What is civil society and why do states repress it?
The first challenge in quantifying and measuring civil society repression is defining civil society itself. “Civil society” can be defined broadly as “the arena of uncoerced collective action around shared interests, purposes, and values” (Howell et al., 2008, p. 91), and includes both formal and informal organizations, as well as more amorphous phenomena like social movements and protests.
Civil society need not be explicitly political or rights-focused—any sort of voluntary association can be considered a civil society organization (CSO), including bowling clubs, self-help groups, faith-based organizations, and labor unions. These CSOs allow citizens to come in contact with each other, build social capital, and aggregate individual preferences (Putnam et al., 1994). NGOs represent a subset of CSOs and are often distinguished by narrower charitable purposes or specific tax designations; for instance, trade unions and professional organizations typically do not self-identify as NGOs, but as CSOs (Tomlinson, 2015, p. 124). Despite these (minor) definitional differences, the terms CSOs and NGOs are often used interchangeably (Salamon et al., 2004), creating additional challenges for consistent quantification.
Though less immediately threatening to political legitimacy than armed rebels, large-scale popular protests, or opposition electoral movements, voluntary CSOs can pose significant challenges to governments—both democracies and autocracies—and potentially trigger repressive responses. The relationship between states and civil society varies substantially depending on regime type, political context, and the nature of civil society activities. CSO–state relations can be categorized into three general types: CSOs can reinforce, oppose, or substitute state power (Heiss, 2019a).
First, CSOs can strengthen and bolster the state. In democracies, associational life “increases civic engagement, embeds norms of reciprocity into society, and helps improve democratic governance” (Heiss, 2019a, p. 561). In autocracies, CSOs can lend legitimacy to repressive governments (Brass, 2016) and keep dictators in power longer, transforming NGOs into “mediators between the people’s demands and the administration’s offers” (Néfissa & Abd al-Fattah, 2005, p. 8).
More threatening, however, is the notion that civil society can stand in opposition to—or even replace or substitute—the state. By participating in civil society, citizens are empowered to “stand up to city hall” (Jamal, 2007, p. 4) and find avenues for coordinating collectively to pursue their community interests. CSOs can act as watchdogs against government overreach and abuse, and their activities provide avenues for anti-government mobilization (Heiss, 2019a). This oppositional role is particularly evident in human rights advocacy.
For instance, domestic human rights CSOs can appeal to transnational networks of human rights NGOs, who then exert pressure on democratic governments, who in turn confront the repressive government, creating a boomerang effect that can improve human rights (Keck & Sikkink, 1998). With repeated boomerang-like pressure, domestic civil society and its international supporters can reshape human rights norms within a country, creating an upward spiral of norm socialization that leads to the institutionalization of improved human rights policies and practices within an offending state (Risse & Sikkink, 1999). Advocacy CSOs can also shame and pressure repressive states through international public opinion (Kelley, 2017; Murdie & Davis, 2012; Smidt et al., 2020), provide direct support for democratization efforts (Bush, 2015, 2019), and directly consult on domestic policy reforms (Heiss, 2019b).
Given potential threats to political stability, states work to direct, co-opt, or control CSOs to ensure regime survival, neutralize potential opposition, and reshape civil society in their favor. Regulations and repression are central to this strategy. Because NGOs and CSOs are more formally institutionalized, repressive states can limit these organizations in more targeted, nonviolent, and bureaucratic ways that avoid the potential consequences of violent crackdown (Chaudhry, 2022). Administrative crackdown—also known as NGO restrictions, anti-NGO laws, or the global crackdown on civil society (Carothers & Brechenmacher, 2014; Chaudhry, 2022; Chaudhry & Heiss, 2022a; Dupuy et al., 2021; Heiss, 2017)—allows regimes to impose legal and regulatory barriers to CSO entry, advocacy, and funding in ways that can seem innocuous at face value, but provide governments with significant leverage and control over civil society. Importantly, not all CSO-related regulations are inherently restrictive. Most countries regulate and oversee formal CSO activities in order to routinize behavior (Bloodgood & Tremblay-Boire, 2017), encourage transparency and accountability (Chaudhry et al., 2025), and improve CSO access to the policy-making process (Teets, 2014). Civil society regulations become repressive when governments use laws to limit, co-opt, or capture civil society activities and limit their potential influence on regime preferences.
