Introducing the “Religious Minorities at Risk” Dataset

1 Introduction

In recent years, religion has become increasingly contentious. Violence with religious overtones has been increasing (e.g. Walter, 2017) and so has the discrimination of religious minorities (e.g. Fox, 2016). But how do religious minorities react to discrimination and marginalization? Does marginalization explain grievances that then generate (violent) mobilization of religious minorities? And why do some religious groups protest violently, while others remain peaceful?

Unlike for ethnic groups (Cederman, Wimmer, & Min, 2010), previous research cannot find a straightforward answer to these questions and is unable to establish a causal chain running from deprivation to conflict (e.g. Basedau, Fox, Pierskalla, Strüver, & Vüllers, 2017; Fox, Bader, & McClure, 2017). There are theoretical and methodological explanations for this failure. In theoretical terms, we need to distinguish between violent and nonviolent forms of mobilization. Recent studies have focussed exclusively on violence. Moreover, deprivation and grievances might vary according to intensity and form. Existing studies looked solely at religious discrimination but not at deprivation and grievances in political and economic terms. Methodologically, we lacked the necessary fine-grained data on religious minorities’ level of deprivation, on their grievances, on their capacities as a group and on their engagement in violent and nonviolent forms of mobilization.

This paper introduces a new dataset that may fill some of these gaps, specifically by providing pertinent worldwide information on relevant characteristics of these minorities and why they may mobilize.

2 The “religious minorities at risk” dataset

We compiled data for 771 religious minorities worldwide for the period between 2000 and 2014. Following the Religion and State Project (RAS)^[1], a religious minority is defined as “a religious community that either forms the numerical minority in a given country or is politically or otherwise marginalized.” We used the list of the RAS-Minority data set (RASM) as a starting point for data collection.^[2]

The dataset includes detailed information on a wide range of variables relevant to study the causal chain running from minorities’ deprivation to their mobilization. We first created (dependent) variables that capture minorities’ mobilization, with and without the use of physical violence, in several ways. We started by rhetoric in terms of calls for mobilization and then counted the frequency and size of events in terms of participation. We also generated variables that capture the motivation for the mobilization as stated by the actors themselves. For violence, we also included the intensity, in terms of lethality, and distinguished between several forms of violence: armed conflict, communal conflict, terrorism and “spontaneous” riots.

A range of other variables gather potential causes of mobilization and they are inspired by a theoretical framework that expects actors to mobilize according to their motivation and capacity to do so (see Huber & Basedau, 2018). Regarding motive we look at objective deprivation^[3] in terms of economic inequality and the exclusion from political representation. Religious discrimination was gathered by the Israeli part of the project (see Fox, 2018). We were keen to collect fine-grained data on subjective grievances – as grievances require a nuanced conceptualization. We distinguish between religious, political and economic grievances and document how they vary in terms of intensity. Past mobilization, before the period under investigation, may also show intrinsic motivations of the groups in the long run. In addition, we have data on whether groups were victims of violence by the state or other groups.

For religious minorities’ capacity and opportunity, we first look at structural variables like the overlap of the religious identity with ethnic differences as well as the regional concentration of the settlements. Both should increase the ability to mobilize, as should the size of the group relative to the total population. We searched for organizational features and compiled information on the existence and number of organizations that represent the religious minorities. As the literature theorizes that competition between organizations and elites can increase escalation through outbidding (De Juan, 2009; Toft, 2007), we compiled data on rivalries between religious groups and organizations.

Generally, we tried to conform to best practices in the collection of data (Davenport & Moore, 2015; Salehyan, 2015), which, however, need to be balanced with practical concerns. Regarding sources, we defined a number of prime sources such as the International Religious Freedom Reports and the Religion and State Project. In addition, we used secondary literature, if available, and defined several key words for a systematic Google search.^[4] Generally, we followed the principle of “positive coding”, which means that we relied on indications in the respective sources that something was present. As a principle, most variables are formulated as “are there reports…?”, which means that the coding of a zero does not necessarily mean that the respective phenomenon is not present.^[5]

In order to control for differences in data availability and the quality of sources, we introduced precision codes to indicate whether we assume information (in general) to be scarce for the religious minority or the country at hand. The precision codes, which can be valuable for the process of supervision and data cleaning, also indicate several reasons for any lack of precision, such as “indirect conclusion” or “conflicting information in sources”. After finishing the raw coding, we engaged in a round of data cleaning and additional coding, where we used the precision codes to identify some aspects that require fixing or refinement. Coding was performed by several coders and we are very indebted to them.^[6] All codings were put down in code sheets and very thoroughly inspected by the supervisors.

3 Descriptive results

Our dataset provides plentiful descriptive information that can be analyzed over time as well as across the different world regions and religious denominations. In the following, we summarize the most striking findings according to two clusters: (i) Religious minorities’ engagement in peaceful and violent forms of mobilization and (ii) grievances that may capture minorities’ motivation to mobilize.

