Tracking Religious Freedom Violations with the Violent Incidents Database: A Methodological Approach and Comparative Analysis

Abstract

Measuring and comparing religious freedom across countries and over time requires reliable and valid data sources. Existing religious freedom datasets are either based on the coding of qualitative data (such as the Religion and State Project or the Pew Research Center), on expert opinions (V-Dem or the World Watch List) or on surveys (Anti-Defamation League). Each of these approaches has its strengths and limitations. In this study, we present the Violent Incidents Database (VID), a complementary tool designed to collect, record, and analyze violent incidents related to violations of religious freedom based on media reports and other public sources. We critically describe the criteria and process for selecting, coding and verifying the incidents, as well as the categories and indicators used to classify them. We also compare the VID with other existing religious freedom datasets and show how the VID provides a complementary picture of the nature and dynamics of religious freedom violations. We offer a preliminary analysis of the data collected through the end of 2024 with selected figures for data visualization. We conclude by discussing anticipated improvements for the VID as well as its potential applications for policy makers, advocates, and practitioners.

1. Introduction

In 2011, Thomas Schirrmacher wrote an opinion article on the challenges of counting the number of Christian martyrs in which he concluded, “What we need is a database in which for any year we could enter all the known, larger cases [of religious persecution] so that at the end of the year we not only have a usable estimate, but rather a situation where given the list, everyone can investigate the estimate’s resilience” [1]. This statement of need inspired the development of the Violent Incidents Database (VID), which collects, records and analyzes violent incidents concerning violations of religious freedom of all faiths, as input for both research and policy-influencing efforts. The VID is publicly accessible online at www.violentincidents.com (accessed on 9 May 2025).

At present, the VID is the only comprehensive data collection effort that systematically tracks religious freedom violations involving physical violence in its multiple dimensions with an event-based focus: individual and collective, state and non-state actors, religious and non-religious motivations, and in all spheres of life. The VID collects data concerning all faiths and, where possible, records the religious affiliation of both actor and victim. For example, in many incidents, Christians may be victims, but in others, they are the perpetrators. The VID also includes geographic information that allows for subnational analysis, which can surface important regional differences within a country.

Sadly, many civil society organizations working for justice in the field of religious freedom do little to collect comprehensive data. They are generally very good at discussing issues, raising awareness in the media and on social networks, diagnosing social situations, and making recommendations for public policy, but very few of them engage in the tedious, time-intensive, expensive, and sometimes dangerous task of documenting incidents. Nevertheless, effective political advocacy depends on objective, up-to-date, and reliable information, which often means documenting and counting incidents of religious freedom violations, which are a subset of human rights [2].

In this study, we present the methodology of the Violent Incidents Database (VID) as a complementary tool designed to collect, record, and analyze violent incidents related to religious freedom violations based on media reports and other public sources. Our understanding of religious freedom or Freedom of Religion or Belief (FoRB) is based on Article 18 of the UDHR and the ICCPR1. We compare the VID with other existing religious freedom datasets and show how the VID provides a complementary picture of the nature and dynamics of religious freedom violations. We also critically describe the criteria and process for selecting, coding, and verifying the incidents, as well as the categories and indicators used to classify them. We offer a preliminary analysis of the data collected through the end of 2024 with selected figures for data visualization. We conclude by discussing anticipated improvements for the VID as well as its potential applications for policy makers, advocates, and practitioners.

2. Comparison with Other Religious Freedom Datasets

The growing interest in academia in the documentation and measurement of religious freedom has led to the development of an increasingly rich corpus of religious freedom monitoring instruments, ranging from qualitative monographs and narrative reports to surveys and quantitative tools. Religious freedom monitoring developed into an entirely new field of study (Birdsall & Beaman, 2020 [3]). After experimenting with very rudimentary ratings of religious freedom, academics started to develop increasingly sophisticated datasets to track freedom of religion or belief (FoRB) [4]. As an illustration, Katherine Marshall’s comprehensive working paper “Towards Enriching Understandings and Assessments of Freedom of Religion or Belief: Politics, Debates, Methodologies, and Practices” [5] discusses 31 different instruments.

These datasets have in common that they present ordinal data (in contrast to event-based data like the VID). They can be categorized into three categories depending on their methodologies2. First, there are the socio-metric tools. These are based on the coding of narrative sources such as the International Religious Freedom reports of the US State Department. Coding of narrative sources, like other types of textual content analysis, involves developing and assigning codes to specific ideas, facts, or recurring themes found in the textual data collected. These codes are later used in analysis. The code systems enable researchers to compare various occurrences of the same code across cases or across time regarding a single case as well as discover correlations between codes that can provide meaningful insights.

