April 12, 2021

This is an early draft of a working paper.

Access the code and full analysis notebook at https://stats.andrewheiss.com/canary-ngos/

Canary in the Coal Mine? Civil Society Repression as a Predictor of Human Rights Abuses

Suparna Chaudhry

Lewis & Clark College

Andrew Heiss

Georgia State University

Many countries across the globe have increasingly cracked down on NGOs over the last few decades. These attempts to repress civil society organizations—many of which are sanctioned by law—represent a bureaucratic form of repression that could indicate more severe human rights abuses and democratic backsliding. This is especially the case for electoral democracies, which unlike autocracies, may not aggressively attack civic space. In this paper, we explore if crackdowns on NGOs predict broader human rights repression. Anti-NGO laws are among the most subtle means of repression and attract the least least domestic and international condemnation, as many audiences mistake this repression as regulation and are not personally affected by it. These laws may also make it easier for states to prevent and deter future large-scale popular mobilization. Using original data on civil society restrictions over the last three decades, we test whether NGO crackdown is a predictor of political terror, physical integrity rights violations, and violations of private civil liberties. We find that while formal de jure laws provide little information in predicting future repression, de facto civil society repression predicts worsening respect for physical integrity rights and civil liberties.
Keywords—civil society restrictions, human rights, political terror, civil liberties
Acknowledgments: We would like to thank participants of the 2021 International Studies Association (ISA) convention for their comments and feedback.

Governments across the globe have been cracking down on civil society organizations (CSOs), a phenomenon commonly referred to the closing or shrinking of civic space (Carothers and Brechenmacher 2014; Dupuy, Ron, and Prakash 2016; Chaudhry 2016; Dupuy and Prakash 2018). The last two decades have seen a proliferation of steps designed to limit the influence of CSOs and non-governmental organizations (NGOs), and more than 100 countries have proposed or enacted 244 measures restricting, repressing, or shutting down civil society since 2013 (International Center for Not-For-Profit Law 2021). In Human Rights Watch’s 2016 World Report, Executive Director Kenneth Roth argued that civil society was under more aggressive attack than at any time in recent memory, and over the past decade, the majority of the world has seen a substantial narrowing of allowable civic space. As seen in Figure 1, by the end of 2020, only 21% of countries had open and unrestricted civil societies (CIVICUS 2021; Chaudhry and Heiss 2021).

2020 CIVICUS Monitor civic space ratings

Figure 1: 2020 CIVICUS Monitor civic space ratings

This hostility to civil society is not unique to autocratic governments or even illiberal democracies (Abramowitz and Schenkkan 2018)—long-established liberal democracies have also increased restrictions on civil society in their countries. Over the past decade, India has canceled the licenses of tens of thousands of NGOs, targeted major groups such as Greenpeace International and Ford Foundation, and has banned 5,000 NGOs from receiving foreign funds (Press Trust of India 2015, 2019). In 2013 Canada attempted to revoke the charitable status of civil society organizations for opposing the Northern Gateway pipeline and Canadian mining companies’ initiatives abroad (CIVICUS 2013). In 2019, activists from the US-based nonprofit No More Deaths were arrested for providing food and water to migrants a the US-Mexico border and faced a 20-year prison sentence before their convictions were overturned in 2020 (Cramer 2020).

Anti-civil society measures are occasionally violent, but recently, government repression of NGOs has generally been non-violent. Rather than publicly arrest or beat activists, states implement “administrative crackdown,” or the passage of laws to create barriers to NGO advocacy, entry, and funding (Chaudhry 2016). The use of legal tools instead of violent repression means that NGOs have to adapt their practices to stay within the bounds of permitted activities and operations (Heiss 2019, 2017). While in some cases this has led to the depoliticization of NGOs and avoidance of direct political activism by CSOs, states cannot soften or neutralize all forms of political activities in subtle, legal ways like they have with NGOs. Sooner or later, states need to deal with threats to their stability in a more direct, violent, confrontational manner. Administrative NGO laws and general civil society repression could be an important early warning sign—or a canary in the coal mine—of worsening human rights repression in a state, signaling that a state will continue down the path of further repression. What does state targeting of NGOs tell us about the state’s broader respect for civil society and the overall level of human rights repression? Can the crackdown on NGOs act as an indicator for subsequent increasing repression?

In this paper, we explore the power of anti-NGO repression in predicting human rights abuses, political terror, physical violence, and violations of private civil liberties. We argue that information about NGO crackdown can be a valuable indicator of human rights trajectories. In some cases, such crackdown can act as an early warning signal for worsening repression, suggesting which countries are testing the waters with comparatively “milder” methods of crackdown before enacting more violent forms of repression. To assess if attacks on NGOs and activists can provide us useful information about future repression, we use an original dataset on state crackdown on NGOs using legal repression across all countries from 1990–2013. We differentiate between two forms of civil society repression: official de jure anti-NGO restrictions that are codified in law, and general de facto levels of civil society repression. We propose and test two empirical expectations—that de jure and de factor restrictions predict future political terror, physical integrity rights violations, and government encroachment on private civil liberties.