Both peaceful and violent mobilization among religious minorities increased significantly during the period under investigation, with a somewhat stronger increase in peaceful mobilization since 2010 (see Figure 1 in Appendix). Interestingly, while two-thirds of the reported nonviolent events triggered the participation of several thousand participants, over 75 percent of the violent events were rather small-scale involving the participation of less than 100 individuals. Moreover, we found that the intensity of violence increased over time and that most minorities resorting to the use of force engaged in lethal forms of violence (see Figure 2 in Appendix).

Analyzing religious minorities’ mobilization across the different world regions and religious denominations also provides interesting findings. Both nonviolent and violent mobilization among religious minorities increased in each part of the world during the period under investigation. While nonviolent mobilization was most prevalent among minorities residing in Western democracies and Asia (see Figure 3 in Appendix), violent mobilization occurred most frequently among minorities living in Sub-Saharan Africa and Asia (see Figure 4 in Appendix). Furthermore, the data reveals that mobilization, both peaceful and violent, was by far most prevalent among Muslim minorities than among minorities of other religions (see Figure 5 and Figure 6 in Appendix).

Our data reveals that grievances are widespread. However, they differ quite substantially according to form and intensity. Generally, grievances increased considerably during the period under investigation and over 55 percent of the religious minorities under investigation expressed grievances. Political and religious grievances were expressed more frequently than economic grievances (see Figure 7 in Appendix).

Dissecting religious minorities’ grievances across the world regions and religious denominations reveals interesting insights. Grievances increased over time in all world regions and across all religious denominations. They were most prevalent among religious minorities living in former Soviet Union countries and Western democracies (see Figure 8 in Appendix). In addition, over three quarters of the recorded grievances were among Christian and Muslim minorities (see Figure 9 in Appendix).

4 From deprivation to mobilization?

In order to demonstrate the utility of the dataset, we check whether our new data can confirm the causal chain from deprivation via grievances to mobilization of religious minorities, testing the two links in the chain separately. For the first set of regressions, we use grievances as the dependent variable. In the second set of regressions, grievances are included as an independent variable, with mobilization being the dependent variable. For both links, we use a pooled logit logistic regression model, pooling all group-year observations together, with robust standard errors clustered by minority group, to allow for dependence between the observations within these groups. To address unobserved heterogeneity, we also estimate an OLS model with minority- and year fixed effects, again clustering the standard errors at the minority group level.

For the first set of regressions, we employed dependent variables capturing the existence of subjective grievances. The variable any grievances, as well as the more specific forms for political (political_griev), economic (economic_griev) and religious (religious_griev) grievances, is constructed as binary (0/1). As independent variables, we use variables that reflect objective deprivation. In the political sphere, the dummy variable political exclusion takes value 1 if the minority was excluded from the government. Economic deprivation is captured by economic deprivation, which takes value 1 if members of the minority were on average poorer than the country’s average. Finally, religious discrimination ranges from 0 to 52 and indicates the level of religious discrimination the minority faced, including restrictions of religious practice such as prayer, places of worship, diet, dress, and education.

Table 1 (in Appendix) presents the estimated effects of objective deprivation on the likelihood that a minority group holds grievances. The first two columns report estimated coefficients when the dependent variable is any type of grievance (any grievances). We find statistically significant, positive effects of both religious discrimination and exclusion from the government in the pooled logit model. In the fixed effects model, only religious discrimination remains significant. The remaining columns report the coefficients we estimated when using the respective specific types of grievances as the dependent variable. The relations between the different types of deprivation and the expressed grievances are always positive, but their statistical significance depends on the model specification. Nevertheless, the estimations by and large confirm the causal link running from deprivation to grievances.

For the second set of regressions, the dependent variable is mobilization, which is a binary (0/1) variable, reflecting whether the minority mobilized in a given year. Table 2 in the Appendix shows that grievances positively impact the probability of a minority group mobilizing. The effect of any grievances is positive and significant, both in the pooled logit and in the fixed effects OLS model. Regressing on the different types of grievances separately, we find similarly strong results for political and economic grievances. The effect of religious grievances is only significant in the pooled logit model. We thus find strong support for the assumption that grievances – particularly political and economic ones – increase the likelihood that religious minorities mobilize. Combining our results, we find the most robust evidence for a full causal chain from political deprivation, via political grievances, to mobilization.

5 Conclusion

This paper introduced the new dataset “Religious Minorities at Risk” that provides detailed information on more than 700 religious minorities worldwide. This dataset makes a valuable contribution to existing research as it contains fine-grained information on religious minorities’ level of deprivation, on their grievances, on their capacities as a group and on their engagement in violent and nonviolent forms of mobilization. In doing so, the data enables scholars to study whether deprivation leads to grievances and then to mobilization. Moreover, it can be used to address a variety of other research questions regarding the difference between violent and peaceful mobilization, the consequences of protest or the rationale behind the different forms of violent mobilization such as armed conflict, riots or terrorism.