The main socio-metric tools include the Government Restrictions Index (GRI) and the Social Hostilities Index (SHI) issued by the Pew Research Center, as well as the more elaborate Religion and State (RAS) Project directed by Dr. Jonathan Fox at Bar-Ilan University in Israel. Originally developed by Grim & Finke [6,7], the GRI and SHI offer two complementary ways of measuring religious repression around the world. The GRI evaluates the extent to which governments restrict religious beliefs and practices, focusing on factors such as laws banning certain religious attire, restrictions on proselytism, bias in the registration of religious groups, and government harassment. In contrast, the SHI measures acts of hostility motivated by religion that occur within society itself, including mob violence, sectarian conflict, and religion-related hate crimes. Both indices provide country-level scores, allowing for broad global comparisons and trend analysis.

While the GRI and SHI developed by the Pew Research Center are frequently referenced in both academic and policy discussions [3,4], a significant methodological limitation is their failure to differentiate between religious traditions within a given country. This gap is addressed by the Religion and State (RAS) Project, which systematically collects data on the intersection of religion and government policies worldwide. The RAS Project distinguishes between the experiences of specific religious groups, capturing inter-group variations in treatment, legal status, and societal discrimination. Methodologically, the RAS Project employs a more rigorous data collection process, including a significantly broader range of variables and extensive use of primary and secondary sources, such as national legislation, court cases, media reports, and academic studies. This allows for a more granular, disaggregated analysis of religious freedom conditions across different traditions. In contrast to the GRI and SHI’s valuable but generalized national-level scores, the RAS Project offers a nuanced, group-specific dataset essential for detailed empirical research and more targeted policy analysis [8,9].

The second type of FoRB dataset is expert-opinion-based. Expert opinion-based assessment tools get their input from experts, either by using questionnaires or by probing a select group of experts for their opinions and attempting to reach a consensus. Consensus is pursued through continuous probing and reassessment. Some variations allow the experts to know their colleagues’ opinions and discuss them in a controlled environment, thereby affecting their own opinion [10]. Tools employing this method benefit from an informed data source from which to derive knowledge and formulate strategy; however, it might be limiting. Although the opinions are informed, they are vulnerable to bias, like all humanly produced knowledge. The World Watch List of Open Doors International, for example, includes not only academics but also organization members working in specific countries as its experts, who fill out questionnaires that examine the degree and manifestation of pressure exerted on Christian communities in a territory [11].

Besides the World Watch List, which only focuses on Christians, another expert-opinion-based instrument is the Varieties of Democracy dataset. This quickly became very popular among political analysts and includes a single variable on religious freedom. There also is a small FoRB pilot currently being developed by the Human Rights Measurement Initiative that uses a similar expert-opinion-based methodology.

A third category of FoRB datasets is the use of surveys. With surveys, researchers can glean information from a population of participants either directly through the questions themselves or implicitly by analyzing the respondent’s answers to a few questions and checking for specific themes. In surveys, responses to questions are usually on a scale of agreeing and disagreeing with the statement at either end of a scale. This approach is beneficial because a large population can be probed in a relatively short span of time, and different interactions of variables relating to the population can be analyzed [12]. The Anti-Defamation League Global Index of anti-Semitism examines attitudes towards Jewish people in more than 100 countries and uses surveys to gather responses. Each country then receives a score based on analyzing answers regarding their attitudes towards Jewish people in general and their degree of agreement with Jewish-related stereotypes [13].

This brief comparison demonstrates a gap in FoRB measurement tools. No dataset employs an event-based approach to measure FoRB. There was a short-lived religion pilot that was part of ACLED, but it was later discontinued3. While we do not claim that any singular approach can successfully capture the complexity of the phenomenon by itself, we chose the event-based approach to explore violations of religious freedoms. Thus, the VID is the only tool currently in development that attempts to monitor the infringement of religious freedoms using an event-based approach. The VID’s aim is to complement other instruments and aid FoRB research by providing insight into the nature and dynamics of religious freedom violations.

3. Methodology and Justification of the VID

Limited access to information is common in high-pressure or violent contexts [2]. However, these data are critically important. When incidents are documented, it is this very same written record that becomes the main justification for requesting attention to a specific social problem.

Documentation is particularly important in situations where victims of violence are afraid to report crimes to the police, or when states fail to comply with their duty to register human rights violations. To cite just one example, according to estimates by Ethos (2017), a Mexican think tank, 94% of all crimes in Mexico are not reported. In its report The Human Rights Situation in Mexico, the Inter-American Commission on Human Rights (IACHR) found that “the internal forced displacement has not been documented and analyzed comprehensively by the [Mexican] State, which is the main obstacle facing the comprehensive response that Mexico should give this phenomenon”. The report also observed that the situation “is evidenced by the invisibility of the problem”, which hinders efforts to “adopt the measures necessary to provide an effective response to this phenomenon”. Therefore, one of the most important purposes of documenting incidents, particularly when they concern human rights violations—including religious freedom—is to ensure that a record of specific violations is kept, so as to hold the responsible party accountable and demand compensation for victims.

