Oil and Revolutionary Regimes: A Toxic Mix...Oil and Revolutionary Regimes: A Toxic Mix Jeff Colgan,...

Oil and Revolutionary Regimes: A Toxic Mix

Jeff Colgan Princeton University

November 2008

Paper prepared for the

International Political Economy Society Annual Meeting, Philadelphia

Abstract: The international political economy of oil has multiple, important influences on peace and security. Oil-exporting states, or petrostates, engage in militarized interstate disputes (MIDS) at a much higher rate than non-petrostates. Why are some but not all petrostates inclined to adopt dissatisfied, aggressive foreign policies and to engage in MIDS on that basis? This paper investigates this question by testing a theory that proposes that when revolutionary regimes come to power in petrostates, they have a dramatically higher propensity to engage in MIDS than comparable non-petrostates. This theory is tested with statistical regression analysis using a new quantitative dataset that identifies revolutionary regimes in the period 1945-2001. The results show that petro-revolutionary regimes constitute a special threat to international peace and security. This evidence implies that one of the standard explanations of the relationship between revolution and war, based on a theory of �balance of threat�, is not fully satisfactory. Moreover, the findings challenge the conventional understanding of the link between oil and war. Reader�s Note: This paper is a modified draft version of the quantitative empirical chapter of my dissertation on International Security and Oil-Exporting States. As such, it is meant to be read in connection within a larger whole, but where possible I have tried to give context from the rest of the dissertation in this paper. Thanks are owed to my dissertation committee Robert Keohane, Christina Davis, and Jennifer Widner, as well as to Sarah Bush, Jessica Green, Mark Melamed, and Jordan Tama for comments on earlier drafts of this paper. Financial support from the Bradley Foundation, the Woodrow Wilson School, and the Social Sciences and Humanities Research Council of Canada is gratefully acknowledged.

Oil and Revolutionary Regimes: A Toxic Mix Jeff Colgan, 2008

1

Introduction

Even a casual observation of world politics suggests that petrostates � states in which

revenues from net oil exports constitute at least 10% of GDP � are among the most dangerous

states in the world. Petrostates show a remarkable propensity for militarized interstate disputes

(MIDS), on average, and especially MIDS in which they act as the aggressive party. For

instance, petrostates engaged in MIDS at a rate almost 80% higher than non-petrostates in the

period 1965-2001.1

While some petrostates engage in many disputes, especially as aggressors, not all

petrostates appear to have this tendency. Certain petrostates � Iraq and Libya, for instance � are

aggressive in inter-state conflicts, whereas others � e.g., Saudi Arabia, Kuwait � are not. Why

are petrostates so inclined to adopt dissatisfied, aggressive foreign policies and to engage in

militarized interstate disputes on that basis? Moreover, why are some petrostates inclined to do

so but others are not? These are the two principal questions that motivate this research project.

These questions lie at the intersection of two branches of study: international political

economy and international security. Recent study of international security has moved

increasingly towards examining the roots of war in the domestic politics of the disputing states.2

In petrostates, the global market for oil is a crucial factor in shaping domestic politics. The

political economy of oil has been studied in terms of its impact on democracy,3 regime stability,4

1 For petrostates, the rate is an average of 0.84 MID per year, compared to 0.47 per year for non-petrostates. The figures given here are based on the Correlates of War database and the author�s analysis. Figures exclude major powers (as defined below). 2 B. Bueno de Mesquita et al., 2003; E. Mansfield and J. Snyder, 2005; B. Lai and D. Slater, 2006; M. Tomz, 2007; J. Weeks, 2008; H. Goemans, 2004, 2008; S. Walt, 1996. 3 M. Ross, 2001, 2006; Haber and Menaldo, working paper; K. Ramsey, working paper. 4 T. Karl, 1997; B. Smith, 2004.


2

and civil wars,5 in addition to the various economic aspects of the �resource curse.� Given these

pervasive political consequences, there is a real need to explore how the global market for oil

influences international security by acting via the domestic politics of oil-exporting states.

Studying the foreign policy of petrostates is essential for understanding the dynamics of

international relations in the postwar period. Iraq has been involved in three of the most

significant wars since 1980, including the enormously costly and bloody Iraq-Iran war. Backed

by its oil and gas revenues, Russia pursued an often aggressively expansionist foreign policy

under the Soviets, and again more recently under Vladimir Putin. This fact was dramatically

illustrated by the recent Russian militarized incursion into Georgia. The revolutionary state of

Iran has engaged in direct militarized conflicts with multiple enemies, funded belligerents

against the state of Israel, and appears to be aggressively pursuing a nuclear weapons

development program. Venezuelan President Hugo Chavez has not hesitated to mobilize his

country�s armed forces and wield the threat of war in his efforts revolutionize the international

relations of Latin America. Libya under Qadhafi waged direct warfare with some of its

neighbors and indirectly attacked others by funding insurgents. No picture of global

international relations can be considered complete without understanding the role of petrostates

such as Iraq, Iran, Libya, Russia, and Venezuela in global affairs.

The link between resources and wars has been the topic of considerable speculation and

investigation.6 Much of the literature on resources in the 1980s and 1990s, following in the wake

of the oil price shocks, focused on resource competition and the potential for international war as

a means of conquest for oil.7 Yet with the notable exception of the Iraqi invasion of Kuwait in

1990, it is difficult to identify many international wars in which conquest for oil has played a

5 P. Collier and A. Hoeffler, 2001; J. Fearon and D. Laitin, 2003; B. Smith, 2004. 6 M. Klare, 1999, 2001; T. Homer-Dixon, 1999; S. Pelletiere, 2004; P. LeBillion, 2005. 7 B. Russett, 1981; A. Westing, 1986.


3

determining causal role. Even when a US-led coalition twice militarily occupied an oil-rich

country in the Persian Gulf � Kuwait in 1991 and Iraq in 2003 � the victors chose to continue

buying their oil at market rates rather than simply take it as a prize of war. The view of

petrostates as passive targets of conquest also gives rise to a problematic question: why is it that

so many petrostates are aggressively initiating conflicts? Iraq, Venezuela, Libya, and Russia are

better characterized as instigators of resource-backed aggression than as targets of resource

conquest.

The theory being tested here on the relationship between oil and international conflict

suggests that domestic revolutions play a key role in resolving these questions and explaining the

link between oil and war. 8 As Stephen Walt and others have noted, a major domestic revolution

increases the state�s propensity for international conflict.9 The precise causal mechanisms for

this relationship are still the subject of some debate, however. Some view domestic revolutions

as making the state more attractive as a target of war,10 whereas others suggest that revolutionary

regimes are more likely to act aggressively towards international rivals.11

The proposed theory specifies three broad causal mechanisms that link revolutionary

regimes to militarized interstate disputes. First, revolutions weaken or destroy institutions that

might serve as a constraint on the executive�s power, allowing revolutionary leaders to pursue

aggressive foreign policy preferences. Second, the political dynamics of revolutions generate

incentives for aggressive foreign policy by creating nationalistic fervor, empowering radicals,

and delegitimizing moderate leaders. Revolutionary regimes also often perceive the prior regime

as having been too cooperative with foreign states, particularly those that had influence in

8 For a much more full discussion of this theory, see J. Colgan, PhD dissertation, forthcoming. 9 S. Walt, 1992, 1996; T. Skocpol, 1988; T. Gurr, 1988. 10 S. Walt, 1996; J. Goldstone, 1996; A. Kydd, 2000. 11 T. Skocpol, 1988; T. Gurr, 1988.


4

domestic affairs, and this can motivate revolutionary regimes to engage in international conflict

in opposition to those states. Third, revolutions can also make states attractive as targets of

international conflict, when their rivals perceive them to be temporarily weak or disorganized.

