Scholarly article on topic 'The Pacifying Effects of Local Religious Institutions: An Analysis of Communal Violence in Indonesia'

The Pacifying Effects of Local Religious Institutions: An Analysis of Communal Violence in Indonesia Academic research paper on "Political Science"

0
0
Share paper
Academic journal
Political Research Quarterly
OECD Field of science
Keywords
{""}

Academic research paper on topic "The Pacifying Effects of Local Religious Institutions: An Analysis of Communal Violence in Indonesia"

The Pacifying Effects of Local Religious Institutions: An Analysis of Communal Violence in Indonesia

Political Research Quarterly 2015, Vol. 68(2) 211-224 © 2015 University of Utah Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1065912915578460 prq.sagepub.com

Alexander De Juan1, Jan H. Pierskalla2, and Johannes Vüllers3

Abstract

This paper tests whether local religious institutions have a dampening effect on the probability of communal violence. It argues that a dense layer of institutions strengthens horizontal and vertical contacts and networks within religious communities. Horizontal linkages help to bridge social, economic, and ethnic divisions. Vertical contacts enable religious leaders to stay informed about communal grievances among their followers and to coordinate conflict resolution attempts. In our analysis of more than 60,000 villages in Indonesia, we are able to document a statistically significant and substantively meaningful negative effect of the density of local religious institutions on the probability of mass fighting. This effect is robust to the inclusion of an exhaustive list of confounding variables and alternative measures of violence. We present additional evidence that this pacifying effect of religious institutions is weaker or absent in conflicts that evolve along explicitly religious cleavages.

Keywords

religion, peace, Indonesia, local religious institutions

Introduction

In July 2008, the Indonesian village of Bok experienced a series of demonstrations that threatened to escalate into massive communal violence. Local religious leaders personally visited people involved in the demonstrations and reminded them of their religious obligations to refrain from violence. They used their local moral authority to bring warring parties to the negotiation table and keep them engaged. Tensions were ultimately resolved before violence could escalate, without the need for any formal police intervention (for additional details of the case, see Kingsley 2011). These events illustrate an often-neglected dimension of the role that religion can play in violent conflicts.

Research on violent intrastate conflict has investigated religion as a potentially escalating factor. Notably, quantitative studies have argued that specific demographic structures—such as religious fractionalization or polarization—increase the likelihood of violence (e.g., Ellingsen 2000; Montalvo and Reynal-Querol 2005). Religion, however, is more complex than such analyses suggest. Focusing on religious demography alone disregards essential dimensions of religion that likely affect its role in political violence (Basedau, Pfeiffer, and Vullers 2014). Most notably, religions are organized through

institutions that can influence their impact on conflicts. Yet we know little about how religious institutionalization affects the role of religion in conflicts. We contribute to filling this gap.

Specifically, we are interested in local religious institutions, namely, places of worship—churches, mosques, temples—at the village level. These institutions constitute a main interface between believers and their religion. Contrary to high-level institutions, such as religious umbrella organizations or high-ranking clerical bodies, village-level institutions are closer to the people and often enjoy substantial legitimacy within their respective communities (Appleby 2001). Such local organizations can sometimes be used to incite religious violence by facilitating collective action (i.e., preaching religious hatred, organizing armed movements, and coordinating fighters; De Juan 2009; Toft 2007). We argue, however, that in

'German Institute of Global and Area Studies, Hamburg, Germany

The Ohio State University, Columbus, USA

3University of Konstanz, Germany

Corresponding Author:

Alexander De Juan, German Institute of Global and Area Studies, Neuer Jungfernstieg 2', Hamburg 20354, Germany. Email: alexander.dejuan@giga-hamburg.de

most cases, local religious institutions are an important resource that religious elites can draw upon to prevent an outbreak of violence and settle (especially nonreligious) conflicts peacefully. A dense layer of local religious institutions allows religious leaders to stay informed about various ethnic, social, or economic grievances among their followers and coordinate conflict resolution attempts. Local religious institutions can act, in the words of Fearon and Laitin (1996, 715), as "decentralized, nonstate institutional mechanisms" that facilitate peace.

We test this hypothesis empirically using the example of Indonesia. Our analysis of more than 60,000 villages demonstrates that village-level religious institutionaliza-tion has a statistically significant and substantively meaningful negative effect on the probability of mass fighting. This effect is robust to the inclusion of an exhaustive list of confounding variables, alternative measures of violence, and various estimation approaches.

Our findings have several important implications for the existing literature. First, because we substantiate an argument about the role of local religious institutions in conflict, we extend the theoretical focus of standard conceptions of religion's impact on violence. Moreover, our insights into the pacifying potential of religion contribute to broadening empirical research on its role in violence, which has thus far concentrated on its escalating effects. Second, our argument and empirical evidence also apply to the literature on communal violence and civil conflict more generally. Our focus on local, nonpolitical institutions highlights a factor that has not been studied sufficiently using quantitative approaches. Third, we add to the specific discourse on violence in Indonesia. Many qualitative studies have focused on the role of religion in Maluku and Sulawesi, highlighting its escalating effects, which may have led to a distorted view of the role of religion in Indonesia more generally. Contrary to some qualitative accounts, we are able to show that religion (specifically, local religious institutionalization) has effectively limited the outbreak of communal violence.

Communal Violence and Local Religious Institutions

Research on communal violence1 has emphasized the role of state and nonstate actors and institutions without actively considering the role of local religious institutions. For example, Horowitz (2001) points out that communal violence is closely linked to the role of government security forces. The reaction of the security forces— whether they support or oppose the violence, and their ability to do so—influences the occurrence of communal violence. Relatedly, Wilkinson (2004) argues that political elites instrumentally use communal violence to secure electoral support by manipulating identity groups for

their own interests. Communal violence is thus the product of exploiting social group boundaries for political benefits. In a similar vein, Brass (2003) highlights the importance of ethnic groups' political mobilization in the context of political competition. "Riot networks" sustain the separation of social groups by aggravating real (or construed) conflicts between them in peacetime. Politicians can then use this persistent separation of social groups to escalate small conflicts into large-scale violence (Brass 2003).

Varshney (2002), by contrast, provides an explanation for the potentially peaceful influence of local institutions. He claims that institutions (i.e., trade unions or employer associations) crosscutting ethnic or religious group boundaries decrease the likelihood of communal violence. Such institutions mediate the tensions between identity groups, rebut stereotypes, and highlight commonalities. Using a comparison of six cities with and without Hindu/Muslim violence in India, Varshney (2001, 378) finds that the existence of such "peace systems" can prevent communal violence by decreasing the potential benefits to politicians of aggravating intergroup tensions. Thus, while these studies provide for systematic analyses of state and nonstate institutions and religious violence, they do not focus on the role of features of religious communities, such as levels of religious institutionalization.

The majority of the literature on civil wars and armed conflict treats religion as one identity marker among others, such as ethnicity or language. The basic argument is that religious diversity (i.e., specific religious constellations or horizontal inequalities between religious groups) is prone to interreligious mobilization and increases the risk of violence. Some statistical studies find evidence that religious identities can intensify large-scale violent conflicts (e.g., Akbaba and Taydas 2011; Basedau, Pfeiffer, and Vüllers 2014), while others do not find overall robust associations with violence (e.g., Ellingsen 2000; Fearon and Laitin 2003; Montalvo and Reynal-Querol 2005). Importantly, though, quantitative studies on religion and organized armed conflicts generally do not take the institutional dimension of religion into account. Thus, we also know little about the role of religious institutions from this particular research strand.