A nascent version of the VID was developed in September 2011 to support the information management needs of the World Watch List of Open Doors International. The project was discontinued a few years later because the organization moved to a different data collection system, which is useful for its purposes, but has the disadvantage of not being public. In January 2018, the VID was integrated within the Observatory of Religious Freedom in Latin America (OLIRE, in Spanish), with a regional focus. The VID has since become a flagship project of the International Institute for Religious Freedom (IIRF), with a global focus. The worldwide update for 2021–2024 was funded by Global Christian Relief. Data on Latin America continues to be provided by OLIRE and data for Nigeria is provided by the Observatory of Religious Freedom in Africa.

The Violent Incidents Database has been developed to collect and synthesize information available in order to support religious freedom advocacy efforts. The VID attempts to establish the quantitative impacts of religious freedom violations. We adopt a very broad definition of religious freedom, in line with Article 18 of the ICCPR. We also use a broad definition of violence, operationalized through basic categories like killings, attacks on places of worship, arrests, abductions, displacement, etc. (please refer to Appendix A for more information). There is also an “other forms of violence” category, which can include subjective experiences of violence. We have a category for non-physical violence, though we do not actively search for records in that category. Religion is defined using the self-identification criterion, and we follow the same categories as the Religion and State project. Methodological Appendix A provides specific operationalizations of these terms. We intentionally use broad definitions because overly specific ones would result in discarding many incidents that may have a religious component.

Beyond definitions, it’s important to emphasize that the VID collects data in the broadest sense. Users can make their own selections from the data based on their own definitions of religious freedom, which may be more or less narrow.

The process of recording incidents, writing reports, and publishing about them brings a different quality of attention and can raise obscure and distant atrocities into public awareness. Establishing the quantitative impact of an issue makes it a “social fact” that can be considered [16]. If it is not documented, it is as if it did not exist.

Fundamentally, the VID generates knowledge. Like all knowledge production, the VID goes beyond “facts”. We seek to reduce bias by applying academic tools and training in the design of the database and with VID researchers, but it is impossible to avoid completely. Research is produced by researchers [17]. We view this as an unavoidable part of humanity as Polanyi [18] has argued. For this reason, and to allow users to work productively, the VID offers information with transparency. We largely rely on publicly available sources anyone can access and make these sources available for each record4. The VID makes explicit what is likely already implicit to area specialists and FoRB experts [19]. The VID’s focus on FoRB can also provide accessible and distilled information for policy makers who might not have the same implicit knowledge but are responsible for creating social policy.

In line with the core mission of the International Institute for Religious Freedom to promote religious freedom for all faiths from an academic perspective, the VID provides reliable data to strengthen academic research in the field and to inform public policy. The global expansion of the VID included a three-year period from November 2021 to December 2024. Ten researchers with regional experience and linguistic specialties were hired and trained to monitor assigned countries. These researchers work part-time for specific periods during the year, each focusing on a specific region. They primarily concentrate on reading major news publications and reports related to their assigned areas. We have curated an annotated list of mandatory sources they must consider, along with a secondary, longer list of optional sources. Researchers are also free to browse the web in search of additional materials. The VID researchers submitted incidents along with sources to a supervisor who reviewed and continued training the researchers by giving appropriate feedback. These incidents are then analyzed by a reviewer who double-checks the incidents, verifies the sourcing, and approves the new records for admission to the database.

To ensure the reliability of external information found online, coders were provided with clear instructions alongside a tiered source list prioritizing official reports (e.g., U.S. Department of State International Religious Freedom Reports, USCIRF reports, reputable international NGOs) and globally recognized news agencies. Coders were instructed to rely primarily on sources from this list and to exercise caution with any supplemental web searches, using only sources that were professional news outlets, peer-reviewed research, or known human rights organizations. Social media, blogs, and non-reputable sources were explicitly discouraged unless independently verified through multiple channels.

Researchers selected for coding roles were required to have a prior academic or professional background in religious freedom (FoRB) or human rights, ensuring a baseline familiarity with credible documentation standards. Before beginning independent work, research assistants received training sessions covering source evaluation, data reliability, and consistency expectations. Early in the project, all incoming coder outputs were thoroughly reviewed by senior researchers to calibrate judgment and ensure adherence to the standards. Coders were required to provide citations for every incident and, for larger or more severe incidents, to corroborate information across two or more independent sources whenever possible.

Thus, while coders had some flexibility to find supplemental information, their work was constrained by structured source guidelines, initial training, continual oversight, and source triangulation requirements, ensuring that information incorporated from web searches met the same quality thresholds as official reports and major news outlets.

We do not conduct any factual validation of the incidents we collect, as we do not have the capacity (which would require having access to researchers on the ground worldwide, which would require a massive budget). Instead, we offer the possibility of a posteriori falsification: if a user encounters an incorrect incident, they can let us know, and we may decide to remove it. This has happened a number of times.

Interrater reliability testing is not applicable to our method of data collection in the conventional sense. Our research assistants were not coding subjective impressions or judgments; rather, they were tasked with systematically extracting factual information from a pre-specified set of approved sources. Each extracted data point was then reviewed by senior researchers to confirm that the information was correctly recorded, aligned with our methodological definitions, and properly sourced.