This theory is similar to, but distinct from, the theory advanced by Mansfield and Snyder

regarding the correlation between democratizing states and international conflict.12

The special economic characteristics of oil are hypothesized to enhance all three of these

causal mechanisms. First, because oil income can be easily centralized by the regime, weak

institutionalization and a strong personalization of power are more likely in petrostates. Second,

the international nature of the oil market often exacerbates nationalistic grievances based on

foreign intervention in the domestic oil industry. And third, revolutionary petrostates can appear

to be particularly tempting prizes to international rivals because of their oil wealth. Not all of

these mechanisms will be in place in each case, but one or more of them are likely to be present.

Thus oil and a revolutionary regime combine to form a toxic mix for international security. The

hypotheses derived from this theory are to be tested in this paper.

The first set of hypotheses to be tested relate to the effect of a revolutionary regime on

the state�s foreign policy and propensity to engage in militarized interstate disputes (MIDS). The

second set of hypotheses focus on the role of oil, and its interaction with revolutionary regimes.

Specifically, the theoretical hypotheses to be tested are:

H1a: States led by revolutionary regimes are more likely to engage in aggressive

(revisionist) MIDS than comparable states led by non-revolutionary regimes.

12 E. Mansfield and J. Snyder, 2005. Mansfield and Snyder argue that the process of democratization often sweep away institutions and create incentives for international conflict. The theory tested here shares some of this perspective, but proposes that this is not restricted to regimes that attempted democratization. One consequence of this difference is that regimes in states such as Libya and Iraq which are �revolutionary� but not �democratizing� are expected to have a high propensity for international conflict. For greater details, see J. Colgan, PhD dissertation.


5

H1b: States led by revolutionary regimes are more likely to engage in MIDS as

aggressors (revisionists) than as defenders (targets of revisionism).

and

H2a: On average, revolutionary regimes in petrostates are more likely to engage in

aggressive (revisionist) MIDS than revolutionary regimes in non-petrostates.

H2b: On average, non-revolutionary regimes in petrostates are no more aggressive in

engaging in MIDS than non-revolutionary regimes in non-petrostates.

These hypotheses are tested against a null hypothesis (i.e., that neither oil nor revolutionary

regimes have any effect on states� aggression or propensity for conflict). They can also be tested

against the hypotheses generated from standard accounts of the dynamics between revolutions

and wars, discussed below.

The theory tested in this paper challenges and contributes to the study of international

relations in two ways. First, it challenges a prominent view of the relationship between

revolutions and war, developed by Stephen Walt, as characterized by a change in the balance of

threat or power. Instead, the theory tested here suggests that revolutionary regimes act more

aggressively than non-revolutionary regimes, on the basis of individual- and state-level

characteristics. Second, it disputes the view of �resource wars� as generated primarily by violent

international competition for oil. In fact, the major link between oil and war is not to make

petrostates the target of wars, but rather to make them, when led by revolutionary regimes, far

more aggressive and likely to initiate war.


6

Methodology

The focus of the empirical testing in this paper is quantitative regression analysis. These

regressions use militarized interstate disputes (MIDS) as the primary dependent variable, with a

special focus on whether the state�s role in such disputes was as an aggressive (revisionist)

power or as defensive (non-revisionist) power. While this approach takes advantage of the

strengths of quantitative empirical methodology, as with all such studies some of the richness of

the phenomena under investigation is lost in the process of quantification. For instance, using

MIDS as the dependent variable provides important information about state behavior, but more

information is available about the degree and type of international conflict involving petrostates.

This topic is taken up qualitatively in the case studies presented elsewhere.13

Dependent Variable

Conceptually, the principal dependent variable for this analysis is a state�s propensity to

engage aggressively in militarized interstate disputes (MIDS). Clearly, international conflict is

an inherently strategic interaction, involving more than one actor. Dyadic analyses are often

used to study the interstate conditions under which wars or disputes occur between states,

notably in studying the democratic peace hypothesis. However, the probability of a conflict is

also partly a function of the monadic characteristics of each actor, in much the same way that a

person�s grumpiness at a particular moment affects the likelihood that they will engage in an

13 J. Colgan, PhD dissertation, forthcoming.


7

argument. The present theory seeks to explain a monadic property, the state�s propensity for

MIDS, so the analysis is monadic. Future research may investigate how this monadic property

affects strategic dyadic relationships.

MIDS are a class of international events in which a state threatens, displays, or uses force

against another state(s). This broad class of conflicts ranges in intensity from full-fledged wars

to disputes that involve threats to use force but no actual fatalities. While there is considerable

heterogeneity in these events, they provide considerable information about a state�s interstate

disputes. These events have been coded in the Correlates of War (COW) data set (v3.02). These

data have been widely used by scholars and represent the standard dataset for studying

international peace and conflict.14

Operationally, the main dependent variable is Aggressor-MIDS, which is a count of the

MIDS in each year in which the state has been coded as the revisionist (or aggressor) state. This

follows from the fact that one of the major theoretical claims being tested is that revolutionary

regimes, especially petro-revolutionary regimes, are aggressive rather than simply being the

targets of disputes. The term �aggressor� is used in the same way that the terms �dissatisfied� or

�revisionist� are used in the international relations literature. As the codebook for the COW

dataset explains,

Within the data, the revisionist variable attempts to indicate which states are

dissatisfied with the existing status quo prior to the onset of a militarized

interstate dispute. Both sides of a dispute can be considered revisionist if they

both are dissatisfied with the status quo, but the state that openly attempts to

challenge the pre-dispute condition by 1) making claims to territory, 2) attempting

14 E. Mansfield and J. Snyder, 2005; J. Gowa, 1999; B. Bueno de Mesquita et al., 2003.


8

to overthrow a regime, or 3) declaring the intention not to abide by another state's

policy, was coded as revisionist.15

Some variants of this dependent variable are also used, including MIDS (all MIDS involving the

state), and Defender-MIDS (MIDS in which the state is coded as non-revisionist). The analysis

uses the �revisionist� variable rather than the �Side A� initiator, as the authors of the COW

dataset specifically warn against this latter practice.16 The time period of the analysis is 1945-

2001, based on the availability of data. The unit of analysis is the state-year.

First Explanatory Variable: Revolutionary Regimes

One problem that has plagued the research on revolution and war is that there is no

widely-accepted universe of cases of revolution or revolutionary regimes. Indeed, variations on

the universe of cases are almost as numerous as the number of scholars who have examined the

question.17 This difficulty can lead to selection bias in the empirical testing of theoretical

hypothesis, particularly if the analysis focuses on cases that are supportive of the theory.

Avoiding this bias is facilitated by the construction of a comprehensive set of cases within a

defined time period, subject to a well-specified identification procedure.

Accordingly, a unique dataset was constructed for the purpose of operationalizing one of

the key independent variables in the theory tested in this paper, revolutionary regime. The two

15 From the Correlates of War data project. Jones et al., 1996. �Militarized Interstate Disputes, 1816-1992� Conflict Management and Peace Science, 15(2). p. 178 16 Jones et al., 1996. 17 Stephen Walt focuses his (1992, 1996) research on revolution and war on ten cases of �unambiguous� revolutions, though he suggests in later work (1996) that many other cases exist. According to Kurt Dassel, Charles Tilly identifies nineteen discrete revolutions in his book From Mobilization to Revolution (K. Dassel, 1996). Theda Skocpol focuses on only the �great� revolutions, which appears to limit the universe of cases to fewer than ten (T. Skocpol, 1979, 1988). Robert Snyder identifies twenty-four revolutions during the Cold War for his research (R. Snyder, 1999).


9

principal criteria that must be met for a regime to be considered revolutionary are: (i) the leader

must have used armed force against his own state as an integral part of his coming to national

power, and (ii) once in power, the regime must have implemented radical domestic changes for

the purpose of transforming the organization of society. Regimes and leaders are often used

interchangeably in this analysis, as for the purpose of this dataset, a regime is equivalent to the

period of time that a leader was continuously in power (e.g., four or eight years for a typical US

President).

