Although there has been much debate about whether democratization causes ethnic conflict, and many comparativists have argued about which kinds of political institutions are best for managing communal strife, little large-JV work has addressed these issues. The authors apply a theory of ethnic conflict—the ethnic security dilemma—to derive predictions about the impact of democratization and political institutions on ethnic unrest. They then test these predictions by performing a series of pooled time-series analyses covering all ethnic groups in the Minorities at Risk data set from 1985 to 1998. The authors find that democratization, federalism, and presidentialism may not be as problematic as some argue and that proportional representation tends to reduce severe ethnic violence. They conclude by suggesting some directions for future research.
DEMOCRATIZATION, POLITICAL INSTITUTIONS, AND ETHNIC CONFLICT A Pooled Time-Series Analysis, 1985-1998
STEPHEN M. SAIDEMAN Texas Tech University/ U.S. Department of Defense
DAVID J. LANOUE University of Alabama
MICHAEL CAMPENNI SAMUEL STANTON Texas Tech University
The third wave of democratization has coincided with an increase in the number of ethnic groups in conflict (Gurr, 1993). Even more strikingly, the decline in the proportion of liberal democracies after 1992 (Diamond, 1996) has been followed by a decrease in the number of new ethnic conflicts
AUTHORS' NOTE: We appreciate the suggestions made by the editor and reviewers of this journal. We are grateful to Ted Gurr, Anne Pitsch, Deepa Khosa, and the Center for International Development and Conflict Management at the University of Maryland for providing access to the Minorities at Risk data set and assistance in using it. We also would like to thank Amy Heyman of the World Bank Data Development Group for providing us with the economic data.
COMPARATIVE POLITICAL STUDIES, Vol. 35 No. 1, February 2002 103-129 © 2002 Sage Publications
103
104    COMPARATIVE POLITICAL STUDIES / February 2002
(Gurr, 2000). These apparent trends should give scholars pause before recommending democracy as a solution to ethnic conflict.
However, ethnic conflict varies over time and among democracies, raising a number of questions. If political competition exacerbates communal tensions, do increases in ethnic tensions coincide with democratization? Are the differences in ethnic unrest among democracies due to variations in political institutions? Are presidential systems more prone to ethnic conflict than parliamentary democracies? Does the electoral system matter? Does federalism cause more problems than it solves? This article addresses these questions, which are central both to debates in the comparative politics literature on institutions and to policy makers as they attempt to design the most effective constitutional arrangements.1
PREVIOUS RESEARCH
By addressing how political institutions influence levels of ethnic protest and rebellion, we are stepping into a long and lively debate about the impact of dissatisfaction on protest and the ability of individuals to organize into effective groups. Although we do not derive our arguments from theories of relative deprivation (Gurr, 1970) or policy dissatisfaction (Barnes & Kaase, 1979), we do focus on conditions that shape the security of ethnic groups. We therefore face some of the same problems, including the assumption that groups are acting and responding to institutions as if they are unitary actors. Clearly, assuming that groups act in this manner requires that individuals have been able to solve important collective action problems.
Rather than addressing the vast literature on collective action, we highlight a few salient points. First, scholars have shown that rational individuals will engage in collective dissent if they believe that the group to which they belong will benefit from group efforts more than they, personally, will lose (Finkel & Muller, 1998; Muller & Opp, 1986).2 Political institutions ought to
1. In addition to the literature cited below, Regan (2000), among others, has focused on outsiders' efforts to manage other people's ethnic conflicts. See also Saideman (2001) for a discussion of other kinds of conflict management techniques.
2. This argument is, of course, quite controversial (Green & Shapiro, 1994).
Conversations with Cherie Maestas, Craig Emmert, James Fearon, Arend Lijphart, Ross Marlay, Grant Neeley, Brian Sala, and Matthew Shugart greatly assisted us in our coding and methodology. Will Moore provided very valuable comments. An earlier version of this article was presented at the 1999 Annual Meeting of the American Political Science Association, Atlanta Marriott Marquis and Atlanta Hilton and Towers, September 2-5,1999. Of course, all remaining errors in our project are the authors' responsibility. The views presented here do not represent those of the U.S. Department of Defense.
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT
105
matter in these calculations because some may increase the likelihood of gains for the organized groups, whereas others may increase the likely costs of organization. Furthermore, this dynamic should be especially true if we assume that ethnic groups, by definition, perceive themselves as part of a larger community and that their chances for success (broadly defined) are contingent on how the group is treated (Horowitz, 1985).
Second, we find Lichbach's (1994) arguments to be persuasive: that collective action problems of potential dissenters are not insurmountable and that we ought to move beyond these problems to address other important questions. Thus for the purposes of this research, we assume that collective action problems get solved,3 so that ethnic groups can and do act as groups. The question then focuses on what conditions are associated with greater or less ethnic dissent, both peaceful and violent.
Surprisingly, there has been relatively little large-Af analysis of the relationships between political institutions and ethnic conflict. Instead, most analysts have considered the effects of system type and electoral rules on regime survival (Bohrer, 1997; Power & Gasiorowski, 1997; Stepan & Skach, 1993), on the number of parties or viable candidates (Cox, 1997; Lijphart, 1994; Moser, 1999; Shugart & Taagepera, 1994), on citizen satisfaction (Anderson & Guillory, 1997), on political repression (Davenport, 1997),4 and on economic success (Crepaz, 1996).
One of the few quantitative studies addressing institutions and ethnic conflict is Cohen's (1997) work using the Minorities at Risk (MAR) Phase I data set. He found that federalism increases protest but reduces rebellion and that the use of proportional representation (PR) electoral systems reduces rebellion. However he did not control for past political behavior. Furthermore, Cohen omitted both economic variables and questions of presidentialism versus parliamentarism. His analyses also do not say much about regime type and duration. Still, Cohen provided interesting findings for the period from 1945 to 1989. One of our tasks, then, is to consider whether his findings hold up throughout the 1980s and 1990s.
In our study, we use two different measures of ethnic conflict: protest and rebellion. The first involves making an appeal to government leaders for redress of grievances; the second represents a conscious attempt to destabilize (and in some cases overthrow) the government itself. We think
3. Lichbach (1994) asserted that the collective action problems of dissenters, or the rebel's dilemma, are actually relatively easy to solve.
4. There has been an explosion in studies of repression and dissent, including sophisticated analyses considering the interaction between government decisions to repress and efforts to engage in dissent. For extensive discussions of this literature, see Davenport (2000) and Moore (2000). Due to space and data limitations, we do not address repression here.
106    COMPARATIVE POLITICAL STUDIES / February 2002
that the distinction between nonviolent and violent dissent is a meaningful one, particularly because some institutions may be intended to facilitate nonviolent dissent specifically to prevent the use of force by disgruntled ethnic groups. As we discuss our hypotheses below, we specify when we expect an institution to have a differential impact on protest and rebellion. By focusing on protest and rebellion, we can determine whether these operate as two distinct phenomena.
THE ETHNIC SECURITY DILEMMA
The ethnic security dilemma refers to competition between ethnic groups for control of the government. This view applies the international relations concept of the security dilemma to domestic politics. In international relations, the absence of government, anarchy, leaves states with no one to rely on except themselves. This self-help system causes states to worry about what other countries might do to them. Because states in such an environment must consider what other states can do rather than what they say they will do, any attempt by one state to increase its security threatens its neighbors' security. Thus the dilemma is that any efforts to increase one's security tend to leave all players worse off (Jervis, 1978; Waltz, 1979).