Authors explicitly studying the link between religion and peace acknowledge the important role of religious institutions (e.g., Sisk 2012; Toft, Philpott, and Shah 2011), but generally attribute little causal explanatory power to them. Many existing studies conceptualize institutions as faith-based actors or organizations (Appleby 2001; Bercovitch and Kadayifci-Orellana 2009). They highlight the theological norms of peace for the actions of religious institutions (Ter Haar 2005, 7). Therefore, the peace impact of religious institutions is conceptualized as a result of existing peace norms within the religious

communities (Appleby 2001, 835). Other studies argue that opportunity structures, rather than religious norms alone, affect how religious institutions act in conflicts (Harpviken and Roislien 2008, 364-65). They fail, however, to concretely conceptualize the relevant opportunity structures. Moreover—and similarly to other studies on the potential peace impact of religious institutions—they lack systematic and robust empirical analysis.

The Peace Impact of Local Religious Institutions

Numerous studies illustrate that religion can contribute to either violence or peace (e.g., Toft, Philpott, and Shah 2011). A larger theoretical and empirical debate is still trying to determine the exact conditions under which religion affects violence in either direction. We add to this discourse by studying a new channel through which religion's pacifying potential can be realized. More specifically, we argue that local religious institutions can provide effective local-level conflict resolution, preventing grievances from developing into violent clashes and thereby reducing the risk of communal violence—in particular outside the narrow domain of religious grievances. The remainder of this section will elaborate on this argument.

Religion can contribute to inciting violent armed conflict (e.g., Akbaba and Taydas 2011; Basedau, Pfeiffer, and Vüllers 2014). However, discussions on the role of religion in violence often neglect the fact that, while religious clashes attract much attention, they actually represent a minority of conflicts around the world. Data provided by Svensson (2007) for 1989-2004 show that religious incompatibilities play a role in only around 49 out of 168 armed conflict dyads; parties to the conflict belong to different faiths in only 52 out of 165 cases. This corresponds to numbers provided by Toft (2007): out of 133 civil wars included in her data set, only 42 (32%) centered on religious issues. Finally, Basedau, Pfeiffer, and Vüllers (2014) present comparable figures based on a newly compiled data set on religious violent conflicts: out of a total of 138 armed conflict onsets between 1990 and 2010, they identify only 60 cases in which the warring factions differed greatly in their religious affiliation. More importantly, they record only 41 onsets of armed conflicts in which at least one warring faction had explicit religious aims. These data sets indicate that religion is sometimes associated with violence, but that religious violent conflicts constitute a minority of all armed conflicts.

To get a sense of the role of religion in more decentralized forms of violence, we examine three country cases that are often cited for the high relevance of religion in their conflicts: Nigeria, Pakistan, and Indonesia.2 We rely

on country-specific, newspaper-based violent event data sets to determine the share of violence that may reasonably be attributed to religious ideas and actors. The first data set was compiled by the nonprofit organization Nigeria Watch; it screened fifteen local newspapers and human rights reports from 2006 to 2012 to extract information on 9,255 violent events in Nigeria. Only 567 of these events involved organized Christian or Islamic groups, and only 123 can directly be attributed to religious issues according to the data set. For Pakistan, we investigate the BFRS data set (Bueno de Mesquita et al. 2013), which contains incident-level data on political violence from 1988 to 2011. Out of a total of 17,679 events for which the data set provides information on the actors involved, religious/sectarian militants or informal groups, or religious parties, are mentioned as responsible actors in only 1,656 cases. Finally, we look at our main case: Indonesia. The Sistem Nasional Pemantauan Kekerasan Indonesia (SNPKI)3 is a government and World Bank financed violent event database for several of Indonesia's provinces. The project provides a monthly updated list of violent events (conflict, domestic violence, crime, violence during law enforcement), starting in 1998. In this extensive data set, religion was recorded as a trigger of violence in only 1,361 out of a total of 119,107 events; religious groups or actors were involved in only 1,285 instances of violence.

It is certainly difficult to pinpoint the exact cause or primary identity of the actors involved in armed conflicts or communal violence. Moreover, a lack of information, specific coding decisions, or regional foci in country-level event data sets can lead to the underreporting of religious violence. We believe, however, that the sources highlighted above demonstrate that religious issues and identities represent just one set of potential cleavages among many others that may lead to violence.

While much research attempts to explain such narrow instances of religious violence, much less attention is given to another equally pertinent question: what is the role of religion in all other cases of violence, namely, in conflicts that do not seem to have any religious connotation? We focus on this question. Rather than trying to explain the ambivalence of religion in conflicts, we ground our analysis on the empirical observation that religious escalation of violence is, in many cases, an exception rather than the norm. We argue that if we investigate communal violence in more general terms, instead of focusing on narrow cases of actual religious violence, we will find that religion effectively contributes to peaceful conflict resolution, and reduces the risk of communal violence.

We concentrate on a specific element of religion that we consider particularly relevant for communal conflicts: village-level religious institutions, namely, places of

worship such as churches, mosques, prayer houses, or temples. We apply a rather narrow perspective on religious institutions, and do not include organizations such as religious political parties or civil society actors, or high-level religious institutions such as national religious hierarchies or umbrella organizations. We focus on places of worship because they constitute the core of every religious community: in church services or Friday prayers, communities practice their beliefs. This is where religious elites, such as priests, gurus, or imams, provide religious interpretations to the believers. Contrary to national-level institutions, places of worship are directly accessible to the local communities and thus provide an essential interface between the believers and their organized religion. Wider religious activities, such as choirs, prayer groups, or social welfare activities (e.g., soup kitchens, education, or family services), are often directly linked to local churches, mosques, or other local places of worship. These strong ties are likely to be weaker beyond the congregational level, highlighting the importance of local institutions (Wald, Silverman, and Fridy 2005).

Our main hypothesis is that a high density of local religious institutions decreases the likelihood of communal violence. The two main mechanisms operate via strengthening horizontal and vertical contacts within religious communities.

Religions have specific features that make them particularly well suited to fostering strong horizontal networks (Pickering 2006). Religion is, in many cases, not based on self-selection as people are born into their respective religious communities. It, therefore, tends to create more heterogeneous communities (Ammerman 2003). Moreover, religious ceremonies involve the repeated interaction of individuals from various socioeconomic backgrounds. Finally, religion engenders norms that foster intrareligious cooperation among such different groups (Ammerman 2003). We argue that a high level of institutionalization contributes to fomenting these elements. Local religious institutions are encultur-ated—they are often deeply rooted in their own communities, representing a complex web of relationships cutting across economic, political, and ethnic divisions. Moreover, they provide venues for regular face-to-face interaction and contribute to forging intrareligious contact and cooperation through regular and shared religious rituals.

This reasoning is related to the concept of social capital (Putnam 2000, 65-79) and a prominent argument put forward by Ashutosh Varshney, as discussed above. Putnam (2000, 66) argues that "churches [understood as religious institutions of whatever faith] provide an important incubator for civic skills, civic norms, community interests, and civic recruitment." All of these features of religious institutions, thus, are relevant for social capital

that helps to solve communal problems because of the preexisting networks and relationships. Horizontal networks of religious communities, therefore, are one facet of social capital. According to Varshney (2001), heterogeneous, interreligious civic organizations helped to build networks across religious communities in India and thereby contributed to dampening violent interreligious conflicts. In a similar vein, we argue that religious institutions can create bonds across economic, political, and sometimes even ethnic divisions and thereby reduce the risk of communal conflict across these nonreligious cleavages. Dense religious institutions allow frequent contact between members. This aids the flow of information, the emergence of meaningful reputations, and the sanctioning of out-of-norm behavior through peer social pressure (Fearon and Laitin 1996). Independent of the ethnic demographics within and across villages, dense religious institutions can act as an integrative force that bridges potential gaps even between fairly polarized identity and socioeconomic groups.4 If most conflict-relevant cleavages are nonreligious, as argued above, dense religious institutions should be associated with lower levels of communal violence.

