Institutions,contexts,and ethnic violence in comparative perspective

Abstract

Which combinations of government structures and electoral systems create better frameworks for addressing ethnic violence? Is there any one-size-fits-all institutional solution to violent ethnic conflict? Why or why not? These questions are of substantial importance to scholars and policymakers alike, but the extant literature does not provide a systematic and thorough exploration. In this article, we argue that the effects of political institutions on ethnic violence are moderated by parameters of ethnic configurations. Through a large comparative study, we find that institutions are relevant when ethnic groups are not geographically dispersed, and whether ethnic minorities face a majority group also matters. For concentrated minorities facing a majority, semi-presidential-proportional and presidential-proportional systems are more effective in reducing violence. In cases involving concentrated minorities facing no majority, parliamentary-non-proportional systems are associated with the most intense violence. We conclude that states seeking to alleviate ethnic violence by institutional engineering must take contexts seriously.

Keywords

Government structures electoral systems ethnic violence ethnic configuration multilevel analysis

Which ‘combinations’ of government structures (semi-presidentialism vs. presidentialism vs. parliamentarism) and electoral systems (proportional representation vs. non-proportional systems) create better frameworks for addressing ethnic violence?¹ Are they parliamentary-proportional representation (PR) systems, as many scholars and politicians worldwide have argued?² Why or why not? These questions are of substantial importance because ethnic divisions have been so widespread and have become the world’s most serious source of violence since the 1990s. However, there are not many large comparative studies on this subject that provide satisfactory analysis, and the theories and empirical results of these studies are often incomplete and biased.

Using the Minorities at Risk (MAR) dataset from 1985 to 2006, we revisit these questions by analyzing all ethnic groups at risk in democratic states. While extant studies recognized the important interactive effects between government structures and electoral systems, none of them tested these effects in their models. This is probably because multiplicative models suffer from collinearity problems and are difficult to interpret (Braumoeller, 2004; Kam and Franzese, 2007). Furthermore, past studies did not take seriously whether ethnic minorities are geographically concentrated and whether they face a majority group. Through a large comparative study, we found that combinations of government structures and electoral systems are relevant only when ethnic groups are geographically concentrated. We also confirmed that whether concentrated minorities face a majority group affects whether institutional combinations facilitate checks and balances or the formation of coalition or broad coalition cabinets, which in turn influence ethnic violence.³ Specifically, in cases involving concentrated minorities facing a majority group, semi-presidential-proportional and presidential-proportional systems are more effective in reducing violence because these institutional combinations facilitate checks and balances in this ethnic configuration. In cases involving concentrated minorities facing no majority, parliamentary-non-proportional systems are associated with the most intense violence because these institutional combinations facilitate neither the formation of coalition cabinets nor checks and balances.

The ethnic security dilemma

The ethnic security dilemma refers to competition for control of the government among ethnic groups. Saideman et al. (2002) first applied the concept of ‘security dilemma’ in international relations to ethnic conflict. They argued that the competition for control of the government between ethnic groups could cause a security dilemma. Ethnic groups seek to control the government because they believe that the government is the greatest potential threat to them. Groups fear that should another group control the government, it will use state resources to act against them and even endanger their survival. This fear is mutually reinforcing: one group’s attempt to control the government will evoke others’ fear and counter-actions, and eventually every group is worse off because of the irrational competition. This ethnic security dilemma thesis thus suggests that ‘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’ (Saideman et al., 2002: 107). By ethnic groups, we mean groups that belong to a certain ascriptive category, such as language, race, religion, tribe, and so on.

The implication of the ethnic security dilemma is supported by the global evidence of ethno-political violence, which shows that ethnic conflicts can usually be lessened by some combination of policies and institutions of power sharing (Cederman et al., 2010; Gurr, 1993). In theory, discriminated or mobilized ethnic groups are those who would challenge and fight against the state if these groups’ interests were not appropriately protected in their struggle for state power (Cederman et al., 2010; Elkins and Sides, 2007; Gurr, 1993: 3). Most of these groups at risk are ethnic minorities, whose demographic size is less than 50% of the national population, and only a few of them are majorities (Gurr, 1993).⁴ It should be noted that some ethnic minorities are empowered and advantaged. These groups are likely to lose power, though, thanks to their minority status. When this change happens, these advantaged minority groups are at risk of being retaliated against and having restrictions imposed on their rights (Gurr, 1993). Because of this, in addition to discriminated and mobilized groups, advantaged minorities are also included in our analysis. This article focuses on investigating how institutions influence the intensity of ethnic violence among all these groups at risk. We hereafter use the terms ‘ethnic groups’ and ‘ethnic minorities’ interchangeably to denote these groups at risk. We also use the term ‘majority groups’ to denote advantaged majority groups.

