Measuring Subnational Variation in Freedom of Religion or Belief Violations: Reflections on a Path Forward

Abstract

In recent years, numerous cross-country data collection instruments measuring freedom of religious or belief (FoRB) violations have become available. These have provided important insights, not least of which being the steady increase in religious discrimination across the globe. At the same time, the country-level focus of extant datasets often obscures subnational variation and leaves open important questions about the mechanisms driving FoRB violations in particular contexts. Through this article, we seek to initiate a more systematic discussion of how to collect, analyze, and, where appropriate, incorporate subnational measures of FoRB violations into global datasets. Drawing on original empirical fieldwork in Latin America, we emphasize the need for such efforts by showing that structural forms of religious discrimination observed at the subnational level often go unnoticed in existing global datasets. We then identify key conceptual and measurement challenges that should be addressed in order to better advance scholarship on subnational FoRB dynamics, as well integrate subnational and national data collection efforts. Our findings have implications for how we study and respond to the growing number of FoRB violations around the world.

Keywords:

• freedom of religion or belief
• religious discrimination
• datasets
• subnational level
• Latin America

Prior to the Russian invasion of Ukraine in February 2022, several activists and journalists called attention to the deteriorating human rights situation in the Russian-controlled areas of the Donetsk and Luhansk Oblasts. Rising levels of religious discrimination were among the most pressing of their concerns. This included the outright ban of Jehovah’s Witnesses in the regions and the requirement for almost all religious organizations to undergo “religious expert evaluations” by the Russian-backed, local authorities. In many cases, the latter process prevented faith communities from being able to reregister under revised laws that draw on Russian, rather than Ukrainian, legal codes. These institutional changes, unsurprisingly, also translated into increased reports of societal intimidation, harassment, and violence towards minority religions in the two regions (for example, see “2020 Report on International Religious Freedom” 2020).

Global datasets of FoRB violations, however, have struggled to reflect the shifting levels of religious discrimination in Ukraine, leading to contradictory inferences. The Pew Research Center’s Global Restrictions on Religion (GRR) dataset, for instance, ranks government restrictions on and social hostilities involving religion in the country at approximately the same level between 2010 and 2016. In contrast, the Varieties of Democracy (V-Dem) project points to declining levels of FoRB since 2013 and even more so since 2017.

The case of Ukraine illustrates the challenge of integrating subnational data into the existing tools used by scholars and policymakers to track, study, and respond to FoRB violations around the world. Almost all of these instruments aim to capture country-level trends, even if they do consider local variation to some extent. The benefit of this approach is clear. Global datasets have drawn attention to important, and previously overlooked, trends in human rights violations. Most notably, they have raised awareness of rising FoRB violations around the world since at least the end of the Cold War. At the same time, these datasets continue to face the standard aggregation problems of any data collection effort that relies on composite measures. These methodological challenges matter not just for how we study FoRB violations, but also because they can potentially misinform policy responses.

Through this article,¹ we seek to initiate a conversation about how to collect, analyze, and, where appropriate, incorporate subnational measures of FoRB violations into global datasets more systematically. We enter into this discussion knowing full well that this is not an easy ask. Collecting localized data is laborious, time intensive, and costly. At the same time, the challenges are not insurmountable as advances in several cognate disciplines illustrate. Major survey instruments, such as the World Values Survey, the European Values Survey, and the Latin American Public Opinion Project, have all increasingly incorporated subnational measures. A question that remains unanswered in the study of FoRB violations then is whether the effort to scale down measures is worth it. As such, our primary aim in this article is to demonstrate the utility of such an approach. Clearing this hurdle could lead the way to a more sustained discussion of a second set of questions—how to effectively complement and incorporate subnational measures into existing tools and to what extent do subnational measures of FoRB violations complement or, perhaps, contradict the findings of national measures.

This article proceeds in four parts. First, we summarize existing efforts to document and study FoRB violations at the cross-national and, to a more limited extent, the subnational level. Next, we draw on a sample of Latin American cases to illustrate more methodically the prevalence of subnational variation in FoRB violations, as well as identify patterns obscured by cross-national measures. We then consider short-term challenges that need to be addressed, as well as pathways forward, for incorporating more subnational analysis into the study of FoRB violations. We conclude with a discussion of the implications of such an approach for future research and policy.