Laws prohibiting entry make it difficult to engage in associational life and instead keep the general public oriented towards the state. Prohibitions against advocacy prevent CSOs from lobbying for improved human rights or against regime abuses. Funding laws effectively short-circuit international spiral and boomerang pressure by preventing domestic CSOs from collaborating with international partners and limiting the scope of their programs. CSO regulations also allow states to co-opt domestic civil society, “us[ing] nonprofit regulation as a tool of political control to shore up their continued rule” (Spires, 2020, p. 573) by reshaping CSO programming, strategies, and missions to be more aligned with government preferences (Lian & Murdie, 2023; Teets, 2014).
How has civil society repression been measured?
Quantification of civil society repression is difficult. One challenge is that most instances of state repression and human rights violations are already aimed at elements of civil society. Extralegal disappearances of social movement leaders, violent dispersal of protests, and arrests of community organizers and activists are all examples of states repressing informal components of civil society. Legal restrictions on CSOs are not typically included in common datasets that track human rights violations, such as the CIRI Human Rights Project (Cingranelli et al., 2014) or the Human Rights Measurement Initiative (HRMI) (Clay et al., 2020). The fact that many regulations are benign at face value add further complications—it is not always clear if a civil society-related law might count as restrictive or be used to repress rights. This measurement gap creates a systematic undercount of repression that overlooks how states increasingly rely on bureaucratic tools to control civil society.
Another challenge is that there is a mismatch between codified de jure laws that restrict civil society and the de facto implementation of those laws. Governments often have unenforced laws, either the vestigial remains of past legal regimes, or laws that are designed to be selectively enforced (Cartwright, 2016). NGO laws are no different. For instance, Egypt’s Law 84 of 2002 contained harsh provisions that were inconsistently applied until after the 2011 Arab Spring, when enforcement intensified (Chaudhry & Heiss, 2022b). Similarly, a sponsor of Russia’s 2015 Undesirable Organizations Law described it as “a weapon hanging on the wall that never fires,” explicitly acknowledging its goal to deter CSOs from challenging the state by chilling civil society activity (Kozenko, 2015).
In line with this conceptual split, scholars have developed two general approaches to measuring civil society repression, each addressing different aspects of the regulatory environment. The first focuses on de jure restrictions, cataloging and analyzing formal legal frameworks that constrain civil society activities. The second approach attempts to capture de facto repression, assessing how these laws are implemented in practice and their actual effects on civic space.
De jure repression
Even in the most repressive authoritarian regimes, states often use statutes and regulations to provide legal justification for repression. Since states tend to publish their criminal and civil legal codes in open forums, scholars often analyze these laws to determine the de jure legal environment for human rights. Research organizations like the International Center for Not-for-Profit Law (ICNL) monitor legislative and regulatory developments related to civil society and publish detailed annual summaries of dozens of countries’ de jure legal environments. Other projects collect data on specific attributes of civil society laws, like the Global Register of Nonprofit Data Sources (GRNDS), which provides a standardized schema for describing nonprofit regulations (Bloodgood et al., 2023)
Researchers have used these resources to compile country-year panel datasets documenting the presence or absence of different types of anti-NGO laws. Through close reading of ICNL summaries, UN and US State Department reports, think tank research reports, and national constitutions, Christensen & Weinstein (2013) offer a dataset of specific types of legal barriers to entry, funding, and advocacy for NGOs in 98 countries between 1909 and 2012. Similarly, Dupuy et al. (2016) report data on laws restricting foreign funding for NGOs in 192 countries from 1993 to 2012, while Bromley et al. (2020) count foreign funding laws for 60 countries from 1994 to 2015.
In general, research using these datasets of laws (measured as counts) has tended to focus on either the determinants of foreign funding laws (Bromley et al., 2020) or the effect of these laws on foreign aid (Christensen & Weinstein, 2013; Dupuy et al., 2016; Dupuy & Prakash, 2018). Though Christensen & Weinstein (2013) collected data on legal barriers to entry and advocacy, their focus was primarily on funding barriers alone and they leave these other variables largely unexplored.