The extant literature generally argues that parliamentary and PR systems better mitigate ethnic conflict because parliamentarism better facilitates the formation of coalition or broad coalition cabinets than presidentialism and semi-presidentialism (Lijphart, 2002, 2004, 2012; Linz, 1994), and PR systems tend to give ethnic minorities more representation and influence in policymaking than non-proportional single-member district (SMD) systems (Brancati, 2006; Cohen, 1997; Ishiyama, 2000; Lijphart, 2002; Saideman et al., 2002). These opinions, however, are by no means universal. Cheibub (2007) found that while the rate of coalition formation is lower in presidential systems than in parliamentary ones, the difference between the two is not substantial (40% vs. 43% of all country-years), and the rate in mixed democracies is the highest. Many scholars pointed out that the separation of powers and some power-sharing arrangements under presidential systems can mitigate majoritarianism (Birnir, 2007; Brancati, 2006; Ishiyama, 2000; Saideman et al., 2002).⁵ Moreover, parliamentary systems in conjunction with SMD usually lead to one-party majority cabinets (Gunther and Mughan, 1993), as opposed to coalition cabinets. This fact suggests once again that the effects of government structures cannot be correctly evaluated without considering their interactions with electoral systems. Furthermore, some scholars advocated preferential voting systems, as opposed to PR, for the incentives to accommodative behavior the former provide (Horowitz, 1991), but only a few countries adopt these systems, making their effects hard to evaluate. Many studies thus focused on comparing PR and SMD systems and confirmed that, generally, PR outperforms SMD in mitigating ethnic violence.

Further complicating is the notion that institutional effects are ultimately contextual (Lublin, 2014; Moser, 2008; Reilly and Reynolds, 1999).⁶ As many studies have found, spatially concentrated groups are more likely to engage in ethnic violence than spatially dispersed groups. (Byman, 2002; Cederman et al., 2010: 97; Cohen, 1997; Gurr, 1993; Saideman and Ayres, 2000; Saideman et al., 2002; Toft, 2003). Political institutions should matter for concentrated groups only. Furthermore, whether there is a majority group in a state influences whether institutionally induced coalition or broad coalition cabinets and checks and balances are likely to happen. These parameters of ethnic configurations moderate the effects of political institutions and must also be taken into account.

The effects of institutions on ethnic violence

This section will first discuss the moderating effect of ethnic group concentration, followed by a discussion of the effects of institutions in two theoretically distinct contexts: concentrated ethnic minorities facing a majority group and concentrated ethnic minorities facing no majority group.

The moderating effect of group concentration

The strength of institutional effects is not homogenous for ethnic groups. Indeed, how ethnic groups spread out spatially determines the difference. Past studies found that group concentration had a positive relationship with ethnic violence (Byman, 2002; Cederman et al., 2010: 97; Cohen, 1997; Gurr, 1993; Saideman and Ayres, 2000; Saideman et al., 2002; Toft, 2003). Group concentration leads to violent conflict for two reasons. First, concentration provides a natural boundary to ethnic minorities’ integration, and obstructs their political socialization process. This reinforces ethnic minorities’ discontent, and makes them less likely to perceive the political system as legitimate (Cohen, 1997). Second, if motivated, concentration makes groups more capable of mobilizing and waging a successful fight against states, and makes them less vulnerable if counterattacked (Saideman et al., 2002; Toft, 2003).⁷ Institutions which provide more power sharing are hence needed to reduce concentrated groups’ senses of insecurity and propensities for conflict. Which combinations of government structures and electoral systems do a better job towards reaching this goal depends on whether there exists a majority group.

Concentrated ethnic minorities facing a majority group

The most threatening rival to ethnic minorities for control of the government in a state is an ethnic majority. In states with an ethnic majority, parliamentarism generally will not induce the formation of coalition or broad coalition cabinets, so this government structure is less desirable. The way to achieve more peaceful ethnic relations is to ensure checks and balances by adopting non-parliamentary and proportional electoral systems. These notions will be elaborated upon in turn.

No institutionally induced coalition cabinets

If parliamentarism cannot lead to coalition or broad coalition cabinets, the ameliorative effects of this government structure on ethnic violence are very likely to vanish. In states with an ethnic majority group, it is quite likely that the group will command a majority of parliamentary seats, and form a single-party cabinet.⁸ Occasionally, coalition cabinets may be formed thanks to a majority group’s willingness to share power with minority groups and/or to its lack of a majority of parliamentary seats. However, a majority group can easily change its mind or may still be able to form a minority government alone when no other viable alternatives exist. Even if minority groups win parliamentary elections, either alone (when there are two main ethnic groups with similar population sizes, or when party competition is not solely centered on ethnicity, minority groups would have a chance to win elections) or by forming electoral alliances with other minority groups (when, besides a majority group, there coexist some smaller minority groups), broad coalition cabinets are still not very likely to be formed. This kind of power concentration in minority groups, just like power concentration in a majority group, would also lead to ethnic violence. In sum, when there exists a majority group, parliamentarism generally cannot guarantee the formation and survival of coalition cabinets, especially broad coalition cabinets, as advocated by Lijphart (2012). As to the effects of electoral systems, since coalition or broad coalition cabinets are generally rare in this ethnic and institutional configuration, the electoral system used will not significantly improve the effects of parliamentarism.