Existing Datasets and Research on FoRB Violations

Efforts to catalogue and analyze FoRB violations around the world have grown extensively over the past few decades. During the 1990s, a loose arrangement of faith-based organizations, primarily based in the United States, began to document the persecution of religious, especially Christian, minorities in a number of countries (Hertzke and Philpott 2000; Hertzke 2004). This reporting raised awareness of FoRB violations around the world, but it also remained ad hoc in nature.

Since the turn of the 21^st century, the documentation of FoRB violations has become more systematic. The majority of this research collects country-level data on government-based (GRD) and societal-based (SRD) religious discrimination. The three most widely-referenced, contemporary sources remain Pew’s GRR dataset (see Pew Research Center 2018), the Religion and State (RAS) dataset (see Fox 2019), and the Varieties of Democracy (V-Dem) dataset (see Coppedge et al. 2021).² These data sources cover different time periods and locations and adopt distinct methodologies. Taken as a whole, however, they illustrate the substantial boom in quantitative data collection on FoRB violations around the world over the past two decades.

Existing global datasets primarily focus on country-level measures of FoRB violations, but they do not completely ignore local dynamics.³ The 20 GRD indicators, for instance, aggregate ways that “national and local governments” restrict religion. Additionally, most variables explicitly ask whether the restriction is applicable at “any level of government” (“Codebook for Pew Research Center’s Global Restrictions on Religion Data” 2016, 1). The RAS variables focused on GRD, likewise, tend to represent the practice of national governments but are also coded if a majority of local or regional governments engage in a practice applicable to a measure (Fox 2017, 1). Measures of restrictions on proselytization and the building, leasing or repairing of places of worship also explicitly consider regulations by regional or local governments (Fox 2017, 16).

The inclusion of local dynamics into global FoRB datasets does not come without its challenges. One issue concerns data reliability and validity. States often seek to conceal FoRB and other human rights abuses by limiting, censoring, or manipulating records, media, and other publicly available information (Marshall 2021, 56). Moreover, there is the challenge of reporting bias with more information available for certain countries, namely those with freer media access. On top of that, data verification remains a persistent challenge in countries with authoritarian regimes. Thus, the local dynamics said to inform certain measures in cross-national datasets may not be uniformly identified.

There are also a host of standard aggregation problems inherent to cross-national data collection. The most obvious, perhaps, is a loss of information. Global datasets tend to focus on major or traditional religious traditions, often omitting or collecting less information on non-traditional religious communities or indigenous religions (Petri 2019; Petri and Buijs 2019). Several critics have also raised concerns that the country-level scores of global datasets do not accurately reflect the lived experience of religious communities on the ground or obscure other local realities (Petri 2022; Marshall 2021; Schirrmacher 2016). Another critical issue is Simpson’s paradox, or the fact that relationships at the local level can be completely reversed when analyzed at the aggregate level (see Greco et al. 2019).

In response to these and other concerns, a small, but growing cohort of scholars and practitioners, have begun to shift attention to FoRB violations in specific countries or regions. This work frames its analysis more in terms of contested religion-state relations than explicitly documenting FoRB violations. Several scholars, for instance, have documented religion-regime interaction at the subnational level across multiple religious groups in China and Russia, revealing patterns of cooptation and coercion between local authorities and religious communities (for example, Koesel 2014; Yang 2013). Others have catalogued the regulation of sharia law at the provincial level in Indonesia and Nigeria (Salim 2015; Buehler 2013; Ropi 2017). Still other analysts focus on regionally-based regulation of religious teachings in Indonesia and other parts of Asia (Ropi 2017; Achilov and Shaykhutdinov 2013). Additional scholarship notes recent strains in otherwise long-standing cooperative relations between certain religious communities and the state in the Philippines, Singapore and elsewhere, especially around issues of gender, sexuality, and reproductive health (Buckley 2017; Woods 2018). Finally, at least one recent article explicitly coded and compared GRD measures from the RAS dataset from the 26 cantons that comprise Switzerland (Helbling and Traunmüller 2016). Noting substantial regional variation, the authors concluded, “In terms of index scores, our sub-national comparison of the Swiss cantons is roughly like comparing disparate nations such as France and Luxembourg (8) with Spain, Sweden, and Poland (13) or with Germany and Greece (15)” (Helbling and Traunmüller 2016, 403–404).

In this article, we build on the turn towards studying subnational variation in religious regulation and discrimination by identifying underexplored patterns of FoRB violations in Latin America. Unlike the majority of existing studies, we explicitly highlight regional and local dynamics obfuscated by existing global datasets.