To ensure accuracy and consistency, we implemented several key reliability mechanisms. First, all coders underwent initial training that included specific instructions on how to extract, record, and document incidents according to our coding rules. Second, all entries were subject to a mandatory review process: senior team members independently verified that the information recorded matched the source material and adhered to the coding protocols. Third, any discrepancies identified during this review were discussed with the coders and corrected collaboratively. Fourth, for complex or ambiguous cases, a second source was required, or a senior reviewer adjudicated the final coding decision.

Thus, while traditional interrater reliability statistics (such as Cohen’s Kappa) are not applicable due to the structured nature of the task, we employed layered verification processes to ensure high reliability across all coders, regardless of the country, religion, or incident type being documented.

Data collection parameters were adjusted in 20235; however, they do not differ considerably from the previous structure [20]. To the original categories: geographical location, date of incident, summary, nature of the incident, responsible actor, religion of victim(s), additional information, and web sources, we refined the religious categories and included the religion of the responsible actor. The religion categories have been adjusted to follow the religious minorities codes used in the Religion and State Dataset6. The actor’s religion is often not named in media reports, though we include it where possible7. In many cases, religious affiliation can be inferred by the name of the group claiming responsibility. For instance, the Allied Democratic Forces (ADF) in the Democratic Republic of Congo are a known Islamic terrorist group and are listed as such by the Ugandan government.

Regarding information sources, just like delicious chocolate, not all media and news reports on the internet are of equal value. When training the researchers, IIRF staff developed an annotated source list of reliable sources. These include the United States Commission on International Religious Freedom FoRB victims list, the International Religious Freedom Reports of the US Department of State, the Armed Conflict Location and Event Data Project, Open Doors Analytical, the Country Reports on Human Rights Practices of the US Department of State, Human Rights Without Frontiers, Amnesty International, Global Christian Relief, Forum 18, Human Rights Watch, Bitter Winter, Reuters, Associated Press and the New York Times. These sources largely follow the IIRF’s mandate of promoting religious freedom for all religions and have an established history of providing credible information. The researchers were given a supplemental list of fifty websites and keywords. They were also encouraged to rely on their regional knowledge and linguistic specialties. As the VID attempts to record a broad range of incidents, and since anyone interested can look back and evaluate any record, most web sources are permissible. The IIRF is entrusted with certain records on the condition of confidentiality, though this is a small minority of the total records8. Last year these represented approximately 8% of incidents.

The VID raises the visibility of religious freedom violations. This visibility is instrumental in the recording and enumerating of incidents and establishing of patterns of discrimination for case-by-case and contextual analyses [21]. Increased visibility aids the elaboration of recommendations for legal and policy reform and can inform national and international decision makers, religious communities and civil society organizations.

Inspiration: Event-Based Data Collection

The VID draws inspiration from event-based data collection methods in adjacent fields which have gained momentum in recent years, particularly by scholars examining conflicts, protests, and violations against minorities. Examples of such efforts include databases on conflicts such as: the Armed Conflict Location and Event Dataset (ACLED), the Social Conflict in Africa Database (SCAD) and the Uppsala Conflict Data Program (UCDP); databases and datasets on protests: the Nonviolent and Violent Campaigns and Outcomes (NAVCO) dataset, and the Violent Political Protest (VPP) dataset; and databases concerning minorities such as the Minorities at Risk (MAR) Project [22,23,24,25,26,27,28].

The event-based approach uses discrete occurrences as units of analysis to study a phenomenon. The events are picked according to strict criteria, ensuring that they are relevant and similar enough to allow specific features to be analyzed, with insights gleaned from this analysis paving the way toward globalizations about the phenomenon’s unique characteristics. Examples of the employment of this approach include the ACLED, which collects data on internal conflicts in 50 unstable states, based on location, actor, and date [26].

Although these various projects have similar goals, and sometimes overlap, they differ in the units of analysis examined and/or in the criteria for case selection. NAVCO, for example, examines both violent and nonviolent campaigns (a campaign is defined as “… a series of observable, continuous, purposive mass tactics or events in pursuit of a political objective.” [24], whereas the VPP deals exclusively with violent protests that resulted in at least twenty-five casualties. There can also be differences in scope. For example, while the UCDP seeks to record cases globally, the SCAD focuses solely on Africa [29].

Event-based data collection would not have been possible without the greater access to information offered to researchers in recent decades. News media published online is of particular importance. It brings greater attention to incidents, both local and global, that might not otherwise receive attention. Media sources are also often archived which facilitates selection of cases, comparative research, and time-series analysis [30].

Event-based data are important, both for advocacy and research, because they are based on reported facts rather than opinions held by experts or the measurement of attitudes in the population. This is not to say that media reports are a panacea. Potential issues can arise when relying on media reports as a data source, namely measurement errors and biases related to the way the media conveys its reports and the data within them. Some events might receive more coverage than others. This can lead to oversampling and other errors of representation. Moreover, in cases where the description of the event is also coded, the way an event is described might be biased due to the agenda of the source. This description bias can affect analysis and might lead to wrong conclusions.