Both Glasius et al. (2020) and Chaudhry (2022) have since expanded Christensen and Weinstein’s data to include 96 countries through 2016 (Glasius et al., 2020) or 130 countries through 2013 (Chaudhry, 2022), and both use all three categories of legal barriers—funding, entry, and advocacy—as outcome variables to explore the determinants of anti-NGO laws. Chaudhry & Heiss (2024) also use laws (as counts) to examine their effect on foreign aid and whether additional new anti-NGO laws predict future human rights abuses (2022b), while Fransen & Dupuy (2024) explore the relationship between laws and the size of the international NGO sector in repressive countries.
While counting anti-NGO laws has value in researching legal restriction, binary indicators fail to capture the varying scope, severity, or enforcement of different regulations (see DeMattee, 2019). Additionally, the emphasis on national-level legislation can obscure important subnational variation in regulatory environments, particularly in federal systems where provincial or local authorities may have significant regulatory authority (Ye & Heiss, 2025).
De facto repression
Focusing on counts of laws overlooks the selective enforcement that shapes actual civil society experiences, and data on the de facto implementation of laws allows researchers to explore on-the-ground repression of CSOs. However, measuring de facto repression is more difficult than simply counting laws. The informal nature of many restrictions, selective enforcement patterns, and the difficulty of observing self-censorship or anticipatory compliance all complicate efforts to quantify the actual environment for civil society. Recent attempts have ranged from expert assessments and public opinion surveys to counts derived from human rights reports and event data, with distinct methodological strengths and limits. Most of these newer measures aggregate conditions across diverse civil society sectors, potentially obscuring how states may differentially treat service providers versus advocacy organizations or domestic versus international NGOs.
In the early 2000s, research on associational life under authoritarianism tended to use in-depth qualitative fieldwork and interviews (Henderson, 2002; Jamal, 2007) or broader public opinion surveys. The World Values Survey and regional projects like the Arab Barometer, Afrobarometer, and Asian Barometer all include questions about perceptions of the openness of civil society and individual participation in CSOs (Jamal, 2007). Though rich in detail, these surveys are conducted irregularly, thus making them unsuitable for cross-country longitudinal analysis.
Around the same time, CIVICUS—a research and advocacy organization focused on civil society—launched its Civil Society Index (CSI), which used expert surveys to rate the legal environment of 30–50 states along four dimensions: the sector’s structure, external environment, values, and impact (Heinrich, 2004). In the 2010s, the CSI evolved into the CIVICUS Monitor (CIVICUS, 2024a), which uses expert surveys and quantitative human rights indicators to generate scores on a scale of 1–100 with five levels of civic space: open, narrowed, obstructed, repressed, and closed. CIVICUS Monitor scores are based on a weighted average of common measures for freedom of association, assembly, and expression (CIVICUS, 2024b) and are thus conceptually broader than just civil society repression. However, the project has wide coverage with annual scores for 196 countries since 2018, and could prove useful for future research on the repression of civic space more generally.
Researchers have also extracted details of civil society repression from broader human rights datasets. Among the rights measured by the CIRI Human Rights Project (Cingranelli et al., 2014) is a three-category measure of assembly and association freedom based on US State Department reports, capturing how easily citizens can participate in trade unions, cultural organizations, or domestic and international NGOs. Each year, countries receive a score—severely restricted, limited, and unrestricted—based on the severity and universality of civil society restrictions. CIRI data is available for nearly 200 countries annually, with scores beginning in 1981, making it ideal for cross-sectional analysis. However, with only three possible outcomes the data fails to pick up minor changes in the civil society regulatory environment (Bakke et al., 2020).
Similar limitations apply to data from HRMI, which uses Bayesian item response theory models based on expert surveys to generate more granular measures of recognized civil and political human rights, including the right to assembly and association (Clay et al., 2020). HRMI only offers data on associational rights for 47 countries from 2017 to 2023, but has plans for broader coverage in the future.