Using multilevel Bayesian modeling, we find that formal de jure laws provide very little additional information when attempting to predict future repression. However, we find that accounting for more general civil society repression does boost our models’ predictive power. Increasing de facto civil society repression predicts an increase in the probability of moderate political terror and results in worsening violations of physical integrity rights and civil liberties.

Our paper makes an important contribution in predicting the onset of repressive state policies. While structural factors such as levels of democracy, economic development, electoral competitiveness can identify countries that are likely to have more or less respect for civic freedoms, these factors are also slow to change over time. As such, they may not be particularly suitable for forecasting short-term trajectories, especially for moderately repressive countries. On the other hand, state repression of NGOs is more dynamic and provides information about changing levels of human rights standards and repression in subsequent years.

Below, we first look at existing literature on repression and political violence, outlining state motivations to repress NGOs. We then elaborate on why attacking NGOs may be an important predictor of subsequent repression. Using both original data as well as data from several existing datasets, we build an ensemble of predictive models and explore (1) the marginal effects of civil society repression on different forms of broader human rights repression, and (2) the out-of-sample predictive performance of each model. We conclude with a discussion of our findings and their implications for assessing the repression of human rights.

Government repression and NGO restrictions

Repression is the use of coercive action against an individual or organization, within the territorial jurisdiction of the state, for the purpose of imposing a cost on the target as well as deterring specific activities and/or beliefs perceived to be challenging to government personnel, practices, or institutions (Goldstein 1978, xxvii; Poe and Tate 1994; Davenport 2007b; and Ritter and Conrad 2016).

Early literature on repression focused on the role of structural factors such as levels of democracy, economic development, electoral competitiveness in predisposing a state to increased use of repression. Most notably, democratic forms of government—and democratization in general—has a pacifying effect on state repression (Davenport and Armstrong 2004; Krain 1997; Richards 1999). Democracies have a negative effect on repression largely because of the increased probability of sanctions against authorities for undesirable behavior (Davenport 2007a, 47).

More recent research focuses on how the behavior of various actors shapes government response. One of the most consistent findings in the political violence literature is that authorities repress to counter ongoing challenges to the status quo. In other words, dissent increases repressive behavior (Davenport 2007b). However, dissent and repression have an endogenous relationship: state authorities also frequently engage in preventive repression to undermine or restrict groups’ abilities to dissent (Ritter and Conrad 2016; Conrad and Ritter 2019). In addition to attacking groups willingness to organize against states, such preventive repression also damages states’ capacity to impose costs (Davenport 2007a, 47). NGOs are one actor that can pose challenges to political authorities, and state repression of these actors, especially through legal restrictions or administrative crackdown can also be considered a form of preventive repression (Chaudhry 2016).

A robust, strong civil society is widely considered as an important component on democratic transition and consolidation (Putnam, Leonardi, and Nanetti 1994). Politically, as important civil society actors, NGOs enable citizens to voice their grievances, seek redress, and carry out lobbying and advocacy activities to further their interests. They help keep government accountable, monitor human rights violations and corruption. From an economic perspective, NGOs provide services when the state falls short. As a result of these roles, NGOs can wield real influence, which can be threatening to states. State crackdown on NGOs has become pervasive as these organizations have the ability to influence electoral politics (Dupuy, Ron, and Prakash 2016; Weiss 2009), aid mobilization (Murdie and Bhasin 2011; S. R. Bell et al. 2014; Boulding 2014), and threaten a state’s economic interests (Dietrich and Murdie 2017; Lebovic and Voeten 2009; Nielsen 2013).

There is no doubt that violence is often effective at curbing dissent. But it can also backfire, leading to widespread protests, decreasing leaders’ legitimacy and increasing their criminal liability (Kim and Sikkink 2010; Sikkink 2011). In contrast to violence, administrative crackdown has fewer adverse consequences—citizens may see it as regulation rather than repression, resulting in little to no counter-mobilization. It may also circumvent concerns among state security or bureaucratic agents about illegitimate orders based on violent repression (Stephan and Chenoweth 2008). There are also fewer international consequences since laws—unlike violence—rarely invite condemnation or threats of aid withdrawal (Carothers and Brechenmacher 2014).

We argue that crackdown on NGOs foreshadows future government transgressions. Autocratic leaders want to avoid accountability, and NGOs often stand in conflict with that goal. Civil society organizations, including NGOs, can help shape the narrative of leaders’ performance both domestically and internationally. Monitoring by NGOs and subsequent shaming can include significant external revenue loss to states in the form of loss of aid, trade, and foreign direct investments (Keck and Sikkink 1998; Barry, Clay, and Flynn 2013). Government attempts to repress NGOs can signal the beginning of a deteriorating human rights situation. Such repression reflects the government’s willingness and capability to use measures that can make it avoid accountability. Once a state targets NGOs, over time it may grow less hesitant to apply repression more widely and more severely.