In Nepal, for example, local religious actors and institutions have traditionally played an important role in intracommunity mediation and informal conflict resolution. This holds true for Muslim, Hindu, and Buddhist communities alike. The specific pacifying potential of religious institutions rests on horizontal networks based on integrating religious beliefs, symbols, and activities that transcend nonreligious cleavages. Resolution of ethnic, resource-based or economic conflict is made binding through religious oaths in the name of the common God and touching of common sacred books or symbols. Festivals, discussions of classical religious texts, and performing religious rites provide occasions for religion-based interaction and reconciliation across nonreligious divides (Dahal and Bhatta 2008).

The second mechanism operates through strengthening vertical networks between religious leaders and believers. Religious elites often enjoy a moral credibility that is unmatched by other local actors (Powers 2010, 331), and local institutions give such elites direct access to believers (Traunmuller and Freitag 2011) and strengthen their capability to monitor, mediate, and police communal conflicts (Fearon and Laitin 1996). Priests, imams, gurus, or religious teachers meet their constituents in daily services, Friday prayers, or prayer groups, which provides the means to publicly condemn violence and call for peace and restraint (Wald, Silverman, and Fridy 2005). Furthermore, close relations between religious elites and local believers can serve as an early warning system for social tensions. Believers will inform religious leaders about perceived injustices or growing

conflicts through local institutions. Therefore, religious elites know very early about growing tensions and can act to prevent the same. Assuming, in line with Fearon and Lai tin (1996), that religious communities and leaders generally prefer peace over war, these institutional ties support local religious elites' monitoring, mediation, and policing activities.

In Sri Lanka, for example, religious committees and leaders, such as mosque committees, the church, parish councils, and their subsidiary bodies, have been particularly active in solving disputes within their communities. They focus on land disputes, interethnic conflicts, or youth violence. They use their local influence and direct contact to the believers to identify emerging conflicts and engage parties in peaceful conflict resolution. Religious leaders provide guidance during sermons; often complaints are addressed during or after common prayers. Local religious committees investigate issues that have been raised, summon or visit the parties to the dispute, issue warnings, mediate, or arbitrate (Centre for Policy Alternatives 2003).

Thus, vertical ties support the conflict-mitigating role of the aforementioned horizontal networks in that they provide additional capacities for the identification and effective mediation of tensions across various socioeconomic cleavages—provided that religious elites do not face explicit incentives to foment violence.5

Communal Conflicts in Indonesia

According to the Indonesian Ministry of Religious Affairs, approximately 88 percent of the Indonesian population self-identify as Muslim, 9.3 percent as Christian, 1.8 percent as Hindu, and 0.6 percent as Buddhist (Human Rights Watch 2013). There are also an undetermined number of other customary and indigenous religions that the state has been encouraging to join one of the major faiths (Schiller 1996).

While there were cases of interreligious and communal violence before the end of the 1990s (Kim 1998; Van Klinken 2001), the number of clashes between various identity groups increased substantially after the fall of the Suharto regime (Varshney, Tadjoeddin, and Panggabean 2008). The most prominent interreligious clashes have been violent conflicts in Maluku and Sulawesi. Fighting on Maluku began in 1999, pitting Christian Ambonese against Muslims from Sulawesi. The religious dimension of the conflict was apparent from the start: churches and mosques were targeted and burned down, and there were reports of forced conversions and massacres exclusively targeting Muslims and Christians (International Crisis Group [ICG] 2000, 2002; Turner 2003). The violence in Sulawesi, especially in the city of Poso, displayed similar patterns: religious narratives of violence were preached

to people within houses of worship (Van Klinken 2007). Churches and mosques served as military headquarters and communication nodes for coordinating the fighters (ICG 2002).

Whereas such interreligious conflicts have been particularly intense and have received much media and scholarly attention, they have by no means been the only incidences of violence in Indonesia. As argued above, other forms of conflict may have been less severe but all the more frequent. What we briefly demonstrated above, using the SNPKI data, is also confirmed by another data set collated by Varshney, Tadjoeddin, and Panggabean (2008). Their data contain information on violent conflicts in fourteen of Indonesia's twenty-eight provinces for the period 1990-2003. According to their data, ethno-communal (including interreligious) conflicts have accounted for nearly 90 percent of total fatalities but only 17 percent of the incidences. Taken together, other types of collective violence—including economic conflicts, conflicts between the state and communities, intervillage brawls and vigilante killings—have been much more frequent. This is particularly true for provinces other than Maluku and Sulawesi. Java, for example, accounted for the highest number of incidents of collective violence. However, only around 2 percent of these conflicts have been ethnocommunal in nature. These data show us that interreligious conflicts in Indonesia have been rather rare events compared with other forms of collective violence. Furthermore, even in the high period of communal violence in Indonesia, the vast majority of communities did not experience any form of organized violence. Whereas religion has clearly contributed to violence in Maluku and Poso, focusing on these specific conflicts may distort our understanding of the role of religious institutions in violent conflict overall.

The existing literature on communal violence in Indonesia has put forward several arguments to explain the patterns of violence. Bertrand (2004) argues that the fall of Suharto marked the start of a period in which Indonesia's "National Model" was renegotiated, which redefined the balance between various identity groups and led to various forms of violence. Braithwaite et al. (2010) argue that Indonesia experienced a period of ano-mie in which security forces failed to stay neutral, which contributed to the spread of violence. Similarly, Wilson (2008) asserts that the security forces' failure to limit communal violence contributed to the overall escalation of violence. Some trace local violence directly to the meddling of Jakarta elites (ICG 2000; Turner 2003), but Bertrand (2004, 6) and work by Van Klinken and others deemphasize the role of centrally guided security forces in communal violence and highlight the importance of local competition for resources in triggering violent clashes (Van Klinken 2007). An important recent study

by Tajima (2013) argues that neither of these explanations can fully account for the spatial and temporal patterns of communal violence. He argues that localities that depended heavily on state security institutions experienced more violence in the aftermath of authoritarian breakdown, as compared with communities that could rely on informal, nonstate security institutions to manage local conflict.

While these studies offer insights into the causes and timing of communal violence in Indonesia's posttransition period, we aim at understanding the role of local religious institutions in these conflicts. Anecdotal evidence suggests that religious institutions have, in many cases, been involved in conflict resolution. They have reportedly played a significant role in de-escalating large-scale conflicts such as ethnic violence between Dayaks and Immigrant Madurese in Kalimatan (ICG 2001), during anti-U.S. protests in Jakarta (Schröter 2003), in phases of instability in Lombok (Kingsley 2011), and in the autonomy conflict in West Papua (Rutherford 2006).

We believe that they can be successful in preventing and peacefully resolving local-level disputes, which form the bulk of the conflicts in Indonesia (e.g., intra-and intervillage brawls, conflicts over land issues, vigilante activities). Reports prepared by the World Bank show that in Indonesia, religious leaders play a significant role in local-level conflict resolution. In many regions, they have substantial bargaining power within their local communities that enables them to bring conflicting parties together and mediate effectively. The individual authority and legitimacy of religious institutions allows them to intervene in conflicts and effectively prevent escalation into mass fighting (World Bank 2004, 2008).