Posen (1993) first applied the security dilemma to ethnic conflict. He considered the competition of groups in the absence of the state as empires collapse. He focused on the translation of military-strategic variables to internal conflicts. There is, however, a key flaw in this approach. Because it starts with the existence of anarchy within states, it fails to explain why governments collapse,5 limiting how widely we can apply Posen's theory. We revise the ethnic security dilemma to facilitate its application beyond collapsed empires and failed states.6
Our view of the ethnic security dilemma starts with the idea that the government of any state is the greatest potential threat to any group inside its boundaries (Rummel, 1994). It usually takes a state's resources to commit genocide, and the fear of group extinction is an important element of ethnic identity and group conflict (Horowitz, 1985). Groups may fear that others control the government and may use its resources (the army, the secret police, the courts, economic influence) against them. Thus the search for security motivates groups in divided societies to seek to control the state or secede if
5. Snyder and Jervis (1999) developed the logic of Posen's (1993) ethnic security dilemma further, noting but not really correcting this limitation of Posen's conceptualization.
6. For a version of the ethnic security dilemma focused on domestic politics, see Saideman (1998).
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT 107
the state's neutrality cannot be assured. Obviously these efforts can exacerbate the situation, because one group's attempts to control the state will reinforce the fears of others, so they respond by competing to influence and even control the government. The ethnic security dilemma, then, essentially occurs when the efforts of one group to control the government cause all to be worse off. The competition between groups creates the risk that a relatively neutral or harmless government will fall into the hands of one group that could dominate the others.
To be clear, the purpose of our article is not to test the ethnic security dilemma. Rather, the concept is useful for developing a set of expectations concerning which institutions might matter and what are their likely effects. Ethnic groups will be more secure if they have access to decision makers, if they can block harmful government policies, and if they can veto potentially damaging decisions. Thus such factors as electoral systems, the power of executives and legislatures, and federalism may serve to limit or enhance the threat felt by various groups. Institutions that promote power sharing or provide some level of representation or self-government for minority groups are therefore assumed to be more likely to reduce conflict.
Closely related to these institutional concerns is the question of uncertainty. Because the tendency toward protest and violence is inversely related to group security, we assume that groups are more likely to act up when they are uncertain about their position and prospects for the future. Thus we expect more ethnic conflict during periods of institutional upheaval in which new regimes are installed and elections are held for the first time. In addition, because authoritarian regimes often provide a certain level of stability, sometimes enforced ruthlessly, it is possible that ethnic unrest will, ceteris parabus, actually be greater in democracies.
Based on these general issues of inclusiveness and uncertainty, we can develop a number of hypotheses about the relationships between political institutions and ethnic protest and rebellion. We will divide these hypotheses into two classes: hypotheses regarding regime type and hypotheses related to specific political and electoral institutions.
QUESTIONS OF REGIME TYPE
As noted above, we expect that ethnic protest is more likely in democracies because democratic competition creates insecurity. But there are additional reasons to expect a relationship here. Under democratic regimes, the costs of protesting are less and the perceived benefits are greater (protests might influence politicians). However, does regime type affect the use of vio-
108    COMPARATIVE POLITICAL STUDIES / February 2002
lence? The core notion of the ethnic security dilemma is the fear that another group may attempt to control the state at one's expense. Although the capture of a government by a hostile ethnic group can occur in both authoritarian and democratic regimes, the temptation is perhaps greater in a democracy in which leaders often compete with each other to be the best defender of their ethnic group. Ethnic outbidding is a well-known dynamic in ethnically divided democracies (Horowitz, 1985; Rabushka & Shepsle, 1972; Rothschild, 1981).
Competition with other elites causes politicians to be more nationalist than their rivals. This intragroup competition interacts with intergroup competition to cause groups to fear one another and thus to throw their support behind the most nationalist politicians. This fear can be quite rational; the competition among Group A's nationalist politicians, after all, increases the risks for members of Group B, especially if they support less nationalist politicians (de Figueiredo & Weingast, 1999). Democracies make this kind of process more likely because of the necessity to play to a larger audience. In authoritarian regimes, leaders may compete with each other for the support of the relevant constituency, but because this constituency is smaller, it is easier to target more specific inducements to prevent the defection of supporters and to attract the other officials' constituents. If it is easier to bribe or coerce a few people, then appealing to ethnic identities is less necessary. Because competition between politicians is likely to exacerbate fears and competitive pressures among ethnic groups, we expect that ethnic protests and rebellion are more likely in democracies.7 Therefore we can posit our first hypothesis:
Hypothesis 1: Ethnic protests and rebellion are more likely in democracies than in authoritarian regimes.
We must also draw the contrasts between old and new political systems. The coincidence of democratization and the rise of ethnic conflict in the early 1990s suggest that political change, regardless of its direction, causes ethnic conflict and/or worsens existing divisions. That is, alterations in political structures may cause more ethnic unrest, even when the polity is moving toward greater democracy. The ethnic security dilemma suggests that transitions are likely to intensify ethnic group insecurity, because groups become uncertain about their futures and are tempted to act preemptively. In addition, the state is more vulnerable to capture by one group at the expense of others during transitions. Moreover, older regimes, democratic or not, are more
7. Focusing on repression might also lead to the same expectation (Davenport, 1995). Likewise, collective action to organize protests and rebellion is probably easier in democracies than in authoritarian regimes.
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT 109
likely to have worked things out, either satisfying the demands of competing groups or discouraging dissent through enduring repressive institutions.8 Hence, the following hypothesis:
Hypothesis 2: Ethnic protests and rebellion are more likely in states with younger political institutions.
Perhaps it is not the mere fact of political change that exacerbates ethnic strife but rather the direction of the change—toward democracy (Snyder, 1999). The ethnic security dilemma suggests two distinct logics with conflicting conclusions. First, security dilemmas suggest that political competition can increase ethnic strife, and political change increases uncertainty. Thus the combination should increase ethnic insecurity even more. One security-oriented logic suggests that democratization is likely to increase ethnic strife (Saideman, 1998).
However, as events in the former Soviet Union testify, ethnic conflict can also cause democratization. That is, new democracies often represent some form of settlement to the conflicts of the past. If the conflict and competition for political power among ethnic groups leads to genuinely democratic political competition, this might reduce rebellion, if not protest, because groups will have more say over their political fortunes. Given that rebellion is quite costly, groups in new democracies might try to work through the system before they become dissatisfied and engage in serious dissent (Gurr, 2000). Although competition can exacerbate ethnic conflict, our assessment is that increased access will ameliorate it, at least in the short term.
Hypothesis 3: Young democracies are less likely to experience severe ethnic strife.
There is one remaining possibility: The outbreak of ethnic conflict in Eastern Europe and the former Soviet Union coincided with the first meaningful elections in many years. Because authoritarian regimes can decide to hold elections before a real transition to democracy occurs (as was the case in Yugoslavia), first elections and young democracy are not the same thing.9 Indeed, the first elections a country has may or may not lead to democracy, as the experience of many countries demonstrates. If first elections cause ethnic conflict, as the early 1990s suggest, then we need to pay special attention to them (Turner, 1993). There should be more dissent during a first election, because this is often the first opportunity to make clear what a group desires.