In exploring the impact of state crackdown on NGOs on subsequent human rights conditions, we look at both de jure measures—or formal legislation that seeks to repress NGOs—as well as the de facto implementation of these measures. This distinction is important. While states may pass anti-NGO laws, these laws can also lie dormant. For instance, Russia passed its 2015 Undesirable Organizations Law with the purpose of targeting and eliminating specific foreign-connected NGOs it deemed dangerous. One of the law’s sponsors, Aleksandr Tarnavsky, described the law as a preventative measure that would not affect the majority of NGOs working in Russia. Rather, the law would be a “weapon hanging on the wall that never fires” and stand as a warning to potentially uncooperative NGOs (Kozenko 2015).

Simply noting the passage of (or counting) these laws may therefore miss the effect of their de facto implementation. Moreover, the implementation of these laws may not be limited to its legal character. The government may also engage in harassment of NGOs and activists to dissuade NGOs from operation or expressing themselves. This may include measures such as detentions, short-term incarceration, as well as beatings, threats to families and destruction of valuable property.

Finally, counting official laws can also undermeasure civil society repression because governments can repress NGOs without passing or relying on formal laws. As prior research in Egypt, Ethiopia, and Russia points out, “rather than systematically enforcing all repressive legal measures-often impossible owing to capacity constraints,” governments may choose to use extra-legal means to intimidate, harass or prosecute a few NGOs or activists. These cases, especially during moments or crises or contention, “highlight to the wider NGO community-and citizenry-that the government is willing and capable of using the law to repress dissent” (Brechenmacher 2017, 94).

Based on these purposes of civil society restrictions and the different methods states use to implement them, we propose two empirical expectations about the relationship between civil society repression and general human rights:

Data and modeling approach

To test these empirical expectations, we use a new dataset of civil society laws, as well as measures from several other larger datasets focused on human rights, political terror, and democratization. Table 1 provides a summary of our key variables.

Table 1: Outcomes and predictors
Measure Category Description Source
Political Terror Scale (PTS) Ordered category ranging from 1 (least terror) to 5 (most terror) Gibney et al. (2019)
Physical violence index 0–1; higher values indicate less violence Coppedge et al. (2020); v2x_clphy
Private civil liberties index 0–1; higher values indicate greater respect for civil liberties Coppedge et al. (2020); v2x_clpriv
Latent human rights values ≈ -3–5; higher values indicate greater respect for human rights Fariss, Kenwick, and Reuning (2020)
Total legal barriers De jure laws Count of all anti-NGO legal barriers Chaudhry (2016)
Barriers to advocacy De jure laws Count of laws restricting NGO advocacy Chaudhry (2016)
Barriers to entry De jure laws Count of laws restricting NGO entry, including registration Chaudhry (2016)
Barriers to funding De jure laws Count of laws restricting NGO funding sources Chaudhry (2016)
CSO repression De facto repression ≈ −3–3; higher values indicate less repression Coppedge et al. (2020); v2csreprss

Measuring de jure and de facto civil society repression

To measure de jure repression, we use data on anti-NGO laws from Chaudhry (2016) which includes official NGO laws for 162 countries between 1990 and 2013. We create indexes of the total number of laws passed in each country based on three broad categories of laws related to barriers to entry, funding, and advocacy, and we also sum each of these indexes to create a count of total legal barriers. We also create a yearly indicator variable to show if a country increased its count of anti-NGO laws (models using this outcome are included in the online appendix).

  • Barriers to advocacy (2 possible barriers): NGOs are restricted from engaging in political activities, and advocacy restrictions differ if the NGO receives foreign funding
  • Barriers to entry (3 possible barriers): registration is burdensome, NGOs are not allowed to appeal if denied registration, and entry requirements differ if the NGO receives foreign funding
  • Barriers to funding (5 possible barriers): NGOs need prior approval to receive foreign funding, NGOS are required to channel foreign funding to state-owned banks, NGOS face additional restrictions if receiving foreign funding, and NGOs are prohibited from receiving foreign funding
Number of legal barriers to NGO activity per country over time (de jure legislation) and average level of CSO repression across democracies and non-democracies (de facto implementation)

Figure 2: Number of legal barriers to NGO activity per country over time (de jure legislation) and average level of CSO repression across democracies and non-democracies (de facto implementation)

As mentioned previously, simply counting the number of NGO laws does not necessarily capture the restrictiveness of the general civil society environment in a country. Governments might repress civil society without formal laws or pass restrictive laws that they can keep in reserve for later emergencies. To measure the general level of civil society repression in a country, we use V-Dem’s CSO repression indicator (v2csreprss in Coppedge et al. (2020)), which captures the extent to which governments harass, deter, liquidate, and arrest members of civil society organizations. Figure 2 shows each of these types of legal barriers per country over time, as well as average levels of civil society repression by regime type. While anti-NGO laws have increased steadily since 1990—particularly laws related to foreign funding—there has not necessarily been a corresponding rise in civil society repression, which has remained fairly constant across different regime types. This divergence is likely indicative of the split between de jure laws and their de facto implementation—a rise in laws is not necessarily tied to direct CSO repression.