The basis for this dataset was the Archigos database (v.2.8.1) of state leaders developed

by Goemans et al. For each state in each year (state-year), only one leader is selected; where

there are multiple leaders in a given year, the leader in office at the end of the year (31 Dec) is

used. Each state-year is given a dichotomous 1/0 coding for �revolutionary regime.� This

coding is based on the two principal criteria plus two additional filters, as described below.

Greater detail for all of the coding rules is given in the codebook.18

The coding for the first criterion is based primarily on whether or not the leader used

armed force against his own state at any time prior to coming to office as an integral part of

coming to national influence, and ultimately, state leadership. This specification implies that a

leader such as Adolf Hitler, who led an attempted (but failed) coup to rise to national influence,

and was subsequently elected to office, are considered as having used force and thus a potential

revolutionary. The term �used armed force� implies leadership in the act. An individual who is

a relatively low-level functionary of a revolution or coup is not considered to have led it. It is

possible for more than one leader to have �led� the use of force in any particular event, but the

leadership is restricted to its senior leaders. Thus both Lenin and Stalin can be considered to

18 J. Colgan, 2008. Codebook for Revolutionary Regimes Dataset, unpublished.


10

have �used armed force� as leaders of the Russian Revolution, but not Khrushchev, even though

the latter fought in the Red Army at a young age.

The second criterion is judged on the basis of how much the social, economic, and

political institutions and practices were changed during the leader�s tenure, relative to the social,

economic, and political institutions and practices that existed in the regime prior to the leader

coming to power. In all cases, the focus is on domestic policy, rather than foreign policy. Seven

possible areas of change have been identified: the method of selection and power of the national

executive; the structure of property ownership; the relationship between state and religion; the

official political ideology, if any; the official state name; the status of women; and the presence

of a revolutionary council or committee. These areas were identified inductively by studying the

distinguishing characteristics of well-known revolutions.19 Dramatic changes in policy in at least

three of the seven categories are required for the regime�s policy to be considered revolutionary.

For example, the Iranian revolution in 1979 changed the relationship between state and religion

(political dominance by clerics), the power and selection of the national executive (replacement

of the monarchy by a Supreme Leader), the status of women (inequality in inheritance law and

segregation of the sexes), and the official name of the country (changed to the Islamic Republic

of Iran), as well as many other changes. From a social science perspective, some of these

categories are easier to objectively measure (e.g., changes in the official name of the state),

others are more directly connected to the substantive changes of interest (e.g., a radical change in

the relationship between religion and the state), but all of them are believed to provide

information about the degree to which the state is transformed under the regime. In all cases,

changes are measured relative to the relevant prior regime, rather than some external standard.

19 The revolutions used to determine these categories were: Iran (1979); Cuba (1959); China (1949); Ethiopia (1974) and Nicaragua (1979).


11

Thus a change from monarchy to theocracy is equally revolutionary as a change from theocracy

to monarchy.

In addition, two filters are applied to try to screen out false-positives (i.e., regimes that

might be classified as revolutionary, but which do not in fact systematically share an elevated

propensity for MIDS). The first filter applies to the founding regime of a state. While founding

regimes are almost certainly revolutionary in some sense of the term, they are not considered

revolutionary regimes in this dataset for two reasons. First, the focus of the theory to be tested is

on a regime that transforms an existing society, and independence movements that create a new

state are specifically excluded. Second, and more practically, it is often difficult or impossible to

judge the degree to which a founding regime changes the institutions relative to a �prior regime�

when no such regime exists. (Some colonies did have prior regimes with similar territorial

boundaries as the states that replaced them, but this is frequently not the case.) For the dataset,

this means the regime under the first (but only the first) leader is coded as non-revolutionary.

This is true of states exiting colonialism where the state did not previously exist in its modern

form: thus the first leader of Nigeria is part of a founding regime, but the first postwar leader of

Ethiopia is not.

Second, leaders who are installed by foreign powers after a major international war are

not considered revolutionaries. Foreign-installed leaders often do not have a free hand to control

their state�s policy, especially in the realm of foreign affairs. Moreover, the theoretical factors

identified about the domestic politics of revolutionary movements are unlikely to apply to a

leader who is appointed by a foreign power. Thus even if General Douglas McArthur�s rule in

Japan after WWII was revolutionary in comparison to the domestic politics of pre-war Japanese

Empire, the theory developed here would not expect Japan to have an aggressive foreign policy


12

under McArthur. Similarly, leaders installed by the Soviets in Eastern Bloc countries are not

expected to have increased aggression in their foreign policies.

When the two principal criteria are met, and neither of the filters applies, the state-year is

coded as being a revolutionary regime. All state-years are also coded with a dichotomous

variable called �ambiguous� to indicate borderline cases or cases where the information available

for coding was poor. This variable can be used to provide robustness testing to ensure that

marginal cases are not driving the results.

Second Explanatory Variable: Petrostates

The second key explanatory variable for this analysis is whether or not the state is a

petrostate. To be classified as a petrostate, the state�s gross revenues from net oil exports must

constitute at least 10% of annual GDP. There are 33 states that have met this requirement for at

least one year in the period 1945-2001. However, a number of them are marginal cases at best,

as their oil export income barely exceeds the 10% threshold and tends to do so for only short

periods of time when oil prices are high. For instance, Mexico�s gross revenues from net oil

exports were equal to approximately 11% of GDP in the period 1980-1983, but averaged just

3.4% of GDP over the period 1965-2001. These marginal cases are closer to non-petrostates

than to the rest of the petrostates, as the oil export revenues in states such as Saudi Arabia,

Venezuela, and Nigeria constitutes 30-70% of GDP. As such, an additional criterion is applied

to the definition: the state�s revenues must have exceeded 20% of GDP for at least ten of the

years in the period under analysis. With this stricter definition of a petrostate, five marginal

cases are further screened out: Malaysia, Mexico, Egypt, Tunisia, and Norway. (However, these


13

states will be included in tests for robustness.) This leaves 28 states, as listed in the Appendix,

which are treated as petrostates for the present analysis.

This operationalization of petrostates generates a dichotomous variable for the purposes

of regression analysis. However, there are several ways to operationalize the concept of

�petrostate�; indeed, the best methodology has been the subject of some debate.20 Two other

methods are by measuring the revenue from petroleum exports as a percentage of GDP, and by

measuring the per-capita revenue (or rent) associated with oil exports; each has its advantages

and disadvantages.21 These other two operationalizations will be used to replace the

dichotomous variable to test the robustness of the empirical findings.

The third and most important explanatory variable is the interaction between petrostate

and the revolutionary regime, called Petro Revolution. This is also a dichotomous variable,

which takes the value of 1 when both Petrostate and Revolutionary Regime are equal to 1;

otherwise it is 0. The effect of introducing this variable into the regression is to isolate the

specific political dynamics of revolutionary petrostates. The variable Petrostate thus focuses on

the effect of oil politics in non-revolutionary petrostates.

In the data, there are 839 state-years in which Petrostate has a value of 1; of these, Petro

Revolution has a value of 1 in 159 state-years and 0 in the remainder. These observations are

based on 28 petrostates over the time period 1945-2001. While there are five states that are

classified as petrostates during the entire period under analysis, many states are classified as

petrostates only for part of the time period under analysis, as oil is discovered or runs dry. This

20 See M. Ross 2006. 21 Variants of these measures have also been used; see M. Ross, 2001 and K. Ramsey, 2006 (working paper). Both of these approaches generate continuous measures, with the implicit assumption that a state that has a higher value is �more of a petrostate� than a state that has a lower value. The drawback to this approach is that it is not always clear this assumption is warranted.


14

is an advantage, because it allows the variation over time within a state to be exploited for the

purpose of analysis.