8. Indeed, this raises a problem for our analyses: Political change may be endogenous.
9. In our data set, first elections are not perfectly correlated with democracies, although they are negatively correlated with regime duration; first elections tend to occur in younger regimes.
110    COMPARATIVE POLITICAL STUDIES / February 2002
We should also expect more protests during first elections because they are often characterized by irregularities. Regarding rebellion, ethnic insecurity is similarly likely to rise in a year in which a system faces its first election. A first election increases uncertainty about who will rule. Whoever wins the first election may change the rules so that there is no second election or may work to ensure that future elections will not be competitive. Thus competition among groups should be intense, and groups will have more to fear. Therefore,
Hypothesis 4: When a country undergoes its first election under the current political system, ethnic protests and rebellion are more likely.
POLITICAL INSTITUTIONS AND CONFLICT
In democracies, the rules determine how much voters and elections matter. The constitutions of countries determine how much power is held by the winners and whether those who come in second (or worse) have any power, influencing how severe the ethnic security dilemma may be. There are three basic dimensions under which analysts consider the ability of particular institutions to manage ethnic conflict: the nature of the executive, the type of electoral system, and the distribution of power between the central government and subunits.
PRESIDENTIALISM VERSUS PARLIAMENTARISM
Comparativists have engaged in a lively debate about whether presidential or parliamentary systems are better for political stability. Some argue that presidential systems are better because of greater accountability, greater identifiability, and the existence of mutual checks (Shugart & Carey, 1992, pp. 44-49; Shugart & Mainwaring, 1997).
The ethnic security dilemma theory suggests that presidential systems are superior in reducing conflict. According to the ethnic security dilemma, ethnic groups are more insecure, and thus most likely to engage in violence and preemption, when they cannot block policies that might hurt them. The division of powers between president and legislature allows each to serve as a check on the other, even when the same party dominates both branches. Parliamentarism, on the other hand, can be quite threatening to minority groups if they cannot get significant representation and especially threatening if one party tends to gain control with no need for coalitions. In presidential systems, the parties in the assembly may choose which of the president's
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT      111
policies they will support. Ethnic groups may be safer in presidential systems because there are more points within the system to block unfavorable actions.10
Hypothesis 5: Ethnic conflict is more likely in parliamentary systems than in presidential systems.
ELECTORAL SYSTEMS
Ultimately, the implications of presidentialism or parliamentarism are not clear without considering the electoral system. The debate over electoral systems has been livelier than the one over presidentialism versus parliamentarism (Lardeyret, 1991; Reynolds, 1995). The choice of electoral system is crucial for shaping political outcomes. Plurality systems tend to exaggerate the power of the strongest parties, whereas proportional representation is designed to produce what the name implies—a distribution of seats proportional to the votes received (Lijphart, 1994).
What does the ethnic security dilemma suggest about electoral systems? We propose that plurality systems increase group insecurity, because political change can be quite dramatic, and exclusion often results. If groups cannot block objectionable policies, then they may have to engage in either protest or violence to have influence. In a PR system, on the other hand, coalitions can change, but ethnic groups may gain representation and could play a significant role either as a coalition partner or in opposition to the government. Therefore,
Hypothesis 6: Ethnic conflict is more likely in systems characterized by plurality than those with PR.
FEDERALISM
Whereas the two previous sections refer to institutions in democracies, federalism can exist in both democratic and authoritarian regimes. For instance, in Tito's Yugoslavia, significant political power resided at the republic level. Indeed, some argue that Yugoslavia disintegrated because of its federal design, as did two other federations—Czechoslovakia and the Soviet Union (Roeder, 1991; Saideman, 1998; Snyder, 1999).
Lijphart (1977), among others, argues that federalism is a useful tool for managing ethnic strife. Segmental autonomy is an essential ingredient for consociationalism. By giving groups some control over their own lives, there may be less conflict about who controls the central government. Although
10. For a recent and thorough discussion of veto players, see Tsebelis (2000).
112    COMPARATIVE POLITICAL STUDIES / February 2002
federalism is just one form of segmental autonomy, it is more widely practiced and facilitates other power-sharing institutions, including bicameralism, requirements of supermajorities for constitutional change, and the like (Horowitz, 1993; Lijphart, 1991).
Accordingly, the ethnic security dilemma suggests that federalism reduces ethnic strife. Federalism gives many groups more control over outcomes than they would otherwise have. There are two caveats. First, federalism often puts smaller minorities at risk, as the titular nationality may dominate the republic at the expense of other groups. Before the disintegration of the Soviet Union, non-Georgians felt threatened in Georgia.11 Second, federalism facilitates collective action, as individuals can influence their local government with their participation. We therefore expect that federalism will increase protest but decrease rebellion, concurring with Cohen.
Hypothesis 7a: Ethnic protest is more likely in systems characterized by federalism.12 Hypothesis 7b: Ethnic rebellion is less likely in systems characterized by federalism.
ACCOUNTING FOR OTHER FACTORS
Before moving on, we need to address other dynamics that may be responsible for ethnic conflict. By including these, we can better determine the impact of institutions. Conflict, for example, may be more severe in poorer countries than in rich ones (Lipset, 1991). Rich countries may be able to buy off ethnic conflict; poor countries are less able to accommodate demands. Conflict is also more likely when the economy is in decline. Economic downturns cause people to lose their jobs, put pressure on politicians, and force the government to divide a smaller pie. Consequently, in bad economic times, people are more likely to blame other ethnic groups for their problems, and we should expect more strife during such periods.
11. Our dependent variable of rebellion focuses solely on violence between a group and its host country, not among groups, so conflicts between the titular nationality and minorities within the republics do not show up in our analyses. Communal conflict, violence between groups, is coded for a much shorter time period by the Minorities at Risk Project (Gurr, 1999) and has not been subjected to the same reliability assessments. Still, in initial tests we found that federalism was positively associated with communal violence, suggesting that federalism works in that it focuses conflict away from the center, causing minorities within the subunits to be more dependent on the central government.
12. We do not distinguish in our data set between nonethnic federalism (as in the U.S. case) and ethnofederalism (as in the Soviet case). This is one direction for future research.
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT      113
These factors do not distinguish between groups within a country,13 because they account for how well the country is doing as a whole. Groups frequently vary in how well or how badly they are doing economically and politically. One would expect that relatively disadvantaged groups would be more likely to conflict with the government than those that are doing relatively better. We therefore include in our analyses indicators for economic and political differences.
Two variables are included to deal with conventional arguments about ethnic conflicts. One addresses the identities themselves. Scholars have argued that groups that are different along many dimensions are more likely to fight with each other than when cross-cutting cleavages exist (Lipset, 1960). By considering cultural differences, we can address these claims.
We must also consider whether or not ethnic groups are concentrated. First, studies of ethnic conflict show that more concentrated groups are more likely to engage in conflict (Ayres & Saideman 2000; Byman, 1997; Gurr, 1993; Saideman & Ayres 2001). Second, a few of our variables interact with group concentration, particularly electoral systems and federalism. If a group is dispersed, then plurality electoral systems are more problematic, because members of that group may never be able to gain representation. Concentrated groups, however, may retain political viability in a plurality system because they can win pluralities where they are concentrated (the Scots in Britain, African Americans). We examine whether concentration interacts with the other variables by including a subsequent set of analyses, focusing on groups that are largely dispersed versus those that are highly concentrated.