Over the years, the growing influence of religious institutions compared with local state institutions and traditional authorities has increased their role in Indonesian village-level governance (Duncan 2009). Religious affiliation and networks have long played an important role in careers in business and politics, which increases the social leverage of religious institutions. For various reasons, traditional institutions in Indonesia (adat), that is, customary laws that govern interpersonal relations and can be used to regulate conflicts, have lost their influence over the population (Acciaioli 2001). Thus, in many areas, people's focus was shifting from traditional to religious norms and institutions: mosques and churches became more important than adat (Duncan 2009). Religious institutions' substantial and growing influence over their constituencies—compared with alternative local-level institutions—has enabled them to effectively engage in informal conflict resolution within their areas of influence.

Statistical Analysis Variables and Data

We now turn to a more systematic treatment of our argument. To substantiate our claim that the density of local religious institutions has pacifying effects on communal violence, we utilize data on Indonesian villages from the peak of communal violence in the early 2000s. Other statistical analyses of communal violence in Indonesia have focused on a comprehensive explanation of the violence (Barron, Kaiser, and Pradhan 2009), and have noted, among other factors, the potential relevance of local religious institutions. Our analysis is distinct in that our theoretical argument and empirical investigation explicitly focus on the effect of local religious institutions and trace the mechanism in more detail.

We use the Indonesian village (desa), the lowest administrative level, as our unit of analysis. We draw on the 2003 Indonesian village-level census (Pendataan Potensi Desa [PODES]) for the majority of our covariate information. The village census is conducted regularly by Indonesia's statistical agency (Badan Pusat Statistik [BPS]) and covers a number of issue areas, ranging from basic infrastructural information and economic activities to religious institutions and political affairs.6 The village census is implemented by enumerators from local BPS branches in collaboration with local village heads. The 2003 iteration of the PODES covered more than 67,000 villages and, for the first time, also contained questions on communal violence. This rich source of information allows us to trace the effect of local religious institutions on communal violence, while controlling for a number of important alternative factors that have contributed to the outbreak of communal violence in Indonesia. More importantly, it allows us to effectively compare villages in regions that have been identified in the literature as hotspots of communal violence (e.g., Ambon and Poso) with those in regions that escaped widespread mass fighting.

To measure communal violence, we create a simple dummy indicator for each village that takes a value of 1 if any incidents of mass fighting were reported for the previous year in the PODES 2003 survey, and 0 otherwise.7 This variable includes all forms of violent incidents, ranging from religious or ethnic violence to riots and mass brawls. Based on this indicator, only about 7 percent of all villages experienced communal violence across the archipelago in the early 2000s. In additional analyses, we further distinguish between different types of violence. The PODES offers additional subcategories: communal fighting, ethnic fighting, and clashes with state security forces. While this classification is not related to any theoretical conceptualization from the

literature on communal violence, it offers us some limited opportunities to assess the effect of religious institutions on more specific forms of violence. We create additional dummy variables to capture each category individually.

To test whether local religious institutions have a pacifying effect on communal violence, we rely on information about the number of places of worship in each village. The village census records this information for all major religions (Islam, Catholicism, Protestantism, Hinduism, and Buddhism). For our main analysis, we simply calculate the total number of places of worship and normalize by local population size. We log our measure, as we expect diminishing returns in the pacifying effect of religious institutionalization. While the number of places of worship per capita is not a perfect measure of local religious institutions, we believe it is a good proxy for the general capabilities of local religious elites to leverage horizontal and vertical networks for peaceful conflict resolution (see robustness checks for alternative specifications).

We control for a large set of confounding variables. We draw on the wealth of information in the PODES, as well as additional information on ethnic and religious compositions and the socioeconomic inequality of the village population from the population census.8 We consider structural factors such as the (logged) population size (PODES), the (logged) distance to the district capital (PODES), (logged) population density (PODES), and whether the village was located in an urban area (PODES), as the wider conflict literature has identified these as relevant factors (e.g., Fearon and Laitin 2003). We also include several socioeconomic measures to cover local grievances: local unemployment (PODES), presence of natural resources (PODES), the poverty rate (Tajima 2013), interpersonal inequality based on educational attainment (Tajima 2013), and the degree of inequality between ethnic and religious groups, also measured via educational attainment (Tajima 2013). Furthermore, we control for the political influence of the former ruling party Golkar and the former major opposition party Indonesian Democratic Party - Struggle (PDI-P) at the local level via binary variables that indicate whether either party received the most votes in the village during the last legislative elections (PODES), in case political dynamics influence the relationship between religious institutionalization and communal violence.9 Tajima (2013) provides strong evidence that a prior reliance on state security institutions played an important role for communal violence after the authoritarian breakdown. We use his measure of the distance to the nearest police post to capture how strongly villages depended on formal security institutions to regulate conflict. We include this variable to distinguish the potential effects of local religious institutions from his argument.

As we focus on the effects of religious institutions rather than the religiosity or religious composition of the population itself, we control for the level of ethnic and religious fractionalization of the village population, measured via standard Herfindahl fractionalization indices (Tajima 2013).10 The expectation is generally that localities with higher levels of ethnic or religious fractionaliza-tion are more likely to host groups with intergroup grievances and are thus more likely to experience violence. To single out the potential benefits of religious institutions, we also construct a measure of local religious institutional fractionalization11 and polarization, based on the number of different types of houses of worship.12 Polarization, as opposed to fractionalization, measures the degree to which the distribution of religious institutions resembles two equally sized groups. Extreme polarization is thought to increase violence, because collective action is more likely for two equally sized, antagonistic groups (Montalvo and Reynal-Querol 2005). Summary statistics for all variables are shown in the supplementary online appendix (http://prq.sagepub.com/supplemental/).

For our main analysis, we estimate a standard logit model with standard errors clustered at the district level. We also estimate conditional logit models that include fixed effects at the district level and remove any remaining unobserved factors shared across villages within the same district.

Main Results

We now turn to our main results. Table 1 presents the estimated coefficients and standard errors for our main logit and conditional logit models, using the mass fighting and subtype indicators as our dependent variables.

The first two columns show clear support for our main hypothesis: villages with higher levels of religious insti-tutionalization have a lower probability of having experienced a mass fighting incident. The coefficient is statistically significant below the 0.1 and 5 percent levels in the logit and conditional logit models, respectively. This effect is obtained in the presence of a number of important confounding factors. Of the control variables, we find similar results as prior studies; the strongest results are for the unemployment rate and the distance to the nearest police post. It is important to point out that the conditional logit models control for unobserved factors at the district level that might affect the level of local religious institutions and communal violence.

The other columns display results for the different forms of communal violence classified by the PODES census. For communal fighting, we again find negative and highly statistically significant results for both the logit and conditional logit models. For clashes between ethnic groups and with security forces, we still find a

Table 1. The Effect of Religious Institutionalization on Mass Fighting.