Control Variables

The regression controls for a set of other variables that may affect a state�s propensity to

engage in MIDS, and have been used by previous work in the literature. The analysis includes

(logged) population size, (logged) GDP per capita, and the number of contiguous territorial

borders with other states, as basic characteristics of a state�s likelihood to engage in international

conflict. Previous work has outlined the theoretical justification for these parameters; suffice to

state here that they are proxies either for the degree to which the state is capable of waging war

(population, GDP) or the geographical likelihood of contact and thus friction with its neighbors

(borders).22

Democratic peace theory suggests that democracies may be less inclined to engage in

revisionist MIDS than other kinds of regimes.23 Modern theorists of the democratic peace argue

that this tendency is strongest (and perhaps only present) in a dyad of two democracies.24

However, the basic argument by Immanuel Kant is that, in a republican regime, �those who

would have to decide to undergo all the deprivations of war will very much hesitate to start such

an evil game.�25 This suggests that democracies will be less likely to engage in war in general,

regardless of their opponent. To address this possibility, the state�s composite Polity IV score is

included as a control variable.

22 B. Lai and D. Slater, 2006; A. Enterline, 1999. 23 See B. Russett, 1993; J. Gowa, 1999. 24 B. Bueno de Mesquita et al., 2004; Z. Maoz and B. Russett, 1993 25 I. Kant, quoted in J. Levy, 1988. p. 658.


15

Major powers in the international system tend to take a much more active role than other

kinds of states, even accounting for their large GDP and population sizes. This is addressed by

introducing a dichotomous variable for �major power� status. The analysis follows the COW

dataset�s specification of the states to be treated as major powers: the five permanent members of

the UN Security Council for all years, plus Germany and Japan since 1991. A dummy variable

was also included for the Cold War period (pre-1990), which may have altered the number of

international disputes.

Some cultural dimensions are included. Huntington contends that religious and cultural

factors shape the fault lines and disputes in the international system, and that Islamic countries in

particular have �bloody borders� and �bloody innards� because they have cultural and

demographic features that make them violence-prone.26 To account for this possible effect, the

Muslim percentage of the population is used as a control variable. Each analysis also includes

dummy variables for each geographical region. Eight regions were used, similar to the World

Bank�s classification: North America, Latin America, Western Europe, Eurasia and Central Asia;

Middle East and North Africa; South Asia; Sub-Saharan Africa; East Asia and Pacific.

Data

Data on states� borders, population, and major power status comes from the COW

dataset. Fearon and Laitin�s (2003) dataset provides the GDP per capita data. The World Bank�s

World Development Indicators was used to provide GDP per capita data where this is missing

from the Fearon and Laitin data. Data on states� religious make-up is based on the World

Christian Database. The measure of regime type is based on the Polity IV dataset, as is the initial 26 S. Huntington, 1996.


16

year as a polity.27 Data on oil revenues, prices, and export volumes are drawn from the BP

Statistical Review of World Energy and the US Energy Information Agency.

Considerable effort went into addressing missing data and ensuring that the dataset was

as complete as possible. Reasonably high-quality data exists for most of the variables over this

time period, such as population and state borders. However, other variables are more difficult.

Only a single observation is available for the religious demographics of each country, but since

this proportion is thought to change relatively slowly over time, this value is used for all years.

The GDP per capita data from Fearon and Laitin is complete for 87% of the observations of

interest, but is missing for some of the key states such as Iraq. Consequently the World Bank

data was used to fill in, bring the total completion rate to 94%. With regard to oil income, the

data for years prior to 1965 is generally unavailable to the public. However, a special request

was made for BP data for 1945-1964 to complement the public reports, and this additional data

allowed the analysis to be extended through the full time period.

Empirical Results

Before turning to the analysis of the hypotheses, it may be helpful to describe some of the

basic characteristics of the new dataset on revolutionary regimes being used. Recall that a

regime is defined as the continuous period of rule by a given ruler in each state, so each regime

has only one leader or group of leaders. The dataset contains 1317 regimes/leaders, containing

data on all regimes/leaders from 1945-2003 for 170 countries, for a total of 7578 observations of

27 This score is a proxy for the regime�s �degree of democracy�; it ranges from -10 to 10.


17

state-years.28 Of this total set, 86 regimes/leaders or less than 7% are coded as revolutionary.29

Table 1 provides some descriptive data that compare and contrast revolutionary and non-

revolutionary leaders. The table focuses only on developing countries, to eliminate

heterogeneity introduced by the special historical, political, and economic features of the OECD

states. The data suggest that revolutionary leaders are exclusively male, relatively young, and

inclined to lead strongly autocratic regimes. Their average length of tenure in office (i.e., the

duration of the regime) is markedly higher than non-revolutionary leaders (almost double, 10.2

years compared to 5.7 years). By definition, all revolutionary leaders used force to come to

power; only about a quarter of non-revolutionary leaders did so.

Table 1: Characteristics of Revolutionary Leaders

Having identified the universe of cases of revolutionary regimes, it is instructive to

consider a simple aggregate picture of the relationship between revolutionary regimes and

aggressive foreign policy. Table 2 shows the average number of MIDS per state-year,

comparing revolutionary regimes to non-revolutionary regimes. The first column shows all

MIDS that the state was engaged in. The second and third columns show the MIDS in which the

28 Countries that do not have a Polity score are not coded. The �number of leaders� reported counts leaders who have come to office at two distinct times (with at least a year out of office) as two leaders; the number of unique leaders is slightly smaller. 29 Under a highly restrictive secondary definition, less than 1% of leaders are coded as unambiguously revolutionary. See the discussion under �Robustness Checks� for use of this definition.


18

state was the aggressor and defender, respectively. The final column shows MIDS in which the

state was the aggressor and military force was used (as opposed to display or threats of force).

Table 2: International Disputes of Revolutionary Regimes

The results are revealing. First, consistent with the theoretical hypothesis being tested,

revolutionary regimes engage in more than twice as many MIDS per year as non-revolutionary

regimes on average. This strongly suggests that the null hypothesis is incorrect, as revolutionary

regimes do not behave in a manner similar to non-revolutionary regimes. Second, the difference

between revolutionary regimes and non-revolutionary regimes is driven principally by their

behavior as aggressors rather than defenders in international disputes. Revolutionary regimes act

as the aggressor in more than three times as many MIDS as non-revolutionary regimes, a

discrepancy that grows even larger when one considers only MIDS in which military force was

used.

These results are quite suggestive, but stronger inferences can be drawn from a robust

regression analysis. This analysis is conducted by using Poisson regression adjusted for time-

series panel data, with standard errors clustered by panel to account for possible heterogeneity.30

Panel Poisson regression is used because the dependent variable is a count variable, in which the

30 The model is specified for time-series panel data, such that zero-inflation is not an issue. Negative binomial models were also used for robustness checks, and generally found to be consistent, although in some cases the statistical software used (Stata 9.0) had difficulty converging the iterations of the model.


19

variance is proportional to the mean. The structure of the data makes panel Poisson regression

more appropriate than a standard OLS regression, which assumes homoskedasticity (i.e., that the

variance is constant). Table 3 below shows the results of the analysis.

Table 3: International Disputes of Revolutionary States � Regression analysis

The results show strong support for the theory. States with revolutionary regimes are

associated with considerably more militarized interstate disputes than states without

revolutionary regimes, especially when they occur in petrostates. The dependent variable for all

models in Table 3 is Aggressor-MIDS. Model 1 is a basic model without the new dataset on


20

revolutionary regimes, used as a baseline for comparability to other scholarly studies. Model 2

shows the changes when the new data are added; model 3 is specified with panel fixed effects,

controlling for all state-specific variables that do not vary over time. As expected by hypotheses

H1a and H1b, the coefficient for Revolutionary Regime is positive and strongly significant, with

confidence levels above 99%. This is true in all models, regardless of the specification used,

suggesting a very strong relationship between revolutionary regimes and MIDS.