The VID is certainly not immune to these limitations, as we will discuss. Most of these issues, however, can be mitigated by formulating and following strict and clear procedures for coding. This approach does not neglect the human element but minimizes the risk of errors due to biases, which VID researchers also have [30].

The event-based data collection projects mentioned above inspired the creation of the VID. These important initiatives are quite broad in their definition and scope and lack a commensurate focus on the issue of religious freedom. So far, there has not been a FoRB dataset that is event-based, with the exception of the now-discontinued religion pilot that used to be part of ACLED.

4. Strengths and Limitations of the VID for Research on Religious Freedom

Like any research initiative, the VID data can be misused or used appropriately. In this stage of development, it is important to remember that we are not claiming that the VID presents a comprehensive picture of every nation. When looking at the tables and numbers in the database, it is tempting to read the statistics as a representation of any given country’s situation. However, the information is only what has been recorded in the database by the assigned researcher. The sheer volume of religious freedom violations makes it impossible to claim exhaustive coverage.

The data included in the VID is based on reports published in digital media available on the internet, but there may be cases of underreporting or overreporting. There is no question that the media landscape in a region affects which incidents are “known.” Many incidents are never made public or do not receive sufficient attention from authorities or media (underreporting). Even though we aim to collect data for all religions, we have found some better track violent incidents than others9. Some religious groups see real value in recording incidents and might report on events multiple times or republish other reports. Other religious traditions do not track incidents or might not have networks or funding to report religious freedom violations or advocate for their religious communities.

There is no pro-Christian bias in our data collection, except for the fact that non-Christian traditions are generally less equipped to document religious freedom violations (and some Christian denominations do better than others), and thus, their incidents may be underrepresented in the VID if they do not get into media reports or other types of reports. However, there are important exceptions, such as the data collected on antisemitism (though methodological differences mean that ADL data are not directly usable in the VID) and some data on Hindus. Muslim groups rarely collect data on violence against them, even though they are arguably victims of much violence caused by Islamist groups.

Not all violent incidents appear in news reports, or when they do, do not meet the minimal criteria to be included in the VID. Genuinely terrible violent incidents frequently occur during wars or conflicts, but do not always involve religious freedom violations. Some reports are non-specific regarding the victim or the actor. Other reports do not give enough information on the nature of the event or location but make vague and general assertions. These would not be included as they do not provide enough information to complete a single record10. The same applies to reports that include only aggregate data. Such reports may provide valuable statistical overviews of violations but do not disclose detailed, incident-level data. Since the VID requires individual records to ensure accuracy and traceability, we are unable to include information based solely on aggregate figures. When possible, we seek to contact data providers to request access to detailed records in order to enhance the completeness of the VID.

However, more information and events recorded complements existing datasets and will be of great value in this field, provided the information is used appropriately. We are working to incorporate high-quality sources as mentioned in the annotated source list in the previous section and transparently present the data we collect. We have also built an online self-reporting form allowing anyone to report incidents11. For self-reported incidents, we look for a public news source on the internet or supporting evidence to substantiate the claim. These incidents then go through the review and checking system as described in the methodology12.

There are also times when incidents reported in the media are incorrect or could be exaggerated for a particular constituency (overreporting). We do not have the capacity to verify all incidents listed, though we do have a quality control and vetting process, described above. If reports are flagged up as being false or incorrect, we retroactively correct entries with errors or remove them (a posteriori falsification). This has already happened through the self-reporting form. The original incident was removed thus demonstrating the efficacy of the reporting form as well as a posteriori falsification13. If anyone finds a case is missing or was erroneously reported, the IIRF team can be contacted14.

As said already, our data collection can never be exhaustive or fully comprehensive. The VID is an ambitious project but can only include data based on what is available, but it should not be viewed as a comprehensive record of everything that occurs. In the next phase of the VID, we hope to implement some automation of data collection, which will hopefully address the issue of human limitations, but we will always be constrained by the availability of public information.

5. Preliminary Results

The Violent Incidents Database (VID) encompasses nearly 15,317 records up to 31 December 2024. The number of countries featuring at least one recorded incident has quickly grown. The VID began with a focus on Latin America with OLIRE as a main partner for that data. The VID also partners with the Observatory for Religious Freedom in Africa for data related to Nigeria. An increase in funding facilitated the expansion of VID coverage to the rest of the world, beginning in November 2021. Much of the analysis illustrates the type of information and analysis possible using VID data, but it is not comprehensive of all violations in any particular country or region.

Since November 2021, the IIRF has been diligently curating the VID sources from reputable and validated data providers. Drawing on these sources, the VID documents incidents of various forms of violence associated with religious affiliations. Each incident cataloged entails at least one victim, with classifications spanning a range of categories: including killings, attempts to destroy or defile religious structures, closures of religious establishments, arrests or detentions, abductions, sexual assaults, forced marriages, physical or mental abuse, property damage targeting religious adherents, forced displacement, and non-physical forms of abuse related to religious beliefs. As of our latest assessment, we have identified 1725 incidents occurring in the year 2023, affecting an estimated total of approximately 1,887,000 individuals. In the year 2024, we collected 2956 incidents, with a total of 421,351 victims15.