Mass Communal Ethnic Security

Logit Fixed Effects-logit Logit Fixed Effects-logit Logit Fixed Effects-logit Logit Fixed Effects-logit

Log(Population 0.0137 0.0219 0.0443 0.0680* -0.250** -0.0628 -0.0786 -0.181*

density) (0.0266) (0.0213) (0.0324) (0.0264) (0.0909) (0.0889) (0.07) (0.0727)

Poverty rate 0.00843 -0.169 0.17 -0.211 0.573^ -0.539 -0.166 -0.0541

(0.185) (0.109) (0.194) (0.138) (0.31) (0.475) (0.35) (0.435)

Natural resources 0.177 0.129 0.313 0.15 0.569 -0.114 0.0601 0.224

(0.337) (0.293) (0.38) (0.34) (0.546) (0.89) (1.043) (1.095)

Urban 0.0617 -0.00518 0.0402 0.000696 0.146 -0.109 0.152 -0.304

(0.0966) (0.0638) (0.107) (0.0745) (0.317) (0.376) (0.302) (0.273)

Golkar party -0.38 0.00161 0.259* 0.0308 -0.147 -0.133 _1 49|*** -0.522*

(0.24) (0.0635) (0.127) (0.0766) (0.314) (0.349) (0.412) (0.265)

PDI-P -0.423 -0.0754 0.284* -0.0664 -0.299 -0.37 -0.9901 -0.205

(0.293) (0.0705) (0.132) (0.0874) (0.33) (0.363) (0.589) (0.273)

Vertical inequality 0.183 1.243*** 1.668* 1.843*** 0.638 1.527 -5.654 -2.839

(0.974) (0.346) (0.686) (0.396) (1.385) (1.653) (4.003) (1.765)

Ethnic inequality -1.107* -0.412 -0.544 -0.334 -1.775 -0.23 0.513 1.228

(0.483) (0.381) (0.428) (0.459) (1.17) (1.416) (1.346) (1.46)

Religious inequality -0.196 -0.467 -0.605 -0.394 1.559 0.299 -0.719 -0.849

(0.678) (0.457) (0.598) (0.534) (1.399) (1.419) (1.943) (2.03)

Ethnic fractionalization 0.507* 0.216 -0.13 -0.0268 2.410*** 1.022 1.097* 0.562

(0.219) (0.143) (0.215) (0.179) (0.54) (0.622) (0.455) (0.5)

Religious -0.895* -0.235 -0.0972 -0.108 0.106 0.488 -2.305* -0.35

fractionalization (0.383) (0.249) (0.345) (0.297) (1.022) (0.891) (1.068) (1.017)

Log(Population) -0.00832 0.553*** 0.283*** 0.546*** 0.317 0.750*** -0.122 0.713***

(0.135) (0.0394) (0.0632) (0.0483) (0.213) (0.175) (0.2) (0.134)

Log(Distance district -0.0118 -0.0676** 0.0691 -0.0356 0.0479 -0.136 -0.0427 0.0856

capital) (0.0757) (0.0239) (0.0522) (0.0302) (0.123) (0.0903) (0.162) (0.0889)

Unemployment 1.676*** 1.368*** 1.311** 1.202*** 3.293*** 3.234** 2.492** 2.184*

(0.445) (0.292) (0.416) (0.353) (0.847) (1.1) (0.836) (0.974)

Distance police -0.00628* -0.00785*** -0.009841 -0.0132*** 0.00495 0.00624 0.00895 0.00811 ^

(0.00313) (0.00195) (0.00564) (0.00281) (0.00678) (0.00632) (0.0067) (0.00468)

Religious institutional -0.381 -0.403 -0.235 -0.477 0.328 -0.64 1.23 1.756

polarization (0.367) (0.337) (0.385) (0.407) (1.269) (1.166) (1.386) (1.475)

Religious institutional 1.344* 0.619 0.779 0.531 0.485 0.619 -0.696 -2.472

fractionalization (0.669) (0.601) (0.691) (0.726) (2.27) (2.036) (2.536) (2.75)

Log(Religious -0.518*** -0.101* -0.397*** -0.09421 -0.645** 0.104 -0.600*** -0.0389

institutional density) (0.0804) (0.04) (0.0562) (0.0487) (0.207) (0.192) (0.168) (0.146)

Constant -5.421*** -8.504*** -13.42*** -6.723***

(1.237) (0.582) (1.31) (1.911)

Observations 53,466 51,322 53,466 48,958 53,466 12,914 53,466 21,186

AIC 23,947.4 19,314.2 16,820 13,872.2 1,574.2 992.4 2,568.2 1,708.2

BIC 24,116.3 19,473.5 16,988.9 14,030.5 1,743 1,126.8 2,737 1,851.5

Clustered standard errors in parentheses. AIC = Akaike information criterion. BIC = Bayesian information criterion. +p < .10. *p < .05. **p < .01. ***p < .001.

negative effect in the logit estimations, but the coefficient for religious institutionalization loses statistical significance in the conditional logit models. The somewhat weaker findings for the conditional logit models are likely due to the much smaller number of cases, as districts with no violence are automatically dropped from

the analysis, or because ethnicity correlates more strongly with religion than other cleavages.

Our results indicate that religious institutions are a statistically relevant factor in explaining variation in communal violence across Indonesian villages, but we have not yet shown the substantive importance. Figure 1 plots

-8 -6 -4 -2

Religious Institutional Density

Figure 1. Probability of mass fighting as a function of religious institutional density.

the simulated effect of the density of religious institutions on the probability of a mass fighting incident and the associated 95 percent confidence interval.13

The graph shows the substantive importance of religious institutions for peace. Moving from the sample minimum to the maximum, the probability of violence drops from approximately 30 to 1-2 percent. Importantly, we obtain this effect while controlling for a number of existing alternative explanations. This fairly large effect implies an important new insight for understanding communal violence in Indonesia and the role of religious institutions in violence more generally: religious institutionalization reduces communal violence.

Robustness Checks

We implement a number of robustness checks to analyze the sensitivity of our main findings. For example, as an alternative to standard logit, we also estimate rare-events logistic models (King and Zeng 2001) without affecting our main finding (see Section 2 of the supplementary online appendix).14 We also add additional observable covariates that might operate as omitted variables. We include variables that capture the presence of customary adat institutions, the existence of a village representative board, and a neighborhood association. Including any of these variables has no effect on our findings about religious institutions (see Section 3 of the supplementary online appendix). Furthermore, we estimate logit and conditional logit models with fixed effects at the subdistrict level (see Section 5 of the supplementary online appendix). Moving from fixed effects at the district to the subdistrict level controls for any unobservable characteristics shared between villages in the same neighborhood.

Across both models, we still find that religious institu-tionalization has a statistically significant and negative effect on communal violence.

We also substitute our main measure with the logged number of religious schools, as an alternative approach. This variable equally shows the pacifying effects of religious institutions on communal violence (see Section 4 of the supplementary online appendix). Next, we disaggregate different religious traditions, calculating the religious institutionalization measure for each of the major religions separately. We still find a consistently negative and statistically significant effect of religious institutions across religious traditions. The effect is most pronounced for Islamic places of worship, but this is simply because Islam is the dominant religion in Indonesia and has the most houses of worship (see Section 9 of the supplementary online appendix).

There might also be a problem of reverse causality, as in many ethnoreligious clashes in Indonesia, places of worship were targets of violence and were destroyed. If higher levels of violence reduce the number of places of worship, the negative relationship in our results might be due to reverse causality. We explore three different strategies to address any remaining endogeneity concerns.

First, we add information on the number of places of worship from the 2000 PODES, which took place before the large spike in communal violence in the early 2000s.15 We reestimate our three main models using the level of religious institutionalization pre-2000 and can strongly confirm our initial findings.

Second, we exploit information on local forms of violence from a different data source. The SNPKI provides location identifiers down to the subdistrict level and allows us to calculate the total number of local violent

events for different time periods. While not as geographically disaggregated as our village-level data, these data give us the opportunity to utilize a different measurement of local violence and more directly exploit overtime variation. Because our PODES data provide information on religious institutionalization from 1999 and 2002, we calculate the total number of violent events in a subdistrict for 2000-2003 and 2003-2005. Moreover, these data also allow us to measure levels of violence in the pre-2000 period. We aggregate our local religious institutionaliza-tion measure to the subdistrict level and match it to the observations included in the sample of the SNPKI database. The SNPKI database only covers thirteen of Indonesia's thirty-three provinces, with varying time coverage. Hence, the number of observations is drastically smaller than our village-level data set. We estimate models with district fixed effects and the logged number of total events as a dependent variable. The district fixed effects control for a number of confounding factors, and we also include the level of pre-2000 violence for each subdistrict. Again, we can confirm the pacifying effects of religious institutionalization (see Section 6 of the supplementary online appendix).