Checks and balances

When ethnic minorities face a majority group, presidential and semi-presidential systems serve as superior alternatives. In these systems, ethnic groups have two election opportunities; one for a president and the other for a legislature.⁹ As long as ethnic minorities win either of these elections, either alone or by forming electoral alliances, there are two power centers. Ethnic groups can check and balance their ethnic rivals and have weaker incentives to engage in ethnic violence (Birnir, 2007; Brancati, 2006; Ishiyama, 2000; Saideman et al., 2002). Taiwan’s experience, for example, shows how important it is for minority groups to have two chances of election. The Kuomintang (KMT) and other pan-blue parties, whose pro-China supporters are concentrated in northern Taiwan, have always won a majority of seats in Taiwan’s Legislative Yuan.¹⁰ If Taiwan had not changed from an indirect to a direct presidential election and from parliamentarism to semi-presidentialism in the late 1990s, the Democratic Progressive Party (DPP), whose pro-Taiwan supporters are concentrated in southern Taiwan, would not have gained access to executive power by winning the presidential elections in 2000 and 2004. Without such institutional changes, the DPP would very likely have been locked into its status as the opposition party for a much longer period of time. Such prolonged exclusion from power would be a recipe for ethnic violence.

There is one caveat, however. When different groups control legislative and executive power, the mitigating effect of semi-presidentialism is less likely to apply to the president-parliamentary type of semi-presidentialism. President-parliamentarism during divided governments would not mitigate ethnic violence because both a president and a legislature can unilaterally dismiss a government in these systems, and hence lack any incentive to cooperate and compromise in order to maximize influence over the government (Elgie, 2011).¹¹ Minority groups would feel insecure because of the inter-branch endless and mutually destructive fight against each other.

When ethnic majorities win both elections, both sub-types of semi-presidentialism, i.e. premier-presidentialism and president-parliamentarism, are equally desirable as presidential systems. In these periods of unified governments, legislatures either are very unlikely to be dissolved (in both sub-types of semi-presidential systems), or have fixed electoral mandates (in presidential systems). Individual legislators of a majority party have checks and balances over their government, and can choose which of the president’s policies they will support (Brancati, 2006; Saideman et al., 2002), just as many American examples have shown, and many of the legislators of the KMT have done during the presidency of a former president, Ma Ying-jeou. Likewise, the ‘moral hazard’ inherent in presidential and semi-presidential systems gives presidents considerable leeway to reach out to different ethnic groups through their policies in pursuit of reelection (Brancati, 2006; Samuels and Shugart, 2010). It can be expected that minorities will feel more secure in presidential and semi-presidential systems when they face a majority group. These systems ensure checks and balances, and unfavorable actions against ethnic minorities are more likely to be blocked by either the executive or the legislature (Birnir, 2007; Saideman et al., 2002).

The mitigating effect of presidential and semi-presidential systems on ethnic violence can be strengthened by combining them with proportional electoral systems. Many large-N studies have found that for demographically concentrated and dispersed groups alike, or for all the ethnic groups, on average, PR outperforms SMD in reducing ethnic violence (Brancati, 2006; Cohen, 1997; Ishiyama, 2000; Saideman et al., 2002). This effect should be related to the fact that PR systems have a higher proportionality of votes to seats, and, therefore, are generally regarded as fairer, although a few case studies have challenged this longstanding viewpoint (Moser, 2008). When an ethnic group has a fairer share of seats, it is more likely that it will regard the system as legitimate. The higher proportionality of votes to seats in this system also provides minorities with reasonable, if not better, representation to check and balance a majority group in presidential and semi-presidential systems. In this article, we group single nontransferable vote (SNTV) and highly proportional mixed-member proportional (MMP) and mixed-member majoritarian (MMM) systems with pure PR systems and label this group as ‘proportional’ systems. These systems are more proportional than all the other non-proportional electoral systems and should have positive effects on mitigating ethnic conflict.

Concentrated ethnic minorities facing no majority groups

When ethnic minorities face no majority groups in a state, all government structures and electoral systems have their own advantages when combined into certain institutional configurations.

Coalition cabinets

When ethnic minorities face no majority groups in a state, broad coalition cabinets may still be rare, but multi-ethnic coalition cabinets become more possible in parliamentary-proportional systems than in parliamentary-non-proportional systems. Proportional systems increase the possibility that more ethnic groups will gain representation in an assembly and more parties will be needed for the formation and survival of a majority-coalition cabinet. More groups can influence policymaking by joining cabinets, and feel more secure. In contrast, if non-proportional systems are employed, then parliamentarism is not preferable. Parliamentary-non-proportional systems tend to favor only a few groups who join coalition cabinets at the expense of others who do not. This tendency exists because, under non-proportional systems, fewer groups will have legislative representatives, and even one group could seize a majority of legislative seats. Consequently, fewer or even one party will be needed to form a majority-coalition cabinet. In either case, however, groups outside cabinets are less likely to influence policymaking.