RESEARCH DESIGN AND MEASUREMENT
Our starting point is the MAR data set (Gurr, 1999), which provides our dependent variables and our group-specific independent variables.14 The MAR dataset, Phase III (Version 899), has as its unit of analysis individual ethnopolitical groups and includes 275 groups in 116 countries.15 The data set
13. Although it would be best to use actual economic statistics for each group—for instance, a group's unemployment—such data are simply not available for most groups.
14. This data set determines our time frame, because yearly codings of protest and rebellion are available only from 1985 to 1998. Our data set is available on request.
15. Specifically, minorities at risk are defined as those ethnic groups that as groups gain from or are hurt by systematic discriminatory treatment compared with other groups and/or groups that are the basis for political mobilization for the promotion of the group's interests. Our data set contains 264 groups due to missing data. The Minorities at Risk data set considers groups in every country, including only those that meet these additional criteria relating to size (Gurr, 1993, 2000), except advantaged majorities.
114    COMPARATIVE POLITICAL STUDIES / February 2002
is not organized for time-series analysis, nor does it contain data on elections, political institutions, or countries' economic well-being. We added additional variables to the data set and organized it for pooled cross-sectional analysis. Thus our units of observation are each relevant ethnic group within a country for a given year. As an example, U.S. African Americans will represent 14 data points (1 for each year) and will be labeled "African Americans 1985," and so forth. Once we have done this for every group, we refer to our cases as "group/years." Our sample thus includes 3,696 group/years.16
Because our primary interest is the influence of political and electoral competition on ethnic unrest, we consider two indicators of ethnic conflict: protest and rebellion. One might argue that these indicators should be treated as a continuous variable, ranging from the most benign form of protest to the highest form of rebellion. Our hypotheses, however, suggest that different political institutions may affect protest and rebellion separately. Therefore we will consider each as a distinct phenomenon.
The MAR data set creates an ordinal-level scale in which low numbers represent low levels of activity and high numbers indicate the strongest possible actions in each category. For example, a nonviolent protest of more than 100,000 people would achieve the highest protest score; an ethnic group engaging in protracted civil war would receive the highest score on the rebellion variable, with the highest level of protest/rebellion during a given year serving as the score for the entire year. Moreover, protest and rebellion are coded only if they involve ethnic group-based activity (e.g., even a massive antinuclear protest would not be coded). In our sample, protest ranges from 0 through 5, whereas rebellion ranges from 0 through 7. We should note that the modal category in each case is zero: In 37.9% of all group/years, the ethnic group in question did not protest; rebellion is coded 0 for 73.4% of all cases.
Our independent variables fall into three categories: measures of regime type and duration, indicators of political institutions, and control variables.
16. In our analyses below, our total number of observations is lower than 3,696—roughly 2,560 for our analyses of all groups. The missing data are largely due to variables related to rebellion and protest before 1991, regime type, and economic conditions. In the Minorities at Risk data set, several groups are not coded for dissent before 1991 because they did not fit the definition of minority, such as Russians in Estonia (thus our real data set was never actually 3,696). In the Polity data set (Gurr & Jaggers, 1999), many countries are coded as undergoing transitions, experiencing interruptions (e.g., foreign occupations), or experiencing complete collapse of governments. We treat these cases as missing. This presents a potential bias of removing some of the most intense conflicts in the set, because the most severe ethnic wars often destroy the government in the process (Somalia as a case in point). This is not too problematic for our study, because we are trying to say something about how institutions drive behavior, and in these most severe cases, institutions no longer matter. Finally, some observations are missing economic information.
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT
115
For regime type, we use two variables from the Polity98 data set (Gurr & Jaggers, 1999). One of these variables, democracy, ranges from 0 through 10 and indicates "the general openness of political institutions."17 The other, autocracy, is comprised of a 0-to-10 scale, indicating "the general closedness of political institutions."18 Thus our indicator of regime type is calculated simply by subtracting the autocracy score from the democracy score of each country for each year, resulting in a scale from -10 to 10 that measures the relative openness of political competition.19
For regime duration, we recode Polity98's indicator for durability of the regime20 into a dichotomous variable, coding regimes that have endured for more than 20 years as 1 and younger regimes as 0. We choose the passage of 20 years because it should generally take a full generation before institutions bind behavior.21 We specify two additional dummy variables to address our hypotheses about new regimes. The first accounts for whether a state is a young democracy and is scored 1 for each case in which the country in question has maintained a democratic form of government (measured as a regime type score greater than 5)22 for 10 years or less and 0 if not.23 A second measure, first election, is scored 1 for any year in which a country is holding its first election under the current regime and 0 for other years.24
Next, we consider measures of institutions. Each country with a federal system of government receives a score of 1 on our variable "federalism" (0 otherwise).25 We develop two indicators measuring the extent to which a
17. Openness is a composite of indicators for whether the executive recruitment is competitive and open, whether the executive is constrained, and whether "there are relatively stable and enduring political groups which regularly compete for political influence at the national level." These definitions, and the ones in the next note, are from Polity98's Web site.
18. Closedness is a composite of indicators for whether party competition is suppressed and whether parties are restricted, whether executives are selected, whether the executive selection is a closed process, and whether executives are relatively unconstrained.
19. This is the conventional approach for using Polity data (Ward & Gleditsch, 1998).
20. Polity98 codes durable as an "indicator of polity durability based on the number of years since the last regime transition or since 1900 (whichever came last in time)."
21. We used other specifications of durability, including 10 years, and received similar findings.
22. The Democratic Peace debate makes extensive use of Polity, and scholars have generally used either six or seven as their minimum definition of democracy.
23. If we use democracies that are 5 years or younger, the results do not change.
24. We coded timing of elections using Keesings World Record Online Edition: 1960-1999. We checked the elections data with Journal of Democracy's Election Watch (1990-1999) and with International Institute for Democracy and Electoral Assistance (1997).
25. We coded federalism—on the basis of whether substate units had substantial decisionmaking power—using various editions of the Central Intelligence Agency's (1986-1998) World Factbooks, World Almanacs (1984-1997), and Europa World Yearbook (1996). We double-checked some of the less clear cases with regional experts.
116    COMPARATIVE POLITICAL STUDIES / February 2002
country's electoral system allocates legislative seats in proportion to each party's share of the vote. For our analyses of democracies, we code countries' electoral system in the following manner: representatives chosen by majority vote = 1, those chosen by a plurality system = 2, those selected using semiproportional representation systems = 3, and legislators chosen by true proportional representation = 4.26 We also recognize, however, that many patently nondemocratic governments hold elections that are competitive only on paper (or hold none at all). Clearly, in such systems, the type of election process or system of representation is irrelevant, because the process does not actually determine who governs. Hence we create a new variable to separate democratic PR systems from those employed by nondemocracies. To create proportional democracy, we multiply a dummy variable indicating whether or not a proportional representation system is used (1 if it is, 0 if not) by a second dummy variable that is scored 1 for countries that are at least minimally democratic, and 0 otherwise. (For our purposes here, we will measure as minimally democratic those regimes that receive a score of at least 1 on our regime type indicator.) The resulting variable isolates countries that have PR systems and are also at least minimally democratic. All countries without PR systems as well as all nondemocracies are scored 0. We develop a dummy variable, parliamentary, indicating whether a political system is parliamentary (1) or presidential (0). All mixed systems are coded as either presidential or parliamentary, based on whether each government's design favors one or the other.27 In general, president-parliamentary systems are coded as presidential and premier-presidential as parliamentary.28
Our third category of independent variables will be our control variables. We include two economic measures for each year and country in question: gross national product per capita and change in gross domestic product (GDP) per capita from the previous year. The first measure identifies the rela-
26. We used Reynolds and Riley (1997) for our electoral system data. They define systems as semiproportional, "which provide, on average, results which fall some way in between the proportionality of PR systems and the disproportionality of plurality-majority systems" (p. 148). Their focus on outcomes rather than rules might be disturbing except that nearly all of their cases coded as semi-PR are countries with parallel electoral systems. Parallel systems are those in which some seats are determined by PR and others by plurality-majority rules but in which the PR is not aimed at compensating for the plurality-majority rules' disproportionality (which distinguishes Germany's system from others).