Subnational Variation in FoRB Violations in Latin America

In this section, we draw on original empirical fieldwork to provide illustrative examples of FoRB violations in Latin American cases (i.e. Mexico, Cuba, El Salvador, Colombia and Venezuela) that are often missed by global datasets. We focus on this region because it represents a hard test for our argument. Latin American consistently stands out in all three datasets we have discussed as a region with relatively low levels of FoRB violations. Yet, even here, we observe substantial subnational variation. We, of course, cannot provide a comprehensive audit for the entire region due to space limitations.⁴ Rather, the examples show both the prevalence of subnational variation and the degree to which these trends can be overlooked if only relying on existing global FoRB datasets.

Local Patterns and Subnational Variation

Turning first to Mexico, we observe that the main type of FoRB violations comprehensively documented in global datasets is the country’s strict form of secularism that was implemented during the Mexican revolution (1910-1920) (De La Torre, Hernández, and Zúñiga 2017; Grayson 2002). As Anthony Gill (2008, 115) writes: “Mexico represents perhaps the most extreme case of state control over religion”. In 1992, the most anticlerical articles of the Constitution were amended, but the levels of government regulation of religion remain high and are atypical for the region, as the RAS dataset and the GRI confirm. Drawing on his RAS dataset, Jonathan Fox (2018, 130) categorizes Mexico’s religion policy as “separationist,” which he defines as “Minimal support for religion. The state has a negative attitude toward religion and relegates it to the private sphere.” This characterization is generally accurate at the federal level.

However, the reality at the subnational level could not be more different. In numerous personal interviews with government officials at the state and municipal levels conducted by one of the authors over the past decade, the influence of Catholic ministers on decision-making is frequently denounced, as well as its consequences in terms of favoritism of Catholicism and discrimination of non-Catholic minorities. Because of the strong religious/secularist polarization in Mexico it is not always possible to determine the veracity and scope of these allegations. At the very least, though, they suggest that the wall that separates religion and the state in Mexico may be less robust in some regions of the country than others. More research on this subnational variation could even lead us to alter the country-level measures for religious support and religious discrimination in Mexico. Consideration of our subnational observations means the RAS Project’s characterization of Mexico’s religion policy as “Separationist” would have to be changed to “Multi-Tiered Preferences 1,” to acknowledge that “there exists one or more tiers of religions which receive less benefits than the preferred religion but more than some other religions.”

In addition, registered political parties in Mexico receive large amounts of discretionary public funds. The personal observations and interviews over more than 10 years of one of the authors suggest that it is a common practice for the state and municipal chapters of these political parties to channel part of this money to religious organizations (mainly Catholic and Protestant denominations) in exchange for their electoral support. Unfortunately, these exchanges are not taken into consideration in FoRB datasets. This is an indirect though substantial form of religious support at the subnational level. If the prevalence of this finding could be confirmed, Pew’s variable (GRI.Q.20) “Do some religious groups receive government support or favors, such as funding, official recognition or special access?” would need to be scored 0.60 instead of 0.12 and the RAS Project’s measure of “Direct general grants to religious organizations” would go from 0 to 1.⁵

Two more types of religious freedom violations, corresponding to “social hostilities,” can be observed in Mexico. The first is the religious intolerance in rural indigenous communities in the south of Mexico; the second are FoRB abuses in areas with a strong presence of drug cartels. It is interesting to observe that the former is comprehensively documented in cross-country, religious freedom datasets and the sources it uses, whereas the latter only shows up sporadically. Indeed, it appears that the high scores of Mexico on Pew’s Social Hostilities Index can be attributed to the frequent mentions of issues that arise around religious minorities in indigenous communities in the US State Department's International Religious Freedom reports.

Regarding religious discrimination in the context of organized crime, Petri (2020) and Petri and Glasius (2022) have shown that when drug cartels take over essential functions of the state, as is the case in northeast Mexico, they effectively regulate aspects of religion, either because they view religion as a source of revenue or to defend their interests, contradicting the conventional wisdom that organized crime is not particularly concerned with religion. Some exceptions aside (Freston 2018; Sotelo 2017; Gómez Chico Spamer et al. 2018), the issue for a long time was unnoticed in FoRB research. Presumably as the result of intense advocacy by faith-based organizations (Bartolini Esparza 2019, 174), it now gets more coverage in the US State Department's International Religious Freedom reports than before. However, it is not scored on Pew’s SHI because none of the 13 questions that compose this index account for this activity.