India provides a real-world example about the effects of parliamentary-non-proportional systems. In India, there are six geographically concentrated minority groups that are politically active. Under the non-permissive SMD system, not only have the electoral outcomes been disproportional, but also the average time since the groups have had access to government has been much longer than other ethnic minorities in other countries (Birnir, 2007). These facts explain the violent ethnic relations in India to a certain extent. Conversely, another parliamentary democracy, South Africa, has had more peaceful ethnic relations since its transition to democracy. The spatially concentrated groups in South Africa also wait for a shorter time on average than their counterparts in India to have access to government (Birnir, 2007). From the perspective of this article and many other studies (Reynolds, 1995), the fact that South Africa and India have adopted PR and SMD systems, respectively, explains this contrast.¹²

Checks and balances

The inherent characteristic of checks and balances of presidentialism and premier-presidentialism also makes these government structures valuable for ethnic minorities facing no majority groups in a state. President-parliamentarism is problematic in this ethnic configuration, though, because divided governments are more likely to be the norm in states without majority groups, and because president-parliamentarism during divided governments is detrimental to the mitigation of ethnic conflict.

Presidentialism and premier-presidentialism would induce checks and balances regardless of which electoral systems are used. When there are no majority groups in a state, a single party is less likely to command a majority of legislative seats. A legislature would become even more fragmented and weaker if proportional systems are used in this ethnic configuration. However, past studies have found that high legislative fragmentation has a compensatory effect in presidentialism and semi-presidentialism: it will exert stronger pressure on parties and presidents to form coalition governments in order to govern (Cheibub, 2007).¹³ Conversely, when non-proportional systems are employed, although the systems may hurt some groups’ representation and access to legislative power, they will ensure stronger checks and balances between a less fragmented legislature and a president. Under either condition, there would be meaningful checks and balances placed on a president by a legislature. This result is surely better than the one produced under parliamentary-non-proportional systems, which are less likely to provide ethnic minorities with both coalition cabinets and checks and balances.

Propositions

We should address other factors which affect ethnic violence, and include these in our analysis to better test our propositions. Since Saideman et al. (2002) conducted a relatively comprehensive controlled comparison on this subject, we refer to this article to determine other relevant factors. For example, federal systems may mitigate ethnic violence since these systems allow ethnic groups to have more control over their own affairs. Older regimes should be less likely to experience ethnic strife. The longer regimes survive, the more likely it is that these regimes have successfully passed the challenge of transition periods. Regimes that have already held their first meaningful elections may also be associated with less severe conflict. This is because first elections usually arouse considerable uncertainty and fears among ethnic groups. Some may worry that their ethnic rivals will win and rule unfairly, and perhaps even brutally, over them, and therefore they decide to act preemptively.

Economic variables also matter. For example, richer countries tend to have less severe levels of ethnic violence. This is because rich democracies have more resources as well as better overall capabilities than poor countries to accommodate the demands of ethnic groups and to prevent insurgents from being able to survive and prosper. Furthermore, because people are less likely to have distress and consider other groups responsible for it in periods of economic growth, economic development is less likely to be associated with violent ethnic conflict.

Group-specific factors affect ethnic violence as well. It is generally argued that fewer political, economic, and cultural differences between dominant and minority groups mean a lower likelihood of ethnic violence (Gurr, 1993). Compared with dominant groups, if minority groups are doing relatively poorly politically and economically, or are relatively distinct along many ethnic cleavages, these minorities are more likely to fight against the state. By controlling for these factors, we can address the notion that not all groups are the same in their propensities to fight.

Based on the above discussions, we can derive the following propositions. Because presidentialism and semi-presidentialism except presidential-parliamentarism under divided governments have similar effects, ceteris paribus, we will use presidential systems to denote these systems hereafter.

Proposition 1: In cases involving a concentrated minority group facing a majority group, ethnic violence is less intense in presidential-proportional systems than in other systems.

Proposition 2: In cases involving a concentrated minority group facing no majority group, ethnic violence is more intense in parliamentary-non-proportional systems than in other systems.

Proposition 3: In cases involving a dispersed minority group, ethnic violence is neither more nor less intense in any one political system than another.

Table A1 in the Online Supplementary Materials presents the 16 possible combinations of institutional and ethnic factors together with the corresponding sample size of groups and countries. The table serves as a useful visual aid to our propositions.

Research design

We take the Minorities at Risk (MAR) dataset from 1985 to 2006 as our starting point, and analyze all ethnic groups at risk in democratic states.¹⁴ The unit of analysis of this dataset is an ethnic group at risk for a given year.¹⁵ However, the Fisher tests augmented by 0–3 lags show that our dependent variable provided by this dataset, rebellion, is not stationary, and the independent variables of interest in this article rarely change over time. We thus reorganize the dataset by removing its original panel structure to make it purely cross-sectional. We also exclude from analysis cases of president-parliamentarism under divided governments because these cases work differently from others of semi-presidentialism as aforementioned.¹⁶ The remaining transformed data includes 153 groups in 65 democratic states. Since our dataset is essentially nested, we then perform a hierarchical linear model on it to avoid inflation of the Type I error rate (Raudenbush and Bryk, 2002).