Models 4 and 5 introduce the second key variable, Petrostate, and the interaction variable

Petro Revolution. As expected, the interaction variable Petro Revolution is positive and strongly

significant, indicating that revolutionary regimes in petrostates are even more warlike than

revolutionary regimes in non-petrostates. This finding holds in model 4, which replicates the

specification structure of model 2. The finding also holds in model 5, which uses fixed effects to

control for all state-specific variables that do not vary over time. The striking finding in model 5

is that, even when the analysis focuses only on within-state variation over time, the interaction

variable is found to be strongly associated with Aggressor-MIDS. As expected by the theory, the

coefficient for Petro Revolution is positive and strongly significant, again with confidence levels

above 99%. Note also that this result is obtained even though the models include regional

dummy variables, and thus controlling for variation in the amount of international conflict

between regions (e.g., controlling for the high levels of conflict within the Middle East). Thus

there is considerable evidence to support the most important hypothesis, H2a.

The hypothesis H2b suggests that among non-revolutionary states, petrostates are no

more aggressive than their counterparts in non-petrostates. Again, the results provide support for

this hypothesis. Perhaps surprisingly, the coefficient on the variable Petrostate is negative,

indicating an inverse correlation between non-revolutionary petrostates and aggressor-MIDS.


21

This finding is statistically significant in model 4, but not significant in model 5 once fixed

effects are introduced. Thus these results are not conclusive, but they suggest that non-

revolutionary petrostates are at least no more aggressive to engage in MIDS than comparable

non-petrostates. This finding is striking because overall, the opposite trend holds: petrostates

engage in more MIDS on average than non-petrostates. The implication is that the overall

above-average rate of international disputes is driven almost entirely by the small subset of

revolutionary petrostates, which are extremely aggressive; the rest of the non-petrostates do not

engage in many MIDS. This finding provides further support for H2a and suggests that the

conventional explanation for the relationship between oil and international conflict is not fully

satisfactory.

The results also provide insight on the importance of the control variables in determining

the war-proneness of a state. Consistent with earlier findings in the literature, the number of

territorial borders is an important factor: the more contiguous neighbors a state has, the more

opportunities it has for international disputes. Population size is associated with increased war-

proneness cross-nationally, but this effect becomes statistically insignificant when state fixed-

effects are added (models 3 and 5). The Muslim percentage of the population, income per-

capita, Cold War years (pre-1990), and the state�s status as a major power are not associated with

propensity for MIDS in any of the models. Finally, the state�s Polity score (from -10 to +10) is

inversely associated with MIDS in model 1, consistent with the hypothesis that democracies

engage in fewer aggressive disputes. However, this relationship disappears (i.e., is not

statistically significant) in models 2-5, when the data on revolutionary regimes are used. This

latter result suggests that it may be worth investigating the extent to which revolutionary regimes

are related to the democratic peace thesis.


22

The magnitude of these statistical effects is worth considering. Figure 2 below shows the

effect of the variables Petrostate, Revolutionary Regime, and Petro Revolution on the onset of

aggressive MIDS for a typical state. In this comparison, a �typical state� is considered, in which

population, GDP per capita, Polity score, Muslim percentage of the population, and the number

of state borders are all set to the average of the data sample.31 The behavior of this typical state

is considered under four possible conditions. In the first condition, the state has no oil and is led

by a non-revolutionary regime. As indicated by the bar on the left of Figure 2, such a state could

be expected to aggressively engage in approximately 0.14 MIDS per year, or about one every

seven years. In the second condition, the state becomes a petrostate, but is still led by a non-

revolutionary regime; Figure 2 shows that such a state will aggressively engage in roughly 0.11

MIDS per year, or about one every ten years. Thus as discussed previously, non-revolutionary

petrostates appear no more aggressive (and perhaps slightly less aggressive) than comparable

non-petrostates.

The most important findings are shown in the two bars on the right hand side of Figure 2.

As expected, non-petrostates with revolutionary regimes aggressively engage in MIDS more than

comparable states with non-revolutionary regimes; they do this at a rate of about 0.18 per year,

or 32% more per year than comparable state with non-revolutionary regimes. However, this

difference pales in comparison to the effect of revolutionary regimes in petrostates. Such states

aggressively engage in MIDS at a rate of about 0.37 per year, which is more than twice the rate

of comparable non-petrostates. As Figure 2 makes clear, the combination of a revolutionary

regime and an oil-exporting economy is a toxic mix for international peace and security.

31 Thus for the state being compared in Figure 2, the GDP per capita is $2,110, population is 8.3 million, Polity score is -0.6, the state has 5.5 external borders, and Muslims make up 24% of the population. Other values, such as the geographic location of the state, the presence of the Cold War, and the state�s status as a major power are not adjusted in this analysis, meaning that they vary according to the data within each category of Figure 2.


23

Figure 2: Effect of the Combination of Oil and Revolutionary Regimes on MIDS32

These results suggest that the theory under investigation provides powerful insight into

the behavior of petrostates. Indeed, when one moves past the quantitative results and considers a

broader view of foreign policy behavior, the relationship between petro-revolutionaries and

aggressive conflict becomes even more evident. This topic is taken up in case studies, but some

examples can be highlighted here.33 Saddam Hussein led a revolutionary movement, seized

power by force, and then used his country�s oil income to centralize power and build a powerful

military apparatus. Under his leadership, Iraq invaded both Iran (1980) and Kuwait (1990).

Moreover, Iraq engaged in a number of lower-level militarized disputes with other states,

including the Tanker Wars of the 1980s and violations of UN sanctions throughout the 1990s. In

32 Error bars are not shown, but differences between the revolutionary regimes (red bars) and the non-revolutionary regimes (blue bars) are statistically significant, as are the differences between petro-revolutionaries and non-petrostate revolutionaries. 33 See J. Colgan, PhD dissertation, forthcoming.


24

Libya, Muammar Qadhafi�s revolutionary regime aggressively engaged in four separate border

wars with neighboring Chad, and a variety of militarized disputes with other countries such as

Uganda and Sudan. In addition, though not captured in this dataset, Qadhafi�s regime supported

terrorists and insurgencies in at least thirty countries around the world, ultimately leading to the

US bombing of Libya in 1986. In Iran, the revolutionary regime under Khomeini did not initiate

the Iran-Iraq war, but it did aggressively continue the war many years after Saddam Hussein

declared his willingness to negotiate for peace in 1982. Further, Iran�s continuing

aggressiveness and pursuit of nuclear weapons in the 21st century can be attributed to the

revolutionary hardcore members of its regime. In Venezuela, President Hugo Chavez�s

Bolivarian Revolution started when he came to power in 1999, largely too late to be captured in

this dataset (which ends in 2001). However, Chavez�s aggressive foreign policy actions against

multiple countries in the Americas, ultimately bringing Venezuela to the brink of war against

Colombia in 2008, is very much consistent with the expectations of the theory being tested here.

One important petrostate that is not well-described by this theory is Russia. Russia�s

democratizing revolution under President Yeltsin in 1990-1991 did not generate an overly

aggressive regime. However, this is understandable once the details of the revolution are

understood. The hypothesized causal mechanisms, present in other cases of revolutionary

regimes, were largely absent in Russia during the 1990s. Rather than enabling the

personalization of power, the revolutionary movement in Russia was towards democratization

and decentralization of power.34 Moreover, many revolutionary regimes (e.g., Iran) are

aggressive because the domestic regimes they overthrew were perceived as too conciliatory with

34 President Yeltsin�s regime was quite centralized and even autocratic by Western standards. However, in comparison to the Communist regime that preceded it, Yeltsin�s regime was considerably more democratic. Moreover, the personalization of power was considerably less severe under Yeltsin than it has become in Vladimir Putin�s regime.