Turning now to Cuba, FoRB datasets generally do a good job at describing the religious freedom situation in the country and point to the high levels of government involvement in religion. However, two observations can be made about the accuracy of the measurements of the religious freedom datasets when considering subnational variation. The first is that although Cuba is a unitary state with a high degree of administrative centralization, the intensity of surveillance and administrative restrictions are reportedly higher in the eastern half of the island.

The second observation regarding subnational variation in Cuba is the subtlety of several forms of religious discrimination, which often goes beyond legal and policy aspects of FoRB. This includes things like the frustration that is caused by the government tactic of bureaucratic discouragement, the general legal insecurity due to the inconsistent, and therefore unpredictable application of regulations throughout the territory and in time, the intimidation and the cultivated mistrust between people that are caused by permanent surveillance, and the continuous minor forms of harassment. These elements are hard to quantify which explains why they are overlooked by FoRB datasets. When looked at individually, these threats may be negligible but taken together they create a culture of fear that is paralyzing and effectively restricts religious expression in different spheres of society (Petri 2020, 210).

In El Salvador, as in Mexico, the interface between organized crime and religion is a source of religious discrimination. However, it only receives a lateral mention in the 2020 US State Department's International Religious Freedom report: “gang activity continued to create security concerns at a national level, which affected the general population, including members of religious groups, but was not based on religious discrimination.” This hasty conclusion contradicts the findings of fieldwork conducted in and around the city of San Salvador (Brenneman 2014; Flores Orozco 2014; Orellana 2017; Arauz Cantón and Petri 2018).

In Colombia, organized crime also interferes with religious freedom (Flores Chiscul and Petri 2019), but we would like to focus here on the religious freedom situation in indigenous territories. The RAS dataset points to a univocal conclusion: government involvement in religion in Colombia is limited, especially in comparison to Mexico. Not only has Colombia no official religion, the levels of religious discrimination and religious regulation are low. Putting some sporadic elements of favoritism of the majority religion aside, it can be concluded that there is no noteworthy religious discrimination in the country. This statement holds true when observing the FoRB situation at the national level, but completely overlooks the atypical situation of the resguardos indígenas [indigenous reserves] where religious freedom for minorities is not guaranteed. Indeed, when applying the Religious Regulation Index of the RAS dataset, which is a composite measure of 29 variables, to the Nasa indigenous reserves, a subnational area where one of the authors of this article conducted extensive fieldwork, we obtain a score of 62 points, which contrasts with the 2 points of this index that considers the national level.⁶

Although Colombia is a unitary state (with three administrative levels: national, departmental and municipal), it grants far-reaching self-government rights to its indigenous communities, which, according to the constitution “may exercise jurisdictional functions within their territorial scope, in accordance with their own rules and procedures” (art. 246). This implies, among other things, that indigenous governments have the competency of public order (which they exercise through a guardia indígena [indigenous guard]), as well as the faculty to implement their own justice system. This fuero especial indígena [special indigenous jurisdiction] includes the possibility to order punishments according to their own usos y costumbres [customs and habits] (Ballón Quintanilla 2015: 96; Duarte 2009: 229). Fieldwork conducted between 2010 and 2016 among the Nasa ethnic group in the resguardos indígenas of the southwestern highlands of Colombia revealed that this legal system allows the existence of severe violations of religious freedom (Petri 2020; see also Arlettaz 2011), such as aggression as a result of conversion, violent assaults against attenders of religious services, restrictions of faith-based education and bans on proselytism. Again, because of their national focus, the dynamics in indigenous communities are unnoticed in the major FoRB datasets.