The Varieties of Democracy project has emerged as perhaps the most comprehensive source for measuring de facto conditions (Coppedge et al., 2025). Using a sophisticated Bayesian IRT measurement model based on thousands of expert surveys, V-Dem provides hundreds of democracy-related indicators and indices for 202 countries from 1789 to 2025, including several focused specifically on civil society regulations. V-Dem’s Core Civil Society Index (CCSI) is a weighted average of indices related to the strictness of CSO entry and exit requirements, the openness of public participation in CSOs, and the severity of government repression that is targeted specifically at CSOs (Bernhard et al., 2015). Many studies on civil society repression now rely on both the CCSI and its more specific index of civil society repression (Chaudhry et al., 2024; Springman et al., 2022).
Even with V-Dem’s extensive coverage, it still collapses many details of civil society repression into single values. For instance, its civil society repression measure asks if CSOs face violent deterrence, incarceration, financial liquidation, denial of social services, fines, censorship, or burdensome paperwork requirements—an exceptionally wide range of possible forms of repression. Similar to the approach of counting anti-NGO laws, Bakke et al. (2020) collect data on a range of specific anti-NGO activities in 149 countries between 1994 and 2014, including ten types of subtler actions like bans on travel, censorship of NGO publications, and surveillance of civil society activists. This approach provides a richer picture of the civil society regulatory environment, but with substantially less coverage than what V-Dem offers.
Quantifying de jure and de facto civil society repression
In my own research, I am interested in the downstream consequences of civil society repression. How do bilateral and multilateral donor agencies respond to increased civil society repression in recipient countries? How do CSOs adjust their strategies and programming when facing new laws and more restrictive environments? How do individual private donors think about civil society repression—do they care if their favorite nonprofit is kicked out of a host country? Instead of focusing on just one form of repression, I am most interested in the intersection of the two. How do CSOs and donors navigate the uncertainty inherent in the selective enforcement of formal laws? Exploring both types of civil society repression simultaneously yields fascinating insights, but to answer these questions, I need reliable measures of both de jure and de facto repression. As seen earlier, capturing both types of repression in a nuanced, cross-national, longitudinal way is difficult. To illustrate this complexity, let me explain how my collaborators and I settled on the de jure and de facto data we have used for several projects (Chaudhry & Heiss, 2022a, 2022b, 2024).
The data on barriers to CSO entry, funding, and advocacy in 98 countries initially collected by Christensen & Weinstein (2013) has proven valuable as a foundation for ongoing data collection on formal NGO laws—both Glasius et al. (2020) and Chaudhry (2022) use the same categorization of laws, while Bromley et al. (2020) use many of the same sources. Each of these data collection projects ends its coverage somewhere between 2012 and 2016, and each covers a different set of countries and laws: Bromley et al. (2020) look at nearly every country but only examine foreign funding laws, while Glasius et al. (2020) and Chaudhry (2022) look at a wider range of laws in 100–130 countries.
In an attempt to create data for a full range of CSO laws in every country up through the 2020s, I collected all of the original sources used by these other projects: ICNL annual reports, USAID’s NGO Sustainability Index, US State Department reports, country notes from the United States International Grantmaking project, and annual reports from Freedom House, Human Rights Watch, CIVICUS, NGO Monitor, and the Carnegie Endowment for International Peace, World Movement for Democracy, and Global Integrity. I compiled all these sources into a large, structured, searchable corpus of text, with the hope of identifying and coding laws in an automated way.
Many of these data sources lend themselves to possible automatic coding—reports from ICNL, the US State Department, and CIVICUS all generally follow a similar structure, which could allow for better algorithmic discovery. I created rudimentary text classifiers to identify regulations based on Christensen and Weinstein’s original typology, but these text-as-data methods were unable to reliably identify laws due to substantial cross-country variation in legal language. In recent years, I have periodically attempted to use this corpus of reports with large language models (LLMs) to automatically classify laws, but most LLM-based classifiers have been designed for tasks like spam identification and sentiment analysis (Wang et al., 2024), not careful legal analysis. Thus, as with previous projects, Suparna Chaudhry and I are in the process of manually expanding de jure data.
