 328 82 as sea as s. .83 a: 5% 2 sin. a: 8...: =. 332: 5.55
3558 Ecruceu Ex: «:5 25:0 «.5 32 3 SEE» 58 £55 «36$: BABE—on: Co Eon-om u... E ionic—9:25 .0 Even... ASLEENZDV
. t1 4 a: I It]! .11 Int: :- 0::
Saw YES 2: P 55:3. Enetntmceu 33 A; :20 2 ESE—gen 7.82 on. an 5:» 82.63.. 18% no: bouqm 9E8
32: engages» .9; 332.83 52.. Bea-=8 5.83 nos.
.55 .32 EEEEEQU =5qu~6 5:85 Stank 3 Con—=8. - 2 Sou—55:8 32.3 09.5th 8.3 Epc Bogota—.3 51950 ”Agkgv
l . . 1 31.31 a Fat 10:1”.—
coiauo. Siame— - 85> .52 2. v5 33% «8. 5 Bang
:93 .8 Susana .32 .BtuEvt 5 485$ A; E0235 96 88> .38 2. 503.3 85.85". 0.t. 3:255 iuuuflnflm “PMOmmDmdxn—v
:53me
and .8 32°qu Noo— éutesv. 5 Snack 3 anus—unl— Eeoanoaoxlov vovoo coca—E. claim Aim—<5
a 5h . . . .80. man qucoum
32:2 .32 .e.§£ 52:5? 933.: 3 e 332. . .. .: EB? . .. .. a .s ace; .3ao_§_ 32: ”8mm:
#8— .... .u . t r 42.5% n. . r . nick—Eu...
.38 No.32 iota» cocaimfi Entau 3 SE: - do Eocene. B- 23 £3.53 =1 .6 unseen-om :25. .95 .2520 "AQHOEOEDV
dun—mu
Noo— _oo=E can 2.: 3 E- 3331.— .8nd .03 32° ce- ci AER—«m .23 voice:-
UBC. 425m 35:5 2: \o buEtd< 32:55. .33 3 o. .028 .35 cum—Eu on: E gonzo—Eu mo E022. it. .05 EEO—Em merged—Eu: “EDZ<2V
.35»:
32 59:2 v:- o== .3 2a BSME 6an £3. fiat—ducal: Ego—Em
03-... 438.». this as \e 38:“? Ekatefi a: TV E cube—Eu 2a: 58 5 312.335.... :52!— ofi. com—m antic-SEE ”EXEDEZV
.vuufim 32 Mon
LonEo>oz 533x A133 AISLES ~3322u=ou inau— uocne vovoU <E<z 8.. :3 =0: ”AP—.03
83% 85 Reason»: 233: £5;

202.589 use;

2 03-...

 

26

Thus, the endogenous variable in the demand equation is percent of unemployed
workers (UNEMPTPCT) within each congressional district. Moreover, because
labor unions and manufacturing groups were strongly protesting the passage of
NAFTA, union membership (UNIONTPCT) and manufacturing employment
(MANUF) are controlled for in the model. Both of these variables should have a
negative effect on the likelihood of a congressmen voting in favor of the free trade
agreement.

However, it is also important to consider the role that proponents of
NAFTA had in influencing legislators’ policy preferences. Typically most studies
merely control for interests on one side of the issue. Thus, I depart from other
studies by explicitly controlling for the existence of interests for the free trade
agreement. I do so by controlling for export manufacturing employment
(MANUFEXP), which represent interests in favor of NAFTA. The existence of
manufacturing jobs related to exports is hypothesized to have a positive effect on a
representative’s likelihood of voting for NAFTA. Lastly, the level of resources
that exist to compensate workers who find themselves enduring adjustment costs is
controlled for in the empirical model. Such resources have been referred to as
safety nets. They are general resources used to offset financial injury associated
with unemployment (Bates et al., 1991).

Congressional Supply Variables. Recall that representatives are postulated

to be risk averse in terms of pursuing activities associated with losing economic

27

resources for their re-election efforts. As such, their preferences on trade policy
will, in part, be influenced by groups contributing to their election war chests.
Thus, if labor groups represent a significant proportion of a representatives
campaign funds, the likelihood of the representative voting for NAFTA will
decrease. As such, the influence of contributions by labor political action
committees (LABTPCT) are controlled for in the model. It is postulated that
representatives who receive funds from PAC's will be less likely to vote for the
free trade agreement. To test for the effect that party affiliation has on a
congressman's decision to vote for the agreement, a dummy variable representing
party afliliation (PARTY) is used in the model.

NAFTA's passage was one of the most important issues on the president's
agenda. Clinton was widely known by most Americans to support NAFTA's
passage. However, what effect did the president's position on NAFTA have on
congressional decision making? Surprisingly, presidential influence has been
ignored by most explanations of trade policy outcomes. If a president has a lot of
support within a particular congressional district, it is unlikely that a representative
would publicly criticize the president's policy decisions. However, just because a
particular district supports the president, a member of Congress may not be more
likely to adopt the policy position of the executive on trade policy matters. Again,
the executive has an institutional free trade policy bias, whereas, the legislative

branch does not. Thus, presidential support (PRESUPPORT) within a

 

28

representative's district may not have influence on a representative's position on
NAFTA.41

In addition, the vulnerability of a congressmen's seat is analyzed. It is
hypothesized that the more vulnerable representatives are, the less likely they are
to risk upsetting their constituency interests. Vulnerability is assessed by two
factors. First, the model controls for the total proportion of votes received by the
representative in the last election (VOTEPCT). Secondly, the model controls for
experience in Congress by examining whether a member is new to the house
(FRESHMEN).

Some representative's voting preferences on NAFTA were undeniably
bounded by their own ideological predispositions. As such, controlling for the
ideological predispositions of representatives may help explain N AFTA's vote
outcome. Traditionally Democrats have been associated with protectionist trade
policies, while Republicans have been viewed as advocates for free trade. Thus,
liberal representatives will be more likely to support restrictions on free trade.
Conversely, more conservative legislators will be more likely to vote against
restrictive trade policies. Thus a measure of a representative’s ideology on
economic issues (IDEO) will be used in the model. This variable registers the
liberal voting orientations of legislators on economic policies. The more liberal a

representative, the less likely the member is to vote for the free trade agreement.

 

“This is contrary to Bartels (I991) argument in his analysis of votes on defense appropriations.

29

MSLhQQ

To capture the preferences of congressmen for NAFTA, roll call votes for
each representative on the agreement are used.42 Unfortunately, a more complete
measure of congressional preferences cannot be obtained due to the manner in
which the agreement was considered on the House floor.43 Congress considered
NAFTA under the fast track procedures. This procedure effectively prevents any
amendments from being considered or added onto the agreement. Moreover it
requires that a vote be taken within 90 days after the agreement is formally
submitted to the House by the president.“4 Due to the dichotomous categorical
nature of the dependent variable, an alternative estimate technique from ordinary
least squares is required (Greene 1991; Maddala 1983, 1993). As such, the

empirical model will take the form of a probit below:

 

42Congressional Quarterly Weekly Reports, November 20, I993, pp.3224-3225.

”Due to the lack of amendments, and hence roll calls for such amendments, superior measures of representative
preferences developed by Krehbiel & Rivers (1988) cannot be obtained. See Bartels (1991) for an example
of Krehbiel & Rivers technique. Bartels uses their technique to analyze congressional decisions on defense
expenditures.

“See Congressional Quarterly Weekly Reports, September 4, 1993, p.2335; October 2, 1993, p.262]. Arnold
(1990) and Bach & Smith (1988) note that restrictive rules are used as tactical devices in the House to protect
the content of bills. Generally, restrictive rules have not been found to favor either attentive or inattentive
groups. However it is a strategy that can be used to benefit either concentrated interests or diffuse interests
(Arnold 1990, p.131). Fast track procedures have been used to draft and secure approval for the 1979 Trade
Act and the 1988 Canadian Free Trade Agreement. Interestingly, the Israeli free trade pact was not considered
under fast track However, this is attributed to the fact that no opposition was anticipated in the House nor the
Senate (Congressional Quarterly Almanac, 1985, p.260).

3O

Y1(V0TE) = atl + XI(MANUF)
+ X2(MANUFEXP)
+ 3(UNIONTOTPCT)
+ X4(UNEMPTPCT)
+ 5(IDEO)
+ ,(PARTY)
+ 7(PRESUPPORT)
+ 8(LABTPCT)
+ 9(SAFETY)
+ X10(V0TEPCT)
+ XMFRESHMAN) + 6,

Empirical Results

Descriptive statistics are provided in Table 1.2 for all of the variables used
in the empirical model. In addition, Tables 11.3 and 1.4 provide a listing of the
results obtained from the OLS and probit estimations.

Propositions (l)-(3) need to be assessed by two sets of analysis. The first
part in testing their veracity, involves establishing the nature of unemployment risk
within congressional districts. Throughout the previous discussion of the
theoretical model, an argument was developed identifying those groups which
perceived themselves to be most vulnerable to a loss of unemployment. It is
important to stress that the anxieties associated with joblessness stemmed not only
from groups worried about NAFTA’s passage, e.g. protectionist groups, but also
groups concerned about the possibility that NAFTA would not pass, e. g. anti-

protectionist groups.

31

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 1.2
Descriptive Statistics
Variable Min Max Mean Std. Dev.

VOTE 0 1 0.5379 0.4991
MAN UFEXP 0.9 380.0 129.45 108. 13
MAN UF 5.40 1746.0 670.90 491 .13
UN lONTOTPCI‘ 0.0256 0.3733 0.1638 0.07 36
[DEC 0.00 91.00 47.442 29.076
PARTY 0.00 0.00 0.402 0.490
PRESUPPORT -143200 86310 -35086 33214
LABTPCI‘ 0.00 0.48 0.1018 0.1094
SAFETY 32.00 3166.0 978.78 841.97
UNEMPTPCI‘ 0.019 0.412 0.172 0.066
VOTEPCT 0.50 1.00 0.65 0. 1 1

0.00 1.00 0.262 0.440

 

 

 

 

 

 

 

 

 

 

 

 

 

 

32
Table 1.3
Ordinary Least Squares Model
Independent Variable‘ Coefficient Std. Error
MANUF 0.0004'” 0.0001
MANUFEXP —0.0017'" 0.0002
UNlONTOTPCI‘ 0.7142'" 0.0462
NOB=43S Onetail: (p<. 10)‘ (P<.05)”(P<.01)°" R1: 127

 

 

 

 

‘ Dependent Variable: Total Percentage of District Unemployment

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 1.4
Probit Model
Independent Variable' Coefficient Std. Error
MANUF -0.0029‘” 0.0009
MANUFEXP 0.0150'” 0.0041
UNIONTOTPCI‘ 6.2682” 1.2400
SAFETY 0.0001 0.0001
LABTPCT -3.7358”' 0.9009
FRESHMAN 07703 0.7399
VOTEPCT 0.4637 0.6247
PRESUPPORT - 0.00001 0.00002
PARTY 0.4761'" 0.1942
[DEO -0.0004 0.0007
CONSTANT 0. 3243 0.5291
NOB=435 Onetail: (p<.10)' (P<.05)"(P<01)'”
74% cases conectly‘
-2LnA-X’ = 135.68”

 

 

 

' Dependent Variable: NAFTA Vote
’ This is compared to the restricted model with only a constant term in which 54% of the votes were correctly predicted.

 

33

With respect to the former category, these groups consisted mainly of
manufacturing industries competing with imports and unions identifying with
labor groups, while the latter consisted mainly of manufacturing groups dependent
upon export markets.

To test the assertions identifying the groups associated with unemployment,
an OLS was estimated, which examined the percent of unemployment within each
district UNEMPTPCT, as a function of the percent of manufacturing employment
MANU F, percent of union members UNIONTOTPCT, and manufacturing
employment associated with export industries MANUFEXP. The results reported
in Table 3 suggest support for the arguments developed above. First,
manufacturing employment and union membership within a congressional district
have a positive relationship with district unemployment. Moreover, they are both
statistically significant predictors. Secondly, manufacturing employment
associated with exports is negatively associated with district unemployment.
Likewise, this variable is also statistically significant. Thus, it is possible to
conclude support for the argument suggesting that specific groups are more
vulnerable to unemployment risk than others.

As such, it is possible to explicitly test propositions (1)-(3). Table 1.4
reports the results from the probit model. Recall that proposition (1) anticipates
that representatives with a large manufacturing interest group base will have a

negative propensity to vote for NAFTA. Indeed support is confirmed for this

 

34

MANU F is found to be both negative and statistically significant. Proposition (2)
anticipates the opposite effect for manufacturing groups associated with export
industries. Specifically, representatives from districts comprised of manufacturing
groups dependent upon exports markets are less likely to vote for NAFTA. The
results obtained from the probit estimation render support for this proposition as
the coefficient estimate for MANUFEXP is positive and statistically significant.
Proposition (3) suggests that representatives from districts with a large percentage
of labor union membership, will have a negative propensity to support NAFTA.
Indeed, the results from the analysis support this proposition as the coefficient
estimate for UNIONTOTPCT is negative and statistically significant. Proposition
(4) is the last hypothesis specifying the demand side of the theoretical model
examining factors responsible for constituency trade policy preferences. It
suggests that safety nets help groups or individuals in transition from one job to
another. Thus, the argument contends that members of Congress will have a
higher propensity to vote for NAFTA, if they have resources at their disposal to
give to groups or individuals vulnerable to adjustments (e.g. temporary
unemployment) resulting from the NAFTA. The estimates obtained from the
analysis give some confirmation to this proposition. The estimated coefficient for
SAFETY is positive, however, it is not statistically significant.

The remaining propositions (5)-(10) specify the supply side of the

theoretical model examining factors responsible for congressional responsiveness

35

to constituency preferences. Specifically, proposition (5) states that members who
receive a large percentage of their total campaign contributions from labor political
action committees will have a negative propensity to vote for NAFTA. Indeed,
support is rendered for this proposition as the coefficient estimate for LABTPCT is
negative and statistically significant. Propositions (6) and (7) attempt to ascertain
the effect that electoral vulnerability and experience within the House have on a
representative’s propensity to vote for the NAFTA. With respect to the latter
proposition, the argument developed in this research contends that the percent of
votes members received from their districts in the last election should be positively
related to their propensity to vote for NAFTA. As such, the model anticipates that
representatives who received lower percentages of support, will have a lower
propensity to vote for NAFTA, while members who received a higher percent of
support, will have a higher propensity to vote for NAFTA. Minor support is
rendered for this proposition as the coefficient estimate for VOTEPCT is positive,
but it is not statistically significant. Proposition (7) suggests that new members of
Congress will have a lower propensity to vote for NAFTA because of their
vulnerability stemming from lack of tenure in office. However, the coefficient
estimate for FRESHMEN, albeit negative, is not statistically significant. Thus,
only meager support is found for this proposition.

Proposition (8) examines the influence of presidential support in a

member’s district and what effect this subsequently has on the representative’ 3

 

 

36

propensity to support NAFTA. Again, because of the institutional bias towards
free trade that permeates the executive office contrasts greatly with Congress’s
protectionist bias, more often than not, preferences for trade policy will be
divergent between the two institutions. Thus, it is unlikely that presidential
support in a district will increase the propensity for a representative to vote for
NAFTA. Some minor support is rendered for this hypothesis as the estimated
coefficient estimate for PRESUPPORT is negative. However, it is not statistically
significant.

Proposition (9) examines the effect of partisan affiliation of a member and
its influence on the NAFTA vote. The proposition contends that Democrats are
more likely to vote against the NAFTA. Interestingly however, the estimated
coefficient on PARTY suggests the opposite effect and was statistically
significant. This results suggests that Democrats were more likely to vote for
NAFTA. What possibly accounts for this counterintuitive result, especially since a
majority of Democrats actually voted against NAFTA?"5 One answer lies in the
heterogeneous nature of congressional districts that Democratic members
represent. Research suggests that Republicans tend to represent more homogenous
districts, while Democrats represent more heterogenous districts (Rohde, 1994).
As a result, Democrats have more divergent interests to represent within their

respective districts, more so than their Republican counterparts. As such, using a

 

”Recall that the vote of NAFTA was split 234 for and 200 against. 40% of the vote in support of NAFTA
cam from the Democrats, while over 75% came from Republicans.

37

party control variable to predict propensity to support free trade may not be the
best means of testing what influence partisanship has on trade policy vote
outcomes. The model may need to be complicated in order to take into account the
heterogenous character of congressional districts and what effect this has not only
on congressional preferences, but on constituency preferences as well.

Lastly, proposition (10) seeks to determine what the ideological
predispositions of representatives on economic policy have on trade policy. The
proposition suggests that members with liberal ideological ratings will be more
likely to vote against NAFTA. Minor support was rendered to this proposition.

Specifically, IDEO is found to be negative, however it is statistically insignificant.

Conclusion

The general theoretical framework examining congressional responsiveness
to constituency demands received some support from the model analyzing the vote
on NAFTA presented in this research. Particularly, this research contributes novel
findings to three literatures: congressional decision making, majoritarian politics,
and the political economy of trade policy matters.

First, the analysis reveals that congressional responsiveness on trade policy,
seems to be less a function of political matters related to party, tenure in office,

etc., and more a function of the economic risk that dominant constituent groups

38

confront within a representative’s district. This suggests that economic factors
responsible for constituency demands, both protectionist and anti-protectionist,
have a significant effect on congressional decision making more than convential
political factors. This finding supports Kingdon’s (1989) and Amold’s (1990)
insights about the influence of major industrial groups and attentive publics on
congressional responsiveness. Specifically, as the argument concerning
constituency demands suggests, industry associated with manufacturing (related to
both imports and exports), labor and unions groups, are attentive to trade policy
debates. As such, these groups subsequently find representation of their demands,
as members respond to their economic risk conditions.

Secondly, to the degree that particular attentive groups tend to dominate the
process, the majoritarian politics premise gains support in the realm of trade policy
matters. However, this does not suggests that Congress is dominated by an
overwhelming protectionist bias. Again, attentive groups refer to both opponents
and proponents of free trade. The inclusion of controls for both protectionist and
anti-protectionist groups suggest that they each have a significant influence on
congressional decision making.

Third, the theoretical framework and empirical results support the
conclusion that the political economy of trade policy, to a large degree, evolves
around the potential economic risk to constituents vulnerable to unemployment.

Because groups associated with unions, labor, and manufacturing are vulnerable to

 

39

unemployment, they will seek protection. This supports Wallerstein’s (1987)
contention that unemployment has a direct effect on the demand for protection.

Concluding, the examination of the NAFTA vote yields some interesting
results that make future research in this area an important consideration. Three
particular topics merit attention, which have the potential to yield important
insights into the literature investigating the politics and process of trade policy
matters. First, because trade policy is an issue important to constituents’ interests,
as exhibited by the politics stemming from the NAFTA vote, it is important to
recognize it as a policy arena full of divergent interests which are likely to get
more polarized in the future. Because it is a salient issue riddled with competing
interests, the politics of trade policy formation are likely get more competitive. If
the United State’s dependence on export markets continues to grow, anti-
protectionist forces will continue to secure their policy preferences. However, if
current pleas by politicians in favor of isolationist policies gains increasing
support, advocates of protectionism may find their preferences represented in a
new era of US. trade policy strategy.

This leads to the second point of interest which is that the institutional
variance among the executive and Congress warrants further investigation. If
interest group preferences become more politicized, will institutions currently
charged with authority and decision-making on trade policy matters change in

order to accommodate efficient responsiveness to policy demands? What will

4O

happen to the United States general trade policy orientation if partisan politics
over trade policy matters heightens in the future? As parties become more
homogenous, as reflected in intraparty voting patterns in Congress (Rohde, 1991),
party organizations will serve as stronger cues for constituents to obtain
representation for their preferences on trade policy matters. Will a new era of
protection by demand ensue? An examination of these issues should provide some

answers to these important questions.

CHAPTER 2

MODELING THE DYNAMICS OF POWER, TRADE AND CONFLICT:
FLUCTUATIONS IN THE ECONOMY,
ALTERNATWG DISTRIBUTIONS OF POWER AND
THE PROSPECTS FOR PEACE

Introduction

Over the last several decades, political scientists have been intent on
examining whether recurring patterns exist among macro level economic (e. g.
prices, production, etc.) and social phenomena (e.g. international conflict). The
contention that long cycles of political and social phenomena exist over time in the
international system continues to evoke controversy and hence, remains an

enduring puzzle for international relations scholars.l

 

'It has been six decades since Nikolai Kondratieff (1935) first published his thoughts on the existence of
long waves in the international economy over time. Long waves (or more generally, Kondratieff cycles) are
defined by alternating economic phases. Generally speaking, they are comprised of an expansion phase and
a stagnation phase. Some of the literature refer to the former as an upswing and the latter as a downswing
(Goldstein 1985). The transition point from an expansion phase to a stagnation phase is called a peak, while
the point from a stagnation phase to an expansion is called a trough. The dates of the transition points are
used to define the historical dating of the long wave phases, that Kondratieff posited to repeat roughly every
fifty years. As Goldstein (1988) notes, the longwave phases occur across national boundaries. That is, they
are macroeconomic phenomena. Therefore they are systemic-level phenomena and should be modeled
accordingly.

41

42

Of particular interest to scholars in this research tradition has been the
examination of cycles related to both economic activity (Goldstein 1988) and
political order within the international system (Gilpin 1981, Wallerstein 1983, and
Modelski 1987a), and their association with the onset of major power war. During
the last decade, this literature has been engaged in debates over three central
issues: (1) empirically substantiating whether long cycles exist in the international
system; (2) determining the most appropriate cycle phase dating scheme; and (3)
establishing the precedence of economic long waves or political cycles as
generating mechanisms responsible for the onset of conflict in the international
system.2

The present research seeks to advance a single deductive framework that is
capable of addressing these controversies in this literature, while at the same time
broadening and synthesizing the theoretical and empirical contributions of scholars
in this research tradition. This is accomplished first, by integrating political cycle
and economic long wave theories into one theoretical framework, which explains
change in conflict within the international system.3 Similar to Pollins (1995), I
believe that these processes are best understood in an integrated framework.

However, to facilitate a broader understanding of the dynamics inherent within the

 

2Economic long waves consists of series relating to prices and inflation, production, etc., while political
cycles relate to series representing leadership, hegemony, the distribution of power, etc.

3Pollins (1995) makes a similar argument to develop his co-evolving systems model. The present work
seeks to build upon Pollin’s theoretical integration, but is distinct in that the framework advanced here is a
more general model seeking to explain changes in the trend associated with armed conflicts over time
within the international system. This argument will be fully developed in a subsequent section.

43

international system, I contend that the distribution of power is an integral part of a
general theory capable of delineating the mechanisms responsible for generating
change in systemic conflict.4

Secondly, by using new theory in econometrics on cointegration, the model
developed in the present research seeks to redirect the current preoccupation with
determining the existence of cycles, duration, and the timing of their phases, into a
broader conceptualization. Specifically, the conceptualization considers whether
these phenomena are characterized by stochastic trends, and if the trends
associated with each of these factors may be in common over time. As I will
demonstrate, the proposed shift in focus affords a more accurate conceptualization
about the interrelationship between the processes over time. Moreover, it yields a
new unifying insight into the controversies concerning the precise dating of cycle
phases and whether the existence of cycles can be empirically substantiated. The
model itself has broader applicability: it has the potential to incorporate other
systemic factors, both economic and political, in the analysis of international
conflict or in studies seeking to explain system transformation.

To facilitate an examination of the proposed framework, it is necessary to
briefly return to an discussion of the enduring controversies within this literature.

The discussion highlights the main issues in dispute within the literature and

 

‘In this research, the analysis of conflict extends beyond major power war. This is done for reasons that
will be elaborated upon in the development of the theoretical model. In short, to depict the overall level of
hostility in the international system, conflictual events short of war need to be taken into account. This will
aid in the development of a general indicator reflecting overall systemic conflict. Pollins (1995) makes a
similar argument in his co—evolving systems model.

44

emphasizes how the framework advanced in this research can address these issues.
Specifically, I contend that the theoretical and empirical framework advanced in
this research resolves these disputes. In addition, the framework integrates and
extends the numerous contributions of scholars in this research tradition. The
discussion briefly highlights both theoretical and empirical recommendations that
are advanced in this study, while development of the full arguments for each
critique are developed in the appropriate sections.

Theoretically, this literature has traditionally not integrated the analysis of
economic and political cycles into a unified causal theory capable of explaining the
onset of systemic conflict in the international system. This has negated the ability
to specify and examine explicitly the causal mechanisms responsible for
generating change in system conflict. Recent work by Pollins (1995) has sought to
address this issue. Examining global leadership cycles and economic long waves,
Pollins contends that there exists a tendency in traditional research to isolate the
effects of these phenomena when examining how they are correlated with the
occurrence of conflict.