Finally, the situation in the Venezuelan-Colombian border area, where guerrillas have a strong presence, is also an important region for additional study of subnational trends in FoRB violations. Based on the fieldwork conducted for a 2021 USIP-USAID study on the causes and consequences of FoRB violations, four types or organized crime groups can be distinguished in Venezuela: colectivos (criminal groups acting on their own or in collusion with the government, allegedly to intimidate political opponents including religious leaders), drug trafficking networks (most of which are allegedly run by high ranking government and military officials), groups linked to Hezbollah and other Iranian-backed Islamist groups (we have no information about the implications of their presence in the country for religious freedom) and Colombian leftwing guerrillas (which the Venezuelan government explicitly welcomed into the country and that have a presence, especially in rural areas and on the country’s border).⁷ A number of religious leaders we interviewed alleged that some of these groups have been used by the government to intimidate and subdue the opposition, and there are indications that they also pose a threat to religious leaders known or related to dissidents or government critics, without anywhere for them to turn to for protection. While the above-mentioned study drew on a non-probability sample, the interview data call into question the main conclusion of most FoRB datasets that there are almost no social hostilities involving religion in the country. Perhaps unsurprisingly, a deeper focus on peripheral communities—both along the border and within the interior of the country—reveal a more complicated picture.

Overlooked Patterns Lead to New Questions

While each of the examples given in the preceding section deserve a more elaborate discussion, viewed as a whole they point to significant local factors and subnational variation that is rarely captured in global FoRB datasets. The persistence and prevalence of these patterns, especially in a region that is portrayed as having relatively high levels of FoRB, suggests the value of paying more attention to localized religious freedom measurements. In this section, we present three key patterns highlighted by our Latin American sample that the growing subnational research program would do well to further consider in order to advance our understanding of FoRB violations: administrative and geographic factors, access to information, and the types and roles of non-state actors.⁸

Regarding administrative and geographic factors, one of most significant areas where subnational variation can be observed is in federal states that devolve significant regulatory powers to local governments. Above, we described the case of Switzerland, where each canton evidently makes different choices when it comes to areas of religious policy. Although it is not technically federalism, the indigenous territories with broad self-government rights, such as in Colombia or, to a lesser extent, Mexico, could also fall within this category, because they set their own religious policy, by law or as the result of a de facto political arrangement.

Similarly, subnational variation is also likely to be observed in democratic countries that contain “authoritarian enclaves” or “subnational undemocratic regimes.” It would be interesting to explore to what extent, for example, the “subnational undemocratic regimes” that have been identified by Harbers and Ingram (2014) in Mexico correlate with localized FoRB violations.

The examples we presented also suggest that subnational variation of FoRB can be expected in territories that are so remote that the presence of the central state is weak or absent, such as some remote areas in Mexico, but also in Colombia and Venezuela. Though not always, remote areas often correlate with the presence of organized crime, which, as we have argued, tends to interfere with religious freedom. The variation between east and west Cuba follows other explanations but could also be linked to the relative isolation of the eastern cities due to its geographical distance with Havana, the country’s capital. This raises questions about territories in other states that remain far from the central administration and enjoy either de facto or de jure devolved powers. India’s Northeastern Region, for instance, would be one area for future research to explore in more depth.

Connected to these administrative and geographic factors, our sample shows there often are issues related to the availability of data to track subnational variation. If data on subnational FoRB violations would be readily available, it would be more likely to get detected in global FoRB datasets. This is, unsurprisingly, a major challenge, particularly in remote areas where access to information is difficult. In areas dominated by organized crime, data collection is also dangerous, not to mention these organizations may try to prevent information about their activities from being publicized. It is challenging to study the government systems of indigenous communities because they are mainly based on oral traditions, meaning there is no database that can be consulted to check their religious policy. And in authoritarian regimes, such as Cuba or Venezuela, the high levels of polarization, the partiality of the available information and its compartmentalization and censorship by the government make it difficult to obtain objective information about the religious freedom situation (see Glasius 2018).

Finally, our study of Latin American countries suggests that. Cross-national datasets often overlook these organizations because they tend to focus on the effects of state policies. As such, they are less well suited to identify structural forms of religious discrimination at the subnational level that have nothing to do with the quality of national legislation (see Petri and Glasius 2022). Moreover, extant datasets vary in the extent to which they explicitly track local aspects and they often fail to explain which local FoRB violations drive a particular measure’s score. The types and roles of non-state actors in restricting religious freedom, therefore, remains an important avenue for future subnational research.

Future Challenges

FoRB datasets have been instrumental in raising visibility of violations worldwide and to get the issue included in both domestic and foreign policies. As we hope is clear, our aim is not to criticize these datasets for insufficiently detecting subnational variation; to be fair, that is not what they were primarily designed to do. Moreover, some of the shortcomings of FoRB datasets we have identified are applicable to cross-national, quantitative analysis in general. Rather, our aim is to encourage more systematic consideration of local levels of FoRB and subnational variation.