Because MAR’s measure of rebellion encodes only those conflicts that involve ethnic group-based activity, it is perfect for our purposes. It is an 8-point scale, with 0 indicating no violent conflict reported and higher values representing more intense forms of violent conflict.¹⁷ There is a single score for each ethnic group in each year.¹⁸

The MAR dataset also provides other group-specific independent variables, including cultural, political, and economic difference indices, and group concentration levels. We have added additional country-specific variables to the dataset. These additional factors include: government structures, electoral systems, fractionalization, first elections, regime duration, federal systems, GDP per capita, and change in GDP per capita. These variables can be classified into three categories: measures of political institutions, parameters of ethnic politics, and control variables.

For political institutions, we construct the variables of government structures and electoral systems based on our reading of various sources.¹⁹ Parliamentary is coded as 1 for parliamentary systems and 0 otherwise.²⁰ But for a few exceptions, we assign a score of 1 to Proportional system for pure PR, SNTV, and MMP systems and a score of 0 for other systems.²¹ The exceptions include the following cases. Ecuador, Niger, and Panama, while using mixed systems with unlinked tiers, are classified as proportional systems in our analysis. This is because the percentages of legislators elected under PR in these states are significantly higher than those of other countries with unlinked tiers, ranging from 63% to 90%.²² Hungary, Italy, and Philippines are classified as non-proportional systems as their mixed systems with linked tiers have much lower percentages of legislators elected under PR than other countries with linked tiers, ranging from 15% to 25%.

Next, we consider parameters of ethnic politics. MAR provides the measure of group concentration levels. This measure takes on four possible values, with 0 indicating the most dispersed distribution and 3 the least. We then created a dummy variable of Concentration recoding original values of 0 and 1 into 0 as well as 2 and 3 into 1, with 1 and 0 representing concentrated and dispersed distributions, respectively.²³ Many scholars have provided their indices of ethnic fractionalization (Alesina et al., 2003; Fearon, 2003), which help determine whether there is a majority group in a state as a whole. In this research, we use a fractionalization score of 0.6 as a cut-off point as previous research shows a nearly exact relationship between a fractionalization score less than 0.6 and the existence of a majority group in a state (Fearon, 2003).²⁴ The ethnic index constructed by Alesina and his associates is used to create the country-specific variable of Majority with 0 and 1 indicating the absence and existence of a majority group, respectively.²⁵

We also control for other relevant factors. There are two dichotomous control variables. Any federal system receives a score of 1 on the variable of Federal system (0 otherwise).²⁶ A First election is coded as 1 for the year in which a country had its first election and 0 for all the other years.²⁷ Enduring regime, GDP per capita, Change in GDP per capita, and Differences indices are also included as control factors.²⁸ Enduring regime is coded as an indicator of polity durability based on the number of years since the last regime transition or since 1900 (whichever occurred more recently).²⁹ GDP per capita and Change in GDP per capita capture relative wealth and whether the economy is improving or declining, respectively. The MAR dataset provides three differentials indices. The Political differentials index measures access to power and to civil service, recruitment, voting rights, etc. and ranges from no negative differentials (0) to serious differentials (4). The Economic differentials index measures differentials in economic status and positions among groups, and ranges from a value of −2 for most advantaged groups to 4 for most disadvantaged groups. The Cultural differentials index focuses on how distinct groups are in terms of their differences in ethnicity/nationality, language, religion, and the like, with higher values indicating more diversity. Because the MAR dataset does not include these index variables since 2004, we use 2003 indices of each group as proxies for their 2004–2006 data.

Descriptive statistics for all variables used in our quantitative analysis are shown in Table A2 in the Online Supplementary Materials.

Since one of the 16 possible combinations of institutional and ethnic factors (i.e., dispersed groups facing no majority groups under presidential-non-proportional systems) has no observations in our data, we must set the constant term to zero so that there are only 15 relevant parameters to be estimated in our model, as shown in Table 1.

Table 1.

Hierarchical linear model: rebellion, 1985–2006.

Variable	b	SE
Fixed effects:
Parliamentary (PAR)	1.61	0.88
Proportional system (PRO)	1.03	1.03
Presence of majority group (MAJ)	1.36	1.02
Concentration (Con)	1.06	0.80
PAR*PRO	−1.41	1.16
PAR*MAJ	−1.64	1.07
PAR*Con	0.36	1.00
PRO*MAJ	−1.05	1.15
PRO*Con	−0.32	1.10
MAJ*Con	0.12	1.00
PARPROMAJ	1.42	1.43
PARPROCon	−0.97	1.55
PARMAJCon	−0.77	1.38
PROMAJCon	−0.95	1.28
PARPROMAJ*Con	2.70	1.96
(Control variables not shown here)
Constant	0.00	—
Variance Components:
Country-level
Constant	0.00	—
Concentration	0.68*	0.15
Individual-level	0.83*	0.06

Table entries are ML estimates.

p < 0.05; 2-tailed tests. For full estimation results, please see Table A3 of Online Supplementary Materials.