25

the US and other foreign powers, but in Russia, this was not the case. The Communist regime

was a classic Cold War antagonist in the international system, and by opposing it the

revolutionary regime was thus disposed to avoid international disputes rather than engage in

them. It was not until the nationalistic backlash in the late 1990s, continued and inflamed by

President Putin, that Russia became much more aggressive, leading to military wars with both

Chechnya and Georgia.

The theoretical hypotheses being tested are thus strongly supported by the empirical

evidence. It is worth noting that the number of petrostates in the world is relatively small, and

thus the number of petro-revolutionary regimes under analysis is also small. This should make

us cautious in reaching conclusions. Still, the data includes 159 observations of petro-

revolutionary state-years. This analysis benefits from variation across countries as well as

variation across time within countries, in which both the nature of the regime and the state�s oil

income can change. The analysis of both cross-national and cross-temporal variation suggests

that petro-revolutionary regimes have a dramatically higher propensity for international conflict

than comparable non-petrostates.

Robustness Checks

Several further tests are revealing about the robust relationship between petro-

revolutionaries and Aggressor-MIDS. Table 4 below shows six additional models, each of which

is a variation of model 4 above. Model 6 introduces a lagged dependent variable, to try to

control for any on-going disputes, which could be perpetuated through a process of �war begets

war.� This variable is positive and highly significant, but reduces the significance of the three


26

key explanatory variables only slightly. Given the tendency of lagged dependent variables to

wash out the significance of all other variables, it is striking that the principal findings of the

analysis are sustained. Model 7 introduces a dummy variable for each year, to try to eliminate

any possible system-wide temporal effects on conflict. Model 8 uses a variation of the

dependent variable, focusing only on aggressor-MIDS in which international force was used (as

opposed to including displays and threats of force). In model 9, the COW measure of National

Capabilities was added to the regression.35 As Table 4 illustrates, the correlation between

revolutionary regimes in petrostates and Aggressor-MIDS is remarkably robust.36

Model 10 investigates an alternative, broader definition of a �revolutionary regime.� It is

important to know whether the transformative domestic policy of a revolutionary regime is

actually important as an indicator of foreign policy behavior; one could imagine that all regimes

that have used force to come to power are similarly inclined to behave aggressively in their

foreign policy. Model 10 replaces the key explanatory variable, Revolutionary Regime, with a

variable for all regimes that used force to come to power. (Revolutionary Regimes are a subset

of this group.) In this model, the Petro Revolution variable (which is the interaction between

regime type and oil) is also replaced by the appropriate interaction term. It is clear from the lack

of statistical significance on these two variables that Revolutionary Regimes are indeed a special

set of regimes. This finding suggests that Gurr�s proposition that elites that come to power by

35 This was done in an attempt to determine whether the impact of oil was solely acting through an increase in military capability. However, as I have argued elsewhere (Colgan, working paper), the cross-national time-series datasets on military expenditure and capability are prone to significant measurement error. There inclusion here is for purposes of comparison to existing scholarly work; in the event, the results are robust with or without including this variable. 36 Only in Model 8 did the significance of the coefficient for Petro Revolution dip below 95% confidence, and this is due to an increase in the standard error due to the relatively small number of events in this specification. Even so, the significance remains with confidence above 90% in all models.


27

use of force are more inclined to use violence internationally is, on its own, insufficient to

explain the tendency of revolutionary regimes to engage in MIDS.37

Table 4: International Disputes of Revolutionary Petrostates � Robustness checks

37 T. Gurr, 1988.


28

Conversely, if one narrows the definition of �revolutionary regime,� its effect appears to

grow even stronger. Model 11 is a re-analysis using a more restricted version of the

Revolutionary Regime variable which is coded 1 for a small subset of revolutionary regimes, and

0 for all others. This subset is arguably the �unambiguous� cases of revolutionary regimes,

representing just eleven regimes or less than 1% of the total dataset.38 Using this specification,

the statistical coefficients simply gets stronger (more positive and more statistically significant).

In addition to the analyses presented in Table 4, the results were probed to ensure that no

single country case was driving the result. All of the state-years associated with a number of the

most important country cases (e.g., Iraq, Iran) were dropped, one country at a time, and the

regressions re-tested. The results did not materially change, and the key findings remained

robust. Moreover, to ensure that the Iran-Iraq war was not driving the result by itself, all of the

observations involving either Iran or Iraq during the period 1980-88 inclusive were dropped from

the analysis. In this latter specification, the results were somewhat weaker, but remained

significant at the 90% level.

A series of further robustness tests were conducted. First, all of the tests conducted in

Table 4 were repeated using fixed-effects models. Second, to test for the statistical stability of

the result, each of the control variables was dropped from the analysis, one by one, and the

regressions re-tested. Third, the definition of revolutionary regimes was modified in various

ways using the coding for Ambiguous cases. For instance, in one specification, all regimes coded

as ambiguous were temporarily coded as revolutionary regimes; in another specification, they

were all temporarily coded as non-revolutionary regimes. Fourth, the two variants of the

38 The 11 �unambiguous� revolutionary leaders/regimes are: Mao (China), Castro (Cuba), Khomeini (Iran), Pol Pot (Cambodia), Ortega (Nicaragua), Banti (Ethiopia), Qadhafi (Libya), Kerekou (Benin), Ngouabi (Congo), Al-Bashir (Sudan), and Ne Win (Myanmar). These regimes are selected based on the frequency with which they have been identified as revolutions by other major scholars (e.g., Walt, Hutington, etc.). Each of the regimes also had revolutionary societal transformation under at least four of the seven categories identified above.


29

definition of petrostate were each substituted for the dichotomous measure Petrostate, and the

regressions re-tested. Fifth, the models were re-tested when the sample was restricted to the

developing countries only (i.e., non-OECD). In none of these robustness tests did any of the

results change materially.

As a final robustness check, the specification of the model was changed from a panel

Poisson count model to a logit model, in which the dependent variable is set to 1 if the state

engaged in any Aggressor-MIDS in a given year and 0 otherwise. This specification focuses the

analysis on the incidence of international disputes rather than the number of them, to ensure that

the results obtained are not driven solely by the observations in which states engaged in multiple

Aggressor-MIDS. Under this specification, none of the results changed materially, again

supporting the robustness of the principal findings in Table 3.

Alternative Explanations

The empirical analysis conducted thus far provides strong support for the principal

hypotheses tested in this paper, namely that states led by revolutionary regimes engage in more

militarized international disputes than non-revolutionary states, especially when they occur in

petrostates. In addition to this finding, the data available allow us to consider the two hypotheses

derived from one of the conventional explanations for this behavior. Specifically, they are:

AE1: States led by revolutionary regimes are more likely to be involved in MIDS than

comparable states led by non-revolutionary regimes, but in those disputes they are more

likely to be the defender than the aggressor.


30

AE2: MIDS initiated after revolutions, when the revolutionary state is either aggressor or

defender, should be less likely to yield victory for the aggressor than MIDS between non-

revolutionary states.

One standard explanation for the relationship between revolution and war suggests that

domestic and individual-level factors do not explain the propensity of revolutionary regimes to

engage in international disputes because this explanation assumes that the revolutionary regime

is the principal aggressor. Stephen Walt, in particular, argues that the empirical evidence does

not support this latter premise.39 However, the data presented here suggests precisely the

opposite. As Table 2 above shows, revolutionary regimes are far more likely to be aggressors

than their non-revolutionary counterparts. Moreover, the difference between revolutionary

regimes and non-revolutionary is more pronounced in disputes when the state acts as the

aggressor than as the defender. This strongly suggests that the hypothesis AE1 must be rejected.

This is not to suggest that revolutionary and non-revolutionary regimes are identical

when it comes to the number of disputes in which they are the defender. In fact, as Table 2

shows, revolutionary regimes engage in 70% more Defender-MIDS than non-revolutionary

regimes, on average. Further regression analysis using Defender-MIDS as the dependent

variable rather than Aggressor-MIDS, confirms that revolutionary regimes are strongly correlated

with the number of Defender-MIDS. There are at least two plausible causal explanations for this

tendency. First, it could be that the aggressive behavior of the revolutionary regime provokes

counter-aggression by other states. Second, it could be that revolutionary states are seen as weak

(and thus attractive) military targets in the years immediately following the installation of a

revolutionary regime. Likely it is a combination of both factors at work.