We chart out, in this section, two key pathways for advancing our understanding of local FoRB dynamics—further developing the subnational research program we document above and “calibrating” the methodologies of global FoRB datasets in order to make them more sensitive to subnational variation. Our recommendations, of course, come with their own practical challenges. Moving forward, more consideration will need to be given to how to evaluate these tradeoffs and determine which approaches would be most effective for advancing our understanding of FoRB.

The first, and primary, way to increase our understanding of local FoRB dynamics is to further advance the burgeoning research program on subnational variation. To do so, it is recommended that scholars consult lessons from similar developments in cognate disciplines. The civil war literature stands out, in particular (see Zhukov, Davenport, and Kostyuk 2019). Recent debates in that subfield underscore, for instance, the need for consistency across comparative units of analysis (e.g. state, province, county, or city). Often this decision is data-driven and lacks clear theoretical motivations. However, comparable spatial and temporal units are critical for drawing generalizable inferences.

The civil war literature also cautions against spatial or temporal fragmentation in subnational research. Because scholars often focus on a single country or region, the generalizability of findings is not always clear. However, the subnational FoRB research program will only advance if it is able to avoid this self-contained study of particular countries or areas. This will require careful attention to how to balance the strong internal validity of in-depth subnational analyses with the desire to still demonstrate the broader applicability of findings.

Finally, our Latin American sample points to a particularly useful set of cases for exploring subnational variation in other regions—federal states, states that grant autonomy to indigenous groups, and states that contain remote areas with a weak state presence. These represent most-likely cases for subnational variation due to lower levels of central authority. Exploration of these cases would not only provide additional evidence of subnational variation, but also offer important opportunities to identify previously overlooked patterns and the underlying processes driving FoRB violations in particular contexts.

Subnational studies of FoRB, of course, need not carry the full responsibility of advancing our understanding of local FoRB dynamics. Analysts can also work to “calibrate” the sources that inform global datasets to be more sensitive to subnational variation. For instance, they can broaden the number and types of sources that are used. To date, FoRB datasets rely to a large degree on the International Religious Freedom reports of the US State Department. Although this is a unique source because of its global coverage and the speed of its updates, Thomas Schirrmacher has criticized the reliance on this source in particular because, in his view, it does not always provide a comprehensive overview of the (local) particularities in a country. Moreover, he argues that the cases cited in the IRF reports should be taken as illustrations rather than as exhaustive lists of FoRB violations (Schirrmacher 2016). In the same vein, Katherine Marshall (2021) and others assert that quantitative indexes should be more open to input from a larger variety of sources in order to reduce their blind spots on subnational realities (see also Birdsall and Beaman 2020; Petri 2022).

The RAS project already integrates this perspective, as it is not restricted to a set list of 19 sources like the Pew Research Center. Broadening the number of sources consulted would inevitably be labor intensive and time consuming (especially if sources in different languages are considered, and sources that only cover particular countries or regions). However, even a minor effort in this realm could go a long way in accounting for the main instances of undetected subnational variation.

Global datasets can also cast their net wider by measuring key variables that subnational studies identify as important drivers of FoRB violations (e.g. organized crime). The Pew Research Center’s Social Hostilities Index (SHI) and the societal module of the RAS Project already do so, but these indicators could be unpacked to a larger extent (Petri 2019, 79–80). Broadening the set of variables is not only related to subnational variation but would also make an important contribution to broader research and policy debates.

Finally, cross-national datasets that focus more on subnational variation might consider employing country experts who can reflect on the preliminary scores of the indicators of religious freedom datasets, and highlight matters related to subnational variation, as well as other findings based on fieldwork they conducted themselves or know about. Country expert reviewers can help mitigate the risk of uneven assessments of particular countries. They may also become increasingly useful because there may be more empirical data available for certain countries and issue areas than for others. It is also possible that the more you look, the more you find. In part, this is related to the presence of research groups, including faith-based organizations, and the success of their advocacy efforts. Another common problem in empirical research is that by zooming into a particular subnational area, issues that are minor tend to get exaggerated. Expert reviewers are not a perfect fix to these dilemma, as some may have their own bias. However, we recommend these as a way of validating the data rather than generating it.

It is, of course, beyond the scope of this article to provide bespoke solutions to all of challenges of subnational data collection. We hope by raising awareness to both the need for and risks of collecting subnational data, however, that our understanding of local FoRB dynamics will advance.