Results

Table 1 presents key maximum likelihood estimates of the hierarchical model. Full results are available in Table A3 in the Online Supplementary Materials. No control variables significantly affect ethnic rebellion (p < 0.1), although the signs of their estimates are mostly as expected. This result is quite similar to the ones reported by Saideman et al. (2002). In their analysis, all but three control variables are insignificant.

That said, the most important question is whether the combinations of government structures and electoral systems demonstrate the expected effects on ethnic violence. Table 1 shows that none of the 15 coefficients of ethno-political systems are statistically significant at the 0.05 level (2-tailed tests). This, however, is purely an artifact of setting the constant term to zero. Had we set any one of the 15 core variables to zero, all the estimated coefficients would be shifted by a constant, and the statistical properties of the estimates would change accordingly. In other words, the statistical properties of our core variables’ estimated coefficients are inconsequential. What are consequential are the statistical properties of the differences of the 15 ethno-political systems, and these properties do not depend on which variable (the constant or one of our core variables) is set to zero.

We can now start to examine whether the marginal effects of the institutional configurations are significant, given a certain ethnic context.

Concentrated ethnic minorities facing a majority group

The top section of Table 2 presents the tests associated with Proposition 1. Two out of the three test results are consistent with our theory that a multiplicity of power centers and proportional representation make the presidential-proportional system the best institutional design to reduce conflict for concentrated ethnic minorities. With the parameters of ethnic politics thus fixed, presidential-proportional systems clearly outperform parliamentary-proportional and presidential-non-proportional systems in managing ethnic violence. Specifically, a geographically concentrated group that faces a majority group in a presidential-proportional system can maintain a level of rebellion that is less severe by 1.30 units than in a parliamentary-proportional system, and by 1.29 units than in a presidential-non-proportional system. Given that rebellion is measured by an 8-point scale, a 1.29–1.30-unit decrease is both statistically significant and substantial. The decrease in severity of rebellion associated with a change from parliamentary-non-proportional to presidential-proportional is 0.85 units in the theoretically expected direction, but it is not statistically significant.³⁰

Table 2.

Marginal effects of ethno-political systems on rebellion.

	Marginal effects	$χ^{2}$ (1)
Proposition 1
H_1a: PRE-PRO vs. PAR-PRO	−1.30	6.06*
H_1b: PRE-PRO vs. PAR-non-PRO	−0.85	1.23
H_1c: PRE-PRO vs. PRE-non-PRO	−1.29	6.95**
Proposition 2
H_2a: PAR-PRO vs. PAR-non-PRO	−1.67	4.31*
H_2b: PRE-PRO vs. PAR-non-PRO	−1.27	5.28*
H_2c: PRE-non-PRO vs. PAR-non-PRO	−1.98	9.90**
Proposition 3
H_3a: PRE-PRO vs. PAR-PRO	0.01	0.00
H_3b: PRE-PRO vs. PAR-non-PRO	0.01	0.00
H_3c: PRE-PRO vs. PRE-non-PRO	−0.02	0.00
H_3d: PAR-PRO vs. PAR-non-PRO	−0.38	0.34
H_3e: PRE-PRO vs. PAR-non-PRO	−0.58	0.58

p < 0.05; **p < 0.01.

Concentrated ethnic minorities facing no majority groups

The middle section of Table 2 presents the tests associated with Proposition 2. The results of all three tests are consistent with the Proposition, which predicts that the parliamentary-non-proportional system is the worst institutional design for concentrated groups facing no majority. Indeed, given such an ethnic group configuration, parliamentary-proportional, presidential-proportional, and presidential-non-proportional all outperform parliamentary-non-proportional. Specifically, a geographically concentrated group facing no majority in a parliamentary-non-proportional system maintains a rebellion level that is higher by 1.67, 1.27, and 1.98 units, respectively, than in a parliamentary-proportional, presidential-proportional, or presidential-non-proportional system, respectively. All the differences are statistically significant. In particular, the difference between parliamentary-non-proportional and presidential-non-proportional, short of a full 2-unit change, is the most dramatic, and it is highly significant at the 0.01 level.

Given that five out of six theoretical predictions are empirically and statistically supported, our findings provide rather strong evidence for Propositions 1 and 2. In cases involving a concentrated minority group facing a majority group, ceteris paribus, ethnic violence is generally less intense in presidential-proportional systems than in other systems. This is because presidential-proportional systems have the advantages of multiple power centers and proportional representation to enhance the possibility of minority influence in policymaking. In cases involving a concentrated minority group facing no majority groups, ceteris paribus, ethnic violence is most severe in parliamentary-non-proportional systems. This is because all the other systems provide either coalitional politics or checks and balances, whereas parliamentary-non-proportional systems are less likely to do so. Powerless minorities in these systems are forced to engage in violence to make their voices heard.