Specifically, Pollins argues that more often than not, establishing the
precedence of economic long waves, or leadership cycles, as generating
mechanisms responsible for the onset of conflict in the international system, has

taken priority over developing an integrative approach to the study of these

45

phenomena and their interrelationship with armed conflict.5 To address this void,
Pollins (1995) develops a system-level model, which views the incidence of armed
conflict as influenced by the interaction, or co—evolution, of economic long waves
and leadership cycles. However, he, along with traditional research in this area,
does not examine how these processes are interrelated, nor their temporal
relationship with systemic conflict.6 The research in this paper demonstrates that a
true “co-evolving” argument necessitates the development of theory capable of
delineating how the processes under consideration are connected. Thus, a co-
evolutionary theory involves (1) determining the independent effects of cyclical
phenomena upon one another, and the possibility of reciprocal relationships (i.e.
feedback) upon one another over time; (2) contemplating what structural changes
are fundamental to generating change in the international system; and (3)

rigorously delineating the temporal ordering of the processes characterizing the

 

’For example, Modelski’s (1987a) work on leadership suggests that both political and economic structures
comprise distinct realms, and as such should be analyzed separately. Moreover, recent work by Modelski &
Thompson (1995) examines leading sectors and global power as separate, independent, co-evolving
systems. Pollins (1995) discusses this latter point to demonstrate a conceptual shift in his model from
traditional long cycles research. He is one of the first scholars to promote an integrative approach.

“Again, this is a general criticism warranted to the cycles literature. Developing a theory capable of
explaining the interrelationship between economic waves and leadership is not the purpose of Pollins
research. Specifically, he states:

“While these processes themselves may well be inter-related in ways yet to be specified, I
contend that each makes its own contribution to the generation or suppression of armed
conflict, in keeping with the particular phase each cycle is in at a particular time.”
(Pollins, 1995, p.9)

As such, the co-evolving systems model that Pollins promotes does not explain how these processes
influence one another (i.e. how they are integrated) over time. Pollins notion of integration is a mechanism
in which to match cycle phases associated with leadership cycles and economic long waves. This is done to
see how the joint ranking of these cycles are ordered to determine the cumulative effect that they have on
the likelihood of armed conflict.

46

system and their relationship with systemic conflict. This necessitates an
examination of exogenous assumptions made within a co-evolving theoretical
model.

In this research, a theory is developed that addresses each of these issues.
The theory contends that economic activity and commerce over time, tend to
impact the concentration of power among states in the international system. When
this occurs, the configuration of states in the system is altered. This subsequently
changes the international structure. To the extent that this change produces
uncertainty about the security of states in the new system structure, then the onset
of systemic conflict will also be affected.7

This theoretical framework affords several innovative extensions into this
literature. First, the model integrates the influence of the distribution of power and
international commerce into this literature. All too often the influence of trade is
overlooked as a determinant in the nature and magnitude of systemic conflict.8
Currently studies that examine the influence of interstate commercial activity on
conflict yield mixed conclusions. In addition, this framework specifically

delineates the exogeneity of armed conflict, to the endogenous processes

 

7Again, note the theoretical argument developed in this research uses a more general measure of conflict
similar to Pollins (1995). Because the analysis of major-power war is too restrictive for the purposes of the
following research, conflicts short of war are also analyzed. This will be further elaborated upon in the
development of the model. Note, because this research uses a more general measure of conflict, it is
inappropriate to directly compare the analysis in this study with most of the traditional cycles research that
analyses major power war.

“This is especially true in studies investigating systemic level behavior. Notable exceptions are the work of
Gowa (1989, 1994), Gowa & Mansfield (1993), and Mansfield (1994).

47

responsible for generating changes. These endogenous processes are the
distribution of power, commerce and economic activity. These exogeneity
assumptions posited can subsequently be tested in the empirical model developed
in this research.

Empirically, the methodological approach advanced in this research, not
only has the potential to resolve existing debates over empirically substantiating
the existence of cycles, their duration, and the determination of their dating
schemes, but can advance a common framework that is capable of (1) empirically
demonstrating how cyclical phenomena co-evolve, by using new theory in
econometrics on cointegration, to see if these phenomena share a common trend
over time; and (2) assess what phenomena are formally generating dynamics in the
international system by conducting exogeneity tests within a cointegrating
framework. Briefly each of these points will be discussed to delineate the
particular controversy in the literature and how the proposed research seeks to
address the issue.

As previously mentioned, one of the central interest to scholars in this
literature is empirically substantiating whether long cycles exist in the international
system.9 That is, proponents maintaining the existence of cycles, find that over
time, systemic phenomena (e. g. leadership, economic waves, etc.) tend to rise and

decline over time in repeated patterns, and that particular sub-periods tend to be

 

’For a discussion of recent literature on this issue see Beck (1991b), Goldstein (1991), Williams (1992) and
Williams et al. (1994).

48

correlated with the onset of major power war. However, establishing whether
cycles exist has been bedeviled by variances in the series under investigation, the
time frames under analysis, and the techniques used to verify their (non)existence.
Moreover, a related issue in dispute pertains to the dating schemes that depict the
timing and duration of various cyclical phenomena.lo Most studies examining long
cycles have tried to fit a common cyclical pattern among the series of interest, in
an attempt to establish whether cycles exist and to establish their duration.ll
However, cycle dating schemes typically encounter criticism because they are
considered subjective. Moreover, any nonmonotonic series (e. g. a series which
drifts in one direction then changes and drifts in the opposite direction) has
cyclical properties that are randomly distributed throughout time, rendering
substantive significance to cyclical periods tenuous at best.12

The frameWOrk utilized in this research goes beyond substantiating the
(non)existence of cycles, by examining the relationship among the trends
associated with cyclical phenomena (e.g. economic waves, power concentrations,
systemic conflict, etc.). If these trends are co-evolving, then their empirical
determinants will suggest that they share a common trend over time. The focus on

common trends negates the issue of empirically substantiating cycles, in favor of

 

loGoldstein (1988), Modelski & Thompson (1988), Pollins (1995), Thompson (1992), Wallerstein (1983)
all specify dating schemes for various cyclical phenomena.

"Imposing a cyclical structure on the series of interest, is considered a restrictive assumption about the
functional form of the data.

”Pudaite (1991) and Beck (1991b) make similar arguments in their analysis of the long cycles literature.

49

deriving some theoretical expectations for why and how these series are related to
one another over time. This conceptualization extends a beneficial insight as it
affords an unrestrictive specification about the relationship between these
constructs over time. Previous attempts to specify periodicities have produced
conflicting conclusions in the literature, resulting in endless debates over the
empirical existence of cycles.

Another empirical criticism in this literature is the lack of attention to the
assessment of exogeneity assumptions in most of the long cycles literature.
Exogeneity is fundamental in the determination of causality between two or more
variables. Typically most of the literature ignores this issue, both theoretically and
empirically.” The purported research elucidates a theoretical model and empirical
test capable of assessing causal mechanisms between conflict and economic
phenomena in the international system. Error correction models in general, and
cointegration in particular, imply causality. Generally speaking, for a pair of series
to have a long-run equilibrium relationship, by definition there must be some form
of causation between them. In other words, the series need to have the necessary

dynamics to form a long-run relationship to begin with (Granger, 1988a). This

 

”Goldstein (1988) is one of the few scholars that has addressed the issue of causation. Included in his
empirical analysis is a series of Granger causality tests. However, as will be discussed in a subsequent
section, Granger causality tests are not adequate tests for causation, especially in the context of a
hypothesized cointegrating relationship (Granato & Smith 1994a, 1994b).

 

COT

and

I‘DlSc
c311$th
( 1983.,-
Ericsst
Suture
”“7 felt
in the d
013’ Pa
and hen

50

research begins to explore the necessary dynamics and restrictions necessary for
cointegration in a partial system (Johansen 1992a, 1992b).”

By taking the proposed theoretical and empirical approach, this research
yields insights into several enduring puzzles in the long cycles literature and
illuminates a more general picture of the relationship between power, trade,
economic activity, and systemic conflict. Drawing upon several related, albeit
distinct literatures, this paper advances a more general theory and test in support of
the significance of economic waves and political cycles in explaining conflict in
the international system. This, in turn, results in the development of an empirical
model, which is tested, and subsequently finds support for a set of theoretical
propositions. Specifically, the empirical analysis finds that the distribution of
power, conflict, trade and economic waves in the system, are independently
characterized by stochastic trends. Moreover, when these trends are analyzed
together, empirical evidence suggests that they share a common path over time.
This leads to the development of another set of tests, which demonstrate that
conflict in the system alone responds to changes in the distribution of power, trade

and economic waves.

 

"Discussions of the issues involved in assessing causality and diagnostics tests appropriate to determine
causality within a cointegration framework have only surfaced in the last several years. See Granger
(1988a, 1988b), Johansen (1988, l992a, l992b), Granger & Lee (1989), Johansen & Juselius (1990). and
Ericsson (1992) for a discussion in economics and Ostrom & Smith (1992) and Durr (1992) in the political
science literature. Generally speaking, cointegration implies symmetry between any series forming a long-
run relationship, complicating for exogeneity conditions within a statistical model. It will be demonstrated
in the development of the statistical model proposed that under certain restrictions, it is possible to assert
that particular series may have asymmetric effects on an equilibrium relationship (i.e. cointegrating vector),
and hence, may be deemed weakly exogenous to the dependent process under analysis.

51

The general framework developed in this analysis, should prove sufficient
to explore other types of social phenomena related to power, trade, and war in the
international system. However, before discussing the theoretical framework
advanced in this research, a brief discussion about the current literature will
precede in order to demonstrate where existing research stands and where the

present research departs and seeks to make a contribution.

Long Cycles in Review

Long Cycles are conceived by its proponents as a systemic theoretical
explanation for the recurrence of war over the last several centuries. It is not a
theory about the causes of major war per se. Instead, it is based on the notion that
behavior in the global political system has a time pattern or rhythmic
characterization.ls Within political science, this literature is most notably
associated with the work of Modelski (1978, 1981, 1987a, 1987b), Thompson
(1983, 1988) and Goldstein (1985,1987, 1988).16 Most of this research extends the

initial insights of Kondratieff (1935) in an attempt to uncover the interaction

 

I’Pcrhaps it is the focus on the empirical attributes of these series that render long cycle critics some validity
in their assessment of this literature as atheoretical.

"This list of research is by no means exhaustive. For extensive annotated bibliographies see Barr (1979),
Goldstein (1988), Modelski (1983, 1987a), Rosecrance (1987) and Thompson (1988).

52

between periodic major wars in the global system and changes in the behavior of
prices and production indexes in the world economy.'7

The origin of the Long Cycles literature has been largely attributed to the
original investigations of Kondratieff (1935). His research consisted of examining
numerous price, production and consumption series for several advanced industrial
nations, in an effort to determine whether long waves of economic phenomena
exist and what time periods they might possibly comprise. In his analysis, he finds
some support for the existence of longwaves and concludes that approximately
three waves exist which can be traced back to the late 1780's.18 Interestingly,
Kondratieff surmised that long waves were behavioral regularities in the
international system that could be generalized to other economic and social
phenomena, such as conflict. Specifically, he contended that rising phases of long
waves were correlated with high tension, which subsequently increased the
probability of conflict. However, he didn't offer any theoretical arguments for his
assertions, other than suggesting that rising phases of the long wave tended to be

characterized by high levels of interstate tension over markets and raw materials.

 

I”The analysis of long cycles and political phenomena has been organized into several theoretical
frameworks. The leadership cycle (Modelski 1978) and the power transition school (Organski 1958,
Organski & Kugler 1980) and the world-system school (W allerstein, 1974). All of these studies examine
war in a slightly different manner. Specifically, leadership studies focus on the effect of war in establishing
new world orders, while the power transition studies focus on changes in power and it’s effect on patterns of
war over time. The world systems perspective examines the problem of war and hegemony through a
Marxists perspective.

"A wave (or cycle more generally) is comprised of an upswing and a downswing. Kondratieffs cycles date
roughly: ( 1) 1830-1860 (up), 1860—1890 (down); (2)1890-1914 (up), 1914—1940 (down); (3)1940-1968
(up), l968-to be determined (down). Kondratieff used moving averages and trend deviations to analyze on
wholesale prices, interest rates, wages, foreign trade and production data series from Britain, France,
Germany and the United States.

53

Thus his work is perhaps the first theoretical link in this literature that examines
cycles of macroeconomic conditions and conflict behavior in the international
system.19 However, Kondratieff did not make any explicit attempt to deal with
exogeneity issues other than to suggest that recurring events were the efiects of
long waves and not responsible for causing them.20

In the late 70's and early 80's Modelski began publishing some of his
research on long cycles. This work is largely attributed to begin the new upswing
on long cycles research in political science. In his work, Modelski is the first
scholar to extend Kondratieffs notion that behaviors occur in repeating patterns in
order to uncover recuning phenomena related to global leadership. He
hypothesized that long-term fluctuations in political and economic systems within
the international system are characterized by similar processes that are correlated
in some manner. As such, an increase in the demand for resources and capital in
one system, would subsequently decrease the supply available to the other

1

system.2 Moreover, this process is cyclical and can be characterized by four

 

”Although Kondratieff is largely characterized as the original scholar to link economic and war cycles,
Goldstein (1988) cites Johan Akerman (1932), a Swedish economist, as the first researcher to associate war
as the central cause of long waves.

20Here Kondratieff seems to make an explicit exogeneity assumption. That is, he draws a causal distinction
between long waves and other phenomena. Specifically, he suggests that long waves contribute explanatory
or predictive power for studying recurring phenomena (e. g. war) in the international system. However he
never explicitly addresses the issue.

2|Modelski conceives the international system to be comprised by distinct realms, Le. a political system and
an economic system. However, he seems to suggest that acquiring resources in these distinct realms is a
zero sum process. This may not adequately capture the dynamics of the interrelationship between the
political and economic systems. Especially if one takes the view that they maintain some long-run
equilibrium relationship as the current research contends.

54

distinct phases (1) global war; (2) the emergence of the world power; (3)
delegitimation of power; and (4) deconcenetration of power. The occurrence of
another global war starts a new cycle. Modelski (1981) also models the
association between world power phases and Kondratieff's economic waves.
However, he does not explain the explicit link between economic and power
cycles. Specifically, he does not address issues associated with causality. In fact,
most of this research virtually ignores exogeneity assumptions about what is
driving these cyclical processes over time. In some sense, Thompson & Zuk's
(1982) research is one of the first attempts to causally link cycles of conflict and
economic phenomena together by examining the timing of wars and the timing of
upswings and downswings of Kondratieff waves.

By examining the timing of Kondratieff price waves, Thompson & Zuk
(1982) seek to determine whether war inaugurates or reinforces upswings in
Kondratieff waves and whether war inaugurates the down swing of the long wave.
Using various descriptive statistics, they find little support for the notion that war
inaugurates upswings, but they do find evidence that Kondratieffs upper turning
points coincide with the termination of war.22 They conclude that there "can be
little doubt that wars and their associated economic disruptions lead to price level

increases " (1982, p.635). In order to test the extent to which wars are responsible

 

22Specifically they examine wholesale prices at war peaks and calculate a five year mean preceding the war
peak, and its corresponding percentage increase. Moreover, they examine price levels at different turning
points of the Kondratieff, by analyzing data on upswing annual percentage changes and the prewar annual
percentage changes.

55

for inflation, they use an ARIMA impact assessment model (Box & Tiao, 1975).
From their analysis they find significant increases in inflation during war, followed
by a period of readjustment that lasts about twenty years. They conclude by
stressing the need to study the long term impacts of wars because of its
fundamental impact on political and economic systemic processes. This research
is valuable in its assessment of war having an long term effect on aggregate
economic conditions. However this research stops short in making any explicit
exogeneity assumptions. Instead, Thompson & Zuk (1982) suggest that their
approach is similar to Modelski's long cycle of world leadership. Specifically,
they contend "that Modelski's argument that global wars and Kondratieff long
waves reflect a common underlying process-alternating changes in global order
priorities-offers an attractive way out of the causality corner" (p.640, 1982).23
However, I contend that the notion of "process-alternating" implies a distinct
causal, albeit nonconstant, relationship between economic and conflict phenomena
over time. The model purported in this research is able to capture the long term
relationship of war and economic phenomena as a nonconstant relationship and to

address the causal assumptions implied by such a specification.

 

23Emphasis added. However, I contend that they offer more causal anecdotes than they give themselves
credit for or at least more than they are ready to acknowledge. This fact is illustrated in some of their
writing. For example, when they comment, "...we should not forget that major wars have reinforced rather
than initiated the price upswing. It is quite conceivable, then, that changes in global-order priorities
precede the advent of global wars and are therefore responsible for the upswings" (emphases added)
(Thompson & Zuk, 1982).

 

CC

EX:

”10d:

56

Goldstein's (1988) Long Cycles, is the most extensive, thorough, historical
accounting of the long cycles literature in both econorrrics and political science.
He argues "cyclic time" varies from one period to another ( i.e. from one cycle to
another) and that strict periodicity is less important than the realization of a
repetition of behavior over time. As such, Goldstein does not think it necessary to
constrain his analysis of cycles to a "fix" period. He suggests that fixed period
cycles are not realistic and therefore develops a "base dating scheme" to allow
more flexibility in his analysis for cycles.“ In order to address the issue of
exogeniety, Goldstein conducts a series of granger causality tests to assess which
series may be causing the other.25 In his analysis he finds support for the notion
that war leads to inflation. Although Goldstein does make one of the first explicit
attempts to empirically verify exogeneity assumptions, there are several issues
complicating his analysis. Foremost is that granger causality tests are not useful in
determining "causation" per se. They are useful only in determining what
variables perform best in their predictive capabilities (Granato & Smith, 1994a,
1994b).26 Moreover, although Goldstein renders valuable insights into the
correlates of various economic and conflict series over time, this does not

explicitly develop a general model per se.

 

2‘ His base dating scheme is essentially a series of cyclic up and downs (of varying periods) that have been
spliced together from the works of different scholars examining long cycles.

2’See Freeman (1983) for an excellent review of granger causality tests applied to political statistical
models.

7“"These issues will be elaborated upon in a subsequent section dealing with cointegration and causality.

57

After the original work of Modelski, Thompson and Goldstein, subsequent
research has become engulfed in analyzing the periodicity of cycles. However,
most proponents of longwaves do not theoretically purport the notion of "strict"
periodicity. This misnomer, i.e. the notion that there can be variance in the
duration of cycles, has compelled critics, such as Beck (1991b) to suggest that
"[t] here is a working oxymoron in the notion of a 'fluctuating' cycle," and to
empirically substantiate the existence of cycles. Beck does so by utilizing spectral
analysis.27 As Beck's interest is primarily methodological, he examines whether
there are long cycles in war severity, and whether those cycles are related to
inflation. His main contention is that fixed periodicity methods, particularly
spectral analysis, are the appropriate tools in which to empirically verify the
existence of long wave relationships. As such, he takes issue with Goldstein's
"base dating scheme" methodological approach, which is primarily descriptive.
His arguments and evidence are compelling. He finds no evidence of cyclic
behavior in any of the series. Interesting however, is the fact that he does find
some support for the notion that there exists an interrelationship between the
variables. However, he stresses that this interrelationship has nothing to do with
cyclicity per se.

Nevertheless, a few scholars have posited that some series associated with

systemic behavior are characterized by irregular periodicities and therefore reflect

 

27To some extent this is not a novel enterprise. Adelman (1965) is one of the first to conceive a test of long
cycles using spectral analysis.

58

nonlinear dynamics, which when modeled appropriately have cyclical properties.
These models find support for the existence of cycles, even though the
periodicities and the duration of cycles may vary throughout time. Acknowledging
that there is disagreement in the literature on the precise nature of the periodicity
of cycles, Sayrs (1993) specifies a nonlinear model of the war cycle where
upswings and downswings do not occur at regular intervals. Utilizing a markov
specification she finds evidence of cycles.28

Most recently, several multivariate models have been specified to analyze
the confluence of various political and economic series over time. However, these
studies do not make explicit exogeneity assumptions and theoretical assumptions
about the relationships between the variables under analysis. Specifically,
Goldstein (1991) utilizing vector autoregression analysis (VAR) finds that periods
of severe war follow phases of robust economic growth, which subsequently lead
to phases of stagnant growth.29 Moreover, he finds that wars cause long waves in
prices and inflation, which tend to subside in the downturn phase of the war
cycle.3o However, Williams (1992) utilizing Bayesian vector autoregression

(BVAR), an approach he believes is superior, seeks to test a model similar to

 

2“Richards (1993) similarly uses a nonlinear estimation technique to determine the existence of cycles in
systemic power concentrations over time. She finds evidence for irregular chaotic cyclical dynamics in
concentration patterns in the international system. The issue of power concentrations will be discussed in
future research.

2"VAR is a multiple time series generalization of an autoregressive model (Maddala 1992). See Sims
(1980) and Freeman et al. (1989) for an in-depth discussion of this type of time series modeling approach.

3° In addition, Goldstein (1988) concludes from his analysis that production waves lead war and price
waves.

59

Goldstein’s. Specifically, Williams contends that there is not sufficient theoretical
justification to impose restrictions on empirical models of long cycles.3| Although
he acknowledges that VAR is an approach that is capable to overcome
specification problems in time series models, it has methodological problems.
Particularly he notes that VAR's are unparsimonious and potentially unstable
through time.32 In his analysis, Williams, using BVAR, finds little support for the
Goldstein's model, and hence, long cycle theory.

An examination of the long cycles literature suggests that several deficits
exist, which need to be addressed, in the development of a general theory of that is
capable of examining the dynamic interrelationship between the economy and war.

First, it is crucial to develop a theoretical foundation. Distinct correlations have
been uncovered which describe the general patterns of conflict and economic
series over time. However, discerning how these various stories fit together to
comprise a theory about the pattern of political and economic phenomena over
time remains unclear. Most of long cycles research does find some correlation
between various economic and political series. The question then becomes, what
drives the patterns observed in the international system? How are they related to

changes in the distribution of power in the international system? How does

 

3'BVAR analysis essentially requires that prior distributional assumptions be adopted prior to estimation.
Technically this involves specifying a prior distribution for the coefficients in a VAR. Generally, in each
equation comprising the (VAR) system of equations, the coefficient of its own lagged variable has a prior
mean of one, while all other variables have prior means of zero. Moreover, the variance of the prior is
specified to decrease as the lag length increases. For a thorough description of this approach in political
science see Williams (1992), and in economics see Doan et a1. (1984) and Littennan (1986).

32See Pagan (1987) for a comparison of a VAR model to other types of time series models.

60

international trade fit into this picture? Answer to these questions will help
provide direction towards a general explanation for systemic conflict, power, the
economy and trade that is capable of examining, not only what factors are
responsible for changing the structure of the international system, but is also able
to specify exactly how change occurs and the repercussions this has for the timing
of conflict.

Second, and integral to accomplishing the first task, is addressing the issue
of exogeneity. Exogeneity assumptions need to be examined for both theoretical
and empirical reasons. Theoretically, exogenous conditions are explicit to any
theoretical model seeking to explain what drives change in the system, in addition
to the timing of these changes. Empirically, exogeneity assumptions are necessary
when specifying a causal multivariate model. The model purported in this
research suggests that the distribution of power and systemic conflict share an
equilibrium relationship. Specifically, changes in systemic conflict can be
understood as both (1) shifting international trade and economic conditions in the
system and (2) reactions to changes in the distribution of power. Empirically this
involves testing for cointegration among the series of interest and subsequently
specifying an appropriate error correction model. Within this type of framework,
causality is assumed and will be examined to determine if proposed exogeneity
assumptions are valid conceptually and empirically. Using theory, a valid

empirical specification can be derived that will render an appropriate test of the

61

interrelationship among economic and conflict phenomena in the international
system. However, before the model is delineated, it is first necessary to
disseminate the dynamics of systemic conflict that are captured in the proposed

theoretical model.

Explaining the Dynamics of Systemic Conflict

This research purports a theoretical explanation that attempts to develop a
more general understanding of the interrelationship between systemic conflict,
power, the economy and trade over time. In order to do so, it is essential to move
beyond the question of whether cycles exist. At some level, the more appropriate
question to investigate is why they may or may not exist. It seems clear from the
previous discussion of the long cycles literature, both economic phenomena and
systemic conflict do not have fixed, conjoint, cyclical periodicities. In the strictest
technical sense then, long cycles do not exist. However, the more interesting
conclusion is somehow these phenomena have demonstrated some
interrelationship. To disserrrinate the dynamics of systemic conflict, the individual

effects of the economy, trade, and the distribution of power will first be explored.

62

The International Economy

Generally speaking, aggregate prices in the international system have
tended to follow similar patterns of systemic conflict over time. That is, the rise
and fall of prices correspond with the rise and decline of conflict in the
international system. However, what theoretical reasons explain their alleged
correlation over time? Initially, the effects of war on prices is fairly indisputable.
The outbreak of war is associated with aggregate shocks in the supply and demand
for goods. As a result, prices tend to rise when a shortage of goods and services
cannot meet demands in the market place (Goldstein 1988, Beck 1991b).
However, the notion that prices lead to war, in a causal sense, is not so clear.
Although prices change when supply and demand shocks occur during a conflict, it
is quite another issue to suggest that price changes are responsible for causing
conflicts.33

However, once conflict has occurred, price changes tend to reinforce
conflict in the system. That is, sudden changes in the supply and demand for
goods can promote tensions among states. Moreover, the escalation of prices tend

to reinforce tensions as inflation makes it more difficult for nation-states to secure

 

3"The explanation developed in this research does not purport an explanation that suggests the relationship
between prices and conflict is symmetrical, because the argument for prices causing war is tenuous at best
(Goldstein 1988, and Beck 1991b). Thus to some degree, the relationship between prices and systemic
conflict may be asymmetrical. This technical issue will be used in the development of an argument that
suggests prices should be treated weakly exogenous to systemic conflict. If this assertion proves true, then
the most appropriate specification of the long-run relationship hypothesized may be cointegration in a
partial system.