27. The key question is whether "the process of forming the executive is institutionally distinct from the process of filling seats in the assembly" (Shugart & Carey, 1992). Sources included various editions of World Almanacs (1984-1997) and Europa World Yearbook (1996). Matthew Shugart's assistance was very helpful for our coding efforts.
28. We get very similar results if we use a different measure that codes presidential systems as 0, mixed as 1, and parliamentary as 2.
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT
117
tive wealth29 and poverty of different countries, whereas the second indicates whether conditions are improving or declining during the year in question.30
To account for differences among ethnic groups, we use three variables from the MAR dataset. The Economic Differentials Index is a "seven-category scale of intergroup differentials in economic status and positions." This variable ranges from a score of -2 for a group that is most advantaged to 4 where extreme disadvantages exist.31 The Political Differentials Index focuses on access to power and to civil service, recruitment, voting rights, the right to organize, and equal legal protection. This measure ranges in value from no negative differentials (0) to extreme negative differentials (4). The third variable is a Cultural Differentials Index, measuring how distinct groups are, focusing on differences in ethnicity/nationality, language, historical origin, religion, custom, and residence. This variable ranges from no differences (0) to extensive differences (4). We also employ group concentration, measuring the degree to which an ethnic group is geographically concentrated. This variable ranges from most dispersed (0) to least (3).
Finally, we need to deal with problems that are most likely in pooled cross-sectional time-series analysis—specifically autocorrelation and hetero-skedasticity. To deal with the former problem, we include a lag of the dependent variable.32 To address this problem, we perform Prais-Winsten regressions with panel corrected standard errors (Beck & Katz, 1995).
RESULTS
ALL GROUPS
We are now ready to turn to our models of protest and rebellion. We perform two sets of analyses, the first focusing on all groups in the data set and the second dividing between democratic states (i.e., those that receive a score
29. We use GDP per capita rather than GDP because it distinguishes between small countries that are quite wealthy and larger countries (e.g., Singapore vs. China). Likewise, we use change in GDP per capita rather than change in GDP because a country with rapidly growing economy and population is less likely to feel that it is growing economically than one with a growing economy and a relatively stable population.
30. The World Bank Data Development Group provided the economic data to us directly, although it is also available at www.worldbank.org/data.
31. For more information about the Minorities at Risk data set variables, see Gurr (1999, 2000).
32. Others have argued that the best predictor of current ethnic conflict is not last year's conflict but the history of unrest since World War II (Fearon & Laitin, 1999). Thus we substituted the maximum level of protest and rebellion from 1945 to 1984 for the lagged dependent variables and found similar results.
118    COMPARATIVE POLITICAL STUDIES / February 2002
of 6 or above on our regime type variable) and authoritarian states (those with scores below 0).
Our models of all states appear in Table 1. As the table indicates, excluding the lagged measures of our dependent variables, six of our independent variables achieve statistical significance for protest and nine for rebellion.
Regime Type: Duration and Change
As hypothesized, democracies suffer more from ethnically motivated dissent and violence than do authoritarian political systems. Groups are more likely to rebel in democratic systems, presumably because rebellions are easier to organize in more open societies and because repression, which is more prevalent in nondemocracies (Davenport, 1995), may actually succeed in preventing rebellious activity or raising its costs.
We also find that older governments incurred more violence, although not significantly more protest, than younger ones. This coincides with our finding concerning younger democracies, which experienced less violence than other kinds of regimes.33 This finding, particularly given the temporal focus of our study, challenges the notion that newer democracies are more likely to face ethnic violence (Snyder, 1999). The very existence of a new democracy, after all, suggests that fundamental grievances have recently been addressed, thus reducing the incentive for rebellion. By contrast, our analyses indicate that first elections do not have a significant influence on protests or violence.
Institutions: Proportionality and Federalism
Our results indicate that institutions do matter. In particular, democracies with proportional representation systems have much less of both types of ethnic conflict. Under PR, minority ethnic groups are more likely to have at least some representation in the legislature, and their members are thus more satisfied that their concerns are being heard within the existing political arrangements. Thus neither large-scale demonstrations nor violence is required for groups to have some say over their destinies.
Federalism also has an impact, although it influences peaceful and violent dissent differently. Under federalism, activists may find it easier to rally support, because it may appear more feasible to influence a regional government than decision makers at the national level. At the same time, federalism reduces the level of ethnic violence. In a federal structure, groups at the local
33. Regime duration remained statistically significant regardless of whether we included our indicators for young democracy and first elections.
Table 1
Pooled, Cross-Sectional Time-Series Analyses of Protest and Rebellion, 1985-1998, All Groups (Prais-Winsten Regressions With Panel Corrected Standard Errors)
Low Concentration High Concentration
Variable	Protest	Rebellion	Protest	Rebellion	Protest	Rebellion
Lag of dependent variable	.59**	.50**	.56**	.45**	.63**	.58**
Regime type	.02**	.03**	04**	.01	.02**	.02**
Enduring regime	.04	.12*	-.05	.04	.13*	.15*
Young democracy	.03	_ 29**	-.08	.07	.09	-.22*
First election	.12	-.02	.16	.08	.20	.03
Proportional democracy	-.16**	_ 29**	-.22*	_ 14**	-.17*	-.31*
Federal system	.14*	-.20*	.21	-.04	.06	-.29*
GDP per capita	.00001	-.00003**	-.00001	-.00001	.00001	-.0002**
Change in GDP per capita	-.0001	.0001	.0002	.00001	-.0001	.0002*
Cultural differences index	.08*	.10*	.04*	-.04	11**	.15**
Economic differences index	-.02	-.002	-.003	-.03	-.02	-.03
Political differences index	04**	.06**	.02	.01	.06*	.15**
Group concentration	.08**	.13**				
Constant	.11	-.001	45**	.29*	.18	.04
Rho	.10	.46	.13	.22	.07	.41
R-squared	.7562	.4473	.7545	.3343	.7728	.5506
n of observations	2,564	2,582	795	790	1,282	1,315
*/><.05. **/><.01.
120    COMPARATIVE POLITICAL STUDIES / February 2002
level can influence many of the issues that matter dearly to them—education, law enforcement, and the like. Moreover, federal arrangements reduce the chances that any group will realize its greatest nightmare: having its cultural, political, and educational institutions destroyed by a hostile national majority.34
ECONOMIC CONDITIONS AND GROUP CIRCUMSTANCES
We also account for the possibility that ethnic unrest was driven, at least in part, by the economic and group circumstances in each country. In these analyses, it is clear that a country's level of development matters more than how well it has been doing lately. Although the wealth of a host country (GDP per capita) does not influence the level of protest, groups in richer countries engage in less violence. Economic change, by contrast, did not show any similar impact.35
Group differences also matter: Groups that are culturally distinct or politically disadvantaged engage in more dissent, both peaceful and violent. Groups that are more culturally distinct engage in serious ethnic conflict because there are fewer crosscutting cleavages that might bind them to other groups or create other axes around which politics might revolve. Political differences matter because groups that cannot act through ordinary political channels or are limited in their ability to do so engage in larger protests and more violence.