Finally, in the supplementary online appendix, we explore an instrumental variable strategy that exploits geographic determinants of the spread of religion across the Indonesian archipelago. Using 2-Stages-Least-Squares IV regressions, we also find that religious institu-tionalization has a clear negative and statistically significant effect, although we cannot be certain that our set of instruments fulfills the exclusion restriction (see Section 7 of the supplementary online appendix).

Testing Additional Theoretical Implications

We also briefly test several other observable implications of our theoretical argument: the role of norms, various interaction models, and effects on conflict resolution. The objective of these additional analyses is to lend further support to our argument that the correlations reported above can be plausibly attributed to religious institution-alization and the resulting horizontal and vertical networks within religious communities.

An alternative explanation for our finding might be that institutionalized religions can act as capable partners of the state. Lower levels of violence might be caused by religious elites collaborating with the security sector to enforce nonviolence norms. To test for this possibility, we interact our main variable with the distance to the district capital and the nearest police post, as presumably any collaboration between religious institutions and the state is facilitated by geographic proximity. We find no evidence to that effect.

We also investigate a number of interactions to see if the results are consistent with our proposed mechanism.

We start by interacting our institutional density measure with variables capturing village size (e.g., population counts and area). If the mechanism relates to facilitating within-religion access, then the effect should be larger in smaller villages. This is exactly what we find. The interaction with population size is positive (i.e., it weakens the pacifying effect) in both the logit and fixed-effects logit models, reaching statistical significance at the 1 percent level in the latter. The interaction term with area also has the correct sign but just misses statistical significance at the 10 percent level. We also interact our measure with the demographic ethnic and religious fractionalization scores, as well as institutional fractionalization and polarization. Because we argue that religious institutionaliza-tion should have a pacifying effect that is independent of demographic constellations, we would expect to find no strong interactions across village types. Again, this is exactly what we find.16

The SNPKI data also allow us to identify whether religious actors were involved in any violent event. Only using events with religious actor participation as the dependent variable reveals that religious institutionaliza-tion in this case has no significant effect and even has a negative sign. This supports the idea that religious institutions can (and often do) mitigate various conflicts but not necessarily conflicts that are already religiously charged.

To further investigate whether our findings are in line with our theoretical argument, we use other survey items from the PODES. For more than 4,000 villages that experienced violent clashes, the census also asks whether the conflict was eventually resolved peacefully and by what type of actor. Table 2 shows estimates of a logit model that uses the same control variables and our religious institutionalization measure.

We find that for villages that experienced conflict, the ones with higher levels of religious institutionalization were more likely to solve their conflict peacefully (statistically significant below the 5% level, Column 1) and to do so via village-level institutions (statistically significant below the 5% level) rather than through the security apparatus (Column 2). This offers another important layer of evidence for our overall finding that (1) religious institutionalization (of any major religion) has played a major role in limiting communal violence in Indonesia, and (2) religious institutionalization has supported peaceful conflict resolution in villages that have experienced conflict.

We also estimate the effect of religious institutional-ization on other out-of-norm behavior that underlies similar sanctions through religious teachings. In particular, we exploit information from the PODES census on the presence of prostitution, rape, and drug usage in the village. Overall, we find mixed results. While the effect of religious institutionalization is negative for drug usage, the coefficient is not significant for rape and positive for

Table 2. Effect of Religious Institutionalization on Peaceful Conflict Resolution.

Peaceful Logit Peace by village Logit

Log(Population density) 0.00905 0.0220

(0.0713) (0.0491)

Poverty rate 0.409 -0.116

(0.354) (0.350)

Urban 0.487* -0.0696

(0.219) (0.176)

Golkar party 1.518*** -0.278

(0.312) (0.231)

PDI-P 1.447*** 0.138

(0.401) (0.275)

Vertical inequality 0.155 -2.039

(1.038) (1.496)

Ethnic inequality 1.481 -1.397

(1.514) (1.062)

Religious inequality -1.195 -0.0944

(2.337) (1.849)

Ethnic fractionalization -0.332 0.0737

(0.435) (0.352)

Religious fractionalization 1.131 0.639

(0.747) (0.610)

Log(Population) 0.821*** -0.129

(0.128) (0.156)

Log(Distance district capital) 0.0350 -0.0625

(0.105) (0.0797)

Unemployment 0.675 0.368

(1.346) (0.742)

Distance police 0.00783 0.00838

(0.0108) (0.00570)

Religious institutional polarization 0.270 -1.114

(0.930) (0.766)

Religious institutional -1.136 1.234

fractionalization (1.618) (1.353)

Log(Religious institutional density) 0.426* 0.212*

(0.174) (0.0969)

Observations 3,311 2,693

AIC 2,624.9 2,267.8

BIC 2,734.8 2,373.9

Clustered standard errors in parentheses. AIC = Akaike information criterion. BIC = Bayesian information criterion. +p < .10. *p < .05. **p < .01. ***p < .001.

prostitution (see Section 8 of the supplementary online appendix). We believe the absence of a consistent negative effect might be due to weaker norms (and enforcement) with respect to out-of-norm behavior below the threshold of communal violence or measurement bias in the survey—the village census is likely to underreport such activities in comparison with more large-scale incidents.

Discussion and Conclusion

Our research is a first step toward understanding the institutional dimension of religion in communal violence. Our findings challenge the existing conceptualization of "religion" as a mere identity marker or theological idea in quantitative studies. Furthermore, our results call into question the dominant explanation in the literature of religion and peace. Some argue that every religion is generally peaceful, and that any nonviolent action by religious actors is a result of nonviolence norms. We find no empirical evidence of such an explanation and instead show that the level of religious institutionalization explains the impact of religious actors' peace efforts.

While we have examined the impact of local religious institutions in the context of Indonesia, there is reason to believe that similar mechanisms are relevant in other countries. First, as noted in section "The Peace Impact of Local Religious Institutions," religion and religious actors do not constitute the main loci of large- and small-scale conflicts. Hence, the pacifying effects of religious institutionalization are likely to be relevant in a large set of contexts of violence. Second, while our empirical study only analyzes Indonesia, the country's enormous size and diversity strengthens the external validity of our findings. Studies from other contexts such as Nepal and Sri Lanka (section "The Peace Impact of Local Religious Institutions") provide auxiliary support for our findings. In a comparative study of religious institutions in Ghana and Côte d'Ivoire, MacLean (2004) shows that local religious institutions in Ghana had a more strongly pacifying impact on communities, as compared with the Côte d'Ivoire. Members of local religious institutions in Ghana met regularly in subgroups, such as choirs or women organizations. These regular contacts between different families, classes, and genders established strong social linkages within the same local religious community and "made local politics, including political party organization, more active, salient, and openly heterogeneous" (MacLean 2004, 605). The importance of the local community and the leading role of the local religious institutions, as MacLean argues, is one reason explaining why possible ethnic conflicts in Ghana were often prevented at an early stage.

While these examples support the relevance of our finding beyond the Indonesian context, it is equally important to acknowledge possible limitations. Indonesia's ethnic and religious landscape, overall frac-tionalized but often locally relatively homogeneous, might allow for different effects of religious institutional-ization as compared with a decidedly polarized context with overlapping and reenforcing cleavages.