63

the goods and services they need.“ Thus changes in the supply of goods and the
rise in inflation that can ensue may produce tensions, which may serve to reinforce
systemic conflict. Thompson & Zuk (1982) find support for the notion that prices
tend to reinforce the upswing in conflict (war). However, they also argue that
prices are not solely responsible for causing systemic conflict. Moreover, another
reason for this correlation, although not widely discussed is that governments tend
to finance wars, not only through loans, but also through fiat money (Sargent
1986). Fiat money is essentially money borrowed from domestic monetary
authorities. Thus, when the money supply increases, a subsequent rise may occur
in inflation. Thus, to the extent that the onset of conflict is responsible for
shocking normal market conditions, then the anticipated relationship between
systemic conflict and inflation is positive. Thus, it is plausible to anticipate that
inflation and systemic conflict will generally follow similar patterns over time.
Figure 2.1 plots systemic conflict and inflation in the world market over time.

Table 2.1 list the variables used in the analysis and their respective sources.

 

3‘Because inflation is the most common indicator used to study the economy, it will be used in the analysis
that follows. It is simply the change in prices from one period to another, i.e. (Prices,) - (PricesH).

0.2

~-0.1

-0.2

 

 

 

1960

1 940

 

 

1880 1 900 1 920

1860

lllllllllllllllellllIIIIIIIT[IllIIIITIIIIIIIT‘IITI‘ITIIilllllll

 

 

60

0

Systemic Conflict ------ Inflatim

 

I

Figure 2.1 Relationship Between Systemic Conflict and Inflation

 

65

 

 

Table 2.1
Variable Descriptions
Variable Name Source
Systemic Conflict Gochman & Maoz (1984). Militarized Interstate Dispute Data.

Distribution of Power

Trade

Inflation

Revised under DDIR by Stewart Bremer. Binghamton University.
Total number of MlD's recorded for each annual observation. MID's
encompasses all subcategories of disputes included in the data set.
1850-1976.

Modelski & Thompson (1988), Sea Power in Global Politics. bong
cycle concentration index value for each annual observation. 1850-
I976.

Kuczynski (1980). The ratio of exports to total world production (in
constant 1913 dollars) for each annual observation. l850-l976.

Mitchell (1980). The first difference of the aggregate British
wholesale price index for each annual observation. Common log
value used for data analysis. 1850—1976.

 

 

 

66

Generally speaking, Figure 2.1 demonstrates the anticipated pattern. That
is, when inflation rises, systemic conflict tends to increase in tandem. Conversely,
when inflation decreases, systemic conflict declines.35 As such, both processes
will not drift far apart from one another for a long period of time, thus establishing
the theoretical notion that these two processes maintain a positive long-run

relationship. This will be formally tested with the following proposition:

Proposition (1): Inflation in the international economy is positively related to
levels of systemic conflict in the international system.

This implies that as systemic conflict increases, aggregate economic
conditions conducive to hostility will increase, over time, in proportion to the
extent of the change. Similarly, as systemic conflict decreases, aggregate
economic conditions conducive to peace will tend to increase. However, what
influence does interstate commence have on conflict in the international system?

It is to this topic that I now turn.

International Trade
The relationship between international commerce and conflict has been

examined extensively. There are four basic theoretical orientations which examine

 

3’ The positive association between these two variables has been verified in several studies (Beck 1991b,
Goldstein 1988, and Mansfield 1994).

67

the influence of trade on interstate relations.36 The first is generally referred to as
the commercial liberal perspective. Its roots extend into neo-classical assumptions
about the benefits associated with free trade. With respect to interstate
relationships, it suggests that trade promotes interactions and dependencies among
states in the international system. To the extent that such interactions are mutually
rewarding and not hostile, then trade will tend to promote peace and cooperation
among nation states. This perspective stands in stark contrast to classical realism.
Classical realists suggest that the frequency of conflict (war specifically), is not
influenced by trade in any systematic way. Only the distribution of power
influences peace and conflict in the system. Similarly, structural realism contends
that wars occur when trade levels are relatively low as well as when they are
relatively high. The effects of trade on war in this perspective, merely depend on
the structure of the international system. Political-military relationships are more
important than commercial relations. However, mercantilists (e.g. economic
nationalists) argue that trade is an important determinant of conflict in the
international system. However, contrary to commercial liberalists, interstate trade
promotes hostility instead of peace because of the potential gains from trade that
can accrue to a state's enemies.

All of these perspectives take a different stance on the benefits of

international commerce and its ability to stimulate or hinder peace and conflict in

 

3"See Mansfield (1994) for an excellent, thorough discussion of these four different perspectives.

68

the international system. Empirical, studies examining which of these perspectives
is correct are inconclusive. The theoretical model developed in this research
contends that trade is an important factor to consider in the study of the causes of
conflict. Because trade is partially responsible for the rise in wealth and growth of
nations, its role in changing the distribution of power in the system can not be
neglected. Thus, classical realism and structural realism are inadequate, as they
fail to take into account the significance of trade. Moreover, trade seems to be

t.37 The amount of trade continues to

negatively associated with systemic conflic
increase over time without a corresponding increase in conflict. Thus, the
mercantilist perspective seems to be inadequate in its explanation of trade affect on
interstate relations.

As such, to the extent that any conflict serves to deter the free flow of goods
and services, the anticipated relationship between systemic conflict and trade is
posited to be negative. Thus, contemplating the long term systemic effects of
commerce on conflict lends support to the commercial liberal perspective and
hence, the notion that these two variables have an inverse relationship. Figure 2.2
below demonstrates this relationship. This relationship will be formally tested

with the following proposition:

Proposition (2): Aggregate levels of trade are inversely related to levels of
systemic conflict in the international system.

 

37This relationship has also been substantiated by other research. See Mansfield (1994).

69

60

0

 

 

 

1960

1940

\‘I
\\

1900 1920

1880

1860

lllllllllllllllllIlllll[[TTTI[llllllllllllll]lIll]llll]Illl|ll

'\

 

 

400

350—

300—

250—

100 "

Systemic Conflict I

Trade

 

1

Figure 2.2 Relationship Between Trade and Systemic Conflict

 

70

The Distribution of Power

The relationship between the distribution of power and systemic conflict
has been debated extensively by proponents of balance-of-power theory and power
preponderance alike. Balance-of-power theorists traditionally believe that as the
distribution of power becomes increasingly dispersed, the probability of war will
subsequently increase. This stands in stark contrast to power preponderance
theorists who posit the opposite relationship between the distribution of power and
war. That is, as the distribution of power becomes progressively more
homogeneous (e.g. power equality/symmetry among states), the probability of war
will increase. However, the evidence in support of these theories is inconclusive at
best.

Most recently Mansfield (1994) has posited, a compelling argument about
the relationship between the distribution of power and war.38 He suggests that the
relationship between the distribution of power and the incidence of war is
quadratic, thus suggesting that both, the highest and the lowest levels of
concentration decrease the probability of conflict, while intermediate levels of
concentration increase the probability of conflict.

However, most of these studies only consider the case of major power war.
Again, because war is such a rare occurrence, when considering the relationship

between the distribution of power and conflict in the system, the analysis of

 

3"Note however that Mansfield’s (1994) research is specific to the analysis of major-power war. As such,
his notion of conflict varies from the present study.

71

incidences of war among major powers poses a selection bias problem. Again,
consideration of warfare among other states, other than major powers, needs to be
considered, in addition to lower levels of conflict. Thus, when power becomes
concentrated in the system, I anticipate that states are more likely to engage in
conflictual behavior. Indeed this relationship is borne out in Figure 2.3 below. As
power concentrations rise in the international system, a corresponding rise in the
level of systemic conflict ensues. Likewise, when power concentrations decrease,
so does the level of systemic conflict. This will be formally tested with the

following proposition:

Proposition (3): The distribution of power concentration in the international
system has a positive relationship with the level of systemic
conflict.

However, how do systemic conflict, the distribution of power, trade and the
economy fit into a general theory? This will be elaborated upon in the next

section.

72

60

-50
-40
-30
t20
o

 

 

 

 

rill]Irrrlirrrlrrii]rrii]fili[irrrlir1111rrllrlrrjfirrrlir111fi

 

 

1.0

 

1880 1900 1920 1940 1960

1860

 

 

Distribution of Power ------ Systemic Conflict

 

 

Figure 2.3 Relationship Between Distribution of Power and Systemic Conflict

73

Towards a General Co-Evolving Systemic Equilibrium Hypothesis

Initially, I surmise that observations in the international system are
analogous to aggregate behavioral outcomes from the states comprising the

39 Moreover the system is characterized by aggregate economic conditions

system.
that are used to evaluate decisions to engage in international commercial activities,
all of which have a substantial impact on the distribution of power among the
states comprising the system. Specifically, the economy has an enduring effect
upon interstate relations, both in terms of (1) determining the configuration of
power (concentration and inequalities) among the states in the system, and (2)
determining the configuration of interstate relations, of which interstate commerce
is of primary importance. Both of these factors play an instrumental role in the
politics of the international system, and hence, the timing of peace and conflict.
Thus, the international economy has an effect upon how nation states are
configured in the system and how this configuration may change over time. This
is the crux of the theoretical argument proposed in this research. That is,

fluctuations in the economy and commerce ultimately precipitate changes in the

distribution of power in the system."0 Consequently, shifts in the distribution of

 

39This assumption maintains a systemic level of analysis. Moreover, it lends support to the supposition that
behavior in the international system can be observed in the system structure. That is, patterns witnessed in

the system structure are not simply processes, but aggregations of actions taken by states in the system at a
particular point in time.

”This is manifested for example, in the asymmetric gains from trade that accrue to states in the international
marketplace, and general changes in macroeconomic conditions that occur over time. Both serve to
increase or decrease the level of growth and wealth of nations.

74

power, increase apprehension and uncertainty about the level of security in the
system. It is this source of uncertainty that serves to influence the level of conflict
in the international system.41 The question then becomes, what determines when
the distribution of power in the system is likely to cause uncertainty about the
international structure, and likewise, warrant the outbreak of conflict?

Consider the structural model of the international system depicted in Figure
2.4. The theoretical framework illustrated in Figure 2.4 is an amended neorealist
argument about the distribution of power and its affect on the likelihood of
conflict. Specifically, it illustrates how the economy and trade influence the
distribution of power in the international system. Moreover, the distribution of
power has an affect on the likelihood of conflict it terms of the uncertainty that it
generates about the configuration of states in the system. That is, uncertainty
stems from whether the new configuration will promote hostility or peace in the
system. As the distribution of power changes over time, this creates uncertainty
about the structure of the international system, which subsequently affects the

likelihood of conflict.

 

"Moreover, it is possible that some feedback exists among structural conditions in the international system.
With respect to systemic conflict, we know that the ebb and flow of economic activity changes dramatically
when great power war occurs. For example, aggregate prices increase, interstate trade is disrupted, and
state production strategies change when war occurs. However, changes in the economy do not necessarily
compel states to engage in conflict. Nevertheless, changes in the economy have an effect on power
concentrations in the international system, the latter of which does influence the onset of conflict. This will
be elaborated upon further in the discussion of exogeneity.

75

Economy

 

DiStribUtiOn ConfliCt
Power Peace

 

Trade

 

   

\_/

Figure 2.4 The Co—Evolving Systems Theory

76

Thus, from a systems perspective, the argument suggests that states monitor
and respond to the level and direction of changes in the distribution of power, and
other commercial and aggregate economic conditions."2 Indeed, this is what
determines the order and nature of the configuration of states within the system.
To the degree that the level of conflict in the system is balanced with the structural
change stemming from shifts in the distribution of power, then the system has

43 However, it is important to note that as the

achieved an equilibrium state.
structure evolves (e.g. cycles), so too will the appropriate level of systemic
conflict. Thus, the aggregate “system” will never be at rest. That is, there does not
exist one single equilibrium state in the international system. Instead, as the
system continues to evolve (cycle), and the observed changes in commerce, the
international economy, and the distribution of power occur, so too will the
appropriate level of systemic conflict. As such, the structure contains a moving
equilibrium that is attained whenever the level of conflict is an appropriate
response to the expected changes in distribution of power.

However, what happens when the level of conflict is not an appropriate

response to the changes in distribution of power. More importantly, why does this

occur? In any given period, states responding to the general trajectory (i.e.

 

42The reference to state here assumes the aggregation of states which comprise the international system.
This additional assumption maintains a systems level approach.

‘3 Note the definition of equilibrium here is distinct from that typically found in the literature examining
structural transformation. Gilpin (1981), Doran (1989) and others seem to suggest that stability in the .
system is an equilibrium outcome, where stability is equated with peace in the international system. The
definition of equilibrium in this study is different. In this research conflict is part of the equilibrium in the
structure of the international system.

77

evolution) of the distribution of power and other economic and commercial
activity in the system, may misperceive the type of structural change that has
occurred, and thus, engage in conflict. To the degree that these expectations are
incorrect, then states may engage in temporary (transient) conflicts. In the
aggregate, such disturbances will cause disequilibrium between the overall level of
systemic conflict and the distribution of power in the system. However, once the
expectations are adjusted, the system will move towards its appropriate level of
systemic conflict. As states recognize the disequilibrium between the distribution
of power and the level of conflict in the system, then states will adjust their level
of peace and/or conflict.

Given the arguments developed about systemic conflict equilibrium, and
about the distribution of power and how it is affected by the economy and trade,

the following proposition will be tested:

Proposition (4): The trend in systemic conflict will be in common with those
associated with the distribution of power, trade and inflation
in the international economy.

A related issue hinging on the formalization of the general theoretical
framework developed in this research is the implicit exogeneity assumptions made
about the interrelationship between, systemic conflict, the economy, trade and

power distributions. Typically these issues have been ignored in the war-economy

literature. However, given the theoretical arguments developed herein, I contend

78

that systemic conflict can and should be considered endogenous. Although
systemic conflict is believed to share an equilibrium relationship with the
economy, trade and changes in power distributions, the latter processes have been
explicitly delineated to be driving changes in the level of systemic conflict. As

such, the following proposition will be tested:

Proposition (5): Systemic conflict is endogenous to changes in the distribution
of power, trade and inflation in the international system.
Again, it is crucial to emphasize the political process responsible for this
change. That is, the distribution of power changes the configuration of interstate
relations. As powerful nation-states rise and fall, new configurations of interstate
relations subsequently form. This process tends to stimulate conflict among states
as they seek to determine where they are configured in the new system, and most
importantly how they relate to states in the new “equilibrium” within the system.
As such, the dynamics associated with changes in the distribution of power are not
only responsible for triggering out-of-equilibrium behavior in the system, but for
correcting the system as well. That is, the adjustment in the power distributions in
the system, are also responsible for achieving a new equilibrium path. This is a
crucial insight into the development of an error correction model. To substantiate

this contention, the following proposition will be tested:

79

Proposition (6): Disequilibrium among systemic conflict, the distribution of
power, trade and inflation will create equilibrium error.
When this error is eliminated, systemic conflict will move to
its appropriate level.

Now that the precise nature between the constructs has been delineated, it is
possible to specify a model in which to test the hypotheses. However, before a
model can be specified, it is first necessary to determine whether the series in the

model are individually integrated and, whether they together are cointegrated.

A Model of Co-Evolving Cycles

Before the development of an empirical model to capture the hypothesized
relationships between the series pertaining to conflict, the economy, and trade, it is
first necessary to determine whether each of the series under consideration are
integrated. After performing these univariate diagnostic tests, the next step will
be to test for cointegration. This involves determining whether the series being
investigated share a common stochastic trend as proposed in proposition (4).
Initially however, it is necessary to determine whether each series is integrated. A
discussion of the diagnostic steps involved in examining a series for integration
will first be preceded with a brief summary about the properties of stationary (e. g.
nonintegrated) time series in order to discern the intuition behind the methods

involved.

80

Properties of Integrated Series

Integrated variables are a specific class of nonstationary variables. They
have statistical properties which cannot be ignored in the specification and
estimation of time series models."4 Integrated series, as defined by Engle and
Granger (1987), have no deterministic component, but are characterized by a
stationary, invertible, ARMA representation after differencing d times, and are
therefore integrated of order d, denoted xt ~ I(d)."5 Table 2.2 below technically
differentiates between 1(0) and 1(1) processes.

In a less technical sense, theoretically consider a times series x, as a
collection of random variables ordered in time to be characterized by a stochastic

process.46

 

“Although the problems associated with using nonstationary series has been written about and examined in
great detail, numerous scholars continue to evade this technical issue in the specification and estimation of
time series models. At least as far back as Granger & Newbold (1974), econometricians have implored
their readers not to ignore the properties of integrated regressors. Granger & Newbold specifically warned
against the spurious regression problem that would arise when two processes integrated of the same order
(e.g. 1(1)) are regressed upon one another. When this occurs, estimates obtained from an empirical model
will have inflated t-stats because standard errors are typically underestimated, hence resulting in biased
estimates.

45In fact, Wold's (1938) decomposition theorem states that a stationary time series process, with no
deterministic component (e.g. time trend), has an infinite moving average (MA) representation. He proves
that this MA representation can also be represented (by inversion) as a finite auto-regressive moving
average (ARMA) process. Intuitively, 1(0) series have: (I) a finite variance which does not depend upon
time, (ii) have only a limited memory of its past behavior (e.g. the effects of a random shock are transitory),
(iii) tend to fluctuate around the mean (which may include a deterministic trend), (iv) have autocorrelations
that decline rapidly as the number of lags increase. 1(1) series have: (I) have a variance that depends upon
time and goes to infinity as time goes to infinity, (ii) have an infinitely long memory (e.g. the effects of a
random shock will permanently affect the process), (iii) have a tendency to wander widely, (iv) have
autocorrelations that tend to one in magnitude for all time separations (Engle & Granger, 1987).

“Stochastic processes (e.g. random processes) can be either discrete or continuous variables. The former
characterize the series in this study (e.g. price index, exports, etc.), while for example, the recording of an
electrocardiogram characterizes the latter (Maddala 1992, p.527).

81

Table 2.2

Properties of Stationary and Integrated Processes

 

Property‘ 1(0) 1( l )
Variance Finite Unbounded

(01“ 132)-” (grows as tog)
Conditional variance 2

of 02
Autocorrelation function at ,
lag k pg =9 p, =Jl‘(k/l) "IVkaJI"°°
Expected time between Finite Infinite
crossings of y=0
Memory” Temporary Permanent

 

 

 

'Ihis table is duplicated from Banerjee et al. (1994), p. 85.
" A series has a permanent memory if the effect of the shock does not disappear as 1 ~00.

‘In a multivariate context, an integrated process may have some components that do not remain in the
series indefinitely. If a series is integrated, there must be at least one component that will have
permanent effects, but there may be others with temporary memory. For example, a random walk
process would yield an integrated process, but memory would be permanent only for the random walk
component.

82

A series is considered stationary (i.e. not integrated) if there is no systematic
change in the mean of the variable and if no systematic variation exists over time.47
That is, the series has a constant unconditional mean and variance over time. This
is equivalent to implying that the random variables comprising the series are
independent of the effects of time. This is a rather tenuous assumption."8 As such,
stationarity for a series is usually defined in a less restrictive manner where its

mean is constant and its covariance depends only on its lag:

Ea.) = :1

Cov [x(t), x(t+1:)] = Mt)

 

”This is equivalent to conditions associated with strict stationarity (Maddala 1992, p.528). Technically
speaking, conditions associated with strict stationarity are based upon rather strong assumptions which are
tenuous in practice. This is because strict stationarity implies that the distribution of x, is independent of all
t. As a result, not only are the mean and variance considered constant, but all higher order moments.
However, if a series is assumed to be normally distributed, achieving stationarity is not as arduous a
process. As the multivariate normal is characterized by only the first and second moments (i.e. (11.02)), the
properties of a series higher order moments are not considered. As such, in the multivariate case, strict
stationarity and weak stationarity are technically equivalent. For other distributional forms however, this
may not be the case.

“Consider a series representing monthly unemployment rate for a 5 year period. Theoretically, it is hardly
controversial to suggest that the current unemployment rate (x,) is based in part upon last months
unemployment rate (x,,,), or that it can determine future unemployment levels (x,,,) . Although the
unemployment rate changes over time, generally, it does not fluctuate widely from month to month.
However, in a technical sense, each unemployment rate recorded in time is random to the extent that its
value is always "pertaining to chance" (Maddala 1992, p.527). Although implicitly, there is some
dependence implied, technically, the current unemployment rate does not depend, in a deterministic sense
on all previous months unemployment. Thus suggestive of an assumption of independence. Although
series must technically be independent, this may not be a valid assumption given our theories and what they
suggest about the distributional form of a series. In some cases a unit root is implied by many economic
and political models. For example, in the former case the rational use of information by economic agents
justifies the theoretical assumption of an integrated process. Moreover, modeling the budgetary process as
an incremental process is also a justification for an a priori assumption of integration.

83

This is referred to as weak stationarity. Thus if a series is nonstationary (i.e.
integrated), its mean is characterized by some systematic change (i.e. a trend) and
the covariance associated with the series also depends on time."9 In this case, most
of the moments associated with the series distribution will depend on time and
standard statistical assumptions relying on normal distributional properties will be
violated?0 As a result, conducting inference with such a series will be
problematic.51 Thus, because of the properties associated with integrated series,
and the problems that can arise in estimation and inference, it is crucial to conduct

stationary tests on all series comprising a model.

 

‘9 However, another variant of the stochastic process described above, is a random walk. The mean and
variance of these types of time series are also time dependent, however, when taking the first difference, the
series is found stationary (Maddala 1992, p.513). This class of variables is also referred to as difference
stationary processes (DSP). Generally speaking, deterministic trends are only linear projections, thus there
is no change in the variance over time. However, a series with a trend in the stochastic component (TSP)
leads to cyclical movements around the trend, and therefore the variance throughout time is changing.
Fractional integration tests can be used to determine whether a series is of the TSP or DSP class. (See
Granger & Joyeux 1980, Harvey 1991, Diebold & Rudebusch 1989, Dolado et a1. 1990 for a description of
these tests).

”Moreover, in many cases taking the first difference of the variable will not render it stationary. If the
series is characterized by a trend, typically the next step is to determine whether the trend is deterministic or
stochastic in nature. Integrated variables are characterized by unit roots which give rise to stochastic trends,
as opposed to deterministic trends. In the latter, innovations to an integrated process are permanent,
whereas in a stochastic trend they are transient. This has obvious statistical implications, but also
theoretical implications as well. Whether a series has a deterministic or stochastic trend, hence permanent
versus a transient memory process, has implications for the information that one is modeling. Specifically,
whether one is modeling long-run and/or short term components, and whether one or the other is a valid
representation of the theoretical framework under analysis.

"Aside from the issues necessitating diagnostics of the univariate properties of individual time series, there
also exists methodological issues in the multivariate context as well. A widely noted problem in the
multivariate case, necessitating the identification of the order of all variables (i.e. I(d)) in a statistical model,
is the spurious regression problem mentioned previously (Granger & Newbold, 1974). Generally when two
integrated variables (1(1) random walks) are regressed on one another, standard error estimates are biased
downward. Consequently, OLS estimates will be biased. Specifically, the bias produces t-statistics that are
inflated and a R2 that is superficially high; hence, indicating a good fit of the model, when in fact there is not
one. -

84

Diagnosing Integration

Dickey and Fuller (1979, 1981) develop a class of statistics capable of
discerning whether a series is integrated. Generally speaking, the Dickey-Fuller
test (DF) consists of testing a single series against a null hypothesis of integration.

Consider the series x, again. Decomposing it into:

x! [30 + p1: + “r
where: u, ntuH + e,

and [30 is a constant, while [3, represents a linear time trend.52 Assume e, is a
covariance stationary process with a mean of zero. The reduced form of this

equation can be represented more generally as :

xt -_- Y0 + 61: + “xi-I + er
where: y = BOO-ct) + Bloc
6 = 81(1-a)

If the reduced form is characterized by a unit root, then «=1 and 6:0, the latter
suggesting the absence of a deterministic time trend.53 The test statistic used in the

analysis which follows is:

 

 

”This notation follows the discussion found in Maddala (1992, p.583).

”As will be noted later. assuming y=0 i.e. no drift, is not always appropriate. This technical point compels
MacKinnon (1991) and others to derive alternative sets of critical values to take into account the possibility
of drift and its effect on the distribution of a.

85

where a is the OLS estimate of a reparameterized version of the reduced form
equation above and SE(a) is its corresponding estimated standard error.

Specifically, the DF test consists of generating the regression:

Ax, = Yo + 51: + ¢xr-1 + e

I

after which, a test is conducted on the significance of 4).“ However, given the
discussion above concerning the distributional properties associated with
nonstationary series, the asymptotic distribution of this test statistic is nonnorrnal.
Thus, standard t-tests based on normal distributional assumptions are
inappropriate. Appropriate critical values are tabulated by DA. Dickey (1975) and
are reported by Fuller (1976). However, these critical values are tabulated for
y=0, e. g. they assume that there is no drift present in the series. When drift is
present in a series, the distribution of the estimated coefficient on the lagged level
of the series (i.e. (1)) in a DF test depends on the value of the constant (i.e. y).
MacKinnon (1991) argues that if a series contains drift (i.e. y¢0) then a time trend
should be included in the DF regression, and hence, a new set of critical values
should be used that capture the dependence between y and the trend. MacKinnon

provides correct critical values to test hypothesis for DF tests when who.”