Finally, group concentration had a positive relationship with both protest and rebellion.36 Given that the protest variable focuses on the size of protests, it is not at all surprising that heavily concentrated groups are also more likely to be involved in protest movements. Likewise, group concentration facilitates rebellion for two reasons. First, much ethnic violence is centered on separatist conflicts, which require a particular territory to be claimed (Fearon & Laitin 1999, Saideman & Ayres 2001). Second, groups that are dispersed
34. One might also argue that the combination of high protest and low rebellion in federal systems may be a function of the data set's construction (see Note 11). In other analyses, federalism was not statistically significant when we replaced the lag of 1 year with an indicator of the highest level of rebellion between 1945 and 1984. Federalism was also sensitive to the inclusion of the young democracy variable, losing significance when it was dropped.
35. Dropping GDP per capita from our analyses causes regime type to lose statistical significance and causes change in GDP per capita to have a significant impact on rebellion.
36. A history of violence between two groups may cause both groups to concentrate and to engage in further conflict. We tested this by substituting an indicator of the maximum level of postwar rebellion for our lag variable and found similar results.
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT
121
may be inhibited from engaging in violence because they are more vulnerable if counterattacked.
So far, we have treated group concentration as an independent factor influencing ethnic unrest. It is also possible, however, that group concentration may interact with other variables. Federal systems, for example, may be more effective in reducing unrest in areas where minority groups inhabit specific regions of a country (thus allowing for federal boundaries to correspond with ethnic populations). In addition, other factors may have a greater impact on protest and violence when the concentration of ethnic groups makes it easier to organize actions against the state. Thus Table 1 also presents the results of our model after separating between areas with high and low group concentration.37
As expected, the impact of federalism on rebellion is significant only in cases in which groups tend to live together in the same region of the country. Furthermore, our measures for new democracies, economic development, and economic growth significantly influence violence only among the more highly concentrated groups, further indicating the importance of concentration. Proportional representation systems, on the other hand, are effective at reducing ethnic tensions regardless of group concentration levels, presumably because the beneficial effects of PR accrue to minority groups regardless of their geographical distribution.
DEMOCRATIC REGIMES
Up to this point, we have looked at group/years for all of the countries in our sample, including both democratic and nondemocratic regimes. Because some data simply cannot be coded for authoritarian regimes without stretching the important concepts, we need to focus specifically on democracies to determine the impact of particular institutions. Selecting only for such democratic countries (those with a regime type score of 6 or higher) leaves us with a sample size (i.e., total number of group/years) of just below 1,350. Although we are primarily concerned with democracies, we also include our sample of autocratic regimes for purposes of comparison. The sample size for this group is 1,838.
For the analysis of democracies, we specify the same variables employed above, with three critical exceptions. First, we use our ordinal measure of electoral system instead of our dummy variable for proportional democracy. This allows us to compare across different types of democratic electoral insti-
37. We distinguish between highly dispersed groups (group concentration of 0 or 1) and highly concentrated ones (group concentration of 3). Analysts are working to improve the concentration indicator. In future work, we will use these new data.
122    COMPARATIVE POLITICAL STUDIES / February 2002
tutions in terms of their relative proportionality (i.e., among majority, plurality, quasi-PR, and fully PR systems). Second, focusing only on democracies permits us to compare the impact of parliamentary and presidential systems. Finally, we drop our indicator of young democracy from these analyses, because all states included here are, by definition, democracies.
Table 2 presents the results for our analysis of ethnic conflict in all countries that meet our definition of democratic. Interestingly, the impact of regime duration on rebellion, but not protest, remains positive and significant. This analysis, supporting what we found earlier with our young democracies variable, suggests that older democracies are more likely to suffer from ethnic violence. We also find that first elections in democratic systems are not as problematic as feared. Thus the results presented in both tables challenge the current pessimism about democratization (Saideman, 1998; Snyder, 1999). Future work must consider why older democracies face more ethnic conflict. Gurr's (2000) recent findings that the condition of minorities in new democracies have improved throughout the 1990s suggest that access might shape outcomes more than uncertainty.38
Once again, the analyses demonstrate that institutional choice does matter. Ethnic strife decreases as the proportionality of a country's electoral system increases. On the other hand, the analyses in Table 2 indicate that the type of government is less important than much of the literature suggests. The dummy variable for parliamentary systems does not reach statistical significance. These results suggest that arguments about the inherent superiority of parliamentary systems and the flawed nature of presidential systems may be exaggerated (Lijphart, 1991; Linz, 1990).
Federalism, which appeared to exacerbate protest behavior and ameliorate rebellion in Table 1, has no significant effect for our sample of democracies. On the other hand, our analysis demonstrates that federal arrangements in autocracies have a clear impact on ethnic protest and rebellion. Why does federalism seem to matter more (increasing protest, decreasing violence) in nondemocracies? It is possible that autocracies, lacking free and open elections, need other institutional arrangements to address the grievances of ethnic groups. Federal structures may provide such an arrangement. The higher incidence of protest indicates that federal systems may be succeeding in creating a localized "escape valve" for ethnic pressures.
Our economic control variables produce somewhat different results from those in Table 1. In democracies, GDP per capita decreased rebellion but did not produce a significant result for protest. In authoritarian states, however,
38. One reviewer suggested that this result implies something about older democracies: that citizens in older democracies have a larger set of skills and network tools to engage in dissent.
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT
123
Table 2
Pooled, Cross-Sectional Time-Series Analyses of Protest and Rebellion, 1985-1998, Regime Type-Specific Analyses (Prais-Winsten Regressions With Panel-Corrected Standard Errors)
Democracies Autocracies
Variable	Protest	Rebellion	Protest	Rebellion
Lag of dependent variable	.53**	40**	49**	.80**
Enduring regime	.09	34**	-.06	.01
First election	.001	-.02		
Electoral system	_ 07**	-.20**		
Parliamentary	.06	.16		
Federal system	.09	-.16	.55**	-.25*
GDP per capita	.00001	-.00003**	.00003*	-.00003**
Change in GDP per capita	.0001	-.00002**	-.0001	.004
Cultural differences index	.15**	.04	.01	.10*
Economic differences index	_ 07**	-.07	-.01	.04*
Political differences index	.08*	.08	.04	.02
Group concentration	.09*	.21**	.10**	.03
Constant	.38	.58	.12	-.17
Rho	.11	.61	.32	-8.60
R-squared	.7957	.3800	.5432	.7895
n of observations	1,346	1,348	1,000	838
*p<.05. **/><.01.
higher levels of GDP were associated with less rebellion but more protest. Furthermore, economic change (change in GDP per capita) produced no statistically significant results for nondemocracies but did have a negative relationship with rebellion in democracies: Rebellion in democratic states increased as per capita wealth declined.