The VID is dynamic and is continuously updated. The VID has an intermittent production cycle that depends on the output of the research assistants, the quality control review process, and the technical inclusion in the production database. We aim for all approved records to be added to the production database monthly. To date, we have only removed two records. Since the volume has been so low, we have not developed a procedure yet. Overall, 2021 to 2024 have approximately 6000 additional probable incidents pending expert review. As a result, figures are subject to change over time as the database continues to develop and expand.

We meticulously document the perpetrator(s) responsible for each incident where available (see Figure 1). This includes both broader categories which are predefined and an open field where a specific actor can be included for any given record. This categorization allows for helpful cross-cutting analysis. Certain regions in the world experience religious freedom violations from a narrower group of actors. This figure shows how the current information needs to be contextualized.

Given that the VID has collected data in Latin America since 2015 and the global expansion began in 2021, it follows that organized crime would contain the most records. The finding that organized crime is the most common source of religious violence may seem counterintuitive, but it can be explained by two reasons. First, we have a longer history of collecting data in Latin America, a region where organized crime is indeed a significant source of violence, including violence against religious people. (Before the VID started collecting data, organized crime as a source of religious freedom violations was generally overlooked by existing scholarship, highlighting the empirical value of the data we collect.) Second, in some countries where one might expect religious violence, reported incidents are surprisingly low. For instance, despite common assumptions, countries such as Saudi Arabia and Somalia do not exhibit the high frequency of religiously motivated violence that might be anticipated—presumably because government discrimination is already so high that violence is “unnecessary”. In other cases, like North Korea, Afghanistan, and parts of China, we have good reason to believe there is significant religious violence, but much of it does not appear in public records. Both factors confirm that the database is not comprehensive and is influenced by the availability of data and our regional experience. The data on Nigeria is extensive compared to other countries because we have a reliable partner there, the Observatory of Religious Freedom in Africa.

For comparative analyses in the context of FoRB, the ability to categorize incidents can also facilitate overarching classifications of state and non-state groups (Figure 2). This analytical perspective is valuable for documenting incidents and counting affected individuals. When looking at the total number of victims impacted by incidents, we observe a significant reduction in incidents attributed to unknown instigators. This trend suggests that instances of anonymity are typically associated with lower victim counts.

Over the last 12 months of 2024, a certain normalizing process seems to have occurred with the rapid expansion of the dataset with respect to one key metric: the government/non-government distinction for the perpetrator has remained relatively stable. A further comparison into the future will reveal whether or not this is a coincidence but for now we can observe that whether measuring the implication of state actors in FoRB violations at an incident count level or a victim count level, the proportions are also somewhat similar.

However, we know that victim counts can distort the overall picture, which is why the VID records discrete cases against individuals while also acknowledging incidents that impact a larger number of individuals. This approach presents methodological challenges that are important to keep in mind when aggregating data. Certain incident categories are less likely to involve mass events, while others may indeed involve substantial numbers of victims (e.g., incidents categorized as ‘forced to leave home’). An alternative method, which offers greater ease of visualization, involves analyzing the frequency of assigned categories to incident records irrespective of the number of victims (see Figure 3). It is also important to state that certain events, such as those related to sexual abuse, are infrequently reported, despite potentially occurring with relative frequency.

Looking at incident categories, and not individuals, allows numerical analysis to incorporate mass events without obscuring smaller or more frequent events (Figure 4). Therefore, despite the VID documenting 446 incidents involving victims being forcibly displaced since 1 November 2021 (of which 175 occurred in 2024), the total number of victims in this category amounts to 384,310 individuals for the 3-year period. The largest singular occurrence within this category affected 200,000 people from the forced displacement of Rohingyas in Myanmar on 18 May 2024, before most of their buildings were set on fire.

In addition to enumerating incidents, calculating median values provides valuable insights into the typical scale of victimization recorded in the VID (see Figure 5). The data reveals that most incidents involve relatively small numbers of victims: even accounting for recent increases in mass displacement incidents, the median number of victims remains only nine. This highlights the VID’s focus on systematically documenting both small-scale and large-scale violations. The visualization also shows that the majority of incidents typically affect between 1 and 94 individuals. Nevertheless, exceptional cases—such as incidents categorized as “Forced to leave Country” and “Forced to leave Home”—record maximum victim counts of 120,000 and 200,000, respectively. These outliers demonstrate the importance of using medians rather than means when analyzing the data, as they prevent disproportionate influence from extreme cases.

In Figure 5, the years 2021–2023 were combined into a single group to create a more stable and representative baseline for comparison. Given that the Violent Incidents Database (VID) was still in its early development phase during those years, the number of recorded incidents was relatively smaller and potentially more volatile if analyzed year-by-year. Aggregating data across three years mitigates fluctuations and enables a clearer comparison with the more complete and rapidly growing dataset for 2024.