 

5‘Note the reparameterization consists of regressing the change of x, on a lag of itself. This is done in order
to determine if the change in the series, i.e. A, is a factor of the characteristic equation of x,.

”MacKinnon (1991) emphasizes the point that several diagnostic decisions have to be made concerning the
specific form that is appropriate to use for the Dickey-Fuller test for any particular time series under
analysis. These issues are discussed in Banerjee et al. (1994).

86

The Dickey-Fuller test essentially involves regressing current changes on
lagged levels of itself. The null of nonstationarity is tested by determining if the
coefficient on the lagged levels is negative and statistically significant. If it is, the
null is rejected, and the series is determined to be stationary (e. g. not integrated).56
Again, if a series is characterized by stationary properties, it has an identifiable
mean and finite variance around its mean. Thus, the mean and the variance are
constant and not changing over time. Intuitively the expectation for (b on the
lagged level above, if the series is stationary, should be negative and significant.
Given that a stationary series exhibits mean reverting behavior, positive changes
toward the mean at one time period, should be followed by negative changes
toward the mean in the subsequent period. On the contrary, in a nonstationary
series, current changes will not be associated with past levels. Because such series
have nonconstant mean and infinite variance, the series will tend to depart ever
further from any given value as time goes on. If this movement is predominantly
in one direction the series is generally considered to have a trend.

Once a Dickey-Fuller regression has been estimated, it is necessary then to
diagnose whether the residuals are characterized by serial correlation. If the error
terms are serially correlated, then the usual suspects associated with

autocorrelation arise (e. g. inefficient estimates, inflated standard errors, invalid

 

“A variant of this procedure is also used in determining whether two series are cointegrated. In this
diagnostic test however, the residuals from the cointegrating regression are analyzed with respect to their
stationary properties. This will be elaborated upon further when discussing cointegrating regressions.

87

covariance matrix for significance testing, etc.), rendering the test statistic
useless.57 There are several tests used to detect serial correlation. The
specification of these test largely depend upon whether or not a lagged endogenous
term is part of the original regression.58 If serial correlation is present, then it
becomes necessary to add additional lagged changes of the series on the right hand
side of the equation (as many as necessary) in order to achieve white noise. The

test is referred to as the Augmented Dickey-Fuller (ADF).

Ax, = Yo + 51: + (bxt-l + :eijt-j + E:
J:

The purpose in adding the additional Ax”. terms is to allow for ARMA error
processes.” If the residuals are white noise from the regression above, tests for
statistical significance will be reliable.

Four variable series are used in the model to test the theoretical hypotheses,
they are systemic conflict, the distribution of power, inflation, and level of world
trade respectively. Table 2.3 provides descriptive statistics for each of the series.

The results of the Dickey-Fuller tests (DF) for these times series are presented in

 

”See Phillips & Perron (1988) for a discussion of this issue and alternative tests for unit roots in time series
models.

”See Appendix A for a discussion of the tests used for serial correlation in the analysis which follows.

”Thus p is selected on the basis of being sufficient large to insure that e, is white noise.

88

Table 2.4. Again, the DF tests the null of integration against the alternative
hypothesis of stationarity. The DF critical values appropriate for this sample are
calculated by using MacKinnon’s formula and are presented in Table 2.5.

The results in Table 2.4 suggest several interesting findings:

(1) The systemic conflict series is an I(l) difference stationary process.
It contains a stochastic trend, rather than a deterministic trend.

(2) The distribution of power series is an I(l) difference stationary
process. It contains a stochastic trend, rather than a deterministic

fiend.

(3) The inflation series is an I( 1) difference stationary process. It
contains a stochastic trend, rather than a deterministic trend.

(4) The trade series is an I(l) difference stationary process. It contains
a stochastic trend, rather than a deterministic trend.
Now that the nature of the trends in each of the series has been determined, it is
possible to test whether these trends are common across the four series forming a
cointegrating vector (Engle & Granger, 1987). This will subsequently enable the

specification of an empirical model and a test of the remaining hypotheses.

Descriptive Statistics

 

 

 

 

 

Series Mean Std. Dev. Min. Max. N
Power 0.44 0.21 0.20 1.00 125
APower .0006 .0772 -0.540 0.493 124
Inflation 0.008 0.039 -0. 193 0.137 125
Alnflation 0.001 0.042 -0.287 0.121 124
Systemic l 1.48 10.19 0.00 52.00 125
Conflict
ASystemic 0.22 5.65 20.00 24.00 124
Conflict
Trade 205.53 43.35 350.00 1 19.00 125
ATrade 1.77 l 1.84 -36.00 30.00 124

 

 

90

Table 2.4

Dickey-Fuller Tests for Nonstationarity

 

 

 

 

Time Series‘ DF" Box-Pierce Q‘ ADF Breusch-Godfrey
Levels
Systemic Conflict‘ -2.937 33.18 -1.75 .418
Power“ 2113 27.21 -2.04 1.306
Inflation -4.051 17.17 -2.25 3.047
Trade -.209 50.26 -.616 1.48
First Difference

A Systemic -14.61 31.86 NA NA
Conflict

A Power 932 31.28 NA NA
A Inflation -l4.39 57.69 -12.76 2.21
A Trade -11.49 49.00 NA NA

 

 

'Thecommonlogged valueisusedfortheprioeseries.

”The numbers reported in the table are the estimated t-statistics from the coefficients associated with the lagged level of the series
for each DF and ADF regression. For a list of the appropriate critical values see Table 2.5.

‘The tests for serial correlation are dependent upon whether the regressions used to test for stationarity include a lagged endogenous
term. By definition. the DF tests does not include such a term and therefore the Box-Pierce Q is appropriate test for serial
correlation. However. the ADF test does include a lagged endogenous term on the right hand side and necessitates the use of a
Breusch-Godfrey test. For a discussion of these tests, the issues involved. and the appropriate critical values see Appendix D.

‘This series contained a significant constant term that subsequently became insignificant in

91

 

 

Table 2.5
MacKinnon's Critical Values
Model‘ Point (96) 4).. SE ch, 4), Critical
Values”
N=125
Constant. 1 -3.4336 0.0024 -5.999 -29.25 -3.48
No Trend 5 -2.8621 0.001 1 -2.738 -8.36 -2.89
10 -2.5671 0.0009 -1 .438 4.48 -2.58
Constant. 1 -3.9638 0.0019 -8.353 47.44 4.03
Trend 5 -3.4126 0.0012 4.039 -l7.83 -3.45
10 -3. 1279 0.0009 -2.418 -7.58 -315
Constant. 1 -3.9001 0.0022 -10.534 -30.03 -3.99
No Trend 5 -3.3377 0.0012 -5.967 -8.98 -3.39
10 -3.0462 0.0009 4.069 -5.73 -3.08
Constant. 1 4.3266 0.0022 -15.531 -34.03 4.45
Trend 5 -3.7809 0.0013 -9.421 - 1 5 .06 -3.86
10 34959 0.0009 -7.203 4.01 -3.55
Constant, 1 4.2981 0.0023 -1 3.790 46.37 4.41
No Trend 5 -3.7429 0.0012 -8.352 -13.41 -3.81
10 -3.4518 0.0010 -6.241 -2.79 -3.50
Constant, 1 4.6676 0.0022 48.492 49.35 4.82
Trend 5 4.1193 0.0011 42.024 -l3.13 4.22
10 -3.8344 0.0009 -9.l88 4.85 -3.91
Constant, 1 4.6493 0.0023 -l7. 188 -59.20 4.79
No Trend 5 4.1000 0.0012 -10.745 -21.57 4.18
10 -3.8110 0.0009 -8.317 -5. 19 4.02
Constant. 1 4.9695 0.0021 -22.504 -50.22 4.79
Trend 5 4.4294 0.0012 44.501 -19.54 4.18
10 4.1474 0.0010 -1 1.165 -9.88 4.02

 

' The different model types depend upon the presence of a constant term and/or a time trend in a Dickey-Fuller specification.
' The critical values are calculated according to the following formula: C(p)=¢°° + ch, T"+ d), T "

 

92

Diagnosing Cointeggg’gn

In order to test whether the three series comprise a common trend, it is
necessary to estimate a cointegrating regression. First, consider n times series
given by x“, x2,,..., xm. In the analysis which follows n=4; therefore we are
interested in obtaining a set of residuals from a regression among the same four

series tested for integration above. To accomplish this, it is first necessary to

estimate a cointegrating regression of the following form:

n

x1: = Yo + 2B} xjt + Z:

[=2
Essentially this involves regressing the levels of systemic conflict on the other
series representing the distribution of power, inflation and trade. If these series are

cointegrated, analysis of residuals Zt produced from their linear combination

should be stationary. That is, the cointegrating residuals,

it : x! _ >1"1r - Bijr

should be 1(0). Engle & Granger (1987) suggest seven alternative tests for
determining if z, is stationary. In the analysis which follows, two of the tests they
recommend will be conducted. The first is the Durbin-Watson statistic for a

cointegration regression (CRDW) originally suggested by Sargan & Bhargava

93

(1983), while the second is simply a DP statistic for the cointegrating residuals
(CRDF) (Dickey & Fuller, 1981).60

First the DW statistic for the cointegrating regression equation above
(CRDW) should approach zero if the cointegrating residuals (i.e. z,) contain an
autoregressive unit root. As such, the test consists of computing a statistic similar

to the conventional DW statistic:

 

The null hypothesis being tested, using the CRDW statistic, is a single unit root.
Thus, the test consists of rejecting the null hypothesis of non-integration if the
CRDW is significantly different from zero, in favor of the alternative stipulating
cointegration.61

Secondly, the DF for the cointegrating regression equation above (CRDF) is

similar to the original DF's used to test for integration in a single series. In this

 

60Critical values for these tests can be found in most economteric texts. For example, see Banerjee et al.
(1994) and Maddala (1992).

"Again, this test is similar to the original Durbin-Watson test in which the null assumes p=1. Hence, the
test is the null that DW=0 against the alternative, which suggests the DW>0. Recall that DW = 2(1-p).
Thus when p=1, DW=0. Engle & Granger (1987) present the critical values of this test for 100
observations in the case where N=2. However, Engle & Yoo (1987) produce expanded critical values for
100 and 200 observations for a system of up to 5 variables. Because the sample used in this research is
over 100, the latter critical values are used.

94

context, the series under analysis is the cointegrating residuals (2,). That is, the

CRDF statistic is based upon the OLS estimation of:
AZ: : Yo + YIZt—l + :YnAZr-r + 6:

where a test of the significance of the coefficient on the lagged level of the
residuals (i.e. y,) is conducted.‘52 If y, is significant then the residuals from the
cointegrating regression are stationary, i.e. [(0), hence indicating that the four
variables in the system form a cointegrating vector. Table 2.6 reveals the
estimates derived from the cointegrating regression.

Although it is not appropriate to conduct inference from the estimated
coefficients, in terms of the magnitude of their effect on systemic conflict, it is
possible to get some initial insights about the hypothesized effects of the
regressors. Specifically, proposition (1) proposed that inflation in the international
economy will be positively related to levels of systemic conflict in the international
system. Indeed the estimates reported in Table 2.6 suggest that inflation does have
a positive effect on systemic conflict. Moreover, proposition (2) proposed that
aggregate levels of trade in the international economy will generally be inversely
related to levels of systemic conflict in the international system. Surprisingly, the

estimates in Table 2.6 again do not give support to this proposition.

 

”Again, the value of p in the above equation is selected on the basis of being sufficiently large to insure that
e, is white noise. Note for a DP p=0, thus the system includes only one lagged endogenous term on the right
hand side. Thus, in the case where p21, the test would be a augmented Dickey-Fuller specification for the

cointegrating regression (e.g. CRADF).

95

Table 2.6

Cointegration Regression and Evidence of Cointegration

 

 

 

 

 

Level of Systemic Conflict
Independent Variable‘ OLS Coefficient” SE T-Statistics
Levels of Power 14.255 4.01 1 3.554
Levels of Inflation 62.547 21.389 2.924
Levels of Trade 0.039 0.018 2.061
Constant -3.413 4.438 -0.769

Goodness-of-Frt Measures Residual Diagnostics (2..)

R’ = .184 SSE = 9.293 CRDF = 4.35 (p<.05)‘
Adjusted R’ = 0.164 CRDW‘ = .529 Box-Pierce Q = 31.014

 

'I'he results from the procedure were produwd from E-Views. MicroTSP version 1.0

'Because of the nonstationarity of the individual regressors. the coefficients reported have nonstandard distributions. This
subsequently makes drawing any inferences from them. in terms of their magnitudes and statistical significance. problematic (Engle
& Granger 1987).

‘The five percent critical values reported by Sargan & Bhargava (1983) have been revised by Engle & Yoo (1987) for a CRDW test
with four series (i.e. N=4) for the following sample sizes ('1') are: T=50 (1.10), T=100 (.65). T=200 (.48).

‘Table 5 lists the critical values for the DP test of the residuals for a specification with n=4, with a constant and no trend
(MacKinnon. 1991). Because the cointegration regression included a constant term, it is subsequently excluded in the DF
regression specification used to analyze the residuals 2,.

96

Proposition (3) contends that the distribution of power in the international system
should be positively associated with systemic conflict. Indeed this relationship is
borne out in Table 2.6. However, because of the nonstationarity of the individual
regressors, the coefficients reported have nonstandard distributions. Thus,
drawing inferences from these estimates, in terms of the magnitude of their effects
or statistical significance, is problematic.

How well does proposition (4) perform? This proposition suggests that the
trend associated with systemic conflict in the international system, will be in
common with those associated with the distribution of power, inflation, and
aggregate levels of trade in the international economy. Hence, this is an explicit
hypothesis that the series form a cointegrating vector. Technically, if the variables
used in the analysis share a common trend, then the residuals from the
cointegrating regression will be stationary. If the variables comprise a
cointegrating vector, this renders support to the proposition that these series
comprise an equilibrium relationship. As evinced by the residual diagnostics of
(2,) reported in Table 2.6, both the CRDW and the CRDF tests indicate that the
series are cointegrated. Specifically, the CRDW statistic estimated from the
cointegrating regression (.529) exceeds the critical value (.48) reported by Sargan
& Bhargava (1983). Thus, it is appropriate to reject the null of no cointegration.

Moreover, the CRDF statistic derived from estimation (4.35) is smaller than the

97

critical values reported by MacKinnon (1991) listed in Table 2.5, hence rendering
the null invalid.63 Thus, we have support for proposition (4).

A related issue is the implicit assumption that there is a single cointegrating
vector among the three series. This is explicitly examined by using Johansen’s
(1988) procedure for testing multiple cointegrating vectors. The results from the
Johansen test are reported in Table 2.7. The estimates from the test reveal that the
four variables together form at most, only one cointegrating vector. The estimate
of the LLR statistic for the null hypothesis that there are no cointegrating vectors
(e.g. rsO) is significant, i.e. (49.848) exceeds the p<.05 level critical value (39.89).
Moreover, the estimated LLR statistic for the null hypothesis suggesting rs 1, is
significant, i.e. (24.708) exceeds the p<.05 level critical value (24.31). Both of
these results suggest support for the existence of at least one cointegrating vector
among the four series.

Thus the results from the analysis in this section also render support to the
idea of modeling systemic conflict as an error correction process. The next
sections turn to the task of developing an error correction model (ECM) and to
testing the remaining propositions. However, preceding the development of the
empirical model will be a discussion about the nature of causality, in an effort to

test the exogeneity assumptions advanced in proposition (5).

 

63Note, the CRDF test statistic is statistically significant at the p<.05 level.

98

Table 2.7

Johansen Cointegration Test

Null Hypothesis‘ 2 LLR” or=.05
Test Statistic

 

 

r50 .190 49.848 39.89
rsl .131 24.708 24.31
r52 .052 7.995 12.53
r53 .014 1.677 3.84

 

 

 

 

 

'Ihe results from this procedure were produced from E— Views, MicroTSP version 1.0

bThese statistics were generated from a vector autoregression. The initial cointegrating regression was
normalized on systemic conflict. The cointegrating regression included an intercept and time trend. Six
lags were included for each of the endogenous variables in the system

99

The Nature of Causality in Cointegrated Systems

The Role of Exogeneity in Modeling

Exogeneity assumptions play a crucial role, theoretically and empirically, in
causal modeling within the field of international relations. However, exogeneity
as a concept is rarely evoked in discussions of theory building or in empirical
specifications. This deficit is most apparent in the Long Cycles literature. This
lack of dialogue serves to reinforce the benign neglect of (1) rigorously
substantiating the veracity of specific claims presumed to depict the order of how
processes and events occur or function in the world and (2) disregarding vital
information crucial to the consistent and efficient estimation of models used to
conduct inference. As Hendry (1993) notes, the word exogenous “connotes ‘being
determined outside of (the model of analysis),’ and yet researchers frequently
attempt to ascribe the status of ‘exogenous’ to a variable per se (as with, say, the
sun’s energy) and then deduce certain inferences therefrom (e. g. the variable is a
valid instrument for estimation)” As will be demonstrated in the discussion which
follows, exogeneity is crucial in determining the conditions under which processes
responsible for generating the independent variables in a single-equation model

can be ignored without compromising the consistent and efficient estimation of

100

parameters in a conditional model.“ Thus exogeneity goes beyond theoretical
issues associated with the temporal ordering of variables into the realm of
estimation and how to achieve efficiency and consistency in empirical analysis
used to make inferences about behavior or processes being modeled. In particular,
this discussion will focus on the implications exogeneity has for estimation of
cointegrating systems in general and error correction models in particular.‘55
Initially it is important to stress that exogeneity and causality in this
discussion, in particular because of the emphasis of these issues on cointegration
and error correction models, have both, theoretical and empirical implications.
Theoretically the issue of how to specify a variable, as an endogenous or
exogenous process, is crucial in terms of adequately assessing the viability of
hypothesis emanating from our theories about how we think the world functions.
For example, what moves cycles of peace and conflict? The very question
suggests that I am interested in explaining the current levels of systemic conflict,
hence endogenizing the process. However, what if the factors hypothesized to
move cycles of peace and conflict are likewise effected by system conflict? What

implications does this feedback have for exogeneity assumptions? Moreover, if

 

“The conditional model generally speaking, is the model, or equation of interest, that contains the
endogenous process under analysis. That is, the equation in which inference is based. The conditional
model is sometimes referred to as the structural model (Granato & Smith 1994a, 1994b). The marginal
model is the equation that creates the process generating a regressor found in the conditional model.

"The issues dealt with in this section are not new, however their application to cointegration and error
correction models have not been widely considered. The following discussion draws upon the insights of
Cooley & LeRoy (1985), Ericsson (1992), Granger (1988a), Granato & Smith (1994a & 1994b) and
Hendry (1993).

101

subsequent analysis reveals sufficient justification for a cointegrating relationship
(i.e. systemic conflict and inflation for example), what implication does this have
for a hypothesized causal relationship? All of these issues have technical
ramifications. Empirically, at stake is whether there is sufficient information about
the process and the parameters of interest, to insure efficient and consistent
estimation?66 Thus, although theoretical reasons may exist for a particular
specification, the technical ramifications of such a specification cannot be ignored.

In some respects cointegration brings up new issues associated with
exogeneity and causality. Specifically, the symmetric relationship that is typically
assumed between variables that comprise a cointegrating vector, e.g. the notion
that they share a common stochastic trend, may pose problems to efficient and
consistent estimation; hence, this has implications for exogeneity assumptions in
error correction models. Technically however, under certain conditions, it is
possible to specify restrictions on cointegrating relationships, specifically on the
long-run component, which afford consistent and efficient estimation, and hence,
valid inferences, in a single-equation ECM framework. However, as will be

demonstrated, this can only be achieved under strong exogeneity assumptions.67 It

 

6“For example, typically in times series models, such as auto distributed lags (ADL), simultaneous equation
models (SEQ), vector autoregressions (VAR), etc., the history of the process under analysis is of interest.
Moreover, other variables (in a static sense, the independent regressors) are also of interest, in addition to
their past histories. Nonetheless, for purposes of parsimony or identification, many of these regressors are
not modeled. However, it is necessary to consider whether the exclusion of this information compromises
the consistent and efficient estimation of the model.

”Typically it might be assumed that weak exogenous conditions are sufficient; however, given the
symmetric assumption in cointegrating relationships, resulting in dependence between parameters of interest
in both the conditional and marginal models. efficient and consistent estimation of the conditional model is

102

is important to stress the theoretical implications these restrictions have for the
processes under analysis. Specifically, a priori, their must be reason to believe
that the processes comprising a cointegrating relationship have an asymmetric
relationship. Technically, the asymmetric assumption implies that only the
dependent process under analysis responds to equilibrium errors in the ECM. In
terms of the theoretical framework developed in this research, the general systemic
conflict equilibrium hypothesis satisfies this technical restriction on the ECM.
Specifically, recall that the theoretical model specified earlier suggests that the
distribution of power conditions the actions and reactions of states in their
decisions to engage in conflict within the system. Thus, systemic conflict alone
reacts to, or more generally, corrects changes in the equilibrium relationship it
shares with the state of the distribution of power in the international structure, not
vice-versa. How then are exogenous assumptions rendered viable, especially in
the context of cointegration and ECM’s ?

Hendry (1993) notes that more often than not, researchers resort to Granger
causality tests to determine the validity of their exogenous assumptions. Granger
causality is the technical condition that specifies the precedence or temporal

ordering, of one variable’s occurrence before another.68 In the bivariate case, two

 

compromised. Thus, asymmetric restrictions on the long-run components have to be specified; hence, the
necessary introduction of Granger non-causality and strong exogeneity.

“The technical statement of Granger causality was initially developed in Granger (1969), although Granger
himself acknowledges the roots of his thoughts derive from technical points originally espoused by Wiener
(1956). See Freeman (1983) for an introduction of Granger causality in political analysis.

103

time series x, and y,, it can be shown that the series x,, fails to Granger cause the
series y,, if in a regression of y, on lagged values of y, and lagged values of x,, the
coefficients estimated for the latter are zero; hence, signifying no significant

effect. Specifically,

’1 k
y, = 2 arm . 21:13.4-.. . u, (i=l,2,...,k)
i=1 i=

If B, = 0, x, fails to Granger cause y,. However, the use of Granger causality tests
to determine the appropriateness of exogeneity assumptions is flawed for several
reasons. Specifically, it is inappropriate and can lead to Type I and Type H errors
when trying to ascertain whether one variable Granger causes another.69
Specifically, this is problematic not only because Granger causality tests do not
examine causality between variables per se, but also, technically speaking, it is
insufficient for establishing a priori, the prerequisite condition of weak
exogeneity.70 A brief discussion of exogeneity, both its technical and theoretical
aspects will proceed in order to fully demonstrate the intuition and importance of

these issues to the model purported in this research.

 

“Generally, Type I and II errors can occur because Granger causality tests (1) lack power and (2) give no
information about consistent estimation because, by design, they do not afford any insight into weak
exogeneity. See Granato & Smith (1994a) for an intuitive technical discussion of these points.

10Granger causality tests, for example in the bivariate case, enable one to determine which variable has a
greater predictive capacity in accounting for the other, in terms of variance explained. Therefore, its ability
to render causality per se is limited in this technical sense. The limitations of Granger's test, in political
analysis particularly, is discussed in Granato & Smith (1994a & 1994b).

104

Defining Exogeneity

Initially, consider the types of exogeneity discussed by Engle, Hendry &
Richard’s (1983), e.g. weak, super, and strong (strict).7| Weak exogeneity is

defined in the following manner:

weak exogeneity: A variable x, is considered weakly exogenous to estimate a set
of parameters A, if inference for these parameters, which are
conditioned on x, , involves no loss of information.

More generally, weak exogeneity “is a relationship linking certain variables to
parameters of interest and is precisely the condition needed to sustain valid
inferences about those parameters in models which condition on contemporaneous
variables” (Hendry, 1993, p.331). According to this definition, weak exogeneity
implies that the marginal model for x, would not depend on a conditional model.
That is, the parameters within a conditional model are independent of a marginal

model and there are no cross equation restrictions between the marginal and

 

"For the purposes of the following analysis, super exogeneity will not be discussed in great detail.
Generally speaking, super exogeneity is the conjunction of weak exogeneity and "invariance." (The
invariance condition is also referred to as variation free. Generally this concept assumes that knowledge
about the value of a parameter provides no additional information on another parameter’s range of value
Ericsson ( 1992)). That is, if x, has been established to be weakly exogenous, and parameters in the
conditional model remain invariant to changes in the marginal distribution of x, , then x, is deemed super
exogenous. Super exogeneity is used to insure valid policy simulations. Specifically, this concept is related
to the Lucas critique which stipulates that coefficient estimates, specifically in a simultaneously equation
model, should not be assumed to be independent of changes in the exogenous variables within a model
(Maddala 1992, p.389). When trying to simulate a forecast, it is necessary to consider the possibility that
optimizing behavior on the part of an agent, will lead to modified behavior, because the agent may know to
expect that change is going to occur (i.e. an agent may know what a forecast is going to be before it actually
takes place). This assumption is fundamental to rational expectations theory. See Lucas & Sargent (1981)
for a discussion of this class of models.

105

conditional model.72 Thus, under weak exogeneity a regressor of interest can be
deemed independent of contemporaneous and future values of the error process in
the conditional model. In effect, weak exogeneity establishes the conditions
necessary for consistent and efficient estimation, and hence inference.73

Weak exogeneity is distinct from strong exogeneity, which is defined in the

following manner:

strong (strict) exogeneity: Is the conjunction of weak exogeneity and Granger
noncausality. If achieved a model is deemed viable for
conditional forecasting. That is, if x, is weakly
exogenous, and not preceded by any of the endogenous
variables in the system, x, satisfies the conditions for
strong exogeneity.