Group differences also produced some interesting results. Under democratic regimes, cultural, political, and economic differences have a significant relationship with protest but not with rebellion (although cultural and economic differences are associated with political violence in authoritarian states). Given that protest is less costly than rebellion and that protest is easier and probably more efficacious in democracies, these findings should not be surprising. What is noticeable, however, is that the existence of extreme economic disadvantages actually decreases the level of protest. Apparently, economically disadvantaged groups lack the resources (e.g., time, organization, and leadership) to engage in massive protest activity. Groups that are politically and culturally marginalized, on the other hand, are more willing and able to express their dissent.
Finally, group concentration plays a role in exacerbating ethnic conflict in democratic regimes, just as it does in our sample of all countries.
124    COMPARATIVE POLITICAL STUDIES / February 2002
IMPLICATIONS FOR THEORY AND POLICY
Our findings present perhaps as many puzzles for future research as they do answers. We find that young regimes, particularly democracies, may not have as severe problems with ethnic conflict as expected. This means that decision makers should not worry so much about increased ethnic conflict if they support democratization at home or abroad. On the other hand, although the process of democratizing does not promote ethnic strife, it is clear that both protest and rebellion are greater under democratic regimes than they are under authoritarian systems.
In theoretical terms, we need to think more clearly about the causal processes at work. To what extent do ethnic demands produce political change, and to what degree can institutional change head off ethnic unrest? More important, what is it about long-lived regimes—even democratic ones—that causes ethnic warfare? This is a puzzle that future work must explore.
We also find that the electoral system seems to be more important than whether the type of government is presidential or parliamentary. Although we found evidence supporting proportional representation as an important institution for inhibiting or resolving ethnic conflict, the results concerning presidentialism and parliamentarism are less clear. The positive results, although they fall short of significance, suggest that parliamentary systems are not clearly superior to presidential ones for managing ethnic strife. More work, theoretical and empirical, is needed to determine whether scholars should recommend to policy makers presidentialism, parliamentarism, or a combination (or whether it matters at all). Furthermore, our findings about electoral systems, although interesting and important, are quite general. We did not pay attention to different laws that provide proportional results, nor did we pay attention to thresholds. There may be significant variance among countries with PR systems that this study does not capture.
Federalism, apparently, has gotten a bad rap. It is not significantly correlated with more violence, but less, at least in less democratic systems. Future work should address variations in federal structures as well as the role of groups that are marginalized within the federal subunits. In any case, our results suggest that policy makers should not avoid federalism, because it is not as harmful as some have argued.
Our control variables also suggest a few policy recommendations. Economic development reduces ethnic conflict, but short-term changes are not so influential. Thus elites should not focus on quick fixes but on strategies that
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT 125
improve the economy over the long run. We also found that political differences exacerbate ethnic conflict, so if the reduction of conflict is the goal, then leaders should not discriminate against minorities politically. Our finding concerning the negative relationship between economic differences and protest clearly requires more investigation.
Despite the finding that extreme cultural differences are related to more conflict, our study indicates that institutions matter. Because regime type and duration, electoral systems, and federalism all influence the severity of ethnic protest and violence, institutional design has important implications for the stability of most political systems. The problem, of course, is that politicians making choices about institutional design worry about not only what is best for their country and what is best for each ethnic group's security but also what is best for themselves. The real challenge ahead is to encourage politicians to adopt constitutional changes that might lead to more ethnic peace even if such modifications threaten incumbents' positions.
APPENDIX Descriptive Statistics
Standard
Variables	n	Minimum	Maximum	Mean	Deviation
Rebellion	3,590	0	7	1.01	1.96
Protest	3,573	0	5	1.43	1.35
Regime type	3,521	-10	10	.68	7.09
Enduring regime	3,696	0	1	.43	.50
Young democracy	3,521	0	1	.18	.38
First election	3,696	0	1	.048	.21
Proportional democracy	3,559	0	1	.21	.41
Parliamentary	2,262	0	1	.46	.50
Electoral system	2,719	1	4	2.78	1.09
Federal system	3,696	0	1	.23	.42
GDP per capita	3,080	290	29,420	4,997.78	5,596.26
Change in GDP per capita	2,989	-.99%	57.2%	.599	3.04
Economic differences index	3,458	-2	4	1.76	2.02
Political differences index	3,486	-2	4	1.59	1.59
Cultural differences index	3,696	1	4	2.72	1.14
Group concentration	3,696	0	3	2.08	1.10
126 COMPARATIVE POLITICAL STUDIES / February 2002 Frequencies of dependent variables
Protest	Frequency	Percentage	Rebellion Frequency		Percentage
None	1,360	38.06	None	2,633	73.34
Verbal opposition	386	10.80	Political banditry	182	5.07
Symbolic acts of	1,074	30.06	Campaigns of	119	3.31
resistance			terrorism		
Small demonstration	522	14.61	Local rebellion	111	3.09
(less than					
10,000 people)					
Medium demonstration	151	4.23	Small-scale	183	5.10
(between 10,000			guerilla activity		
and 100,000)					
Large demonstration	80	2.24	Intermediate-scale	157	4.37
			guerilla activity		
Total	3,573	100.00	Large-scale guerilla	85	2.37
			activity		
			Protracted civil war	120	3.34
			Total	3,590	100.00
REFERENCES
Anderson, Christopher J., & Guillory, Christine A. (1997). Political institutions and satisfaction with democracy: A cross-national analysis of consensus and majoritarian systems. American Political Science Review, 91, 66-81.
Ayres, R. William, & Saideman, Stephen M. (2000). Is separatism as contagious as the common cold or as cancer? Testing the international and domestic determinants of secessionism. Nationalism and Ethnic Politics, 6(3), 92-114.
Barnes, Samuel H., & Kaase, Max. (Eds.). (1979). Political action: Mass participation in five Western democracies. Beverly Hills, CA: Sage.
Beck, Nathaniel, & Katz, Jonathan N. (1995). What to do (and not to do) with time-series-cross-section data in comparative politics. American Political Science Review, 59(3), 634-647.
Bohrer II, Robert E. (1997). Deviations from proportionality and survival in new parliamentary democracies. Electoral Studies, 16(2), 217-226.
Byman, Daniel L. (1997). Rethinking partition: lessons from Iraq and Lebanon. Security Studies, 7(1), 1-32.
Central Intelligence Agency. (1986-1998). Worldfactbook. Washington, DC: Author. Retrieved from http://www.cia.gov/pubs/
Cohen, Frank. (1997). Proportional versus majoritarian ethnic conflict management in democracies. Comparative Political Studies, 30(5), 607-630.
Cox, Gary W. (1997). Making votes count: Strategic coordination in the world's electoral systems. Cambridge, MA: Cambridge University Press.
Crepaz, Marcus M. L. (1996). Consensus versus majoritarian democracy: Political institutions and their impact on macroeconomic performance and industrial disputes. Comparative Political Studies, 29(1), 4-26.
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT
127
Davenport, Christian. (1995). Multi-dimensional threat perception and state repression: An inquiry into why states apply negative sanctions. American Journal of Political Science. 39(3), 683-713.
Davenport, Christian. (1997). From ballots to bullets: An assessment of how national elections influence state uses of political repression. Electoral Studies, 16(A), 517-540.
Davenport, Christian. (Ed.). (2000). Paths to state repression: Human rights violations and contentious politics. Lanham, MD: Rowman & Littlefield.