We conclude by discussing some limitations of our argument. An influential literature on communal violence deals with the potential conflict-escalating role of religious

elites. Two prominent cases of interreligious violence in Indonesia, Maluku and Sulawesi, also illustrate that religious institutions sometimes fail to mediate (or even contribute to) communal violence. We believe this to be the case for two reasons. First, the peace impact of religious institutions possibly declines once a manifest interreligious conflict begins. This is also supported by our additional robustness checks that show that the pacifying effect of religious institutions is absent for conflicts in which religious actors were involved. The interreligious conflict in both provinces has lasted for several years and has been integrated into the belief systems of the various religious groups. This has resulted in harsh in-/out-group differences between local religious groups. Moreover, effective policing by local religious institutions was inhibited by the social distance between the religious groups and the below-average levels of religious institutionalization. It would, therefore, be interesting to look more closely at the degree and role of religious institutionalization in Maluku and Sulawesi before the outbreak of violence to understand the historical evolution leading up to the conflicts.

Second, some of the main violent actors did not come from the local villages in Maluku and Sulawesi. Religious institutions are likely limited in their ability to substantively influence outsiders, who—similar to recent migrants—are not as heavily integrated into the local religious institutions and are therefore less likely to be affected by horizontal and vertical linkages. Future analyses should further examine the interplay of outsiders and local religious institutions, and more generally, under which conditions religious elites get drawn into an escalating role for conflict.

Finally, while this paper focuses on the role of religious institutionalization, it is worth asking whether the same pacifying effect can be generated by other, nonreligious institutions with similar characteristics. We believe this to be an important future test of our argument. If other nonstate institutions that strengthen horizontal and vertical contacts among members exert a pacifying effect on conflicts, this would provide further evidence for institutional rather than norm-specific mechanisms. Such institutions would need to be sufficiently inclusive to act as an integrative force for a variety of conflict issues (e.g., water or school committees would be too narrowly defined) while being sufficiently salient in communities' daily life to ensure strong-enough horizontal and vertical ties. From such a perspective, traditional ethnic and tribal institutions as well as local nongovernmental organizations (NGOs) may constitute promising subjects of further studies along the lines of the one presented in this paper.

Acknowledgment

We are grateful to Jeremy Menchik, Jana Krause, Matthias Basedau, Carlo Koos, Annegret Mahler, members of German Institute of Global and Area Studies' (GIGA)

Forschungsschwerpunkt 2 (FSP-2) Research Program, three anonymous reviewers, and Jason Maloy for helpful comments and suggestions.

Authors' Note

Replication data for this article can be viewed at https://janpierskalla.wordpress.com

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

1. We define all forms of nonstate violent incidents (i.e., religious or ethnic violence, riots, mass brawls) as "communal violence." This is the most practical approach due to our main theoretical argument and dependent variable (see below). According to this definition, ethnic violence is a subcategory of "communal violence."

2. The cases ofNigeria and Pakistan are only used for descriptive purposes; our main analysis focuses on Indonesia.

3. http://www.snpk-indonesia.com/.

4. As long as ethnic and religious identity do not perfectly correlate.

5. A higher degree of religious institutionalization might also increase competition between religious institutions, which might in turn lead to violence between religious groups. This would counteract the pacifying effects of religious institutionalization. We test for this possibility in our robustness checks by disaggregating religious institution-alization by denomination.

6. The questionnaire is fielded by enumerators of the statistical office at the subdistrict level and relies on information from the village head.

7. The Pendataan Potensi Desa (PODES) questionnaire does not provide information on the number of incidents in each village.

8. Tajima (2013) implements a very insightful analysis of communal violence in Indonesia that also draws on PODES data. He also provides important information on population characteristics from the population census. We use his replication data and match this information to our data set. As the two data sets lack a common village-level identifier, we use information on the district ID, subdistrict ID, population size, altitude, and distance to the nearest police post to match villages.

9. No major legislative elections took place during the PODES survey. The last national legislative elections, as well as elections to district-level representative bodies, took place in 1999. Some villages may have had elections for village head positions, but we are unable to identify those in our data.

10. The fractionalization index is defined as 1 — £where s2 is the population share of group i.

11. Again, calculated as a Herfindahl fractionalization index. Here, instead of using the population share for each religion, we use the share of houses of worship of the total

number as s.

12. The polarization measure draws on work by Esteban and Ray (1994). Polarization is defined as 4X si (1 - si).

13. For the simulation, all control variables were set to their respective means, medians, or modes. We used the CLARIFY routine in STATA to obtain our estimates (Tomz, Wittenberg, and King 2003).

14. Note that "rare" in the context of maximum likelihood logit estimation refers to the absolute counts of each outcome type. As in our sample 4,821 villages out of more than 60,000 experienced a mass fighting incident, violence is somewhat rare in relative terms, but this is not problematic for a logit model.

15. Combining information across different waves of the PODES is not straightforward, as the number of villages, subdistricts, and districts changes. Matching on ID codes produces about 75 percent correct matches. Our results are substantively identical when we use the whole sample (with some mismatched villages) and when we use correctly matched villages.

16. While the interaction with ethnic fractionalization in the village is positive and significant, simulations reveal that any difference is substantively meaningless. For example, the effect of increasing religious institutionalization from the 25th to the 75th percentile at low and high levels of ethnic fractionalization is a reduction in probability of 0.026 and 0.0256 percentage points, respectively.

References

Acciaioli, Greg. 2001. "Grounds of Conflict, Idioms of Harmony: Custom, Religion, and Nationalism in Violence Avoidance at the Lindu Plain, Central Sulawesi." Indonesia 72:81-114.

Akbaba, Yasemin, and Zeynep Taydas. 2011. "Does Religious Discrimination Promote Dissent? A Quantitative Analysis." Ethnopolitics 10 (3-4): 271-95.

Ammerman, Nancy T. 2003. "Religious Identities and Religious Institutions." In Handbook of the Sociology of Religion, edited by Michele Dillon. Cambridge: Cambridge University Press. Pp.207-224.

Appleby, Scott. 2001. "Religion as an Agent of Conflict Transformation and Peacebuilding." In Turbulent Peace: The Challenges of Managing International Conflict, edited by Chester Crocker, Osler Hampson, and Pamela Aall. Washington, DC: United States Institute of Peace. Pp. 821840.

Barron, Patrick, Kai Kaiser, and Menno Pradhan. 2009. "Understanding Variations in Local Conflict: Evidence from Indonesia." World Development 37 (3): 698-713.

Basedau, Matthias, Birte Pfeiffer, and Johannes Vullers. 2014. "Bad Religion? Religion, Collective Action, and the Onset of Armed Conflict in Developing Countries." Journal of Conflict Resolution. Published electronically July 23. doi:10.1177/0022002714541853.

Bercovitch, Jacob, and Ayse S. Kadayifci-Orellana. 2009. "Religion and Mediation: The Role of Faith-Based Actors in International Conflict Resolution." International Negotiation 14:175-204.

Bertrand, Jacques. 2004. Nationalism and Ethnic Conflict in Indonesia. Cambridge: Cambridge University Press.

Braithwaite, John, Valerie Braithwaite, Michael Cookson, and Leah Dunn. 2010. Anomie and Violence: Non-truth and Reconciliation in Indonesian Peacebuilding. Canberra: Australian National University Press.

Brass, Paul R. 2003. The Production of Hindu-Muslim Violence in Contemporary India. Seattle: University of Washington Press.

Bueno de Mesquita, Ethan C. Christine Fair, Jenna Jordan, Rasul Bakhsh Rais, and Jacob N. Shapiro. 2013. The BFRS Political Violence in Pakistan Dataset. https://esoc.princ-eton.edu/files/bfrs-political-violence-pakistan-dataset

Centre for Policy Alternatives. 2003. Informal Dispute Resolution in the North East and Puttalam. Colombo, Sri Lanka: Centre for Policy Alternatives.