Strong exogeneity goes a step further and demands that a regressor of interest be
independent of the contemporaneous, future, and past values of disturbances in the

conditional model. Moreover, in the context of a bivariate relationship, it requires

that an endogenous variable y,, does not Granger cause another regressor x,. If

 

72Determining what the parameters of interest are within a model is hardly an innocuous decision (Engle,
Hendry and Richard, 1983; Pratt & Schlaifer, 1984; Maddala, 1992). Parameters may be of interest for
several reasons. Initially they are chosen because they are related to theories that a model is seeking to
validate. However, with respect to determining exogeneity, parameters of interest are selected specifically
for determining whether empirical relationships are constant over a given sample period, hence securing
information about consistency and efficiency in estimation (Ericsson, 1992; Hendry, 1993).

73However, once weak exogenous conditions have been obtained, how does one proceed to test the validity
of this assumptions? Granato & Smith (1994) note that there are not any straight forward tests
unfortunately. An indirect test suggested by Granato & Smith is Dufour's (1982) parameter constancy test.
Tests for parameter constancy can shed light on a lack of weak exogeneity because coefficient estimates that
are unstable over time are often indicative of a structural change in the process generating regressor (i.e. x,
marginal model). When weak exogeneity holds, the structural model (conditional) will be invariant to such
changes; hence, evidence of parameter instability (stability) provides some indirect evidence that weak
exogeneity is invalid (valid).

106

weak exogeneity has been established first, then finding Granger non-causality
from y, to x,, renders empirical support to the contention that x, is strictly
exogenous. Hence, Granger non-causality provides the additional restriction
necessary to achieve valid conditional forecasting.74

The discussion of exogeneity to this point has centered on processes
assumed to be stationary. It has been demonstrated that if the regressors in a
model are deemed stationary, consistent estimation can be achieved under weak
exogeneity.” As such, under conditions of weak exogeneity, it is possible to get
consistent estimates from a single conditional model. However, what if the
stability of the system is in question? What effect will this have on the system
being analyzed? Moreover, what if one or more of the processes under analysis
are integrated (i.e. nonstationary)? What effect does this have on the process

under analysis? What additional technical issues must be addressed?

Exogeneity and Integrated Processes
In order to examine these questions, consider the simply bivariate

relationship below:

 

7‘From this discussion it should be evident that weak exogeneity cannot be analyzed using Granger causality
tests. Granger causality tests are not designed to assess the lack of dependence between the conditional and
marginal models that is necessary to efficiently and consistently estimate parameters of interest. As Cooley
& LeRoy (1985) and Granato & Smith (1994a) stress, establishing Granger non-causality is only consistent
with strict exogeneity, however it does not “unambiguously confirm such an assumption.”

75An additional assumption is that the error in the conditional model are not serially correlated

107

 

 

 

where: 1.1 is the column of means and Q is the covariance matrix. Moreover,

consider the equations below, which represent the conditional and marginal

models respectively:
y, = a + Bx, + V“ v“ ~ lN(0,02)
x, = ky,_l + 621 62’ " IN(0,n)22)
where v = e - [fl]e
- 1 l 2:
I I (022

Regard k to reflect the linear dependence of x, on y,_,. If the parameters of interest
in estimation are Al: (13,02) for the conditional model, and 12 = (k,n)22) for the
marginal model, then it is possible to conceive that x, is weakly exogenous to y, as
the parameters of interest it, do not depend in any way on the marginal model.
That is, each model can be estimated independently. As such, weak exogenous
conditions have been met and estimation can proceed under this assumption.

However, this may not be the case if the stability of the system is in question.

108

Specifically, if the issue of stability in the conditional model arises, then weak
exogeneity is most likely a problem.

When the stability of a system is in question, additional information is
needed from the marginal model about the processes responsible for generating
regressors in the conditional model. The stability of the system can be analyzed by

examining the reduced form of the conditional model:

y, z or + pyH + e“ e“ “IN(0,w2)

where p is the root contained in the conditional model’ s reduced form, which is
equal to (B-k). Recall that the stability of the system, as represented in the equation

above, depends upon the value of p, such that:

lpl <1 the process is stable

lpl =1 the process is characterized by a random walk"5

lpl >1 the process is unstable (explosive)
If the stability of the system is in question, then the parameters of interest to
conduct inference under weak exogenous conditions are no longer independent.
Specifically, information is needed from both A, and 12, because they each contain
information needed to diagnose stability. That is, [3 and k are not independent of

one another, as their joint distribution yields insights about the stability of the

system under analysis. Essentially, establishing weak exogenous conditions will

 

7"This assumes that the process will be stationary in its first difference.

109

be necessary, but not sufficient (Cooley & LeRoy 1983, Ericsson p.257, 1992,
Granato & Smith 1994b).

Moreover, integrated processes, establishing weak exogenous conditions
alone are not viable, strict exogeneity becomes the relevant concept for the case of
nonstationary regressors. As Granato & Smith (1994) note, in the case of
integrated regressors, exogeneity conditions have implications that extend beyond
the issue of consistency, and hence, weak exogeneity. With integrated series, the
distribution of certain coefficients in the conditional model are of concern. As
previously discussed, integrated processes represent a cumulative function of past
information (or history of a process). As such, the distributional properties
associated with estimates and significance testing from such processes are non-
norrnal. We can assume that x, may affect y, through the product of [3 and k.
Namely, the past of y, will condition contemporaneous values of y, if the product
of (B-k)2 1, making the system unstable. That is, if lplz 1, the system will be
unstable, because y, will be characterized by a unit root or an explosive root, both
of which will yield inefficient and inconsistent estimates. Thus, in order to
guarantee that inference will not be jeopardized strong exogenous conditions need
to be specified. Valid prediction of y from the conditional model requires more
than weak exogeneity. It requires the additional assumption of Granger non-
causality. The requisite restriction is that m12=w22=0, i.e. that y, does not Granger

cause 2, (Ericsson 1992, Granato & Smith 1994b). Specifically, Granger non-

110

causality means that only lagged values of x, enter into the marginal model,
likewise lagslof other variables do not enter the marginal model, other than values
of x, itself.

However, what if the two processes modeled are both I(l)? Engle &
Granger ( 1987) would then consider the original conditional model a cointegrating
regression because the stochastic trend associated with the unit roots of each series

form a linear common form that is 1(0). That is,

ér :yr _ 3x:

is a stationary process. In this case, what are the implications for exogeneity and
causality? In order to assure valid estimation of single equations, it is important
that strong exogenous conditions hold. This involves the use of Granger causality
tests. However, the use of these tests when regressor are integrated is problematic.
Specifically, because Granger causality tests require that the levels of each variable
be regressed on the lagged level of all the variables within a given system of
equations (i.e. a VAR), the unit roots associated with a series will also be
estimated. Thus, if the series are integrated, but not cointegrated, the coefficients

will have nonstandard limiting distributions.77 Granato & Smith (1994b)

 

T’See Granato & Smith (1994b) for an intuitive technical discussion about the asymptotic theory of
integrated processes and the ramifications these processes have for exogeneity tests.

lll

demonstrate that a way to avoid “unit root asymptotics” is to estimate Granger

causality tests as a system of error correction equations.78 Specifically,

— “2054-5294) + “2:

From this restricted VAR, the null hypothesis that y, does not Granger cause x,
becomes:

P
Hozdllil = 0 also 25”, = 0
p:

Assuming that y, and x, are cointegrated, a finding of Granger noncausality from y,

to x, renders support to the notion that x, is strictly (strong) exogenous to y,.79

 

"This method was derived by Engle & Granger (1987) and is referred to as an restricted diagnostic VAR.
Moreover, a similar test is utilized by Johansen (1991) in which the restricted VAR is a vector error
correction model (VEC).

7"’As Ericsson (1992) notes, conditional ECM’s assume the existence of weak exogenous conditions, by
excluding the same cointegrating vector from appearing in both the marginal and conditional model. Again,
this assumption could only be valid however, if a priori, ones theory was consistent with such a
specification. As previously demonstrated, the war-economy equilibrium hypothesis is consistent with an
asymmetric specification. Johansen’s (1992) partial system specification affords a test for weak exogeneity
between cointegrated variables. Also see Granger & Lee (1989) for a detailed technical account of partial
system models.

112

A Test of Exogeneity

In light of the discussion above, it is highly unlikely that conventional
Granger causality tests would prove adequate in determining exogeneity conditions
in the system under analysis. As noted previously conventional Granger causality
tests for diagnosing exogeneity assumptions among systemic conflict, prices and
exports are inappropriate, not only because they fail to take into long-term causal
components that tie cointegrated series together (Durr 1992), but also because
when regressors are integrated, the normal asymptotic distribution of these test
statistics are compromised. In order to demonstrate this point, conventional
Granger causality tests were conducted on the series used in the model. Table 2.8
lists the results obtained from estimation.

The results of the conventional Granger causality tests, in which a model 1
consisting of current levels of systemic conflict are regressed on a number of lags
of itself, is compared with a similar model in which the same number of lags of
power are included (along with inflation and trade respectively in separate tests),
do not render support to the hypothesized causal relationships. Recall that Granger
causality tests rely on an F-test to determine the degree to which controlling for the
previous levels of the distribution of power in the system improves the total

variance explained above simply analyzing systemic conflict alone.

113

Table 2.8

Systemic Conflict Granger Causality Tests

 

 

 

 

 

 

Independent Variables (6 lags) F-Stat‘
Current Levels in Systemic ’1
Conflict
Levels of Power 1.526
Levels of Inflation 1.189
Levels of Trade 0.715 11
Current Changes in System
Conflict
Changes in (A) Power 0.892
Changes in (A) Inflation 1.168
Changes in (A) Trade 0.533

 

 

‘ The F—statistic tests the change in total variance explained (R2) when lags of the independent variables
are excluded from another equation that regresses current values of the dependent variable on six lags of
itself plus six lags of the independent variable. The test statistic~ F(6,106), thus the following critical
values are used: p<.01 (2.96), p<.05 (2.18).

114

Thus, if the exclusion of power from the model leads to statistically significant
reduction in the explanatory power, then the tests would indicate that indeed,
systemic conflict is endogenous to power, or more generally that power Granger
cause systemic conflict. As the results from the tests indicate, none of the series in
their levels or first difference form, were found to Granger cause systemic
conflict.80

Nonetheless, when estimating a restricted vector error correction (VEC)
model, a technique more suited to determining exogeneity when integrated process
are involved that form a cointegrating vector, some support is rendered to the
original hypothesized causal relationships. The VEC model is a restricted VAR
designed to use with nonstationary data, which have been found a priori, to be
cointegrated (J ohansen 1991). Typically an unrestricted VAR does not impose
cointegration on its variable. However, the VEC restricts long-run behavior of the
endogenous variables to converge to their cointegrating relationship, while
permitting the short-run dynamics to take there natural course. Again, the
estimation of a VEC model necessitates that the series under analysis form a
cointegrating relationship. Prior estimation of a cointegrating regression suggested
support for cointegration, while a subsequent Johansen test rendered support to the

notion that only one cointegrating vector was present among systemic conflict,

 

80Specifically, the F-tests verify that neither, the distribution of power, inflation nor trade were able to
explain, in a statistically significant manner, more variance in systemic conflict, than systemic conflict’s
own past history.

115

power, inflation and trade. After the number of cointegrating vectors has been
established, the VEC model consists of taking the first difference of each
endogenous variable and regressing them on a one period lag of the cointegrating
equation (e. g. the long-run component of the model) and the lagged first
differences of all the endogenous variables in the system (e. g. hence capturing the
short-run dynamics). Because the model purported in this research examines four
series, i.e. systemic conflict, the distribution of power, inflation and trade, the VEC
is generalized to a four variable system. Specifically, the test consists of the null
that y, does not Granger cause x, (i.e. Granger noncausality against the alternative
of Granger causality), where 1:, here represent the distribution of power, inflation
and trade, in separate tests of the null. To simplify notation, the distribution of
power, inflation and trade are represented as separate regressors x,, while systemic
conflict is represented as variable y in the hypothesis tests that follow. According
to the model, systemic conflict is posited to be endogenous to the distribution of
power, inflation and trade. As such, in each test, we are seeking to determine if
each x, (i.e. the distribution of power, inflation and trade) does NOT Granger cause
y, (systemic conflict). If this assumption is maintained, then it is valid to conclude
that the distribution of power, inflation and trade are strictly (strong) exogenous to

systemic conflict. In this context, the test can formally be expressed again as:

Hozorllil = 0 also 251,, = 0

116

The results of the VER are in Table 2.9. The estimates obtained from the
VER render some support to the notion of an asymmetric cointegrating
relationship among at least two of the three series. First, the estimate of the long-
run component in the VER is statistically significant in the equations representing
systemic conflict and inflation but not in the power and trade equations.81
Secondly, the lagged first differences of systemic conflict are not significant in
either the inflation or trade equations in the system.82 Both results suggest that
systemic conflict does not Granger cause trade, with some mixed support for the

distribution of power and inflation.

 

"That is,(a,[3,= 0) for the equations relating to the distribution of power and trade. Although, again the
long-run component in the inflation equation is statistically significant at the (p<.01) level. Note that the
estimates associated with the cointegrating equation represent the long-run components of the VEC, i.e. the
error correction component for each equation in the system. Thus, if ((118,: 0) in one of the equations, then
the long-run component is insignificant. As (a,B,= 0) in both the distribution of power and trade equations,
then this affords some initial support to the notion that systemic conflict does not Granger cause the
distribution of power nor exports and hence, power and trade are strictly (strong) exogenous to systemic
conflict.

82This statistical results relates to the second part of the null hypothesis, namely that:

P
2:51p: 0

p=l

Thus, as systemic conflict is not statistically significant in either the inflation or trade equations, this yields
some additional support to the notion that systemic conflict does not Granger cause inflation or trade. Note
however, the second lag of systemic conflict is statistically significant in the inflation equation.

117

Table 2.9

Diagnosing Exogeneity: Restricted VEC Test

Endogenous Variablesa

 

 

 

 

A Systemic A Power A Inflation A Trade
Conflict
Cointegrating Equation” -0.056 -0.001 0.009 —0.079
(0.033) (0.001) (0.0002 (0.073)
-1.727 -1 .091 4.634 -1.076
A Systemic Conflict 0252 -0.0003 0.003 -0.003
(-1) (0.094) (0.001) (0.003) (0.003)
-2.689 -0.210 0.559 -l.107
0004 -0.004 0.001 —0.237
(_2) (0.091) (0.001) (0.003) (0.003)
-0.047 -3.097 1.534 -1.133
A Power -12.958 -0.039 0.008 1.601
(-1) (0.094) (0.090) (0.042) (14.459)
—2.014 -0.436 0.189 0.111
(-2) -2.040 0.041 0.092 -0.472
(0.093) (0.091) (0.043) (14.627)
-0.313 0.447 2.181 -0.032
A Inflation -22.93 -0.094 0.011 --64. 12
(— 1) (2.385) (0.001) (0.090) (0.089)
-l.33l -0.390 0.097 -l.65
9.872 -0.316 -0. 128 -37.67
(—2) (2.497) (0.001) (0.094) (0.093)
0.679 -1.547 -l.355 - l . 153
A Trade -0.011 -0.001 -0.000 —0.067
(~l) (2.478) (0.0006) (0.093) (0.092)
-0.266 -1.912 -0.800 —0.701
0.047 -0.0007 0.000 -0.014
(-2) (2.405) (0.0006) (0.090) (0.090)
1.117 -1.159 0.176 -0.149
R2 0.162 0.12 0.371 0.036
_S_E 5.364 0.075 0.035 12.055
'The results from this procedure were produced from E-Views. MicroTSP version 1.0. The estimated

coefficients, standard errors and t-statistics are reported and listed accordingly.

”The cointegrating equation was normalized on systemic conflict.

118

Thus from this analysis, it is possible to conclude that there exists some support for
the notion that trade is strictly (strong) exogenous to systemic conflict, while
power and inflation may be strictly exogenous.83 As such, it is possible to continue
with a test of the last proposition. Specifically this will involve the specification
and estimation of an error correction model (ECM). It is to this task that I now

turn.

Specifying An Error Correction Model

I have hypothesized that short-term changes in systemic conflict can be
understood as both, a response to changes in the distribution of power in the
international structure, trade and inflation, as well as changes in the long-term
levels of these variables. The latter is characterized as a moving equilibrium and
represents the actual error correction portion of the model. An error correction
model (ECM) is able to capture both the long term and short term effects that
systemic conflict, the distribution of power, trade, and inflation have on one

another.84

 

”However, given that the evidence accumulated is not without problems, future endeavors should include to
use a multivariate system estimator, similar to those proposed by Johansen (1988) and Stock and Watson
(1988), in order to obtain consistent and efficient estimates for the coefficients of interest.

l”Davidson et al. (1978) was the first study to introduce error correction models (ECM’s). However,
Granger (1983) was the first to connect cointegration and its application to ECM's.

119

Theoretically and technically, the notion of a moving equilibrium affords a
more precise statement of how these phenomena are interrelated. The previous
focus on cycles proved to be restrictive in a technical sense, as strict periodicity
methods failed to find supporting evidence for long cycles. Moreover, the strict
focus on cycles rendered little theoretical explanatory power about the
interrelationship between systemic conflict, the distribution of power, inflation and
trade in the international system. Theoretically, the equilibrium argument
purported in this research captures the long term interrelationship between the
variables over time. That is, how systemic conflict, the distribution of power,
trade, and inflation are tied closely together throughout time. Moreover, the
explanation purported in this research is able to explain what happens when the
posited equilibrium is disturbed, i.e. the series are forced apart. Specifically, the
model suggests that the equilibrium error is corrected over the long term as
systemic conflict is aimed at a new level that is consistent with the equilibrium
state in the international system.

To test proposition (6), an ECM will be specified. Their are several
methods available to estimate an ECM. The most common form is the Engle-
Granger two-step procedure. It involves estimating the equilibrium errors in a
static linear regression of the level of systemic conflict on the independent
variables, power, trade and inflation. The final error correction model will

incorporate the equilibrium errors in order to estimate the rate of equilibration,

120

when controlling for the short-term effects (i.e. the first differences of the relevant

variables in the system).85

The Engle-Granger Two-Step Method
Engle & Granger (1987) derive a two step methodology in which to

estimate an ECM. It consists of the following steps:

(1) Regress Y on X in levels in order to obtain the cointegration vector
Zr

Y,=or+[3X,+Z,;

N
11

Y,-[3X,-or

(2) Regress changes in Y on past changes in X, and on the equilibrium
errors represented by cointegration vector Z,

AY, = gAXH - hZ,

In order to test the remaining hypothesis, an ECM was estimated. Reports
of the ECM estimation are located in Table 2.10. Proposition (6) is simply a test to
determine whether or not an ECM is a valid representation of the adjustment
process that has been posited to occur when the long-run relationship between

systemic conflict, the distribution of power, trade, and inflation is offset.

 

“This particular approach suggests that the equilibrium errors are corrected equally by all of the variables
(i.e. whether they are dependent or independent) included in the first regression (i.e. the cointegrating
equation).

121

Statistically, if equilibrium error is eliminated from the long-run relationship
hypothesized between systemic conflict, the distribution of power, trade, and
inflation, then systemic conflict is anticipated to move back to its “appropriate” ,
level. Stated another way, the coefficient on the lagged level of the residual (i.e.
Z,) should be between (0) and (-1) if the hypothesized adjustment takes place as
purported in the theoretical model.86 As indicated in Table 2.10, the estimated
coefficient is (-.213). This is well within the appropriate range of values.

The two-step estimation leads to the following error correction

representation:

AConflict, = -l3.651APower,_, - .213Z,_l + e

f

where Z,,, represents lagged equilibrium errors from the first step, i.e. the

cointegration regression. From Table 2.10, Z,,, may be expressed as:

ZH = ConflictH + 14.26*P0wer,_, + 0.039*Trade,_, + 62.547*Inflati0n,_,

Substituting the latter equation into the former yields:

AConflict, = 0.727 - 13.651APowerH

- .213(Conflict,_, -0.039 *Trade,_l +62.547 'trInflation,_l +14.255 *Power,_ I) +6,

 

“ Generally speaking, the equilibrium error coefficient should always be between (0) and (-1). If not some
egregious misspecification of the model has taken place. Moreover, technically the coefficient on the error
correction component should always be negative. This is not to suggest that the error correction component
negatively effects the dependent process in the typical "directional" sense. Rather, it is meant to imply that
once some type of change has taken place in the equilibrium state, re-equilibrating will need to occur (hence
the negative coefficient), whether it be a positive shift or a negative shift

123

According to the model, systemic conflict will be in an equilibrium
relationship vis-a-vis prices and exports (i.e. the economy and trade) when the
equation above is equal to zero. Likewise any shock to this equilibrium
relationship will be corrected by changes in systemic conflict at the rate of 29%
per year, beginning one year after the shock has occurred.87 Thus support is given

to proposition (6).

Conclusion

The theoretical framework examining the co-evolution of political cycles
and economic waves in this research received support. In particular, the
theoretical model and empirical framework advanced in this research has made
several significant contributions, all of which integrate previous work, and extends
new ideas to the long cycles literature. Theoretically, the co—evolving model
developed in this research advances a more general, causal argument capable of
explaining patterns of systemic conflict over time. Specifically, integration tests
conclude that each of the time series under analysis are characterized by stochastic
trends. In addition, a cointegration test empirically demonstrated that the
individual trends among these series are in common. It is this latter result, the

evidence of a common trend among these series (6. g. cointegration), which I

 

"This follows as (.286) is the estimated coefficient for the lagged residuals saved previously from the
cointegration regression and as the series reflect annual observations over time.

124

contend to be crucial in uniting previous cycles research and in making novel
extensions to this literature.

With respect to integrating previous research, this is accomplished in part
by the theoretical approach advanced in this study, which contends that various
cyclical phenomena have interrelationships that are co-evolving. Previous
research studying cyclical phenomena, model them as distinct and non-interacting
processes. These explanations fail to take into account the dynamic relationships
among these series. The theoretical and empirical evidence reported in this
research, demonstrates that these phenomena do affect one another over time.

This research affords novel insights by demonstrating that the cyclical
phenomena in the model are cointegrated. That is, the series are empirically co-
evolving, as the trends associated with each series are in common (i.e. they follow
the same path over time). This empirical finding serves to extend research in this
area as it negates the utility in (l) empirically substantiating the existence of
individual cycles; and (2) in determining the duration of cycles and their most
appropriate dating schemes. In particular, the approach advanced here advocates
thinking about how the trends associated with these phenomena may drift together
throughout time (i.e. follow an equilibrium path), and why these phenomena get
detracted from their common path (i.e. experience deviations from the

equilibrium).

125

Moreover, cointegration not only affords a rigorous test of the correlation
among various cycles by examining the trends associated with these processes, but
sets them within a rigorous empirical framework. This framework is capable of
(1) testing exogeneity conditions specified in the theory, where systemic conflict is
posited to be a strictly endogenous process, effected by exogenous phenomena
related to power, trade and economic activity; and (2) delineating how these
processes effect changes in systemic conflict through the test of an error correction
model (ECM). Both of these technical issues afford scholars interested in the long
cycles literature to rigorously test causal theories concerning the interrelationship
between economic and political phenomena in the international system.
Furthermore, exogeneity tests and ECM’s also serve to extend novel contributions
to this literature.

With respect to the former, specifying exogeneity conditions is fundamental
in any causal theory and is especially crucial in studies examining dynamic
processes over time. Specifically, exogeneity affords rigor in (1) determining the
nature of the endogenous processes being explained in a theory; and (2) restricting
the dynamic effects of exogenous factors in a model by delineating any anticipated
feedback among the processes under investigation. Both of these issues have been
largely ignored in the long cycles literature. Exogeneity tests conducted in this
research found empirical evidence in support of the theoretical argument that

systemic conflict is a strictly endogenous process responding to fluctuations in the

126

distribution of power, trade and the economy. Moreover, this evidence reinforces
the notion that the cyclical phenomena under analysis in this research, are
appropriately modeled in a “co-evolving” framework. Thus, it is possible to
specify co-evolving models that can rigorously specify causal relationships from
which inferences can be made about the dynamic interrelationships among
political and economic phenomena.

Again, the results from the previous cointegration test supported the
proposition that systemic conflict, power, trade and the economy share a common
trend. Moreover, the results from the Johansen test demonstrate that there exists
only one cointegrating vector. Together these results affirm the appropriateness of
modeling the causal relationship among these phenomena as an error correction
process.