Measuring repression

To see how both de jure and de facto civil society restrictions influence general human rights repression, we use four different quantitative measures of repression, each of which capture different dimensions of political and civil violence. First, following Gohdes and Carey (2017), who use the killing of journalists to predict worsening human rights conditions, we use the Political Terror Scale (PTS), which assigns countries to five different categories of repression each year (Gibney et al. 2019), with Level 5 representing widespread political violence, including arrests, political murders, disappearances, and torture. As seen in the left panel of Figure 3, the number of countries at Level 5 has decreased since the 1990s, while most countries are coded as Levels 2 and 3.

Because the PTS captures broad notions of political terror with only five possible categories, we also look at three continuous measures of repression. From V-Dem, we use a physical violence index (v2x_clphy) that measures freedom from torture and political killings specifically, and a private civil liberties index (v2x_clpriv) that measures respect for property rights, labor rights, freedom of religion, freedom of movement, and other non-political human rights. The middle panel of Figure 3 shows that the average level of these two rights indexes has improved over time.

However, looking at political terror, physical integrity rights, and private civil liberties captures specific dimensions of a country’s respect for human rights, but isolating these trends in repression can prove difficult. Schnakenberg and Fariss (2014), Fariss (2014), and others have argued that a country’s latent and underlying human rights environment causes a state to engage in specific actions related to human rights, such as engaging in political violence, arresting activists, or restricting freedom of religion. Measuring this underlying respect for human rights is difficult, however, as it is by definition unobserved. Fariss, Kenwick, and Reuning (2020) use a Bayesian measurement model to generate estimates of latent respect for physical integrity rights. We use the posterior mean of their latent variable (\(\theta\)) as our final measure of repression in order to see how civil society restrictions predict changes in a country’s future underlying human rights environment. The right panel of Figure 3 demonstrates the global increase in average latent human rights values over time.

Key measures of repression over time

Figure 3: Key measures of repression over time

Modeling approach

Our main question in this paper is whether civil society laws and restrictions serve as an early warning signal to other future forms of repression. As such, we are primarily interested in predicting repression and not isolating the causal effect of civil society repression on general human rights abuses. We do not include a complete set of covariates to remove confounding relationships, but instead use a more parsimonious set of explanatory variables that strongly predict repression. Following Gohdes and Carey (2017), we control for a country’s previous level of repression, its level of democracy as measured by V-Dem’s polyarchy index, its logged GDP per capita (measured by the UN), the percent of GDP attributed to trade (measured by the UN), and an indicator for whether the country witnessed more than 25 battle-related deaths (measured by UCDP/PRIO (2020); Gleditsch et al. (2002)). Because changes in the human rights environment take time to be reflected by these larger indexes, we model the effect of lagged explanatory variables on repression one year in the future. We also include the lag of civil society laws or repression to account for serial correlation.

We extend Gohdes and Carey (2017)’s previous modeling approach by accounting for country-level variation in repression with multilevel modeling and we include random effects for each country. Accounting for the country this way results in intraclass correlation coefficients of around 0.7 for each of our three outcome variables, which means that the country-based structure of the data explains more than 70% of the variation in outcome, resulting in model estimates that are arguably more precisely measured. Additionally, as a robustness check, we account for within-country and between-country variation by splitting each numeric explanatory variable into its country-level mean and its variation from the mean, following A. Bell and Jones (2015). The results from these random effects—within/between (REWB) models are nearly identical to those with the untransformed covariates, so for the sake of simplicity here, we only include results from basic multilevel random effects models.

We run multiple models for each of our empirical expectations, based on different combinations of our outcomes and key explanatory variables:

  • Expectation 1a: Anti-NGO laws (total, advocacy, entry, funding separately) signal worsening political terror
  • Expectation 1b: Anti-NGO laws signal worsening physical violence
  • Expectation 1c: Anti-NGO laws signal constricting private civil liberties
  • Expectation 1d: Anti-NGO laws signal decreasing latent human rights
  • Expectation 2a: Civil society repression signals worsening political terror
  • Expectation 2b: Civil society repression signals worsening physical violence
  • Expectation 2c: Civil society repression signals constricting private civil liberties
  • Expectation 2d: Civil society repression signals decreasing latent human rights

We generate our predictions with two families of Bayesian regression models. Because the Political Terror Scale is measured as an ordered category, we use ordered logistic regression to predict future values of the PTS. As V-Dem’s physical violence index and private civil liberties index and latent human rights values are all measured continuously, we predict these values with Gaussian models. We use median values from the models’ posterior distributions as our point estimates and provide credible intervals using the 95% highest posterior density.1 We declare an effect statistically “significant” if the posterior probability of being different from zero exceeds 0.95.