To provide further context regarding this variability in light of median values, let us examine the highest-ranking VID incidents per category based on victim count in the three rankings in Figure 6 (data sourced from 1 November 2021 to December 2024).

With recorded incidents related to physical violence, we see a greater spread of actors, meaning that more marginal perpetrators are no longer able to hide in the shadow of the large numbers of victims in some incidents (see Figure 7).

6. Added Value of the VID

It is important to keep in mind that the VID is intended as a complement to other FoRB datasets, not to replace them. The VID contributes to FoRB research by collecting additional data, highlighting blind spots or spotlighting undetected forms of religious freedom violations. We encourage all FoRB datasets to make use of the VID data as additional input.

First, in ways similar to the Armed Conflict Location & Event Data Project (ACLED) and the ADL’s USA audit, the VID enumerates and categorizes the impact of real-world stories about disheartening infringements on individuals’ fundamental right to religious freedom. In this sense, the VID constitutes an advancement in religious freedom research because it complements the other FoRB datasets by presenting evidence that should not be denied. It is based on reported facts.

Second, the VID is also geographically scalable. It provides insights into subnational variations at the country level all the way to global trends. Location variables can facilitate an analysis of events taking place within a given state, province, or territory. Building on a preliminary study by Klocek & Petri [31], which highlighted the existence of subnational variation not detected by the main FoRB datasets, the VID’s dataset supports much-needed subnational analysis, albeit with the usual caveats of non-exhaustive data. Overall, VID records significant subnational data. Only 335 records do not have it (4.3% of the total current dataset comprising 7722 records).

Nigeria, Nicaragua, Mexico and Colombia are the richest VID sources for violent incidents also defined at the subnational level.

Table 1. Subnational variation in Nigeria, India and China.

Country	State/Department	Incident Count
Nigeria	Kaduna	805
	Niger	600
	Plateau	337
India	Uttar Pradesh	131
	Manipur	41
	Karnataka	36
China	Sichuan	56
	Guangdong	22
	Xinjiang	21

illustrates the richness and contrasting information available at the level of provinces, departments, and states since November 2021, expanding for a sample country, Colombia (

Table 2. Subnational variation in Colombia.

Country	State/Department	Incident Count	Population	Incidents per Capita × 1,000,000
Colombia	Arauca	28	294,206	95.2
	Chocó	37	544,764	67.9
	Cauca	64	1,491,937	42.9
	Norte de Santander	44	1,620,318	27.2
	Valle del Cauca	93	4,532,152	20.5
	Nariño	33	1,627,589	20.3
	Magdalena	28	1,427,026	19.6
	Cesar	20	1,295,387	15.4
	Antioquia	55	6,677,930	8.2
	Cundinamarca	19	3,242,999	5.9

):

Let us focus on Colombia, noting also population statistics and incidents per capita:

We can now examine these ten, greater concentration Colombian departments to highlight the power of subnational analysis with respect to recorded FoRB incident occurrences. These departments are illustrated in Figure 8 below. We analyze subnational variation, therefore, by also factoring in population distribution, which is a vital step in assessing the religion-associated violence at this level.

The data derived from the VID strongly suggests that when examining religious violence incidents in Colombia, other factors than geography are at play since the most frequent incidents-per-capita departments (Arauca, Chocó and Cauca) are not spread across the country.

This intriguing observation presents an avenue for exploration by stakeholders and policymakers within Colombia and beyond who aim to reduce violence in the country. Subnational data allows researchers to extract insights by establishing correlations between their own datasets and other subnational data, such as income levels.

The comparison with FoRB datasets also suggests that government and social discrimination in Colombia are low and medium, respectively. However, this does not account for subnational variation, because the departments of Arauca, Chocó and Cauca exhibit high numbers of violent incidents.

Thirdly, the VID is dynamic. If new information arises around an incident, then this incident can be updated or even removed. Furthermore, anyone can contribute to an incident, even though our team of researchers will verify the data before editing and adding it to the database. This publicly available, event-based, and dynamic repository allows for continuous updates and corrections to provide accurate and up-to-date information. The IIRF has made the VID publicly accessible and searchable16 as a public good which we hope will be used and maintained by everyone in the FoRB community. It is updated with new records throughout the year and can be queried by a variety of fields including date range.

Fourthly, the open-source approach to data collection has beneficial social relevance. Removing impediments for the systematic recording of violations of religious freedom allows for better and more accessible documentation of incidents, with the possibility of deterring further violence.

Fifthly, the VID allows for a greater degree of granularity. It distinguishes between several categories of state and non-state actors, tracks the religions of the victims and the perpetrators where possible, and records the subnational location where the incidents occurred. Such levels of detail are unavailable in other current FoRB datasets.