Conclusion

This article has sought to draw attention to the need for more systematic study of FoRB at the subnational level in order to help scholars and policymakers better track, study, and respond to religious regulation and discrimination around the world. By relying on country-level measures, existing FoRB datasets have increased awareness of the prevalence and persistence of religious persecution. A large and growing cohort of scholars have also leveraged these data to examine the relationship between FoRB violations and a wide range of social, political, and economic outcomes (see Klocek and Bledsoe 2022; Makridis 2020; Grim and Finke 2011). Those studies have been largely satisfied with conducting macro-level analysis. As a consequence, there has been little discussion about the need for or potential value of scaling down FoRB measures to provide a better account of what takes place at the local level. Drawing on our own empirical research and the broader literature on religious discrimination in Latin America, we illustrated the rich variation that has been largely overlooked and how shifting our analytic focus both challenges our understanding of FoRB levels in certain countries (e.g. Mexico, Colombia) and brings to our attention important subnational variation.

Our findings raise important implications for future research on FoRB violations. One of the most obvious is how to better collect information on local FoRB violations. As we point out in the preceding section, in-depth country case studies alone may not prove sufficient. Actors who violate human rights, including FoRB, often seek to conceal their activities. Victims may also be hesitant to speak openly due to concerns about retribution. Researchers conducting fieldwork will need to employ existing (e.g. list experiments) or develop new tools to gain traction on sensitive local dynamics. They will also need to avoid the geographic fragmentation discussed above.

Another lingering question is how, if at all, to incorporate subnational measures into existing datasets. We have flagged several possibilities, which range from augmenting the sources used to score country-level measures in existing datasets to creating new measures that better consider local circumstances. These choices, of course, come with their own tradeoffs. Future studies could explore the costs and benefits of combining various approaches. Additional work could also examine more systematically the degree to which local information may or may not change the country-level scores in existing FoRB violation datasets. Both tasks would further benefit from the identification of a set of cases that might most benefit from subnational analysis, such as countries with federal systems as we have discussed in this article.

Our analysis also raises questions for future policy. While existing datasets have been used by various national governments to “name and shame” severe FoRB violators, the aggregate measures have not necessarily translated into other policy and programmatic recommendations. Subnational data do not offer a silver bullet. However, they could be an important tool for developing context-specific and locally owned policies and practices that promote FoRB.

Additional information

Notes on contributors

Jason Klocek

Jason Klocek is an Assistant Professor in the University of Nottingham’s School of Politics and International Relations, a faculty affiliate of the University of Notre Dame's Center for the Study of Religion and Society, and a senior researcher with the U.S. Institute of Peace’s Religion and Inclusive Societies team. His work has appeared or is forthcoming in the Journal of Peace Research, Journal of Conflict Resolution, Journal for the Scientific Study of Religion, and other scholarly and media outlets.

Dennis P. Petri

Dennis P. Petri, PhD, is the international director of the International Institute for Religious Freedom; founder and scholar-at-large at the Observatory of Religious Freedom in Latin America; Professor in International Relations and Head of the Chair of Humanities at Universidad Latinoamericana de Ciencia y Tecnología and the Latin American Faculty of Social Sciences (UNESCO); and director of the Foundation Platform for Social Transformation.

Notes

1 The authors contributed equally to this manuscript.

2 The GRR dataset adopts a different terminology for GRD and SRD. The former is referred to as the Government Restrictions Index (GRI) while the latter is the Social Hostilities Index (SHI).

3 The V-Dem dataset is the exception. While other variables in the dataset include subnational measures, this is not the case for its single, composite FoRB measure.

4 In a supplementary analysis, we scored subnational variation for a sample of Latin American countries (e.g., Mexico, Colombia, Venezuela, Cuba) on a selection of indicators of the analyzed FoRB datasets applying their own methodologies to the local level. An Online Appendix with this information is available on the authors’ websites.

5 The recoded score is based on the above-mentioned supplementary analysis, which is available on the authors’ websites as an Online Appendix.

6 This recoded score is based on the above-mentioned supplementary analysis, which is available on the authors’ websites as an Online Appendix.

7 For more on the USIP-USAID study, see Klocek and Bledsoe (2022).

8 It is beyond the scope of this article to determine whether more attention to these issues, particularly in remote regions, would confirm or call into question cross-national data. This is also, though, an important avenue for additional inquiry.