Complete results from all the models (including REWB models), along with posterior predictive checks, goodness-of-fit measures, and prediction diagnostics are all available at a companion website at https://stats.andrewheiss.com/canary-ngos/. Our general modeling approach can be summarized as follows:

\[ \begin{aligned} y_{i, t+1} \mid u_i, x_{i, t} \sim&\ \operatorname{Ordered Logistic}(y^*_{i, t + 1}, \sigma^2_m) & \text{[likelihood for ordinal PTS models]} \\ y_{i, t+1} \mid u_i, x_{i, t} \sim&\ \mathcal{N}(y^*_{i, t + 1}, \sigma^2_m) & \text{[likelihood for Gaussian V-Dem models]} \\ y^*_{i, t + 1} =&\ \beta_0 + \beta_1 \text{Civil society repression}_{i, t} + & \text{[model with covariates]}\\ &\ \beta_2 \text{Civil society repression}_{i, t - 1} + \\ &\ \beta_3 \text{Outcome}_{i, t} + \beta_4 \text{Polyarchy}_{i, t} + \\ &\ \beta_5 \log (\text{GDP per capita}) + \\ &\ \beta_6 \text{Trade as } \% \text{ of GDP}_{i, t} + \\ &\ \beta_7 \text{Armed conflict}_{i, t} + u_i + \sigma^2_e \\ u_i =&\ \mathcal{N}(0, \sigma^2_u) & \text{[country-specific intercepts]} \\ \ \\ \beta_0 \sim&\ \mathcal{N}(0, 10) & \text{[prior population intercept]} \\ \beta_{1-7} \sim&\ \mathcal{N}(0, 3) & \text{[prior population effects]} \\ \sigma^2_m, \sigma^2_e, \sigma^2_u \sim&\ \operatorname{Cauchy}(0, 1) & \text{[prior sd for model, population, and country error]} \\ \end{aligned} \]

Results and analysis

We present the results of our models in two stages. First, in order to see how minor shifts in civil society laws and repression influence future human rights, we examine the marginal effects of our main explanatory variables when all other model covariates are held constant. These effects are by no means causal—rather, they demonstrate how much variation in predicted human rights is possible as the legal environment for civil society shifts up and down within a typical country. Second, we present predictive diagnostics for each of our models to examine the improvements in predictions. The results of the ordered logistic regression models can be unwieldy to interpret with plain numbers due to varying intercepts and thresholds between PTS levels. Accordingly, we present graphical results for all our models wherever possible and include complete tables of model results in the appendix (see Tables 5, 6, 7, and 8). For ease in comparing the count of NGO laws with civil society repression, we reverse the x-axis in the marginal effects plots for our second empirical expectation so that increasing the civil society repression represents worse repression rather than less.

Marginal results

How does repression change following increases in NGO laws?

In general, when holding all other covariates constant, the passage of an additional anti-NGO legal barrier is generally associated with a higher probability of more severe level of political terror in the following year. The probability of seeing different levels of political terror varies across different possible counts of NGO laws (see Figure 4). In the absence of formal legal barriers, Level 2 of the PTS—an environment where political imprisonment, torture, and beatings are rare—is the most likely category of political terror across all models, appearing a predicted 60% of the time. Level 1, where a country is under secure rule of law, is the second most likely outcome at slightly under 40%. The other categories of the PTS are extremely rare when there are no anti-NGO laws.

Marginal effects of increasing anti-NGO legal barriers on the probability of specific levels of political terror

Figure 4: Marginal effects of increasing anti-NGO legal barriers on the probability of specific levels of political terror

As new laws are added, though, the distribution of probabilities for each of the PTS levels shifts. Adding new barriers to advocacy, entry, and funding decreases the likelihood of seeing Level 1 and increases the probability of a country turning to either Level 2 or Level 3 (with extensive political imprisonment and trial-free detention). The marginal effects of all forms of NGO laws results in this turn from Level 1 to Levels 2 or 3, but this switching effect is strongest following additional barriers to advocacy. Anti-NGO laws do not predict more dramatic shifts in the PTS. Across all counts of legal barriers, the overall probability of seeing Levels 4 and 5 is minuscule, and the adding new laws does little to shift that probability. This is not unexpected however—the PTS has a natural ceiling at Level 5. Specifically, in addition to measuring the scope and intensity of government violence, the PTS scale also looks at the proportion of population targeted for abuse. Countries that receive a 5 are rare, as this score captures physical abuses routinely perpetrated against citizens not involved in politics, and includes cases where governments engage in “scorched earth” counterinsurgency policies or summary executions based on class or ethnic affiliation (Gibney et al. 2019). The majority of movement across PTS categories should thus occur between Levels 1–3.

Figure 5 shows the marginal effects of additional NGO legislation on V-Dem’s physical violence index and private civil liberties index. Across all four models for different categories of legal barriers, holding all other values constant, neither rights index moves much in response to new NGO laws. Each barrier-based posterior coefficient is essentially zero with narrow credible intervals that include large proportions of both negative and positive values. The marginal effects of additional laws on physical violence appear to show a positive relationship—that additional laws lead to less violence in the future—but the uncertainty in these estimates eliminates such an interpretation. Within the categories of anti-NGO laws, it appears that additional advocacy and entry laws might reduce future violence, but again, these predictions are indistinguishable from each other. The null effect of legal barriers on private civil liberties is even more stark—a typical country is predicted to have a score of 0.69 regardless of the addition of any kind of NGO law. However, when looking at latent human rights values, adding a new NGO law is associated with a -0.013 decline in the level of respect for human rights in the following year, on average. The marginal effect is larger (-0.02) for additional barriers to advocacy and entry, which represent more burdensome restrictions than limits to funding.