Traditional studies of political cycles and economic long waves have
neglected the use of causal models. The analysis reported in this research suggests
that it is not only necessary theoretical, but it is empirically viable. Specifically,
the theoretical argument advanced in this research contends that economic and
commercial patterns over time serve to change the structure of the international
system through redistributing, or alternating, the distribution of power among
states over time. This framework, delineates systemic conflict as an endogenous
process affected by changes in the concentration of power, economic activity and

commerce in the international system. Advancing this particular theoretical

127

argument, extends the previous focus on the correlations between cyclical
phenomena, into a causal framework. Indeed, this is important in order to make
inferences about the exogenous processes within cycle theories (i.e power, trade,
and the economy) and their affect, in terms of their magnitude and significance,
on systemic conflict. The results from the ECM support the proposition that
systemic conflict is the only cyclical phenomena reacting to shocks in the
equilibrium relationship. That is, out-of-equilibrium behavior (e.g. exogenous
shocks that stimulate deviations from the common trend among the series in the
model) affects the level of systemic conflict. However, this change (i.e. deviation
from the trend) corrects itself (i.e. adjusts) back overtime to a new equilibrium, in
which the series follow a similar pattern once again (i.e. share a common trend).
Analyzing cyclical phenomena through the analysis of an ECM, extends
research in this area because questions concerning out-of-equilibrium behavior can
be explored. When cycles of systemic conflict, power, trade and the economy are
not in equilibrium (i.e. following the same path), this framework suggests that it is
because a shock has occurred in the system that has detracted systemic conflict
from its normal trajectory. These shocks derive in large part from the uncertainty
resulting from the alternating distributions of power, and hence the configuration
of states in the system. The latter event may stimulate conflict in the system as
uncertainty increases about the new equilibrium, or, type of configuration (e. g.

peaceful or hostile) that characterizes the international system.

128

Although the ECM yields support for this explanation, two important issues
remain unexamined that merit investigation in future research. First, the
interrelationship between the exogenous processes need to be delineated more
precisely. Specifically, the combination of factors that stimulate uncertainty need
to be assesses. This will in part, involve the analysis of out-of-equilibrium
behavior among the series used in this analysis. Uncovering these dynamics
among the various cyclical phenomena will help discern the timing of systemic
conflict in the international structure.

Secondly, efforts need to be taken to formally model co-evolving theories
similar to the one advanced in this research. Generally, this type of theory is not
conducive to game theoretic or decision modeling, because of the lack of strategic
action among actors per se. However, dynamic models offer one form of formal
modeling that is capable of delineating the interactions among systemic processes
(Pudaite, 1991). The benefit of formally modeling co-evolving theories is the
deductive manner in which the interrelationships among the processes can be
described and logically deduced. This type of research endeavor will stimulate
rigor in the long cycles research tradition. Attention to these research deficits will
serve to inspire more novel contributions to this evolving literature. Moreover,
scholarly attention to these issues will yield contributions that extend our present
knowledge and understanding of the relationship between political cycles and

economic waves in the international system.

CHAPTER 3

REGIME STRUCTURE, LEADERSHIP UNCERTAINTY AND THE
MAINTENANCE OF COOPERATION:
THE GAINS IN MODELING INTERNATIONAL ECONOMIC REGIMES AS
ORGANIZATIONAL TEAMS

Introduction

Conventional models addressing the initiation and maintenance of
cooperation within international regimes have been modeled as public goods
problem (Kindleberger 1981, 1986; Snidal 1985; Conybeare 1987), or more
generally as reciprocal arrangements among states (Axelrod 1984; Axelrod &
Keohane 1985; Cline 1982; and Dobson 1991).1 However, these solutions
typically fail to take into account the organizational complexity characterizing the
structure of regimes. Regimes, especially those with more than two actors, are
complex, both in terms of the distribution and configuration of states that comprise
them, and the level and type of information that is exchanged among the members.

The theoretical framework purported in this research seeks to tackle the

 

'The use of the term regime is not narrowly considered in any particular context other than economic, in
order to accommodate a variety of institutional settings. Thus, general reference will be made to
international economic regimes as examined in Keohane (1980).

129

130

complexity under such conditions by regarding the structure of international
regimes to be similar to organizational teams? By drawing upon the production by
teams analogy, it is possible to ascertain a solution capable of explaining the
maintenance of cooperation among a group of nation-states, especially when
incentives exist to violate such commitments.

Specifically, the role of leadership is examined in terms of its ability to
achieve and enforce cooperation within the context of an international economic
regime. The notion of leadership described in this research varies from the role
that is typically assigned to a hegemon in the traditional literature examining
regimes.3 Hegemonic stability theory typically assigns two types of hegemony to
dominant states, benevolent and coercive. In the former, the hegemon unselfishly
takes on the burden of initiating and maintaining a regime, where in the latter case,
the hegemon, acting out of pure self interest, pressures states to comply with its
own particular preferences. The conception of leadership in this research is
general enough to take into account both types of hegemony. This is reinforced by
the fact that a leader is not simply appointed because of its capabilities. In fact,

leadership can also be a function of states designating an actor with institutional

authority to reward and punish other states comprising the organization.

 

2See Alchian & Demsetz (1972) and Groves (1973, and 1977) for an example of this class of models.

3For example see Kindleberger (1986).

131

As will be demonstrated, the inclusion of leadership is integral to
maintaining cooperation within a production by teams framework. Because
consumption on the part of actors within regimes is conditioned on their level of
production, the maintenance of output is important to the survival of the regime
over time. To the degree that defection is difficult to determine, a leader is useful
because of its ability to distinguish noncooperative behavior. This is especially
important when the structure of the regime is comprised of multiple actors, making
defection from production commitments more likely, and more complicated to
discern.

Using the Bianco & Bates (1990) model, the following research applies and
generalizes their initial insights by (1) using their framework to analyze the
configuration of international regimes, (2) generalizing the model to capture
multiple actors (i.e. the N-person case), and (3) modeling the dynamic effects of
leadership uncertainty by incorporating Bayesian updating into the belief structure
of the actors in order to analyze changes in cooperative behavior. Generally
speaking, the model finds that it is never in a state’s best interest to defect if the
probability of confronting a strong leader is high. In general, as the probability of
strong leadership increases, the players fear punishment more often and tend to
benefit more from cooperation. In addition, the model finds that as the probability
of confronting a weak leader, which is capable of punishing, increases, the payoffs

for defection actually increase. Thus, the results of the analysis suggest that it is

132

not in the best interest for a weak leader to bluff about its type. States are more
likely to defect if they witness a weak leader punish, rather than a weak leader that

continues to reward.

International Regimes As Organizational Structures

Economic Policy Regimes

The theory of international regimes is still developing. It has emerged
slowly over the last decade and grown into a dominant research agenda within the
subfield of International Organization." Although numerous definitions abound,

international regimes are generally conceived to be:

“...implicit or explicit principles, norms, rules, and decision-making
procedures around which actors’ expectations converge in a given
area of international relations (Krasner 1977, p.1-21).”
This definition of regimes, albeit ambiguous, has been used to describe how states
organize to tackle substantive issues. From macroeconomic policy to free trade,

international regimes have been the primary heuristic tool in which to study

collective action by states to form international economic policy.

 

‘See Haggard & Simmons (1987) for a excellent discussion on the development of the theory on
international regimes.

133

Undoubtedly, the need for international coordination in monetary policy,
trade, and finance has corresponded with the increasingly interdependent nature of
the international economy.s Multinational and transnational corporations have
flourished throughout the last five decades and have subsequently expanded
international capital markets. The surge in capital integration has caused
fluctuations in trade balances, resulting in economic policy changes at the national
level. Within the last couple of decades, international trade imbalances have
induced nation-states to alter trade, monetary and fiscal policies in their domestic
economics as well as unilaterally alter their policy strategies in the international
economy. This general type of policy change on the national level has served to
stymie, if not decrease, the level of policy coordination in the international
economic environment, leaving scholars to ponder the causes and consequences of
such trends.

Capital integration, as a general indicator of growing interdependence
between nation states, tends to foster conflicts of interest between a state’s national
objectives, e.g. domestic growth and wealth, and international objectives, e.g. free
trade and stable markets. As a general result, governments unilaterally manipulate
economic policies to reconcile national economic objectives with international
market pressures. As a result, international economic objectives are subordinated

to national economic objectives. This results in political leaders defecting from

 

sInternational economic regimes is used in this research to generally encompass any type of policy
coordination (e.g. trade, monetary, financial, etc.), among nation states in the international system.

134

international regimes, in an effort to maintain domestic support. This subsequently
threatens the survival of international regimes designed to coordinate collective
action in pursuit of common goals. For example, the World Trade Organization
(WTO) represents a collective coordination effort on the part of advanced
industrialized countries to promote and expand free trade in the international
system.6 However, its success as an international regime is called in to question,
as its ability to effectively monitor and sanction defectors from free trade
principles continues to be challenged. As a result, increasingly many scholars
conclude that multilateral policy coordination will be difficult to secure in the next
century.

What type of regime structure and incentives can induce states to maintain
cooperation in international economic regimes? It is the analysis of this general
question that motivates the current research. Fundamental to understanding how
the maintenance of cooperation can be achieved within an international regime, is
analyzing the organizational factors, which characterize the structure of regimes, in
addition to the incentives used to induce states to (1) join such an organization and
(2) participate (e. g. maintain cooperation) over time. This research contends that
this task can be accomplished by drawing an analogy between organizational

teams and international regimes.

 

6'The Bretton Woods system is an example of a monetary policy regime.

135

Organizational Teams

The analysis of international regimes is formalized by drawing upon
literature addressing production by teams and informational economics in an effort
to explain and predict under what circumstances cooperation can flourish within an
international economic regime (Alchian & Demsetz, 1972; Bianco & Bates, 1990).
With the nation-state as the unit of analysis, an international regime is conceived
as a team comprised of multiple actors which have a joint maximizing goal to reap
gains from coordination. Thus, regimes are analogous to teams to the extent that
each state seeks to acquire a good that would not be produced unless some formal
structure existed to secure its production? Thus, regimes and teams are similar, in
that they are formal institutions, each composed of a set of actors with an
established hierarchy among the members, and similar goals in producing a joint
good. An implicit assumption is that states have an incentive to join regimes in
order to obtain benefits they would otherwise not be privy to if they were not
committed to the organization.8 Fundamental to joining an institution then is the
assurance that benefits will be awarded to members. In the teams analogy, the

distribution of benefits is maintained through a hierarchical relationship among an

 

7Thus to some degree the provision of public goods problem is relevant, but not entirely adequate. For
example, consider the WTO. The WTO has exclusivity associated with its members. That is, not every
state in the system is a member. Moreover, some policy orientations within WTO are asymmetrical.
Specifically, developing countries are permitted to violate certain free trade principles because of their
relative disadvantage to other states in the international market place.

1'For example, consider the most favored nation status (MFN) that is accorded to all members of the WTO.
Every state that is a member of the WTO is privy to MFN status and therefore gets the same trade privileges
afforded any other nation in the organization. MFN status is a benefit for member states which comprise
the WTO and is not guaranteed to nonmember states.

136

anointed leader and the respective members (i.e. followers), comprising the
organization.

Thus, the regime itself is analogous to a long-run contractual agreement
between a set of nation-states. Thus, the regime is characterized by long-term
commitment. Informational concerns come into existence during the life of the
regime as the possibility of encountering moral hazard arises.9 Moral hazard is
manifested in hidden action (Kreps, 1990). When nation-states within an
economic policy regime contract to produce a given level of cooperation, it may
not always be possible to ascertain the exact nature of the form of action that each
nation-state has produced. This is the essence of the problem associated with
hidden action. As such, perfect monitoring and enforcement may be an
impossibility. This is because the total aggregate output produced by the regime is
characterized by noise.lo Thus, the structure of the regime needs to be designed in
a manner that will provide a disincentive for member nation states to seek actions
which go against the design of the initial contract. This is where the role of

leadership becomes important.

 

’Informational concerns will arise throughout the duration of the contract and not simply during its
initiation. Morrow (1994) argues that problems of monitoring and sanctioning in a regime can be alleviated
by controlling for informational problems in the initiation phase. However, the dynamic nature of regimes
suggest that continued revisions of regimes will necessitate monitoring and sanctioning activities.

“’That is, it may not be possible for any particular actor to discern when a defection has taken place.
Formally, this is an issue arising from imperfect information (Kreps, 1990). This will be elaborated upon
further in the discussion of the model.

137

Leadership

As the success of an international regime is contingent upon its survival
over time in promoting cooperation, then structural disincentives to defect from
production need to be incorporated among the actors comprising the organizational
team. To achieve this, the theoretical framework proposed in this research
suggests that some form of authority must be appointed that maintains a superior
level of information about nation states (e.g. members of the team) within the
regime. Again, the necessity of a leader to monitor regimes is one of the
fundamental assumptions made by proponents of hegemonic stability theory
(Kindleberger, 1986; Gilpin, 1987).11 However, instead of focusing on the
coercive versus benevolent aspects of the theory, an important issue to consider is
the ability of the leader to credibly pose a punitive threat to potential defectors.
Thus, a related issue to consider is how cooperation can be achieved as an
equilibrium outcome among a set of nation-states, one of which is designated as a
leader, when uncertainty exists about the capabilities of the leader?

The previous question is important to the degree that it sheds light upon the
evolutionary aspects of leadership. Specifically, by formally determining how
cooperation can be maintained within regimes, especially when dominant powers

responsible for the maintenance and survival of the institution begin to lose their

 

l'Moreover, the lack of leadership is associated with increases in noncooperative behavior (i.e. defection),
and hence, the decline of regimes in general. However, the decline of the United States in its role as
hegemon over the last several decades has not resulted in the dismantling of all regimes.

138

power, it will be possible to specify the conditions in which peace and cooperation
can be maintained in the midst of power transitions within the international
system. Thus, the model explores the operations of an international economic
regime under conditions of strong and weak leadership.

As such, the teams model moves well beyond most analyses of international
regimes by accounting for both the actions of a leader and the other members of
the organization in an iterated N-person strategic environment. By accounting for
how members interact with one another and with a leader, the model is able to
demonstrate the conditions where international cooperation within the structure of
a regime can be attained over time.

The presentation of the model will proceed in two steps. First, the basic
form of the game is presented and trigger strategies under conditions of partially
complete and perfect information are discussed. This part of the game draws upon
an iterated production by teams model developed by Bianco & Bates (1990).
Second, incomplete information is incorporated into the game and Bayesian
perfect equilibria conditions are delineated. These conditions are then used to test

several propositions about the evolution and maintenance of cooperation.

139

Toward A General Formal Model

The game involves N actors, where n represents an individual actor, n6 N,
within an international economic regime. The regime is a long term economic
agreement among a set of member nation-states. Thus, there is an assumption of
future interactions between the nation-states which comprise the regime. As such,
the game is characterized by an infinite repetition, discounted formulation. The
nation-states interact over a period of time modeled as a series of rounds in a
game, where each iteration is designated as t, where t6 T. The game consists of a
leader and a set of followers. Moreover, each nation-state moves simultaneously,
without knowledge of the other nation-state’s actions, and the leader moves after
the followers. The model is formalized first by narrowing the strategy space to a
dichotomous choice between cooperation or defection. This affords a simple way
in which to calculate the costs and benefits of cooperating. Consider the
following, where s, is the strategy choice, and B(s,) represents the total benefits
accruing from a vector of strategy choices from N actors comprising the regime,

and c,, represents the costs associated with cooperation:

O (Defect)
1 (Cooperate)

St

St

Where:
B(S,) = a,(s,, + s2, + ...+ SM)

140

Note, the benefits accrued within the regime are a linear function of the number of
nation states that cooperate.12 Thus, general payoffs for each state n, at iteration ( t)

can be represented by:

3(5)

Vm(sm) = -—N— - Cm (Cooperate)
B(S,)

Vn,(sm) = -—N— (Defect)

It is assumed that mutual cooperation makes all actors better off. However, a
unilateral incentive exists to defect such that each follower possesses a dominant
strategy to defect at each iteration of the game. Defection will always be the
dominant strategy given the equation which derives individual benefits. This holds
even if more than one nation state decides to defect. However, the outcome
associated with all actors cooperating is always better than the outcome associated
with all actors defecting. Thus, the free rider problem is home out in this scenario.
Figure 3.1 demonstrates this relationship. Regardless of the number of actors N,
within the regime that can benefit from cooperation, an individual state always has
an incentive to defect. This is also demonstrated in Table 3.1 where the payoffs
associated with an individual defection are always higher than those associated
with cooperation. How then can nation states under such conditions be persuaded

to cooperate?”

 

'2 Refer to Appendix B, section 3.1, for further details.

l3Cooperation in this context means the production of a good. For example, the maintenance of a stable free
international market structure (Kindleberger 1986).

I41

Payoff to Nation (i)

State (i) Defects

State (i) Cooperates

 

 

Figure 3.1 N-Person Cooperative Dilemma

142

Table 3.1

N, payoff when a,=2 and (=1

 

Number of Nation-States Which Select the Strategy s,=1 (cooperate)

Payoffs‘ Z s,=9 Z s,=8 E s,=7 )3 s,=6 E s,=5 2 s,=4 E s,=3 Z s,=2 E s,=l )3 s,=0
s, for N,

 

 

s,,=1 1.0 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate

s,,=0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0
defect

 

 

 

'Note the payoffs are calculated according to the formulas given for the individual benefits for a nation-state pursuing
one suategy over another. i.e. cooperating or defecting.

143

First, the solution to the model lies in the development of mechanisms
which prohibit the unilateral incentive to defect from cooperation in the structure
of the regime. One way in which to deter noncooperative behavior of potential
defectors, is through the manipulation of trigger strategies. Specifically, the folk
theorem maintains that if discount rates exist, which are high enough (d‘:0 < (1‘ <1),
full cooperation can be maintained as a subgame-perfect equilibrium if actors have
the ability to employ trigger strategies.l4 For example, the manipulation of a
trigger in the context of a free trade regime (i.e. GATT), would be an action akin
to the threat of cutting off most-favored nation status (MFN) to states which erect
barriers to trade. Here erecting trade barriers would represent defection from
GATT and withdrawing MFN status would be a form punishment to a recalcitrant
state.

Secondly, because the regime is configured similar to an organizational
team, the presence of a leader charged with the duties of monitoring and
sanctioning followers is assumed.” A dominant nation state is encouraged to take
on the responsibility of leadership through several institutional incentives. The

first, is the prestige of the role and the power distinction that enables the state to

 

l‘Present gain from defecting is weighted more heavily, while future gains from cooperation are weighted
more lightly. As such, if a discount rate (d) is sufficiently high, a game will most likely be one-shot. Thus
any game that depends on a large number of repetitions also relies on the discount rate not being too high
(Rasmusen, 1989).

”The leader is analogous to a state in the international system that has superior capabilities to the other
member states, or more generally a higher concentration of power relative to the other states in the regime.
Separate payoff functions will be specified for the leader because of the costs associated with punishing
defectors.

144

manipulate the production of the good according to its own preferences.” Second,
the leader is able to obtain residual benefits. Consider that leaders, to some
degree, are responsible for distributing the benefits produced by the regime to the
followers, and in so doing, reward and punish.'7 As a residual claimant, leaders
receive a share equal to o, where (0 < o < l), which is a proportion of each benefit
plus all undistributed remaining benefits.

Third, to understand the evolutionary nature in the role of leader and to
capture variance in the type of leader that may exist overtime, leaders are
characterized as either weak or strong. Strong leaders are designated as L, and
weak, Lw. The primary distinction between the two types is that weak leaders face
a real cost associated with sanctioning a recalcitrant member of the regime. Thus
the payoff configurations for a leader vary according to not only the number of
defections that take place, but also according to the weight in which the cost of

punishment burdens the leader. Because weak leaders are more likely to have

 

”For example, consider the good to be the initiation and maintenance of a free trade structure. Without
question, variance in previous leaders (e. g. hegemon) preferences for maintaining free markets. Some
scholars contend that Great Britain’s tenure as hegemon in the late 18th century enabled it to carve a
market structure amiable to its own industrial development by initiating various bilateral agreements (Gilpin
1987; Kindleberger 1986). This contrasts with the United States reign as hegemon, in which it maintained
free trade by pursing multilateral strategies (Yarborough & Yarborough, 1992). An interesting question is
what domestic and international factors explain why this variation exists?

I7Note that the term follower refers to nation states that are part of the regime. For example, in this
conceptualization, the G7 are led by the United States, while Great Britain, Japan, Germany, France,
Canada and Italy are considered to be the followers.

145

fewer capabilities at their disposal than strong leaders, the weight, represented by
a), will generally be larger for weaker nations than it is for strong nations.”

Given these preliminary assumptions about the structure of the model, it is
possible to specify the payoffs for a leader as follows:

Leader’ 5 Payoffs

x

. = b + _
v“(s,) o (s,) n(l -o) b(s,) cox

 

Again, the payoffs associated with cooperating entails a cost parameter. To
the extent that cooperation constrains state behavior, a type of domestic
opportunity cost is levied upon the followers.” As such, payoffs for followers
represent a portion of the sum total of benefits produced by the regime, minus the
costs associated with cooperating, represented by r. The payoffs for followers are:

Follower’ s Payoff

 

(1 —o) beg]
—— — 1:

v,,(s,) = N A, cooperate get reward
= - 1 A2 cooperate get punished
(1 -0) bot)
= ———N——' A3 defect get reward
= 0 A4 defect get punished

 

”Note that the cost of sanctioning defectors affects the strategies of weak leaders. This will be elaborated
upon later in the development of the model.

l"’For simplicity, costs are assumed homogenous and are held constant for each interaction.

146

Thus, the regime, asan organizational team, consists of a leader and
follower. To guarantee that each state plays fair, each type of actor requires its
own respective trigger strategy to insure that cooperation is encouraged and
maintained in repeated rounds. Note however, that the followers need two types
of trigger strategies. The first is necessary to restrict defection by fellow followers

and to insure that the leader distribute benefits accordingly. This is the g-trigger:

g—trigger t = O cooperate

t > 0 cooperate if b(s,) = B" for all (t ‘< t), defect otherwise

It is the grim trigger strategy. It specifies that a follower will cooperate until
another follower defects or in the event a leader fails to reward cooperative
behavior. With respect to the latter, it is conceivable that a leader may at times
defer or deny benefits to member states without fear of reprisal. This is analogous
to a malevolent hegemon argument. Thus, in order to restrict such activities,
followers need this trigger strategy to specifically address retributions stemming
from noncooperative behavior of a leader.

The other trigger for the followers is the s-trigger, which is used as a
strategic maneuver for a follower wanting to defect because it is in their own best

interest.

s—trigger t = 0 cooperate

t > 0 cooperate if b(s,) 2 Bn_,, the leader rewards follower on
all (t ‘< t), defect otherwise

147

Proposition (1): In a game where followers can use the g-trigger and s-trigger,
full cooperation can be sustained as a subgame perfect
equilibrium if and only if d2 (10y/B9) - (Bl0 - B9)/B9?O

The leader also has a trigger strategy. However, the manipulation of the
trigger will vary among different types of leaders, which is determined by the
variance in their capabilities.2| Although followers in the regime have triggers
available to induce members to cooperate over the long run (e. g. the grim trigger),
leaders also play an important role in maintaining cooperation through their own
duties associated with rewarding cooperative behavior and punishing defection.

Generally speaking, the leader serves the role of distributing the benefits produced

by the regime. The strong leader punishes would be defectors at a cost lower that

weighted less than for a weak leader. Below is the trigger for the strong leader:

Ls-trigger t = 0 reward all followers

t > 0 reward follower if s", = l on all (t ‘< t),
punish nation -state n, otherwise.

Since a weak leader will feel the effects of punishing more so than a strong
leader, weak leaders have a separate and distinct trigger strategy at their disposal.
The main distinction between the two triggers is that the weak leader will punish a
defector, only when the benefits of doing so outweigh the costs of sanctioning (e. g.

Benefits-Costs > 0). Moreover, the weak leader merely distributes the residuals,

 

20Refer to Appendix B, section 3.2 for the proof of this proposition.

2'To simplify the analysis, variance in a leader’s capabilities are dichotomized into a strong leader (L,) and a
weak leader (Lw).

148

rewarding all followers and sanctioning no one if the costs of punishing are too

high.

Lw-trigger t = 0 reward all followers

t > 0 reward follower if s"; = l (cooperate), additionally if

ob(s,) + ——-§-—] s on:
n(1-o)b(s,)

 

punish all followers if s"; = 0 (defect), additionally if

 

ob(s,) + —x__] > (ox
n(l-o)b(s,)

Proposition (2): A leader’s trigger strategy can deter defection by a follower if
and only if d2 (lOy/(l-o)Bg) - (Blo - B,,)/Bg?2

Generally speaking, when followers know they face a strong leader, they
can be induced to cooperate. However, under weak leadership, their is a potential
for followers to get away with defection. Moreover, an additional issue that may
arise, concerns the credibility of a leader’s ability to punish. That is, if the
capability of the leader begins to diminish, such that the costs of punishing
outweigh the benefits the leader would obtain from maintaining cooperation in the
regime, how will the behavior of the followers change? These issues will be tested

with the propositions that follow.

 

22Refer to Appendix B, section 3.3, for the proof of this proposition.

149

Proposition (3a): The higher the probability that a follower will confront a
strong leader 0, the less likely the follower is to defect in a
subsequent round.

As a corollary, the following proposition will also be tested:

Proposition (3b): The convergence of defection will not be effected by the
probability of punishment given a weak leader, 7.

Proposition (4a): The higher the probability that a follower will confront a
weak leader (1-0), the more likely the follower is to defect in
a subsequent round.

As a corollary, the following proposition will also be tested:

Proposition (4b): The convergence of defection will be effected by the
probability of punishment given a weak leader, 7. As y
increases, so will defection.