Dispersed minorities

While institutions matter for geographically concentrated groups concerning their propensity for conflict, they do not for spatially dispersed groups. The bottom section of Table 2 shows that all the comparisons we made in testing Propositions 1 and 2 for concentrated groups lose their statistical significance when applied to dispersed groups. Proposition 3 can therefore not be rejected with our tests. There are no empirical grounds to disconfirm the theoretical argument that dispersed groups are essentially more peaceful.

These findings about institutional effects for different group configurations are important in two ways. First, it is necessary to control for ethnic parameters when investigating the effects of political institutions. If we ignore these crucial contextual factors, our estimates of institutional effects may very likely be biased. Specifically, if we do not distinguish between concentrated and dispersed ethnic groups, we might mistakenly conclude that institutions do not matter at all. The absence of institutional effects in cases involving dispersed groups might dilute the significant effects in cases involving concentrated groups, to the extent that the significance is totally lost. Second, our findings provide useful lessons for institutional engineering in ethnically divided societies. We will elaborate on this in the following section by way of concluding this article.

Concluding remarks

This article has identified the context in which political institutions affect the level of ethnic violence. Through a large comparative study, we find that government structures and electoral systems have interactive effects which are further moderated by the parameters of ethnic configurations. Specifically, for concentrated ethnic minorities, semi-presidential-PR and presidential-PR systems better reduce ethnic conflict, whether a majority group exists or not. However, parliamentary-PR, semi-presidential-non-PR, and presidential-non-PR systems also perform well in mitigating ethnic violence when concentrated minorities do not face a majority group.

In sum, our findings demonstrate that political institutions can and do play a role in the multi-faceted phenomena of ethnic violence, but their effects are moderated by the nuances of the number and spatial distribution of ethnic groups in a country. There is no one-size-fits-all solution to ethnic violence. Parliamentarism and PR systems are not always and clearly superior, as contrary to what has long been suggested by Lijphart (2002).

Supplemental Material

Supplementary_Material – Supplemental material for Institutions, contexts, and ethnic violence in comparative perspective

Supplemental material, Supplementary_Material for Institutions, contexts, and ethnic violence in comparative perspective by Feng-yu Lee and Tse-min Lin in International Political Science Review

Footnotes

Acknowledgements

We thank Robert Elgie, Robert Moser, Yuko Kasuya, Yu-shan Wu, Sophia Moestrup, and the anonymous referees for comments and suggestions on various versions of this article. In addition, we are grateful to Yu-shiuan Huang and Li-hong Weng for their very capable assistance.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Ministry of Science and Technology, Taiwan (grant numbers: 101-2410-H-002-121-MY3, 106-2410-H-002-115-MY3).

ORCID iD

Tse-min Lin

Supplementary material

Supplementary material for this article is available online.

Notes

Author biographies

Feng-yu Lee is an associate professor of Political Science at National Taiwan University. She received her PhD from the University of Texas at Austin. Her research interests focus on comparative political institutions, comparative politics, ethnic politics, and constitutional design of executive-legislative relations.

Tse-min Lin is an associate professor of Government at the University of Texas at Austin. He received his PhD in Political Science from the University of Minnesota. His teaching and research interests cover methodology, formal theory, and comparative political behavior.

References

Alesina

Alberto

Devleeschauwer

Arnaud

Easterly

William

Kurlat

Sergio

Wacziarg

Romain

(2003) Fractionalization. Journal of Economic Growth 8(2): 155–194.

Birnir

Johanna Kristin

(2007) Ethnicity and Electoral Politics. Cambridge: Cambridge University Press.

Brancati

Dawn

(2006) Decentralization: Fueling the fire or dampening the flames of ethnic conflict and secessionism? International Organization 60(3): 651–685.

Braumoeller

Bear F

(2004) Hypothesis Testing and Multiplicative Interaction Terms. International Organization 58(4): 807–820.

Byman

Daniel L

(2002) Keeping the Peace: Lasting Solutions to Ethnic Conflicts. Baltimore, MD: Johns Hopkins University Press.

Cederman

Lars-Erik

Wimmer

Andreas

Min

Brian

(2010) Why Do Ethnic Groups Rebel? New data and analysis. World Politics 62(1): 87–119.

Cheibub

José Antonio

(2007) Presidentialism, Parliamentarism and Democracy. Cambridge: Cambridge University Press.

Cheibub

José Antonio

Gandhi

Jennifer

Vreeland

James Raymond

(2009) Democracy and Dictatorship Revisited. Public Choice 143 (1-2): 67–101.

Cheibub

José Antonio

Przeworski

Adam

Saiegh

Sebastian M

(2004) Government Coalitions and Legislative Success Under Presidentialism and Parliamentarism. British Journal of Political Science 34(4): 565–587.

10.

Cohen

Frank S

(1997) Proportional Versus Majoritarian Ethnic Conflict Management in Democracies. Comparative Political Studies 30(5): 607–630.

11.

Elgie

Robert

(2011) Semi-Presidentialism: Sub-Types and Democratic Performance. New York, NY: Oxford University Press.

12.

Elkins

Zachary

Sides

John

(2007) Can Institutions Build Unity in Multiethnic States? American Political Science Review 101(4): 693–708.

13.