At a marginal level, therefore, adding additional anti-NGO laws in countries with low levels of political terror could signal (albeit weakly) that higher levels of political repression are more likely in the future. Limited repression remains the most likely possible outcome, but the chance of seeing robust rule of law decreases rapidly as more laws are passed, all else equal. When looking at other specific forms of political repression, though, such as physical violence and private civil liberties, NGO legal barriers do not change predicted values in any meaningful way. But looking more broadly, civil society laws are associated with a worse predicted underlying human rights environment.

Marginal effects of increasing anti-NGO legal barriers on V-Dem rights indexes and latent human rights

Figure 5: Marginal effects of increasing anti-NGO legal barriers on V-Dem rights indexes and latent human rights

How does general repression change following civil society repression?

The weak (or even non-existent) predictive power of formal NGO laws is likely attributable to fact that de jure NGO legislation is rare and slow-moving and not always explicitly linked to repression—again, states that engage in political repression can do so with or without legal justification. De facto civil society repression, on the other hand, represents the actual implementation of anti-NGO strategies—both legal and extralegal—and might be more indicative of future human rights abuses. The results from our second set of models, using V-Dem’s civil society repression index as the key predictor, confirm this.

The effect of civil society repression on predicted values of the Political Terror Scale mirrors what we found previously with our first empirical expectation. As seen in Figure 6, at the lowest levels of repression, Levels 1 and 2 are the most common predicted outcomes and are each equiprobable at roughly 50%. As de facto CSO repression increases, though, the probability of seeing Level 1 decreases and is replaced with either Level 2 or Level 3. More severe levels of political terror remain exceptionally improbable at any level of preceding civil society repression.

Marginal effects of changing levels of civil society repression on the probability of specific levels of political terror

Figure 6: Marginal effects of changing levels of civil society repression on the probability of specific levels of political terror

In contrast with formal laws, civil society repression does have a noticeable effect on the predicted values of the other rights indexes (see Figure 7). When holding all other model parameters at their typical values, increased civil society repression predicts increases in both the physical violence index and private civil liberties index: a one standard deviation change in civil society repression is associated with a 0.02 change in overall rights. In other words, a typical country with a civil society repression score of 0 is predicted to have a physical violence index of 0.62 and civil liberties index of 0.67. This implies that an increase in de facto NGO restrictions predicts a 3% decline in both protection from physical violence and respect for private civil liberties in the following year. Moreover, worsening civil society repression is strongly associated with a worse future human rights environment, as seen in the right panel of Figure 7

When holding all other factors equal, the de factor implementation of anti-civil society measures thus appear to be a stronger marginal indicator of future repression than the passing of formal de jure laws. Increasing civil society repression in countries with low levels of political terror can again potentially serve as a weak signal of more a more general increase in political violence in the future, while crackdowns on civil society predict more substantial changes in physical integrity rights, threats to civil liberties, and latent respect for human rights more broadly.

Marginal effects of changing levels of civil society repression on V-Dem rights indexes and latent human rights

Figure 7: Marginal effects of changing levels of civil society repression on V-Dem rights indexes and latent human rights


These predicted probabilities and outcomes are conditional on all other model parameters being held at their average values and show the effect of hypothetically changing a single predictor to different values. They do not show how well these models work with actual data. Following Ward and Beger (2017) and Gohdes and Carey (2017), test the predictive power of both sets of our models using out-of-sample prediction. We divide our complete data into a training set and a test set—the training set includes data from all countries from 1990–2010, and the test set includes the last three years of data (2011–2013). We re-run each of our models using our training data and then use the estimated parameters to predict political terror, physical violence, and private civil liberties for the test data. Additionally, we estimate a model for each outcome without predictor variables for civil society laws or repression as a baseline measure of model performance.

Table 2: Counts of correctly and incorrectly predicted cases across different PTS models
Prediction Baseline Total legal barriers included Barriers to advocacy included Barriers to entry included Barriers to funding included Civil society repression included
Wrong 108 108 110 108 109 111
Correct 377 377 375 377 376 374

We assess predictive power in a few different ways, depending on the model used to predict the outcome. For the ordinal Political Terror Scale, Table 2 shows a count of correct and incorrect predictions from each model. Because the physical violence and private civil liberties indexes are predicted with Gaussian models, we measure predictive accuracy with the root mean squared error, provided in Table 3. In the appendix we include lists of country-year cases that see improved predictions when taking civil society laws and repression into account.

Including civil society restrictions—both de jure and de facto—does very little to improve predictions of political terror. As seen in Table 2, no model provides more correct predictions than the baseline model, and some models perform worse. Accounting for anti-NGO activities does improve predictions in eight cases (see Table 4), but it also results in an equal number of incorrect predictions.