Dahal, Dev Raj, and Chandra Dev Bhatta. 2008. The Relevance of Local Conflict Resolution Mechanisms for Systemic Conflict Transformation in Nepal. Berghof Foundation.

De Juan, Alexander. 2009. "A Pact with the Devil? Elite Alliances as Bases of Violent Religious Conflicts." Studies in Conflict & Terrorism 31 (12): 1120-35.

Duncan, Christopher R. 2009. "Reconciliation and Revitalization: The Resurgence of Tradition in Postconflict Tobelo, North Maluku, Eastern Indonesia." The Journal of Asian Studies 6804:1077-104.

Ellingsen, Tanja. 2000. "Colorful Community or Ethnic Witches' Brew? Multiethnicity and Domestic Conflict During and After the Cold War." Journal of Conflict Resolution 44 (2): 228-49.

Esteban, Joan, and Debraj Ray. 1994. "On the Measurement of Polarization." Econometrica 62 (4): 819-51.

Fearon, James, and David Laitin. 1996. "Explaining Interethnic Cooperation." American Political Science Review 90 (4): 715-35.

Fearon, James, and David Laitin. 2003. "Ethnicity, Insurgency, and Civil War." American Political Science Review 97 (1): 75-90.

Harpviken, Kristian B., and Hanne E. R0islien. 2008. "Faithful Brokers? Potentials and Pitfalls of Religion in Peacemaking." Conflict Resolution Quarterly 25 (3): 35173.

Horowitz, Donald. 2001. The Deadly Ethnic Riot. Berkeley: University of California Press.

Human Rights Watch. 2013. In Religion's Name: Abuses against Religious Minorities in Indonesia. Human Rights Watch.

International Crisis Group. 2000. "Indonesia: Overcoming Murder and Chaos in Maluku." Jakarta/Brussels: ICG Asia Report No. 10.

International Crisis Group. 2001. "Communal Violence in Indonesia: Lessons from Kalimatan." Jakarta/Brussels: ICG Asia Report No. 19.

International Crisis Group. 2002. "Indonesia: The Search for Peace in Maluku." Jakarta/Brussels: ICG Asia Report No. 31.

Kim, Hyung-Jun. 1998. "The Changing Interpretation of Religious Freedom in Indonesia." Journal of Southeast Asian Studies 29 (2): 357-73.

King, Gary, and Langche Zeng. 2001. "Logistic Regression in Rare Events Data." Political Analysis 9 (2): 137-63.

Kingsley, Jeremy. 2011. Democracy, Conflict and Islamic Leadership in Eastern Indonesia: A Village Election Case Study. Melbourne: University of Melbourne.

MacLean, Lauren M. 2004. "Mediating Ethnic Conflict at the Grassroots: The Role of Local Associational Life in Shaping Political Values in Côte d'Ivoire and Ghana." Journal of Modern African Studies 42 (4): 589-617.

Montalvo, José G., and Marta Reynal-Querol. 2005. "Ethnic Polarization, Potential Conflict and Civil War." American Economic Review 95 (3): 796-816.

Pickering, Paula M. 2006. "Generating Social Capital for Bridging Ethnic Divisions in the Balkans: Case Studies of Two Bosniak Cities." Ethnic and Racial Studies 29 (1): 79-103.

Powers, Gerard. 2010. "Religion and Peacebuilding." In Strategies of Peace: Transforming Conflict in a Violent World, edited by Daniel Philpott and Gerard Powers. Oxford: Oxford University Press. Pp.317-352.

Putnam, Robert D. 2000. Bowling Alone: The Collapse and Revival of American Community. New York: Simon & Schuster.

Rutherford, Danilyn. 2006. "Nationalism and Millenarianism in West Papua: Institutional Power, Interpretive Practice, and the Pursuit of Christian Truth." In The Limits of Meaning: Case Studies in the Anthropology of Christianity, edited by Matthew Engelke and Matt Tomlinsonpp. New York: Berghahn Books. Pp.105-128.

Schiller, Anne. 1996. "An 'Old' Religion in 'New Order' Indonesia: Notes on Ethnicity and Religious Affiliation." Sociology of Religion 574:409-17.

Schröter, Susanne. 2003. "Religiöse Mobilisierungen in Indonesien." English title: "Religious Mobilization in Indonesia" ASIEN 89:26-46.

Sisk, Timothy, ed. 2012. Between Terror and Tolerance: Religious Leaders, Conflict, and Peacemaking. Washington, DC: Georgetown University Press.

Svensson, Isak. 2007. "Fighting with Faith: Religion and Conflict Resolution in Civil Wars." Journal of Conflict Resolution 51 (6): 930-49.

Tajima, Yuhki. 2013. "The Institutional Basis of Intercommunal Order: Evidence from Indonesia's Democratic Transition." American Journal of Political Science 57 (1): 104-19.

Ter Haar, Gerrie. 2005. "Religion: Source of Conflict or Resource for Peace?" In Bridge or Barrier. Religion, Violence, and Visions for Peace, edited by Gerrie Ter Harr and James J. Busuttil. Leiden, The Netherlands: Brill. Pp. 3-43.

Toft, Monica D. 2007. "Getting Religion? The Puzzling Case of Islam and Civil War." International Security 31 (4): 97-131.

Toft, Monica D., Daniel Philpott, and Timothy S. Shah. 2011. God's Century. Resurgent Religion and Global Politics. New York: W.W. Norton.

Tomz, Michael, Jason Wittenberg, and Gary King. 2003. "CLARIFY: Software for Interpreting and Presenting Statistical Results." Journal of Statistical Software 8 (1): 1-30.

Traunmüller, Richard, and Markus Freitag. 2011. "State Support of Religion. Making or Breaking Faith-Based Social Capital?" Comparative Politics 43 (3): 253-69.

Turner, Kathleen. 2003. "Myths and Moral Authority in Maluku: The Case of Ambon." Asian Ethnicity 42:241-63.

Van Klinken, Gerry. 2001. "The Maluku Wars: Bringing Society Back In." Indonesia 71:1-26.

Van Klinken, Gerry. 2007. Communal Violence and Democratization in Indonesia: Small Town Wars. London: Routledge.

Varshney, Ashutosh. 2001. "Ethnic Conflict and Civil Society: India and Beyond." World Politics 53 (3): 374-98.

Varshney, Ashutosh. 2002. Ethnic Conflict and Civic Life: Hindus and Muslims in India. New Haven: Yale University Press.

Varshney, Ashutosh, Mohammad Z. Tadjoeddin, and Rizal Panggabean. 2008. Creating Datasets in Information-Poor Environments: Patterns of Collective Violence in Indonesia, 1990-2003. Boulder: Lynne Rienner.

Wald, Kenneth D., Adam L. Silverman, and Kevin Fridy. 2005. "Making Sense of Religion in Political Life." Annual Review of Political Science 8:121-43.

Wilkinson, Steven. 2004. Votes and Violence: Electoral Competition and Ethnic Riots in India. Cambridge: Cambridge University Press.

Wilson, Chris. 2008. Ethno-religious Violence in Indonesia: From Soil to God. Abingdon, UK: Routledge.

World Bank. 2004. Village Justice in Indonesia: Case Studies on Access to Justice, Village Democracy and Governance.The World Bank Indonesia, Jakarta.

World Bank. 2008. Forging the Middle Ground: Engaging Nonstate Justice in Indonesia. The World Bank Indonesia, Jakarta.

Copyright of Political Research Quarterly is the property of Sage Publications Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.