Though more automated methods are not quite suitable for collecting de jure civil society repression data, they might show more promise for de facto civil society repression data. An established literature in political science uses automated coding of event data to track civil unrest, popular protests, military conflict, and human rights abuses based on millions of news reports published on the internet. Projects like the Global Database of Events, Language, and Tone (GDELT) or the Integrated Conflict Early Warning System (ICEWS) and its successor POLitical Event Classification, Attributes, and Types dataset, or POLECAT, (Halterman et al., 2023) use computational natural language text algorithms and dictionaries to parse text and determine who did what to whom, where, and when.
The outputs of this kind of analysis can identify specific actors and events like “Egyptian state security forces arrested protestors” and can provide highly detailed data about the prevalence of human rights abuses. Event data has even been used to measure civil society activity; for example, in partnership with the now-defunct Integrated Data for Events Analysis (IDEA), Murdie & Davis (2012) explore the effect of NGO activities on broader human rights by identifying all instances where international human rights NGOs shamed governments from 1992 to 2004. Event data such as these could be used to categorize other types of civil society activity.
Since new projects like POLECAT use open-source data, dictionaries, and classifiers, I have attempted to define CSO actors and activities and generate event data similar to Bakke et al.’s (2020) types of civil society repression. However, as it stands now, the data is not designed for civil society-specific work. For instance, of the 347,439 events in POLECAT’s 2021 data, only 98 identify an NGO as a recipient actor, and these tend to be large international NGOs like Oxfam, Amnesty International, and Greenpeace (Scarborough et al., 2023). As with formal laws, I have been unable to automatically code de facto civil society repression. To produce their dataset, Bakke et al. (2020) closely read hundreds of reports from news agencies, governments, and think tanks. As this is more resource-intensive than closely reading the more limited set of reports of de jure restrictions, I have not attempted to extend their data beyond 2014.
With newer developments in computational methods and text analytics, we can come close to creating new measures that could go beyond simple counts of laws or V-Dem IRT model indexes, but methodologically, we are not quite there yet. However, as described by Rebecca Cordell in this volume, the future for this kind of automated text-based measurement is promising, and newer methods can allow researchers to more carefully identify, extract, and measure a variety of human rights concepts from large corpora of text.
Combining de jure and de facto regulations
I am most interested in exploring the interaction between de jure and de facto regulations. Given the lack of good data on the heterogenous effects of both formal laws and the actual conditions on the ground, for now I have settled on an imperfect—but pragmatic—approach to measuring both forms of civil society repression. I use counts of laws based on Chaudhry (2022) to capture de jure repression and—despite its very broad concept conceptualization—I use V-Dem’s CCSI and its civil society repression sub-index for measuring de facto repression. Neither approach is ideal, but the two datasets offer the widest and most comprehensive coverage of both forms of repression in the greatest number of countries and years.
In particular, I have examined (1) where and why these different measures do not align, and (2) the effects of these different forms of civil society repression on outcomes like political terror and foreign aid. Analyzing these two measures simultaneously provides a far more nuanced and richer understanding of the dynamics of civil society repression.
Suparna Chaudhry and I explore whether civil society repression can act as a sort of “canary in the coal mine”—an early warning signal of more violent political terror or violations of physical integrity rights and civil liberties (Chaudhry & Heiss, 2022b). We build a set of multilevel Bayesian models to predict future values of two general human rights indexes using both NGO law counts and V-Dem’s CCSI. To explore heterogeneity in de jure regulations, we look at both the total count of NGO laws and counts of narrower categories of laws like barriers to advocacy, entry, and funding. We find that, in general, one new anti-NGO law is associated with a negligibly greater probability of seeing higher values of political terror and a relatively insubstantial (5%) decline in predicted latent human rights scores (Fariss, 2014). The association with latent human rights is slightly larger if the new law is a barrier to advocacy, but in general, new laws do little to predict future repression.
The de facto regulatory environment, on the other hand, is substantially predictive of additional human rights abuses. A one-unit increase in civil society repression is associated with a 25% decline in respect for human rights in the following year. Holding all else constant, de facto civil society repression is a stronger and more reliable indicator of future repression than simply the presence of laws.