Table 3: Root mean squared error across models predicting physical violence, private civil liberties, and latent human rights
Model RMSE % change from baseline
Physical violence index
Baseline 0.0462
Total legal barriers 0.0461 -0.32%
Barriers to advocacy 0.0460 -0.41%
Barriers to entry 0.0462 -0.16%
Barriers to funding 0.0462 -0.05%
Civil society repression 0.0462 -0.12%
Private civil liberties index
Baseline 0.0339
Total legal barriers 0.0334 -1.51%
Barriers to advocacy 0.0336 -0.90%
Barriers to entry 0.0334 -1.43%
Barriers to funding 0.0336 -0.95%
Civil society repression 0.0332 -2.07%
Latent human rights values
Baseline 0.1862
Total legal barriers 0.1872 0.56%
Barriers to advocacy 0.1871 0.51%
Barriers to entry 0.1862 0.02%
Barriers to funding 0.1880 0.96%
Civil society repression 0.1857 -0.29%

When looking at the level of physical violence in a country, including civil society laws and repression also does very little to improve predictive accuracy. As Table 3 shows, accounting for legal barriers and civil society repression only reduces the RMSE by at most 0.4% compared to the baseline model. This is likely related to the lack of improvements in predicting political terror—both indexes measure similar phenomena. Similarly, accounting for civil society repression when modeling latent human rights only results in slight changes to the RMSE. However, including civil society-related predictors when modeling private civil liberties does result in small gains in accuracy, reducing the RMSE by as much as 2% compared to the baseline model.

While these gains in predictive power appear negligible, they are somewhat comparable with the findings of previous work. For instance, in their work on the effect of the murder of journalists on predicting future political terror, Gohdes and Carey (2017) find that accounting for violence against the media adds five correctly predicted cases of PTS scores and improves predictions in seven cases.

Conclusion and implications

Strengthening civil society is an important precondition for democratic transition and consolidation. It keeps governments accountable and protects civilians from the worst excesses of an abusive government. When governments restrict civil society, it reduces these organizations’ ability to maintain accountability and protect against abuse. Given these potential knock-on effects, in this paper, we posited that the ongoing global crackdown on civil society could predict worsening human rights in a country interested in political and civil repression. Passing restrictive laws, imposing barriers to NGO advocacy, entry, and funding, and arresting and expelling NGO staff could pave the way for future repression.

We outline two empirical expectations: that (1) official de jure anti-NGO laws and (2) general de facto civil society restrictions would predict future repression in a country, including political terror, physical violence, and violations of private civil liberties. We find little support for the predictive power of de jure restriction. Marginal predictions show that, all else held equal, adding new anti-NGO laws increases the probability of seeing moderate levels of political terror, but does not change predicted values of physical integrity or civil liberties rights violations. When comparing out-of-sample predictions to a baseline model, including de jure restrictions as predictors does not improve our models’ predictive accuracy.

We do, however, find some predictive power when accounting for general civil society restrictions. As with formal laws, rising civil society repression tends to increase the probability of more moderate political terror. Additionally, increasing civil society repression is associated with worsening respect for physical integrity rights and civil liberties in the future. In out-of-sample predictions, including civil society restrictions improves model accuracy by 1–2% over the baseline.

Importantly, the lack of predictive power when modeling political violence and the modest improvements in predictive power when modeling civil liberties likely reflects the purpose of civil society restrictions. Encroachments on private civil liberties such as restricting freedom of association are not nearly as violent as events like government-sponsored murders. The global shrinking of civil society is thus not likely an early warning sign for future violent repression—states typically signal their future repressive intentions with more violent methods, such as the persecution and murder of activists and the media. Instead, civil society is a possible “canary in the coal mine” for impending violations of other civil rights, such as the freedom of religion, labor rights, property rights, and the freedom of movement.