The Maintenance of Cooperation Under Leadership Uncertainty

Ba esian U atin

In order to examine propositions (3a), (3b), (4a) and (4b) it is necessary to
consider how followers obtain information about the leaders capabilities and how
they use this information to determine their course of action in a subsequent
round. Again, when followers know they face a strong leader, they can be induced
to cooperate. Under weak leadership however, there is a potential for getting away

with defection. If followers do not know whether they face a strong or weak

150

leader, they observe the actions of a leader over time to form a belief about the
type of leader they face. This process, i.e. updating of beliefs by the followers
about the type of leader, can be modeled as a Bayesian process. As such, the
followers beliefs will be updated through a number of interactions, that will be
calculated according to Baye’s rule. Generally speaking, followers are unable to
distinguish strong leaders from weak leaders as long as weak leaders punish
defection. This is analogous to a pooling equilibria. As soon as a leader fails to
punish defection however, followers then know they face a weak leader. This is
similar to a separating equilibria, where actions of the types of leaders are
distinguishable.

To estimate the probability of punishment for defection, a follower
estimates the chances that they face a weak or strong leader, 0, and the chances
that a weak leader will punish, 7. These estimates are updated over the course of

the game through the use of Bayes’ theorem, where:

W) 1t,(s|t)

u/(tls) =
W) 1t,(s|t) + p(t’) 1t,(s|t’)

 

The specific assumptions made about the structure of the updating process are:

(1) A leader’s type is determined by nature’s move. A leader is strong with
probability (0), and is weak with probability (1-0).

(2) The game has an infinite number of rounds with no discemable conclusion.

(3) The leader does not punish a follower who cooperates.

151

(4) A follower’s beliefs about the leader’s true type (strong or weak) are based
on the outcomes of the previous round(s), as revealed by the leader’s
sanctions.

(5) The probabilities with the leader’s strategies are denoted as follows:

1t,(punishlstrong)=1
1r,(reward|strong)=0
1t,(punishlweak)=y
n,(rewardlweak)=l-y
A follower’ s beliefs about the leaders type are calculated according to

Bayes’ Rule delineated above. Since it is assumed that prob(leader rewardslleader

is strong)=0, and u,(leader is stronglrewards)=0, then conversely, u,(leader is

weaklrewards)=1. Essentially this means that when a follower observes that the
leader has rewarded a defector, she knows without a doubt that the leader is weak.

On the other hand, the leader does have an incentive to punish a follower when it

is both strong and weak. Its decision to punish or reward is based on a cost-benefit

analysis, where a leader punishes defection when the benefits exceed the costs.

Recall that the costs of punishing defection for a weak leader are represented by a

concave cost function (ox, where x equals the number of defectors being punished,

n) equals the cost of each follower sanctioned for noncompliance, and 0<<o<1. A

weak leader punishes a follower(s) who defects when:

‘1' x > (ox
M] -o) b(s,)

 

ob(s,)

In contrast, a leader is unable to punish defections when the costs of punishment

exceed the benefits, or when:

152

x

< (ox
N(1-o) b(s,)

 

ob(s,) +

Thus, a follower’s beliefs about a leader’s type (i.e. whether the leader is

strong or weak) are calculated as followed:

p(strong) * p(punish 1 strong)
p(strong) * p(punish 1 strong) + p(weak) * p(punish 1 weak)

0
0+y—0y

 

u1(strong leader I punish)

 

u/(strong leader I punish)

p(weak) * p(punish 1 weak)
p(weak) * p(punish 1 weak) + p(strong) * p(punish l strong)

 

uj(weak leader I punish)

v-OY
v-YB+6

ul(weak leader I punish)

A Simulation

A follower can determine its expected payoffs from either cooperating or
defecting from the international regime, utilizing the information obtained about
the leader’s type through the leader’s actions in the previous round. The leader
determines the payoff function and the other followers shape the actual payoff
associated with this function. Recall that a follower can receive one of the three
payoff functions: A, when it cooperates; A2 when it defects; A3 when it cooperates
and is punished; and A4 when it defects and is punished by the leader. The order

of preferences is: A2 > A, > A3 > A4. However, the expected payoff function for

153

cooperating will always be A,.23 The expected payoff function for defection is
based upon the beliefs the follower has about the leader’s type and the leader’s
reaction (reward or punish). Given that defection has taken place in the previous
round (assume here that round two is being played, where t=2), the expected

payoff functions are:

EP(Defect) A2(belief leader is weak) + A3(belief leader is strong)

A2(y - 0y) + A30

EPD I
(“if“) e+v+ev e+y+ev

 

 

Given that A3=0, the equation above reduces to:

A2(Y - BY)

EP(Defect) = 0 + Y + By

 

In order to empirically test how the follower’s beliefs affect the strategies in
an iterated game, we calculate the expected payoffs across various probability
values for 0 and y in the equations generated for the expected payoffs for
followers. Theta, the probability that the leader is strong, is varied from .99 to .01.

Gamma, the probability that a weak leader punishes, varies from 0.8 to 0.2 for

 

”For simplicity it is assumed that in the payoff function A, (i.e. where the follower cooperates and is
punished by the leader) is not realistic. Future research will incorporate this payoff function.

154

each separate value of theta. The values for A2 (benefits received from defecting)
are taken from the second row of Table 3.1. The calculations are located in Tables
3.2 through 3.16.

Several conclusions can be drawn from a comparison of the various
parameters and their respective payoffs. In Table 3.1, the payoffs for defecting
always exceed the payoffs for cooperating, resulting in collective action problems.
In the first model, full cooperation can be maintained by the leader if discount
rates are high enough. When the leader’s true costs for punishment are not
revealed to the followers, payoffs for cooperation often exceed the benefits for
defection. How does the leader’s privileged information encourage cooperation?

First, the presence of a strong leader diminishes a follower’s payoffs for
defecting. As 0 (0.1), the payoffs for defection decrease overall,
(EPldefection)-°0. The values range from a high of 1.8 (0:01, y=.8, ZS,=9), to a
low of .0004 (0:99, y=.2, ZS,=1). At small values of 0 (.01) however, the
payoffs for defecting always exceed those from cooperating. This produces the
same result as model 1. When 0>0.25, it is always better to cooperate if two or
less other followers defect. In addition, payoffs for cooperation are better than
those for defection at low values of y when as many as 5 players defect. The
overall trends for values of 0 from (O~ 1) demonstrate the importance of
leadership. As the probability of strong leadership increases, the players fear

punishment more often, and benefit from a strategy of cooperation.

155

Table 3.2

N, payoff when 6:099 and y=0.8

 

Number of Nation-States Which Select the Strategy s,=l (cooperate)

 

 

 

 

Payoffs 2 s,=9 2 s,=8 E s,=7 Z s,=6 2 s,=5 E s,=4 )3 s,=3 E s,=2 25,:1 Z s,=0
s, for N,
s,,=l l 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate
s,,=0 .Ol .01 .01 .01 .008 0.8 0.6 0.4 0.2 0.0
defect
Table 3.3
N, payoff when 0:099 and 7:05

 

Number of Nation-States Which Select the Suategy se.-1 (c00perate)

Payoffs E s,=9 Z s,=8 )3 s,=7 E s,=6 E s,=5 E s,=4 E s,=3 E s,=2 Z s,=l E s,=0
s, for N,

 

 

 

 

s,,=l 1.0 0.8 0.6 0.4 0.2 0.0 -.2 ~.4 -.6 -.8
cooperate
s,,=0 .009 .008 .007 .006 .005 .004 .003 .002 .001 .000
defect

Table 3.4

N, payoff when 6:099 and 7:02

 

Number of Nation-States Which Select the Strategy s,=l (cooperate)

Payoffs Z s,=9 2 s,=8 E s,=7 E s,=6 E s,=5 )3 s,=4 E s,=3 E s,=2 2 s,=l Z s,=0
s, for N,

 

 

 

s,,=l 1.0 0.8 0.6 0.4 0.2 0.0 -.2 —.4 -.6 -.8
cooperate
s,, =0 .004 .003 .003 .002 .002 .002 .001 .001 .000 .000

defect

 

 

 

 

N, payoff when 6:075 and 7:08

156

Table 3.5

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Number of Nation-States Which Select the Strategy s,=l (cooperate)
Payoffs E s,=9 E s,=8 E s,=7 E s,=6 E s,=5 2 s,=4 Z s,=3 Z s,=2 E s,=l X s,=0
s, for N,
s,,=l l 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate
s,,=0 .38 .34 .29 .25 ..21 .17 .13 .08 .04 0.0
defect

Table 3.6
N, payoff when 0:075 and 7:05

Number of Nation-States Which Select the Strategy s,=l (cooperate)
Payoffs Zs,=9 2s,=8 2sF7 Es,=6 2s,=5 2s,=4 Es,=3 Es,=2 2s,=l Es,=0
s, for N,
s,,=1 1.0 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate
3,, =0 .26 .23 .20 .17 .14 .1 l .09 .06 .03 .00
defect

Table 3.7
N, payoff when 0:075 and y=0.2

Number of Nation-States Which Select the Suategy s,=l (cooperate)
Payoffs 2 s,=9 I} s,=8 23 s,=7 E s,=6 E s,=5 E s,=4 E :53 Z s,=2 2 s,=l E s,=0
s, for N,
s,,=1 1.0 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate
s,, =0 .11 .10 .09 .08 .06 .05 .04 .03 .01 .00
defect

 

 

157

Table 3.8

N, payoff when 0:05 and 7:08

 

Number of Nation-States Which Select the Strategy s,=l (cooperate)

 

 

 

 

 

 

 

 

 

 

 

 

 

Payoffs E s,=9 E s,=8 E s,=7 E s,=6 E s,=5 Z s,=4 )3 s,=3 Z s,=2 E s,=l E s,=0
s, for N,
s,,:l 1 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate
s ,, =0 .80 .71 .62 .53 .44 .36 .27 .18 .09 0.0
defect
Table 3.9
N, payoff when 8:05 and 7:05
Number of Nation-States Which Select the Strategy s,=l (cooperate)

Payoffs 2 s,=9 E s,=8 E s,=7 )3 s,=6 )3 s,=5 E s,=4 )3 sF3 E s,=2 2 sFl 2 s,=0
s, for N,

s,,= 1.0 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate

s ,, =0 .60 .53 .47 .40 .33 .27 .20 .13 .07 .00
defect

Table 3.10
N, payoff when 0:05 and 1:02
Number of Nation-States Which Select the Suategy s,=l (cooperate)

Payoffs Z s,=9 Z s,=8 E s,=7 2 s,=6 E sFS Z s,=4 E s,=3 2 s,=2 E s,=l E s,=0
s, for N,

s,, =1 1.0 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate

s,,=0 .30 .27 .23 .20 . I7 .13 .10 .07 .03 .00

defect

 

 

 

 

N, payoff when 6:0.25and y=0.8

158

Table 3.11

 

Number of Nation-States Which Select the Strategy s,=1 (cooperate)

 

 

 

 

 

 

 

 

 

 

 

 

Payoffs 2 s,=9 )3 s,=8 E s,=7 E s,=6 Z s,=5 2 s,=4 E s,=3 E s,=2 2 s,=l E s,=0
s, for N,
5,, =1 1 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate
s,,-=0 1.27 1.13 .99 .85 .71 .56 .42 .28 .14 0.0
defect
Table 3.12
N, payoff when 6:025 and y=0.5
Number of Nation-States Which Select the Strategy s,=1 (cooperate)
Payoffs E s,=9 E s,=8 2 s,=7 E s,=6 E s,=5 E s,=4 E s,=3 2 s,=2 2‘. s,=l Z s,=0
s, for N,
3,, =1 1.0 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate
3,, =0 1.08 .96 .84 .72 .60 .48 .36 .24 .12 .00
defect
Table 3.13
N, payoff when 0:025 and y=0.2
Number of Nation-States Which Select the Strategy s,=l (cooperate)
Payoffs E s,=9 E s,=8 E s,=7 Z s,=6 E s,=5 )3 s,=4 )3 s,=3 )3 s,=2 )3 s,=l E s,=0
s, for N,
s,,=1 1.0 0.8 0.6 0.4 0.2 0.0 -.2 -.4 '26 -.8
cooperate
3,, =0 .675 .60 .53 .45 .375 .30 ..225 .15 .08 .00

defect

 

 

 

 

 

159

Table 3.14

N, payoff when 8:001 and 1:08

 

Number of Nation-States Which Select the Suategy s,=l (cooperate)

Payoffs 2 s,=9 E s,=8 E s,=7 2 s,=6 E s,=5 E s,=4 2 s,=3 E s,=2 E s,=l 2 s,=0
s, for N,

 

 

 

 

3,, =1 1 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate
s,,=0 1.8 1.6 1.4 1.2 .99 .79 .59 0.4 0.2 0.0
defect

Table 3.15

N, payoff when (#001 and y=0.5

 

Number of Nation-States Which Select the Strategy s,=l (cooperate)

Payoffs Es,=9 Es,=8 ZsF7 Es,=6 Es,=5 Esq-=4 EsF-3 Err-:2 Es,=l Es,=0
s,forN,

 

 

 

 

s ,, =1 1.0 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate
3,, =0 1.8 1.6 1.4 1.2 .98 .78 .59 .39 .20 .00
defect

Table 3.16

N, payoff when 0:001 and 7:02

 

Number of Nation-States Which Select the Strategy s,=l (cooperate)

Payoffs 2 s,=9 Z s,=8 E s,=7 E s,=6 8 s,=5 2 s,=4 E s,=3 E s,=2 E s,=1 E s,=0
s, for N,

 

 

 

s,,=l 1.0 0.8 0.6 0.4 0.2 0.0 -.2 -.4 -.6 -.8
cooperate
s,,-=0 1.7 1.5 1.3 1.1 .95 .76 .57 .38 .19 .oo

defect

 

 

 

 

160

How does y, the probability that a weak leader punishes, affect the
follower’s payoffs for defection? As y increases from 0.2 to 0.8, the payoffs for
defection in comparison to the payoffs for cooperation increase. For example, the
examination of Tables 3.8 through 3.10 illustrate this point when 0:0.5. When,
y=0.8, the payoff for defecting when {s,=9 equals 0.8. When y decreases to 0.2,
the payoff for defection when [s,=9 equals 0.3. Thus it does not pay for a weak
leader to bear great costs for punishing defectors, because it appears that the
followers will not be fooled. They are actually more likely to defect when a weak
leader punishes, than when it rewards. A weak leader’s best strategy is to be
straightforward about its type, and to punish only when the benefits of doing so

exceed the costs.

Conclusion

Most models analyzing cooperation and coordination in regimes have
focused exclusively on the leader, while ignoring the role of followers. The
hegemonic stability literature in particular focuses on the role of the hegemon,
neglecting how followers influence one another. The model described in this
research contributes to this literature by explicitly analyzing how followers are
affected by variances in the capacity of a leader to punish within the institutional

structure of an international regime. The simulation of the formal model found

161

that the payoffs for defecting generally never outweigh the payoffs for
cooperating.

Moreover, the formal model developed goes well beyond typical
explications of two actor models by incorporating multiple actors in an iterated
environment. By conceiving international regimes in terms of a production by
teams model, the dynamics of nation-states seeking to coordinate policy in an
economic regime can be captured and examined under various conditions of
leadership and information structure. Results from the simulation suggest that the
payoffs from cooperation under conditions of strong leadership, do not change for
regimes with two or more states (followers) cooperating. However, for conditions
in which there exists some uncertainty in a leader’ 8 ability to punish (i.e. a weak
leader with a low probability of punishing), there exists little incentive for states to
cooperate. The payoffs from defecting under such conditions outweigh the
payoffs for cooperating. This result seems to generally hold for regimes with four
or more states (followers) defecting.

Generally speaking, the theoretical framework developed in this research
affords a generalization away from traditional hegemonic stability theory and its
effect on the survival of a regime, by illustrating the importance of both leadership
and followers. This is especially crucial as international regimes tend to have
multiple actors which interact. These interactions subsequently affect the nature

and duration of regimes. Although this research does not specifically delineate

162

how follower’s interact, other than formally illustrating the potential threat of
triggering punishment, this is a topic that future research in this area should
explore. Specifically, the role of Bayesian updating among followers can afford
insights into the credibility of punishment as a deterrent upon a state’s defection,
even under conditions of weak leadership. In fact, this insight would serve to
formally explain why it is that cooperation is maintained within international
regimes during periods of hegemonic decline, when hegemonic stability predicts

otherwise.

APPENDICES

163

APPENDIX A

Serial Correlation Tests

Because most series are characterized by AR(1) processes, the assumption
is typically made that most series are characterized by some degree of serial
correlation. Similarly as the univariate properties of a series are analyzed to
determine their cross correlations, residuals from time series regressions are
analyzed to diagnose whether serial correlation is present.

There exists numerous tests for autocorrelation. However, despite the
limitations of the Box-Pierce Q-statistic, most statistical software packages report
this statistic (Maddala 1994, p.542). This statistics is inappropriate when a lagged
endogenous term is present in the equation from which the residuals are extracted.
When a lagged endogenous variable is present, an alternative diagnostic tool needs
to be utilized, such as the Breusch-Godfrey test (Maddala 1994, p.541). Both the
Box-Pierce Q and Breusch-Godfrey are discussed below.

I. Box-Pierce Q

Q = T25? ~x2(Ldof)

2 er...

2 (=j_+_—_l

where r. =

1 r
X 62
1
1:1

The test is the null (Ho: Autocorrelation) against an alternative (H,: white noise).
When L=20, the appropriate critical values are: (. 01) 28. 4, (. 05) 31 .4 (. 10) 37. 6.
Note this diagnostic test is inappropriate in autoregressive models (or models with
lagged endogenous variables).

164

II. Breusch-Godfrey (use with lagged endogenous variable present).

TR2 ~ x2 (P dof)

The test consists of regressing the OLS residuals (saved from regression of
interest) e, on x,, e,,,, e,_,, ..., e,p,(filling in missing values for lagged residuals with
zeros). TR2 is the statistic to calculate, which is subsequently used to compare with
a set of critical values distributed as a chi-squared distribution, using P degrees of
freedom. (i.e. the number of lagged residuals on the right hand side of the
equation). The test consists of a null of (Ho: autocorrelation),against the
alternative of (H,: white noise). When P=l , the appropriate critical values are:
(.01) 2.71, (.05) 3.84, (.10) 6.63.

165

APPENDIX B

Formal Model Proofs

Note: The formal model presented is a generalized version of a model developed
in Bianco & Bates (1990). As such, most of these proofs, albeit generalized to the
n-person (nation state) case, are derived in the Appendix of Bianco & Bates
(1990). Where relevant I make the necessary amendments to fit my model.

3.1
Total benefits derived from nation-states which cooperate in the regime:

B(S,) = a,(s,, + s2, + + s,,”)

Where a, represents a constant rate of cooperation among nation-states. For the
purposes of this model it is assumed that a, is a constant term with a value greater
than 0. Thus, the amount of benefits will be a linear function of the number of
nation-states which cooperate.

3.2
Suppose that we are referring to an economic regime in which G—lO
countries are members.

10C
— ’ (Bro ' B9)

d2 9

 

B9

From Table 3.1 we have the values associated with total sum of benefits produced
by a regime, conditioned on the number of followers cooperating (i.e. B,,,=10,
B9=9, etc.). Total benefits are a linear function of the number of followers that
cooperate. This equation essentially maintains that a nation-state will choose not
to cooperate unless the outcome for cooperating is greater than the outcome for

166

defecting. Moreover, the latter situation will not hold unless the discount factor is
(d2 .01). This is equivalent to the equation belpw.

 

 

 

BN
N-C 2 BN1
l-d N

3.3

If an individual nation-state (n,), along with the other member nation-states,
uses a trigger strategy, the payoff for n, for the entire game is:

 

 

 

However, if nation-state (n,) defects and the leader retaliates, (n,)’s payoff is:

((1 '0) B“)
N

 

Thus a leader is credible only if:

11- B
1——< ~-1
, >(l—d) BN_,

l-d

Likewise, if the leader defects on the first round (t), thereafter, the other member
nation-states would employ their trigger strategies in the following rounds i.e.
(t+1). Thus, this would give the leader a payoff (BN + 0).

167

However, if the leader continues to use a trigger strategy, the payoff from iteration
(t) forward is:

Thus, the trigger strategies for the member nation-states is an effective threat
against the leader if:

 

 

 

This is equivalent to:
d 2 (1-0)

 

LIST OF REFERENCES

LIST OF REFERENCES

Adelman, Irma. 1965. “Long Cycles - Fact or Artifact?” American Economic
Review 55(3):444-63.

Akerman, Johan. 1932. Economic Progress and Economic Crisis. London:
Macmillan. Reprint ed., Philadelphia: Porcupine Press, 1979.

Alchian, Armen A. and Harold Demsetz. 1972. “Production, Information Costs,
and Economic Organization.” American Economic Review 62:777-795.

Anjaria, Shaplendra J ., Zubair Iqbal, N aheed Kirrnani, and Lorenzo L. Perez.
1982. Developments in International Trade Policy. Occasional Paper
No.16, Washington, DC: International Monetary Fund.

Arnold, Douglas R. 1990. The Logic of Congressional Action. New Haven: Yale
University Press.

Axelrod, Robert. 1981. “The Emergence of Cooperation Among Egoists.”
American Political Science Review 75:306-18.

Axelrod, Robert. 1984. The Evolution of Cooperation. New York: Basic Books.

Axelrod, Robert and Robert Keohane. 1985. “Achieving Cooperation Under

Anarchy: Strategies and Institutions.” World Politics 226-254.

168

 

 

169

Bach, Stanley and Steven Smith. 1988. Managing Uncertainty in the House of
Representatives: Adaptation and Innovation in Special Rules. Washington,
DC: Brookings Institute.

Baldwin, Robert E. 1976. "The Political Economy of postwar US. Trade Policy."
The Bulletin, No.4.

Baldwin, Robert E. 1984. "The Changing Nature of US. Trade Policy Since
World War II," in Baldwin, Robert E. and Anne O. Krueger (eds.), The
Structure and Evolution of Recent US. Trade Policy, Chicago: University
of Chicago Press, pp:5-27.

Banerjee, Anindya, Juan Dolado, John Galbraith and David F. Hendry. 1994. Co-
Integration, Error Correction, and the Econometric Analysis of Non-
Stationary Data. New York: Oxford University Press.

Barr, Kenneth. 1979. "Long waves: A Selective Annotated Bibliography."
Review 2(4).

Bartels, Larry. 1991. "Constituency Opinion and Congressional Policy Making:
The Reagan Defense Buildup." American Political Science Review.
Vol.(85)2: pp.457-474.

Bates, Robert, Phillip Brock, and Jill Tiefenthaler. 1991. “Risk and Trade

Regimes: Another Exploration.” International Organization 45: 1-18.

170

Bauer, Raymond, Ithiel De Sola Pool and Lewis Anthony Dexter. 1963.
American Business and Public Policy: The Politics of Foreign Trade.
Chicago: Aldine-Atherton, Inc.

Beck, Nathaniel. 19913. "The Time Series Analysis of War." Paper Presented at
the American Political Science Meetings.

Beck, Nathaniel. 1991b. "The Illusion of Cycles in International Relations."
International Studies Quarterly 35(4):455-76.

Becker, Gary S. 1983. "A Theory of Competition among Pressure Groups for
Political Influence." Quarterly Journal of Economics. Vol.98, pp.37l-400.

Bhagwati, Jagdish and Hugh Patrick. 1990. Aggressive Unilateralism: America ’s
301 Trade Policy and the World Trading System. Ann Arbor: University of
Michigan Press.

Bianco, William and Robert Bates. 1990. “Cooperation by Design: Leadership,
Structure, and Collective Dilemmas.” American Political Science Review
84: 133-147.

Box, G. E. P. and D. A. Pierce. 1970. "Distribution of Residual Autocorrelations
in Autoregressive Integrated Moving Average Time Series Models."

Journal of the American Statistical Association, 65: 1509- 1526.

171

Box, G.E.P. And G. C. Taio. 1975. “Intervention Analysis with Applications to
Economic & Environmental Problems.” Journal of the American Statistical
Association: 70-92.

Box-Steffensmeier, Janet, Laura Arnold and Christopher Zorn. 1995. “Strategic
Position Taking and the Timing of Voting Decisions in Congress: A Study
of the North American Free Trade Agreement.” Ohio State University
manuscript.

Breusch, T. 1978. "Testing for Autocorrelation in Dynamic Linear Models."
Australian Economic Papers, 17: 334-55.

Cassing, James, Timothy McKeown and Jack Ochs. 1986. "The Political
Economy of the Tariff Cycle." American Political Science Review. Vol.80,
No.3, pp.843-862.

Caves, Richard. 1976. "Economic Models of Political Choice." Canadian Journal
of Economics, May(6), pp.278-300.

Cline, William R. 1982. Reciprocity: A New Approach to World Trade Policy?
Washington, DC: Institute for International Economics.

Cline, William R. 1989. "Macroeconomic Influences on Trade Policy,"

American Economic Review, Vol.79, No.2, pp: 123-127.
Conybeare, John. 1987. Trade Wars: The Theory and Practice of International

Commercial Rivalry. New York: Columbia University Press.

172

Conybeare, John A. C. 1991. "Voting for Protection: an electoral model of tariff
policy." International Organization, Winter 45 (1), pp: 57-79. v

Cooley, Thomas F. and Stephen LeRoy. 1985. "Atheoretical Macroeconomics: A
Critique." Journal of Monetary Economics 16:283-308.

Cooper, Richard. 1968. The Economics of Interdependence: Economic Policy in
the Atlantic Community. New York: McGraw-Hill.