This divergence in repression is unique and illustrative of the selective and targeted nature of civil society regulations—these laws do indeed act as “a weapon hanging on the wall that never fires” until needed by the regime. We illustrate this finding by looking at the case of Egypt, tracking its passage of anti-CSO laws alongside its levels of de facto civil society repression over time. Egypt’s civil society environment improved in the years following the passage of Law 84 of 2002 until immediately prior to the 2011 uprising, which led to a worsening of human rights more broadly. Due to the state’s discretionary enforcement power, the law and its implementation were disconnected: the introduction of the formal law in 2002 did not presage immediate changes to general human rights.
In other research (Chaudhry & Heiss, 2022a, 2024), we find that international foreign aid donors also respond differently to de jure and de facto civil society repression. Donors appear to respond forcefully to legal barriers to advocacy, driving as much as a 43% drop in bilateral foreign aid in the year following a new law. However, there is less of a reaction to barriers to entry or funding: there are no substantial differences in aid levels in the following year after these types of anti-NGO laws are adopted. Donors are also less responsive to de facto repression and are hesitant to change aid allocations following a decrease in the general civil society regulatory environment. It seems that donors care most about more visible restrictions on advocacy—other forms of administrative crackdown might be too subtle to elicit changes in foreign aid.
The future of measuring civil society repression
As seen in my attempts at collecting richer details about civil society repression, there is enormous potential for future work, limited mostly by time and resources. Existing approaches to measuring de jure and de facto repression have formed a strong foundation for quantitative research on the phenomenon of closing civic space, and methodological innovation and conceptual refinement can push scholars to more complex research questions in the future.
In addition to potentially working towards automated coding of regulations from government and research center reports, future work on formal laws should explore more heterogeneity in regulatory regimes. DeMattee (2018) presents a typology of formal CSO regulations along two different dimensions: (1) the enabling environment, or whether regulations permit or restrict CSO operations, and (2) the complexity of the regulations, or the extent with which complying with civil society laws is burdensome. These dimensions apply to four specific characteristics of CSO operations—governance, formation, operations, and resources. DeMattee contends that governments pass laws and regulations that expand or contract the allowable space for CSO work, and that introduce or reduce complexity in following those regulations. Future data collection can move beyond the entry, funding, and advocacy categories and instead examine how laws enable or add complexity to specific CSO actions. Focusing on the quality and purpose of CSO laws creates a more multidimensional picture of de jure regulations than simply indicating if specific laws exist.
Adding nuance to the collection of formal laws follows Bakke et al.’s (2020) research on more detailed de facto civil society repression. Their data on specific types of extra-legal civil society repression like travel bans, censorship, and surveillance is a helpful counterpart to the simpler—and more widely available—measures offered by V-Dem, and should be expanded into the 2020s. This will require substantial resources for either hand-coding or automatically coding event data. Future work could develop more sophisticated event data frameworks specifically designed to capture civil society-related repression. This would involve expanding the actor dictionaries in projects like POLECAT to include a broader range of CSOs, not just prominent international NGOs. While my initial attempts with general-purpose text classifiers and LLMs proved disappointing, models specifically focused on civil society-related corpora might yield better results. Again, all these computational projects would require substantial time and resources, but they could lead to broader coverage of de facto civil society environments.
As more detailed data is collected for more countries and years, we can answer richer questions about the interaction between de jure and de facto regulations. For example, data about the specific nature and aims of de jure laws would allow researchers to explore whether certain types of formal restrictions (i.e., entry, funding, and advocacy) or the goals of CSO regulations (i.e. adding complexity or constraining the enabling environment) consistently precede specific types of practical repression (e.g. arrests, censorship, additional paperwork, etc.).
Making advancements in measuring civil society repression is not just a methodological exercise. Civic space continues to constrict globally, including in established democracies like the United States (CIVICUS, 2024a). As states refine their approaches to restricting civil society, researchers must refine their data and methods in turn to better detect the effects of this subtler type of human rights repression.