Fearon

James

(2003) Ethnic and Cultural Diversity by Country. Journal of Economic Growth 8(2): 195–222.

14.

Gunther

Richard

Mughan

Anthony

(1993) Political Institutions and Cleavage Management. In Weaver

R Kent

Rockman

Bert A

(eds) Do Institutions Matter? Government Capabilities in the United States and Abroad. Washington, DC: The Brookings Institution, 272–301.

15.

Gurr

Ted Robert

(1993) Minorities at Risk: A Global View of Ethnopolitical Conflicts. Washington, DC: U.S. Institute of Peace.

16.

Hellwig

Timothy

Samuels

David J

(2008) Electoral Accountability and the Variety of Democratic Regimes. British Journal of Political Science 38(1): 65–90.

17.

Horowitz

Donald L

(1991) A Democratic South Africa? Constitutional Engineering in a Divided Society. Berkeley, CA: University of California Press.

18.

Ishiyama

John

(2000) Institutions and Ethnopolitical Conflict in Post-Communist Politics. Nationalism and Ethnic Politics 6(3): 51–67.

19.

Kam

Cindy D

Franzese

Robert J

(2007) Modeling and Interpreting Interactive Hypotheses in Regression Analysis. Ann Arbor, MI: University of Michigan Press.

20.

Kostadinova

Tatiana

(2002) Do Mixed Electoral Systems Matter? A cross-national analysis of their effects in Eastern Europe. Electoral Studies 21(1): 23–34.

21.

Lijphart

Arend

(2002) The Wave of Power-Sharing Democracy. In Reynolds

Andrew

(ed.) The Architecture of Democracy: Constitutional Design, Conflict Management, and Democracy. Oxford: Oxford University Press, 37–54.

22.

Lijphart

Arend

(2004) Constitutional Design for Divided Societies. Journal of Democracy 15(2): 96–109.

23.

Lijphart

Arend

(2012) Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries. New Haven, CT: Yale University Press.

24.

Linz

Juan J

(1994) Presidential or Parliamentary Democracy: Does it make a difference? In Linz

Juan J.

Valenzuela

Arturo

(eds) The Failure of Presidential Democracy: Comparative Perspectives. Baltimore, MD: Johns Hopkins University Press, 3–87.

25.

Lublin

David

(2014) Minority Rules: Electoral Systems, Decentralization, and Ethnoregional Party Success. Oxford: Oxford University Press.

26.

Massicotte

Louis

Blais

André

(1999) Mixed Electoral Systems: A conceptual and empirical survey. Electoral Studies 18(3): 341–366.

27.

Moser

Robert

(2008) Electoral Systems and the Representation of Ethnic Minorities: Evidence from Russia. Comparative Politics 40(3): 273–292.

28.

Norris

Pippa

(2002) Ballot Not Bullets: Testing consociational theories of ethnic conflict, electoral systems, and democratization. In Reynolds

Andrew

(ed.) The Architecture of Democracy: Constitutional Design, Conflict Management, and Democracy. Oxford: Oxford University Press, 206–247.

29.

Pereira

Carlos

Melo

Marcus André

(2012) The Surprising Success of Multiparty Presidentialism. Journal of Democracy 23(3): 156–170.

30.

Raudenbush

Stephen W

Bryk

Anthony S

(2002) Hierarchical Linear Models: Applications and Data Analysis Methods. Thousand Oaks, CA: Sage Publications.

31.

Reilly

Benjamin

Reynolds

Andrew

(1999) Electoral Systems and Conflict in Divided Societies. Washington, DC: National Academy Press.

32.

Reynolds

Andrew

(1995) The Case of Proportionality. Journal of Democracy 6(4): 117–124.

33.

Saideman

Stephen M

Ayres

R William

(2000) Determining the Sources of Irredentism: Logit analyses of minorities at risk data. Journal of Politics 62(4): 1126–1144.

34.

Saideman

Stephen M

Lanoue

David J

Campenni

Michael

Stanton

Samuel

(2002) Democratization, Political Institutions, and Ethnic Conflict: A pooled times-series analysis, 1985–1998. Comparative Political Studies 35(1): 103–129.

35.

Samuels

David

Shugart

Matthew

(2010) Presidents, Prime Ministers and Political Parties. Cambridge: Cambridge University Press.

36.

Shugart

Matthew Søberg

(2005) Semi-Presidential Systems: Dual executive and mixed authority patterns. French Politics 3(3): 323–351.

37.

Shugart

Matthew Søberg

Wattenberg

Martin P

(2001) Mixed-Member Electoral Systems: A definition and typology. In Shugart

Matthew Søberg

Wattenberg

Martin P

(eds) Mixed-Member Electoral Systems: The Best of Both Worlds. Oxford: Oxford University Press, 9-24.

38.

Shvetsova

Olga

(1999) A Survey of Post-Communist Electoral Institutions: 1990–1998. Electoral Studies 18(3): 397–409.

39.

Toft

Monica Duffy

(2003) The Geography of Ethnic Violence. Princeton, NJ: Princeton University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.44 MB