39 S. Walt, 1996.


31

The second hypothesis, AE2, suggests that MIDS involving revolutionary states should

be less likely to yield victory for the aggressor than MIDS between non-revolutionary states.

This hypothesis follows directly from the explanation based on the confusion in the perceived

�balance of power/threat� caused by revolutions, and Walt specifically calls for research that

tests this hypothesis empirically.40 This paper seeks to answer that call. The claim can be tested

by creating a MID Outcome as the dependent variable, with a value ranging from 1 for

aggressor�s victory to 0 for aggressor�s defeat.41 If Walt�s miscalculation thesis is correct, MIDS

involving revolutionary states should have an average outcome significantly below 0.5,

indicating more defeats than victory. In fact, the average outcome for MIDS involving

revolutionary and non-revolutionary states is essentially identical (0.49 compared to 0.51), and

the small difference between them is not statistically significant. Further evidence rejecting AE2

is provided by a regression analysis using MID Outcome as the dependent variable and the same

set of control variables used in the previous analysis (model 2). The key explanatory variable,

Revolutionary Regime does not have statistical significance in predicting the outcome of MIDS.

Consequently, AE2 must be rejected, suggesting that while strategic miscalculation may play a

role in some instances, it is not a principal cause of the association between revolutionary states

and war.

40 S. Walt, 1996. Also see A. Kydd, 2000. 41 This is coded based on the �outcome� variable in the COW dataset. For each MID, a value of 1 is assigned if the MID resulted in a victory by the state�s side or a yield by the other side of the dispute; a value is 0.5 is assigned if the MID resulted in a stalemate or a compromise; a value of 0 is assigned if the MID resulted in a yield by the state�s side or a victory by the other side of the dispute. If more than one MID occurred in a state-year, the �MID outcome� is the average of the values assigned for each MID.


32

Does Oil Cause Revolutionary Regimes?

This inquiry started with the observation that petrostates engage in more MIDS than non-

petrostates, on average. This investigation has shown that revolutionary regimes are a key part

of the explanation for this fact. The results from the investigation strongly suggest that the

interaction between oil and a revolutionary regime has a dramatic effect, significantly raising the

state�s propensity to aggressively engage in militarized interstate disputes. Still, it is possible

that there is another explanation to the original question. Does a petrostate somehow produce

more revolutionary leaders? Could this also help explain the puzzle?

In one sense, this is part of the answer. In the period under investigation, petrostates have

been led by revolutionary leaders in more years, on average, than non-petrostates have. To be

precise, on average petrostates have been led by revolutionary leaders 19% of the years under

investigation, while non-petrostates have been led by revolutionary leaders just 14% of the

years.42 (Clearly, these numbers represent cross-national averages: some states such as Cuba and

Libya have been led by revolutionary leaders for as much as 80% of the years under

investigation, while others have not seen any revolutionary leaders.) To the extent that there is a

correlational difference in the rate of revolutions and revolutionary leaders between petrostates

and non-petrostates, this contributes to the difference in the rate of aggressor-MIDS.

However, this is not to say that an oil economy causes the emergence of revolutionary

leaders. Instead, the correlation between petrostates and revolutionary leaders may be a spurious

correlation caused by their relationship to other, more fundamental factors. For instance, the

Muslim percentage of the state�s population and the state�s geographic region are strongly

42 These figures are for developing countries only; OECD states are excluded.


33

correlated with the presence of revolutionary leaders. States where Muslims are more than 75%

of the population have an incidence of revolutionary leaders of 21%; states where Muslims are

less than 25% of the population have an incidence of revolutionary leaders of just 11%.

Similarly, states in the Middle East have higher incidence rates of revolutionary leaders

compared to other parts of the world. Obviously, petrostates are disproportionately likely to be

found in the Middle East and have large Muslim populations, but no one would claim that the oil

caused these factors; it is simply coincidence. It is thus plausible that it is the state�s religious

make-up and/or its geographic region that causally affects the incidence of revolutionary leaders,

rather than the presence of an oil industry in the economy.

Testing these claims involves the development of a causal model for the emergence and

incidence of revolutionary leaders, which is far beyond the scope of this project. This paper

treats the causes of revolutions as exogenous to the model. Nonetheless, a preliminary

investigation of the correlational relationship between petrostates and revolutions is possible.

The dataset on revolutionary regimes has been treated thus far as an independent variable in this

paper, but it can also be used as a dependent variable. Logit regression analysis can be used to

examine the correlates of revolutionary regimes, using many of the same correlates that have

been used throughout this empirical investigation. Table 5 below shows the results from this

very preliminary model of the relationship between petrostates and revolutionary regimes.

The results suggest that petrostates are not strongly correlated with revolutionary

regimes, throwing any hypothesized causal relationship into doubt. Model A is a very basic

regression, without using regional dummy variables, the Polity score, and the Muslim percentage

of the population as control variables. This model shows that without those three correlates,

there is a correlation between petrostates and revolutionary regimes, as suggested by the


34

aggregate figures given above. However, the rest of the models show that this is misleading. In

model B, when the full set of control variables is used, the coefficient for Petrostate is almost

precisely zero, and is statistically insignificant. Notably, the coefficient for the Middle East

region (not shown in Table 5) is strongly significant and positive. Also, the Polity score is

negative and significant, suggesting that an (unsurprising) correlation exists between autocracies

and revolutionary regimes. Model C replicates the format of model 1, but this time includes

fixed effects; the results also suggest that Petrostate is not correlated with revolutionary regimes.

Model D is similar to Model B, but excludes the Polity scores as a control variable. This result

shows that even if Polity scores are excluded on the basis that they might be endogenous to

Petrostate, there is still little or no evidence that there is a correlation between petrostates and

revolutionary regimes. While these results do not represent a comprehensive causal model of the

incidence of revolutionary regimes, they do cast doubt on any correlational relationship between

petrostates and revolutionary regimes. In turn, the lack of correlation casts doubt on any causal

role.


35

Table 5: Correlates of Revolutionary Regimes

Thus this prima facie investigation into the relationship of oil and revolutionary regimes

suggests that oil probably does not cause revolutions. This only reinforces the importance of the

conclusion suggested by the analysis above: that it is the interaction between a revolutionary

regime and significant oil income that is the key driver in explaining why petrostates engage in

so many militarized interstate disputes.

Conclusion

States led by revolutionary regimes are significantly more aggressive and engage in more

militarized interstate disputes (MIDS) than states led by non-revolutionary regimes. The

evidence documented in this paper, gathered on the basis of a new dataset that enables the


36

classification of regimes, provides strong empirical support for the theoretical framework being

tested. This paper also shows that revolutionary states are especially aggressive and often

instigate international disputes. These findings are an important corrective to the focus on

international systemic factors and the �balance of threat� that have traditionally been used to

explain the relationship between revolution and war. Significantly, two hypotheses suggested by

Walt were shown to lack empirical support in this analysis. Thus this paper helps advance our

understanding of revolutions and international conflict.

Just as importantly, this paper shed new light on the relationship between oil and

international conflict. Petro-revolutionary regimes constitute a special threat to international

peace and security. The evidence reveals that revolutionary leaders in petrostates are

dramatically more aggressive and engage in more MIDS than other kinds of states, even

compared to states which are led by revolutionary leaders but which do not have large oil export

revenues. The interaction of oil and revolution leads to a state with an unrestrained, warlike

regime capable of harnessing a significant portion of the state�s resources towards international

conflict. In short, the combination of oil and revolutionary regimes is explosive.