In contrast to broader measures of religious freedom such as the Government Restrictions Index (GRI) and Social Hostilities Index (SHI) developed by the Pew Research Center, the VID adopts a more micro-level approach by systematically documenting specific incidents of violence and discrimination motivated by religious identity. These incidents include physical attacks, arbitrary arrests, vandalism of religious sites, and forced conversions. Whereas the GRI and SHI assess the national legal and social environments surrounding religious freedom—offering valuable but aggregated indicators of systemic conditions—the VID captures discrete, verifiable acts of religious persecution, providing a granular, event-based perspective particularly suited for legal advocacy, humanitarian response, and detailed empirical analysis. Moreover, while the GRI and SHI do not differentiate between specific religious traditions, a limitation also addressed by the Religion and State (RAS) Project, the VID similarly offers disaggregated data that underscores the lived realities of religious communities at the ground level. Taken together, these datasets reflect complementary approaches: the GRI and SHI illuminate national patterns and regulatory environments, the RAS Project provides nuanced differentiation between religious groups within those environments, and the VID captures the immediate manifestations of religious hostility as experienced by individuals and communities.

In this sense, the VID makes it possible to discern patterns of religious freedom violations which are helpful for comparative analysis, as illustrated in Figure 9. FoRB datasets typically gauge governmental discrimination, societal discrimination, or both. Grim & Finke already established a connection between social restriction of religion and government restriction of religion, in which the former encourages the latter, and both lead to violent religious persecution [6]. It is reasonable to hypothesize that governmental discrimination aligns with violent incidents instigated by state actors, while societal discrimination aligns with violent incidents instigated by non-state actors. When there is a divergence, this could point either to a gap in the VID, or to a gap in the FoRB datasets. It could also signal the complementary value of the VID.

The classification into high, medium, and low categories was achieved by dividing the country scores for each metric into three equal groups: the top third representing “high” (3), the middle third “medium” (2), and the bottom third “low” (1).

We offer two brief examples. Colombia scores low on government restrictions and medium on social hostilities according to the Religion and State (RAS) metrics. However, upon scrutinizing the violence data extracted from the VID, it becomes evident that violence perpetrated by state actors is rated as medium, whereas incidents involving non-state actors are classified as high. This discrepancy implies that the RAS metrics may overlook certain subtle subnational disparities identified by the VID.

Regarding Somalia, the RAS government discrimination measures are high, but the violent incidents instigated by government actors contained in the VID are low. There may be two explanations for this. The first is that the level of government discrimination is so suffocating that it does not need to engage in any form of physical violence against religious minorities. The second explanation is that there is a data gap in the VID, which is very possible, considering the fact that data availability for this country is a known challenge.

Finally, as the VID continues to expand and develop it will become increasingly representative of reported religious violence throughout the world. If this potential can be realized, the VID would be a true window into violence and drivers of hostility that limit FoRB across the earth.

7. Concluding Remarks

The VID is the only FoRB measurement tool that is events-based, as opposed to all the other tools that are expert-opinion-based. Both types of sources have their place, but the VID adds value by providing additional granularity. We identify the religious affiliations of both victims and perpetrators, distinguish between several categories of state and non-state actors, and include subnational data, which no other FoRB tool does. The VID is already revealing empirical gaps in other FoRB tools. It can also be seen as introducing a new category in FoRB monitoring, which has until now been dominated by Pew’s “government restrictions” and “social hostilities” categories (the latter category probably still being too broad). We add a third category, “violence”, and show that patterns can be identified using these three categories. For example, there are countries where government restrictions are high, but violence is low, or vice versa.

Further, the VID is open-source and dynamic, publicly accessible and searchable. The VID is still in development, but it has already been used in publications by the United States of Peace [32,33] and the US Commission on International Religious Freedom [34,35]. It was presented at the IRF Summit in January 2024 and featured in a Universal Periodic Review report by the UN Human Rights Council on Nigeria, as well as referenced in two country reports by the US State Department. It serves as a powerful tool for monitoring and advocacy because successful advocacy and awareness-raising rely on factual information.

Adding layer(s) of automation to the procedures of data collection, standardization and storage would benefit the VID. As demonstrated by tools used in both academic and nonacademic circles, the added value of integrating machine learning models, or other artificial intelligence technologies into tools and workflows is high. It is our objective to streamline incident capture and harness some of the potentials of AI to improve the thoroughness of research and develop the representativeness of the records contained in the database. This can facilitate data validation and allow for more rapid operations overall. Hence, it merits consideration going forward.

As we have demonstrated in this paper, the VID complements various FoRB tools by tracking specific incidents. The focus on violence, religious freedom, and public sources means that the VID has limitations that must be acknowledged to use the data appropriately. Yet, the VID can offer data and analysis that, when combined with other research tools, can provide novel insights. The VID contribution of tracking the religious affiliation of both the actor and victim and data-rich categories with subnational information offers a unique contribution. Detailed information at this level is invaluable for understanding the nature of religious freedom violations in a given country. Religious freedom is deeply connected to human rights, therefore even those working beyond FoRB issues should pay attention to trends and developments illustrated by the VID. It is this very information which should inform ongoing research and policy analysis. The International Institute for Religious Freedom hopes the VID will become a trusted source for researchers and policy makers and is an important part of promoting freedom of religion or belief for all.