Coughlin, Cletus C. 1985. "Domestic Content Legislation: House Voting and the
Economic Theory of Regulation." Economic Inquiry. July, pp.437-48.

Davidson, J ., D. Hendry, D. Srba and S. Yeo. 1978. "Econometric Modeling of
the Aggregate Time Series Relationship Between Consumers' Expenditure
and Income in the United Kingdom." Economic Journal 88:661-92.

Deardoff, Alan and Robert Stern. 1979. "American Labor's Stake in International
Trade," in Tariffs, Quotas & Trade: The Politics of Protectionism, eds.
Walter Adams et al. San Francisco: Institute for Contemporary Studies, pp:
125-148.

Deese, David. 1994. The New Politics of American Foreign Policy. New York:
St. Martin’s Press.

Destler, TM. 1992. American Trade Politics. Washington DC: Institute for

International Economics.

173

Destler, I.M., John Odell, and Kimberly Elliott. 1987. Anti-Protection: Changing
Forces in United States Trade Policies. Washington DC: Institute for
International Economics.

Dickey, David A. 1975. "Estimation and Hypothesis Testing in Nonstationary
Time Series." Ph.D. dissertation, Iowa State University.

Dickey, David A. and Wayne A. Fuller. 1979. "Distribution of the Estimators for
Autoregressive Time Series With A Unit Root." Journal of the American
Statistical Association 74(366):427-431.

Dickey, David A. and Wayne A. Fuller. 1981. "Likelihood Ratio Statistics for
Autoregressive Time Series With a Unit Root." Econometrica 49(4): 1057—
72.

Diebold, F. and G. 'Rudenbusch. 1989. "Long Memory and Persistence in
Aggregate Output." Journal of Monetary Economics 24: 143-59.

Doan, Thomas, Robert Litterman, and Christopher Sims. 1984. "Forecasting and
Conditional Projection Using Realistic Prior Distributions." Econometric
Reviews 3:1—100.

Dobson, Wendy. 1991. Economic Policy Coordination: Requiem or Prologue?
Washington, DC: Institute for International Economics.

Dolado, Juan J ., Tim Jenkinson and Simon Sosvilla-Rivero. 1990. "Cointegration

and Unit Roots." Journal of Economic Surveys 4(3):249-273.

 

 

174

Doran, Charles. 1989. “Power Cycle Theory of Systems Structure and Stability:
Commonalities and Complementaries.” In Handbook of War Studies, ed.
Manus Midlarsky. New Zealand: Unwin Hyman, pp:83-110.

Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper
& Row.

Dufour, Jean-Marie. 1982. “Recursive Stability of Linear Regression
Relationships.” Journal of Econometrics 19:31-36.

Durr, Robert. 1992. "An Essay on Cointegration and Error Correction Models."
Political Analysis 4(2): 185-228.

Engle, Robert F. and S. Yoo. 1987. "Forecasting and Testing in Cointegrated
Systems." Journal of Econometrics, 35 :143-59.

Engle, Robert F. and C.W.J. Granger. 1987. "Cointegration and Error Correction:
Representation, Estimation, and Testing." Econometrica, 55(2): 251-276.

Engle, R. F., D. F. Hendry and J. F. Richard. 1983. “Exogeneity.” Econometrica
51:277-304.

Epic/MRA Report. East Lansing, Michigan, Vol. 1, No.20, December 15, 1993.

Ericsson, Neil R. 1992. "Cointegration, Exogeneity, and Policy Analysis: An
Overview." Journal of Policy Modeling 14(3):251-280.

Feenstra, Richard. 1992. "How Costly is Protectionism?" Journal of Economic

Perspectives. Vol.6(3), pp.159-178.

175

Fenno, Richard F. 1978. Homestyle: House Members and Their Districts.
Boston: Little Brown.

Findlay, Ronald and Stanislaw Wellisz. 1986. "Tariffs, Quotas and Domestic-
Content Protection: Some Political Economy Considerations." Public
Choice. 50, pp. 221-42.

Fiorina, Morris. 1974. Representatives, Roll-Calls and Constituencies. Lexington
Massachusetts: Lexington Books.

Freeman, John. 1983. "Granger Causality and the Time Series Analysis of
Political Relationships." American Journal of Political Science 27:327-58.

Freeman, John, John Williams and Tse-nrin Lin. 1989. "Vector Autoregression
and the Study of Politics." American Journal of Political Science 33:842-
77.

Frey, B.S. 1984. "The Public Choice View of International Political Economy."
International Organization 38 (1): l99-223.

Fuller, W. A. 1976. Introduction to Statistical Time Series. New York: Wiley.

Gallarotti, Giulio. 1985. "Towards a Business-Cycle Model of Tariffs."
International Organization. 39, pp.155-l87.

Gilpin, Robert. 1981. War and Change in World Politics. Cambridge: Cambridge

University Press.

176

Gilpin, Robert. 1987. The Political Economy of International Relations. New
Jersey: Princeton University Press.

Gochman, Charles and Zeev Maoz. 1984. “Militarized Interstate Disputes, 1816-
1976.” Journal of Conflict Resolution 28:585-615.

Godfrey, L. 1978. "Testing Against General Autoregressive and Moving Average

Error Models When the Regressors Include Lagged Endogenous Variables."
Econometrics 46: 1293-1302.

Goldstein, Joshua. 1985. "War and the Kondratieff Upswing." International
Studies Quarterly 29:411-441.

Goldstein, Joshua. 1987. "Long Waves in Production, War and Inflation: New
Empirical Evidence." Journal of Conflict Resolution 31(4).

Goldstein, Joshua. 1988. Long Cycles: Prosperity and War in the Modern Age.
New Haven: Yale University Press.

Goldstein, Joshua. 1991. "A War-Economy Theory of the Long Wave." In
Business Cycles: Theories, Evidence and Analysis, Niels Thygesen,
Kumaraswamy Velupillai and Stefano Zambelli (eds.). New York: New
York University Press.

Gowa, Joanne. 1989. "Bipolarity, Multipolarity, and Free Trade." American

Political Science Review 83(4): 1245-56.

177

Gowa, Joanne. 1994. Allies, Adversaries and International Trade. New Jersey:
Princeton University Press.

Gowa, Joanne and Edward D. Mansfield. 1993. "Power politics and International
Trade." American Political Science Review 87:408-20.

Granato, Jim and Renee Smith. 1994a. "Exogeneity Inference and Granger
Causality, Part I: The Case of Stationary Regressors." The Political
Methodologist 5(2):24-28.

Granato, Jim and Renée Smith. 1994b. "Exogeneity Inference and Granger
Causality, Part H: The Case of Integrated Regressors." The Political
Methodologist 6( 1 ):23-26.

Granger, C. W. J. 1969. “Investigating Causal Relations by Econometric Models
and Cross-Spectral Methods.” Econometrica 36:424-438.

Granger, C. W. J. 1980. “Testing For Causality: A Personal Viewpoint.” Journal
of Economic Dynamics and Control 2:3229-352.

Granger, C. W. J. 1983. “Co-Integrated Variables and Error-Correcting Models.”
Economics Department Discussion Paper No. 81-13. University of
California, San Diego.

Granger, C. W. J. 1988a. "Some Recent Developments In A Concept of

Causality." Journal of Econometrics 39:199-211.

178

Granger, C. W. J. 1988b. "Causality, Cointegration and Control." Journal of
Economic Dynamics and Control 12:551-559.

Granger, CW. and J. Joyeux. 1980. "An Introduction to Long Memory Time
Series Models and Fractional Integration." Journal of Time Series Analysis
(1)215-39.

Granger C.W.J. and TH. Lee. 1989. "Investigation of Production, Sales and
Inventory Relationships Using Multicointegration and Non-Symmetric
Error Correction Models." Journal of Applied Econometrics 4: 145-159.

Granger, C.W.J. and P. Newbold. 1974. "Spurious Regressions in Econometrics."
Journal of Econometrics 2:1 11-120.

Greene, William. 1990. Economic Analysis. New York: Macmillian.

Groves, Theodore. 1973. “Incentives in Teams.” Econometrica, 41(4):617.

Groves, Theodore and John Ledyard. 1977. “Optimal Allocation of Public
Goods: A Solution to the ‘Free Rider’ Problem.” Econometrica, 45(4):783.

Haggard, Stephan and Beth Simmons. 1987. “Theories of International
Regimes.” International Organization 41(3):491-517.

Hansen, Wendy. 1990. "The International Trade Commission and the Politics of

Protectionism." American Political Science Review, 84(1):21- 46.

Harvey, A. 1991. Time Series Models. OxfordzPhillip Allan.

179

Hendry, David F. 1993. Econometrics: Alchemy or Science? Cambridge:
Blackwell.

Hufbauer, Gary C., Jeffrey J. Schott and Philip L. Martin. 1993. NAFTA Briefing
Book. Washington, D.C.:Institute for International Economics.

Jackson, John and John Kingdon. 1990. "Ideology, Interests Groups Scores, and
Legislative Votes."

Jacobson, Gary C. 1987. The Politics of Congressional Elections. Boston: Little
Brown and Company.

Jacobson, Gary C. 1990. The Electoral Origins of Divided Government. San
Francisco: Westview Press.

J ohansen, S. 1988. "Statistical Analysis of Cointegration Vectors." Journal of
Economic Dynamics and Control 12:231-254.

Johansen, S. 1991. “Estimation and Hypothesis Testing of Cointegration Vectors
in Gaussian Vector Autoregressive Models.” Econometrica 59: 1551-1580.

J ohansen, S. 1992a. "Cointegration in Partial Systems and the Efficiency of
Single Equation Analysis." Journal of Econometrics, 52:389-402.

Johansen, S. l992b. "Testing Weak Exogeneity and the Order of Cointegration in

UK Money Demand." Journal of Policy Modeling, 14: 313-34.

180

Johansen, S. and K. Juselius. 1990. "Maximum Likelihood Estimation and
Inference on Cointegration - With Applications to the Demand for Money."
Oxford Bulletin of Economics and Statistics, 52(2): 169-210.

Kalt, Joseph. 1984. "Capture and Ideology in the Economic Theory of Politics."
American Economic Review. June 1984, 74, pp.279-300.

Kalt, Joseph and Mark Zupan. 1990. "The Apparent Ideological Behavior of
Legislators: Testing for Principal -Agent Slack in Political Institutions."
Journal of Law & Economics. Vol.33, pp.103-13l.

Keohane, Robert. 1980. “The Theory of Hegemonic Stability and Changes in
International Economic Regimes, 1967-1977.” In Change in the
International System, eds. Ole R. Holsti, Randolph M. Siverson and
Alexander L. George. Boulder, Colorado: Westview Press, pp:131-162.

Kemell, Samuel. 1977. "Toward Understanding 19th Century Congressional
Careers: Ambition, Competition, and Resolution." American Journal of
Political Science. Vol.21, pp.669-93.

Kindleberger, Charles. 1981. “Dominance and Leadership in the International
Economy: Exploitation, Public Goods, and Free Rides.” International
Studies Quarterly 25:242-54.

Kindleberger, Charles. 1986. The World in Depression, 1929-1939. Second ed.

Berkeley: University of California Press.

181

Kingdon, John. 1989. Congressmen 's Voting Decisions. Ann Arbor: The
University of Michigan Press.

Kondratieff, Nikolai D. 1935. "The Long waves in Economic Life." Review of
Economic Statistics 17(6): 105-15.

Krasner, Stephen. 1977. “Structural causes and Regime Consequences: Regimes
as lntervening Variables.” In Stephen Krasner ed., International Regimes,
pp: 1-21.

Krehbiel, Keith. 1993. "Constituency Characteristics and Legislative
Preferences." Public Choice 76: 21-37.

Krehbiel, Keith and Douglas Rivers. 1988. "The Analysis of Committee Power:
An Application to Senate Voting." American Journal of Political Science.
Vol.32(4): pp.1151-1l74.

Kreps, David. 1990. A Course in Microeconomic Theory. Princeton: Princeton
University Press.

Kuczynski, Thomas. 1980. "Have There Been Differences Between the Growth
Rates in Different Periods of the Development of the Capitalist World
Economy Since 1850?" In J. M. Chubb and E. K. Scheuch, eds., Historical
Social Research. Monograph Series. No.6. Stuttgart: Klett-Cotta.

Kwiatkowski, D. P. Phillips, and P. Schmidt. 1992. "Testing the Null Hypothesis

of Stationarily Against the Alternative of a Unit Root.”

182

Laverge, Real P. 1983. The Political Economy of US. Tarifi‘s: An Empirical
Analysis. Academic Press: New York.

Lawrence, Robert and Charles Schultze. 1990. An American Trade Strategy:
Options for the 1990's. Washington, DC: The Brookings Institution.

Levy, Jack S. 1983. War in the Modern Great Power System, 1495-1975.
Lexington, KY: University of Kentucky Press.

Litterman, R. B. 1986. "Forecasting With Bayesian Vector Autoregression: Five
Years of Experience." Journal of Business & Economic Statistics 4:25-38.

Lohmann, Suzanne and Sharyn O'Halloran. 1992. Delegation and
Accommodation in the US. Trade Policy: Theory and Evidence. Paper
presented at the Midwest Political Science Association.

Lohmann, Suzanne and Sharyn O’Halleron. 1994. “Divided Government and
US. Trade Policy: Theory and Evidence.” International Organization
48(4):595-632.

Lucas, R. E. And Sargent, T. (Eds). 1981. Rational Expectations and
Econometric Practice. London: Allen & Unwin.

Lutz, James. 1991. "Determinants of Protectionist Attitudes in the United States
House of Representatives." The International Trade Journal. Vol.5, No.3,

pp.301-328.

183

MacKinnon, J. G. 1991. "Critical Values For Co-Integration Tests," in R. F.
Engle and C.W.J. Granger (eds), Long-Run Economic Relationships,
Oxford University Press, pp:267-76.

Maddala, G. S. 1983. Limited-Dependent and Qualitative Variables in
Econometrics. Cambridge: Cambridge University Press.

Maddala, G. S. 1992. Introduction to Econometrics. New York: Macmillan
Publishing Company.

Maddala, G. S. 1993. Introduction to Econometrics. New York: Macmillan.

Magee, Stephen, William Brock and Leslie Young. 1989. Black Hole Tartfifs and
Endogenous Policy Theory. New York: Cambridge University Press.

Magee, Stephen and Leslie Young. 1987. “Endogenous Protection in the United
States, 1900-1984,” in Robert M. Stern (ed)., US. Trade Policies in a
Changing World Economy. Cambridge: MIT Press, pp.145-195.

Mansfield, Edward D. 1994. Power, Trade and War. New Jersey: Princeton
University Press.

Marks, Stephen V. 1993. "Econorrric Interests and Voting on the Omnibus Trade
Bill of 1987." Public Choice 75, pp:21-42.

Mayer, Wolfgang. 1984. "Endogenous Tariff Formation." American Economic

Review, December (74), pp.970-85.

184

Mayhew, David. 1974. Congress: The Electoral Connection. New Haven: Yale
University Press.

McArthur, John and Stephen V. Marks. 1988 "Constituent Interests vs. Legislator
Ideology: The Role of Political Opportunity Cost." Economic Inquiry,
pp.461-470.

McKeown, Timothy J. 1984. "Firms and Tariff Regime Change: Explaining the
Demand for Protection." World Politics, 36, pp.215-33.

Mitchell, B. R. 1980. European Historical Abstracts, 1750-1975. New York:
Facts on File.

Modelski, George. 1978. "The Long Cycle of Global Politics and the Nation-
State." Comparative Studies in Society and History 20:214-235.

Modelski, George. 1981. "Long Cycles, Kondratieffs, and Alternating
Innovations: Implications for US. Foreign Policy," in The Political
Economy of Foreign Policy Behavior, (eds.) Charles W. Kegley, Jr. and
Patrick J. McGowan. Beverly Hills: Sage, pp:63-83.

Modelski, George. 1983. "Long Cycles of World Leadership: An Annotated
Bibliography." International Studies Notes 10:1-5.

Modelski, George. 1987a. Long Cycles in World Politics. Seattle: University of

Washington Press.

185

Modelski, George (ed.). 1987b. Exploring Long Cycles. Boulder, CO: Lynne
Rienner.

Modelski, George and William Thompson. 1988. Seapower in Global Politics,
1494-1993. Seattle: University of Washington Press.

Modelski, George and William Thompson. 1995. Leading Sectors and World
Powers: The Coevolution of Global Politics and World Economics.
Columbia: University of South Carolina Press.

Molyneux, Guy. 1994. "Unified 'Opinion Leaders' Best A Reluctant Public." The
Public Perspective, Roper Publication. J anuary/February.

Morrow, Jim. 1994. “Modeling the Forms of International Cooperation:
Distribution Versus Information.” International Organization, 43(3): 387-
423.

Nelson, Douglas. 1987. "Endogenous Tariff Theory: A Critical Survey."
American Journal of Political Science.

Newhouse, Neil S. and Christine L. Matthews. 1994. "NAFTA Revisited: Most
Americans Just Weren't Deeply Engaged." The Public Perspective, Roper
Publication, January/February.

Nowzad, Bahram. 1978. The Rise of Protectionism. Pamphlet Series No. 24,

Washington, DC: International Monetary Fund.

186

O'Halloran, Sharyn. 1994. Politics, Process and American Trade Policy. Ann
Arbor: University of Michigan Press.

Olson, Mancur 1968. The Logic of Collective Actions: Public Goods and the
Theory of Groups. New York: Schocken.

Organski, A. F. K. And Jacek Kugler. 1980. The War Ledger. Chicago: The
University of Chicago Press.

Ostrom, Charles and Renee Smith. 1992. "Error Correction, Attitude Persistence,
and Executive Rewards and Punishments: A Behavioral Theory of
Presidential Approval." Political Analysis 4: 127-184.

Pagan, A. 1987. "Three Econometric Methodologies: A Critical Appraisal."
Journal of Economic Surveys 1:3-24.

Peltzman, Sam. 1976. "Towards a More General Theory of Regulation," Journal
of Law and Economics, August (19), pp.211-240.

Peltzman, Sam. 1984. "Constituent Interests and Congressional Voting," Journal
of Law and Economics, April (27), pp.181-210.

Perot, Ross. 1993. Save Your Job, Save Our Country. New York: Hyperion.

Phillips, P. And P. Perron. 1988. “Testing for a Unit Root in Time Series
Regressions.” Biometrika, 75:33546.

Pollins, Brian M. 1989a. "Does Trade Still Follow the Flag?" American Political

Science Review 83:465-80.

187

Pollins, Brian M. 1989b. "Cooperation, Conflict and Commerce: The Effects of
International Political Interactions on Bilateral Trade Flows." American
Journal of Political Science 332737-61.

Pollins, Brian M. 1995. “Global Political Order, Economic Change and Armed
Conflict: Co-Evolving Systems and the Use of Force.” (forthcoming)
American Political Science Review.

Polsby, Nelson. 1968. "The Institutionalization of the US. House of
Representatives." American Political Science Review. Vol.62, pp.144-68.

Pudaite, Paul R. 1991. “Models for Long Cycles in War and Production.” In
Explicit Mathematical Models for Behavioral Science Theories.
Unpublished Dissertation, Department of Mathematics, University of
Illinois at Urbana-Champaign.

Pratt, J. W. And R. Schlaifer. 1984. “On the Nature and Discovery of Structure.”
Journal of the American Statistical Association 79:9-21.

Rasmusen, Eric. 1989. Games and Infonnation: An Introduction to Game Theory.
Great Britain: Cambridge University Press.

Ray, Edward. 1974. "The Optimum Commodity Tariff and Tariff Rates in
Developed and Less Developed Countries." Review of Economics and

Statistics. Vol.56, pp.369-77.

188

Ray, Edward J. 1981. "Tariff and Nontariff Barriers to Trade in the United States
and Abroad." Review of Economics and Statistics, 89, pp. 105-121.

Ray, Edward J. 1987. "Changing Patterns of Protectionism: The Fall in Tariffs
and the Rise in Non-Tariff Barriers." Northwestern Journal of
International Law and Business. Vol.8, pp.285-327

Richards, Diana. 1993. “A Chaotic Model of Power Concentration in the
International System.” International Studies Quarterly. 37(1):55-72.

Rohde, David. 1991. Parties and Leaders in the Postreform House. Chicago:
University of Chicago Press.

Rosecrance, Richard. 1987. "Long Cycle Theory and International Relations."
International Organization 41(2):283-301.

Sargans, J. D. and A. Bhargava. 1983. "Testing Residuals from Least Squares
Regression fro Being Generated by the Gaussian Random Walk."
Econometrica 511153-74.

Sayrs, Lois W. 1993. "The Long Cycle in International relations: A Markov
Specification." International Studies Quarterly 37(2): 215-237.

Schattschneider E. E. 1935. Politics, Pressures, and the Tarifi’. New York:

Prentice Hall.

189

Shepsle, Kenneth A. 1979. "Institutional Arrangements and Equilibrium in
Multidimensional Voting Models." American Journal of Political Science,
23, pp.27-60.

Shepsle, Kenneth A. and Barry R. Weingast. 1981. "Uncovered Sets and
Sophisticated Voting Outcomes with Implications for Agenda Institutions."
American Journal of Political Science, 28, pp.49-74.

Sims, C. 1980. "Macroeconomics and Reality." Econometrica 48:1-48.

Snidal, Duncan. 1985. “The Limits of Hegemonic Stability Theory.”
International Organization 39:579-614.

Snyder James M. 1990. "Campaign Contributions as Investments: The US.
House of Representatives, 1980-1986." Journal of Political Economy,
vol.98, no.6, pp:1195-1227.

Stigler, George, J. 1971. "The Theory of Economic Regulation." Bell Journal of
Economics and Management Science. Spring, pp.3-21.

Stock, J. And M. Watson. 1988. “Testing for Common Trends.” Journal of the
American Statistical Association 83: 1097-1 107.

Takacs, Wendy E. 1981. Pressures for Protectionism: An Empirical Analysis.
Economic Inquiry. Vol.19, pp.687-93.

Thompson, Joel A. and Gary Moncrief. 1988. "Pursuing the Pork in a State

Legislature," Legislative Studies Quarterly. Vol 13, pp.393-402.

190

Thompson, William R. 1982. "Phases of the Business Cycle and the Outbreak of
War." International Studies Quarterly 26:301-11.

Thompson, William R. 1983. "Cycles, Capabilities, and War." In Contending
Approaches to World System Analysis, William Thompson (ed.). Beverly
Hills, Calif: Sage Publications.

Thompson, William R. 1988. On Global War: Historical-Structural Approaches
to World Politics. Columbia: University of South Carolina.

Thompson, William R. 1992. "Systemic Leadership and Growth Waves in the
Long Run." International Studies Quarterly 35:25-48.

Thompson, William R. and Gary Zuk. 1982. "War, Inflation, and Kondratieffs
Long Waves." Journal of Conflict Resolution 26:621-644.

Tosini, Suzanne and Edward Tower. 1987. "The Textile Bill of 1985: The
Determinants of Congressional Voting Patterns." Public Choice. 54, pp.19-
25.

Tyson, Laura D'Andrea, William Dickens and John Zysman. 1988. The Dynamics
of Trade and Employment. Cambridge, Massachusetts: Ballinger.

Wade, Larry L. and John B. Gates. 1990. "A New Tariff map of the United States
(House of Representatives)," Political Geography Quarterly, Vol.9, No.3,

pp.284-304.

191

Wallerstein, Immanuel. 1974. “The Rise and Future Demise of the World
Capitalist System: Concepts for Comparative Analysis.” Comparative
Studies in Society and History 16:387-415.

Wallerstein, Immanuel. 1983. “Three Instances of Hegemony in the History of
the World Economy.” International Journal of Comparative Sociology
24:100-8.

Wallerstein, Michael. 1987. "Unemployment, Collective bargaining, and the
Demand for Protection." American Journal of Political Science. Vol.31,
No.4, pp.729-752.

Weingast, Barry R. 1981. "Regulation, Reregulation, and Deregulation: The
Political Foundations of Agency Clientele Relations." Law and
Contemporary Problems, 44, pp.147-77.

Weingast, Barry R. 1984. "The Congressional-Bureaucratic System: A Principal-
Agent Perspective with Applications to the SEC." Public Choice, 41,
pp.147-92.

Wiener, N. 1956. “The Theory of Prediction,” in E. F. Beckenback, ed., Modern
Mathematics For Engineer. New York: McGraw-Hill.

Williams, John. T. 1992. "Dynamic Change, Specification Uncertainty, and

Bayesian Vector Autoregression Analysis." Political Analysis 4:97-126.

192

Williams, John T. and Michael D. McGinnis, and John C. Thomas. 1994.
"Breaking the War-Economy Link." International Interactions 20(3): 169-
188.

Wilson, Rick. 1986. "An Empirical Test of Preferences for Pork Barrel: District
Level Appropriations for River and Harbor Legislation, 1889-1913."
American Journal of Political Science. Vol(30), p.735.

Wold, H. 1938. A Study in the Analysis of Stationary Time Series. Stockholm:
Almguist & Wiknell.

Yarbrough, Beth and Robert Yarbrough. 1992. Cooperation and Governance in
International Trade: The Strategic Organizational Approach. Princeton,
New Jersey: Princeton University Press.

Zupan, Mark A. 1990. "The Last Period Problem in Politics: Do Congressional
Representatives not Subject to a Reelection Constraint Alter Their Voting

Behavior." Public Choice 65: 167-180.

 

"I1111111111111“