The implications for the global political economy are significant. The international trade

of oil injects vast amounts of money into political systems that are ill-suited to receive it, with

significant consequences for the centralization of political power and the build-up of military

capability in oil-exporting states. These facts raise a host of political, economic, and moral

questions about the world�s continued dependence on petroleum as a primary source of energy in

advanced economies.

These findings are especially significant in light of Middle East politics. Observing the

correlation between oil and international conflict, one might be tempted to believe that this is


37

because much of the world�s oil is found in the Middle East, and the Middle East is a violent and

conflict-prone region. On that view, the correlation might be spurious. However, this paper

shows that when the analysis controls and adjusts for regional effects, the effect of petro-

revolutionary regimes remains strong. That is, even relative to other Middle East states, the

revolutionary regimes in Iran, Libya, and Iraq are especially prone to aggressively engage in

MIDS. Moreover, the toxic interaction between oil income and revolutionary regimes can be

seen in the behavior of states outside of the Middle East, such as Venezuela and Russia. On the

whole, the evidence suggests that oil contributes significantly to the violence in the Middle East.

Perhaps surprisingly, the evidence shows that petrostates with non-revolutionary regimes

may have a propensity for MIDS somewhat lower than comparable non-petrostates. This is

consistent with the hypothesis tested here, but contradicts much of the conventional wisdom

about oil and war. Although it has often been alleged that there is fierce international

competition for oil that leads to attempts at militarized conquest of petrostates, in fact the

evidence shows that non-revolutionary petrostates do not face an above-average propensity for

MIDS. This finding is even more striking because the opposite trend holds overall: petrostates

engage in more MIDS on average than non-petrostates. The implication is that the overall

above-average rate of MIDS is driven almost entirely by the small subset of revolutionary

regimes in petrostates, which have an extremely high propensity for MIDS. Overall, the analysis

confirms that many of the standard explanations for the relationship between oil and

international conflict are unsatisfactory.

The potential uses of the new dataset introduced in this paper extend far beyond the

present analysis. To date, scholars have lacked a comprehensive dataset of revolutionary

regimes subject to a well-specified definition that could be used for quantitative analysis. This


38

dataset could be used as a dependent variable for scholars seeking to understand the causes and

conditions under which revolutions occur. It could also be used by scholars interested in the

comparative dynamics of revolutionary movements, such as investigating how long leaders

remain in power under various conditions. Alternatively, the dataset could be used as an

independent variable to investigate an almost limitless set of comparative and international

behaviors, including conflict, trade, and diplomatic policies. Another possibility is for this

dataset to be introduced into existing studies of phenomena such as the democratic peace, to

observe whether this variable plays an explanatory role. Thus this new dataset opens doors to a

wide variety of studies in future research.


39

Appendix


40

References Bueno de Mesquita, Bruce et al. 2004. �Testing Novel Implications from the Selectorate Theory of War.� World Politics 56(3):363-388. Colgan, J. PhD Dissertation, forthcoming (expected 2009). Collier, P., and Hoeffler, A. 2001. �Greed and Grievance in Civil War.� World Bank. Dassel, K. 1996. �Revolution and war: A critique.� Security Studies 6(2):152-173. Enterline, Andrew J. 1998. �Regime Changes and Interstate Conflict 1816-1992.� Political Research Quarterly 51(2). Fearon, J. and Laitin, D. 2003. �Ethnicity, Insurgency, and Civil War,� American Political Science Review, 97:1. Goemans, Hein, and Giacomo Chiozza. 2004. �Avoiding Diversionary Targets.� Journal of Peace Research 41(4):423-443. Goemans, Hein, and Mark Fey. 2008. �Risky but Rational: War as an Institutionally Induced Gamble.� Journal of Politics. Goldstone, J. A. 1996. �Revolution, war, and security.� Security Studies 6(2):127-151. Gowa, J., and T. Schwartz. 1999. Ballots and bullets: the elusive democratic peace. Princeton University Press. Gurr, Ted R. 1988. �War, Revolution, and the Growth of the Coercive State.� Comparative Political Studies 21(1):45. Haber, S. and Monaldo, V. 2007. �Do Natural Resources Fuel Authoritarianism?� Working paper, unpublished. Homer-Dixon, T. F. 1999. Environment, Scarcity, and Violence. Princeton University Press. Huntington, S. P. 1996. The clash of civilizations and the remaking of world order. University of Chicago Press. Jones et al., 1996. �Militarized Interstate Disputes, 1816-1992� Conflict Management and Peace Science, 15(2). Karl, T.L. 1997. The Paradox of Plenty. Berkeley: University of California.


41

Klare, Michael T. 2004. Blood and Oil: The Dangers and Consequences of America's Growing Dependency on Imported Petroleum. 1st ed. Metropolitan Books. Klare, Michael T. 2002. Resource Wars: The New Landscape of Global Conflict. Holt Paperbacks. Kydd, A. 2000. �Trust, Reassurance and Cooperation.� International Organization 54(2):325-357. Lai, B., and D. Slater. 2006. �Institutions of the Offensive: Domestic Sources of Dispute Initiation in Authoritarian Regimes, 1950-1992.� American Journal of Political Science 50(1):113-126. Le Billon, P. 2005. Fuelling War: Natural Resources and Armed Conflict. Routledge for the International Institute for Strategic Studies, London. Levy, J. S. 1988. �Domestic Politics and War.� Journal of Interdisciplinary History 18(4):653-673. Mansfield, Edward D., and Jack Snyder. 2005. Electing to Fight: Why Emerging Democracies Go to War. The MIT Press. Maoz, Z., and B. Russett. 1993. �Normative and Structural Causes of Democratic Peace.� American Political Science Review 87(3):624-38. Morrow, J. D. 1993. �Arms versus allies: trade-offs in the search for security.� International Organization, 47:207-207. Pelletiere, S. C. 2004. America's Oil Wars. Praeger Publishers. Ramsay, K. 2006. �The Price of Oil and Democracy,� Working paper, unpublished. Ross, Michael L. 2001. �Does Oil Hinder Democracy?.� World Politics 53(3):325-361. Ross, Michael L. 2006. �A Closer Look at Oil, Diamonds, and Civil War,� American Review of Political Science, 9. Russett, B. 1981. �Security and the resources scramble: will 1984 be like 1914?.� International Affairs 58(1):42-58. Russett, B. M., and W. Antholis. 1993. Grasping the democratic peace: principles for a post-Cold War world. Princeton University Press. Smith, B. 2004. �Oil Wealth and Regime Survival in the Developing World, 1960-1999,� American Journal of Political Science, 48:2.


42

Skocpol, T. 2000. �Social Revolutions and Mass Military Mobilization.� Revolution: Critical Concepts in Political Science 40(2):147-68. Skocpol, T. 1979. States and Social Revolutions: A Comparative Analysis of France, Russia and China. Cambridge University Press. Snyder, R. S. 1999. �The US and third World Revolutionary States: Understanding the Breakdown in Relations.� International Studies Quarterly 43(2):265-290. Tomz, Michael. 2007. �Domestic Audience Costs in International Relations: An Experimental Approach.� International Organization 61(4):821-40. Walt, Stephen M. 1992. �Revolution and War.� World Politics 44(3):321-368. Walt, Stephen M. 1996. Revolution and War. Cornell University Press. Weeks, Jessica. 2008. �Autocratic Audience Costs: Regime Type and Signaling Resolve.� International Organization 62(1):35-64. Westing, A. H. 1986. Global Resources and International Conflict: Environmental Factors in Strategic Policy and Action. A SIPRI Publication.

Oil and Revolutionary Regimes: A Toxic Mix...Oil and Revolutionary Regimes: A Toxic Mix Jeff Colgan,...

Documents

Transcript of Oil and Revolutionary Regimes: A Toxic Mix...Oil and Revolutionary Regimes: A Toxic Mix Jeff Colgan,...