DeFigueiredo Jr., Rui J. P., & Weingast, Barry R. (1999). The rationality of fear: political opportunism and ethnic conflict. In Barbara Walter & Jack Snyder (Eds.), Civil wars, insecurity, and intervention (pp. 261-302). New York: Columbia University Press.
Diamond, Larry. (1996). Is the third wave over? Journal of Democracy, 7(3), 20-37.
Election Watch. (1990-1999). Journal of Democracy, 1-10, 1-4.
Europa World Year Book. (1996). London: Europa.
Fearon, James D., & Laitin, David D. (1999, September). Weak states, rough terrain, and large-scale ethnic violence since 1945. Paper presented at the 1999 Annual Meeting of the American Political Science Association, Atlanta, GA.
Finkel, Steven E., & Muller, Edward N. (1998). Rational choice and the dynamics of collective political action: Evaluating alternative models with panel data. American Political Science Review, 92(1), 37-49.
Green, Donald P., & Shapiro, Ian. (1994). Pathologies of'rational choice•theory. New Haven, CT: Yale University Press.
Gurr, Ted Robert. (1970). Why men rebel. Princeton, NJ: Princeton University Press.
Gurr, Ted Robert. (1993). Minorities at risk: A global view of ethnopolitical conflicts. Washington, DC: U.S. Institute of Peace.
Gurr, Ted Robert. (1999). Minorities at risk dataset. College Park, MD: Center for International Development and Conflict Management. Retrieved from http://www.bsos.umd.edu/cidcm/ mar/
Gurr, Ted Robert. (2000). Peoples versus states: Ethnopolitical conflict and accommodation at the end of the 20th century. Washington, DC: U.S. Institute of Peace.
Gurr, Ted Robert, & Jaggers, Keith. (1999). Polity98: Regime characteristics 1800-1998. College Park, MD: Center for International Development and Conflict Management. Retrieved from http://www.bsos.umd.edu/cidcm/polity/
Horowitz, Donald L. (1985). Ethnic groups in conflict. Berkeley: University of California Press.
Horowitz, Donald L. (1993). Democracy in divided societies. Journal of Democracy, 4(4), 18-38.
International Institute for Democracy and Electoral Assistance. (1997). Voter turnout from 1945 to 1997: A global report on political participation. Stockholm: Author.
Jervis, Robert. (1978). Cooperation under the security dilemma. World Politics, 30(1), 167-214.
Keesings World Record Online Edition: 1960-1999 (Vols. 31-44). Retrieved from http:// www.keesings.com
Lardeyret, Guy. (1991). The problem with PR. Journal of Democracy, 2(3), 30-35. Lichbach, Mark 1.(1994). Rethinking rationality and rebellion: Theories of collective action and
problems of collective dissent. Rationality and Society, 6(1), 8-39. Lijphart, Arend. (1977). Democracy in plural societies. New Haven, CT: Yale University Press. Lijphart, Arend. (1991). Constitutional choices for new democracies. Journal of Democracy,
2(1), 72-84.
Lijphart, Arend. (1994). Electoral systems and party systems: A study of twenty-seven democracies, 1945-1990. Oxford, UK: Oxford University Press. Linz, Juan J.(1990). The perils of presidentialism. Journal of Democracy, 1(1), 51-60.
128    COMPARATIVE POLITICAL STUDIES / February 2002
Lipset, Seymour Martin. (1960). Political man: The social bases of politics. Garden City, NY: Doubleday
Lipset, Seymour Martin. (1991). The centrality of political culture. Journal of Democracy, 1(4), 80-83.
Moore, Will H. (2000). The repression of dissent: a substitution model of government coercion. Journal of Conflict Resolution, 44(1), 107-127.
Moser, Robert G. (1999). Electoral systems and the number of parties in postcommunist states. World Politics, 51(3), 359-384.
Muller, Edward N, & Opp, Karl-Dieter. (1986). Rational choice and rebellious action. American Political Science Review, 80(2), 471-487.
Posen, Barry R. (1993). The security dilemma and ethnic conflict. Survival, 35, 27-47.
Power, Timothy J., & Gasiorowksi, Mark J. (1997). Institutional design and democratic consolidation in the third world. Comparative Political Studies, 30(2), 123-155.
Rabushka, Alvin, & Shepsle, Kenneth A. (1972). Politicians in plural societies: A theory of democratic instability. Columbus, OH: Charles E. Merrill.
Regan, Patrick. (2000). Civil wars and foreign powers: Outside intervention in intrastate conflict. Ann Arbor: University of Michigan Press.
Reynolds, Andrew. (1995). Constitutional engineering in southern Africa. Journal of Democracy, 6(2), 86-99.
Reynolds, Andrew, & Riley, Ben (Eds.). (1997). The international handbook of electoral systems design. Stockholm: International Institute for Democracy and Electoral Assistance.
Roeder, Phillip G. (1991). Soviet federalism and ethnic mobilization. World Politics, 43(2), 196-232.
Rothschild, Joseph. (1981). Ethnopolitics: A conceptual framework. New York: Columbia University Press.
Rummel, R. J. (1994). Death by government. New Brunswick, NJ: Transaction.
Saideman, Stephen M. (1998). Is Pandora's box half-empty or half-full? In David A. Lake & Donald Rothchild (Eds.), Ethnic conflict: Fear, diffusion, escalation (pp. 127-150). Princeton, NJ: Princeton University Press.
Saideman, Stephen M. (2001). The ties that divide: Ethnic politics, foreign policy, and international conflict. New York: Columbia University Press.
Saideman, Stephen M., & Ayres, R. William. (2001). Determining the sources of irredentism: Logit analyses of Minorities at Risk data. Journal of Politics, 63(4), 1126-1144.
Shugart, Matthew Soberg, & Carey, John M. (1992). Presidents and assemblies: Constitutional design and electoral dynamics. Cambridge, UK: Cambridge University Press.
Shugart, Matthew Soberg, & Mainwaring, Scott. (1997). Presidentialism and democracy in Latin America: Rethinking the terms of the debate. In Scott Mainwaring & Matthew Soberg Shugart (Eds.), Presidentialism and democracy in Latin America (pp. 12-54). Cambridge, UK: Cambridge University Press.
Shugart, Matthew S., & Taagapera, Rein. (1994). Plurality versus majority election of presidents: A proposal for a double complement rule. Comparative Political Studies, 27,323-348.
Snyder, Jack. (1999). When voting leads to violence: Democratization and nationalist conflict. New York: Norton.
Snyder, Jack, & Jervis, Robert. (1999). Civil war and the security dilemma. In Barbara Walter & Jack Snyder (Eds.), Civil wars, insecurity, and intervention (pp. 15-37). New York: Columbia University Press.
Stepan, Alfred, & Skach, Cindy. (1993). Constitutional frameworks and democratic consolidation: Parliamentarism versus presidentialism. World Politics, 46(1), 1-23. Tsebelis, George. (2000). Veto players and institutional analysis. Governance, 13(4), 441-474.
Saideman et al. / DEMOCRATIZATION, ETHNIC CONFLICT
129
Turner, Arthur W. (1993). Postauthoritarian elections: Testing expectations about first elections.
Comparative Political Studies, 26(3), 330-349. Waltz, Kenneth N. (1979). Theory of international politics. New York: Random House. Ward, Michael D., &Gleditsch, KristianS. (1998). Democratizing for peace. American Political
Science Review, 92(1), 51-61. World Almanac 1985-1998. (1984-1997). Mahwah, NJ: Funk & Wagnalls.