Table 4: Country-year cases where including civil society restrictions improved PTS predictions
Country Year Actual Baseline prediction Total NGO barriers included Civil society repression included
Guatemala 2013 Level 3 Level 2 Level 3 Level 3
Kosovo 2012 Level 1 Level 2 Level 1 Level 1
Sierra Leone 2011 Level 3 Level 2 Level 3
Congo - Brazzaville 2011 Level 3 Level 2 Level 3 Level 3
Somalia 2013 Level 5 Level 4 Level 5
Bahrain 2011 Level 3 Level 2 Level 3 Level 3
Bahrain 2013 Level 3 Level 2 Level 3 Level 3
Libya 2011 Level 4 Level 5 Level 4
Table 5: Full results from ordered logistic regression models predicting political terror
Total barriers (t + 1) Barriers to advocacy (t + 1) Barriers to entry (t + 1) Barriers to funding (t + 1) Civil society repression (t + 1)
Total legal barriers 0.157
[-0.014, 0.324]
Total legal barriers (t - 1) 0.023
[-0.143, 0.194]
Barriers to advocacy 0.442
[-0.179, 1.122]
Barriers to advocacy (t - 1) -0.082
[-0.750, 0.590]
Barriers to entry 0.270
[-0.040, 0.584]
Barriers to entry (t - 1) 0.064
[-0.266, 0.370]
Barriers to funding 0.276
[-0.075, 0.601]
Barriers to funding (t - 1) 0.071
[-0.282, 0.399]
Civil society repression -0.384
[-0.637, -0.104]
Civil society repression (t - 1) 0.079
[-0.201, 0.335]
PTS = 2 2.251 2.285 2.268 2.259 2.265
[1.945, 2.548] [1.986, 2.588] [1.967, 2.562] [1.975, 2.566] [1.973, 2.570]
PTS = 3 4.222 4.303 4.258 4.245 4.298
[3.861, 4.620] [3.929, 4.684] [3.880, 4.649] [3.873, 4.631] [3.912, 4.671]
PTS = 4 6.231 6.351 6.285 6.228 6.349
[5.773, 6.750] [5.866, 6.833] [5.810, 6.756] [5.741, 6.720] [5.870, 6.831]
PTS = 5 8.564 8.646 8.632 8.552 8.628
[7.925, 9.197] [8.037, 9.286] [8.011, 9.218] [7.915, 9.169] [7.989, 9.232]
Polyarchy index -2.226 -2.310 -2.358 -2.252 -1.273
[-2.920, -1.532] [-2.985, -1.595] [-3.071, -1.720] [-2.912, -1.534] [-2.210, -0.236]
Log GDP per capita -0.474 -0.436 -0.449 -0.471 -0.429
[-0.627, -0.320] [-0.590, -0.290] [-0.591, -0.299] [-0.612, -0.326] [-0.567, -0.283]
Trade as % of GDP -0.432 -0.433 -0.419 -0.414 -0.407
[-0.742, -0.092] [-0.763, -0.121] [-0.730, -0.081] [-0.728, -0.079] [-0.737, -0.110]
Armed conflict 1.086 1.066 1.079 1.093 1.060
[0.793, 1.406] [0.746, 1.370] [0.768, 1.391] [0.781, 1.399] [0.747, 1.356]
Cutpoint 1/2 -5.817 -5.631 -5.634 -5.913 -5.370
[-7.081, -4.598] [-6.935, -4.431] [-6.855, -4.416] [-7.124, -4.679] [-6.590, -4.203]
Cutpoint 2/3 -1.772 -1.625 -1.589 -1.869 -1.356
[-2.970, -0.583] [-2.922, -0.521] [-2.796, -0.402] [-3.078, -0.706] [-2.536, -0.229]
Cutpoint 3/4 2.289 2.405 2.470 2.199 2.667
[1.105, 3.519] [1.276, 3.641] [1.265, 3.657] [0.947, 3.330] [1.559, 3.857]
Cutpoint 4/5 5.925 6.026 6.084 5.824 6.270
[4.616, 7.097] [4.795, 7.219] [4.878, 7.314] [4.567, 6.994] [5.037, 7.399]
Num.Obs. 3594 3594 3594 3594 3612
R2 0.807 0.806 0.807 0.807 0.806
R2 Marg. 0.710 0.709 0.710 0.706 0.714
LOOIC 5192.3 5218.4 5199.5 5198.2 5252.1
LOOIC s.e. 97.4 97.4 97.7 97.4 97.7
WAIC 5191.7 5217.7 5198.8 5197.4 5251.4
Posterior means; 95% credible intervals in brackets
Table 6: Full results from Gaussian regression models predicting physical violence
Total barriers (t + 1) Barriers to advocacy (t + 1) Barriers to entry (t + 1) Barriers to funding (t + 1) Civil society repression (t + 1)
Total legal barriers 0.001
[-0.002, 0.005]
Total legal barriers (t - 1) -0.003
[-0.007, 0.000]
Barriers to advocacy 0.005
[-0.009, 0.019]
Barriers to advocacy (t - 1) -0.011
[-0.026, 0.003]
Barriers to entry 0.006
[-0.001, 0.013]
Barriers to entry (t - 1) -0.008
[-0.015, -0.001]
Barriers to funding -0.001
[-0.009, 0.006]
Barriers to funding (t - 1) -0.002
[-0.009, 0.006]
Civil society repression 0.026
[0.020, 0.031]
Civil society repression (t - 1) -0.022
[-0.028, -0.017]
Physical violence index (t) 0.966 0.965 0.966 0.967 0.963
[0.953, 0.978] [0.952, 0.976] [0.952, 0.979] [0.955, 0.979] [0.950, 0.974]
Polyarchy index 0.009 0.010 0.011 0.009 0.001
[-0.002, 0.019] [0.000, 0.021] [0.001, 0.022] [-0.002, 0.019] [-0.013, 0.015]
Log GDP per capita 0.001 0.001 0.001 0.001 0.002
[0.000, 0.003] [0.000, 0.003] [-0.001, 0.003] [0.000, 0.003] [0.000, 0.003]
Trade as % of GDP -0.002 -0.002 -0.002 -0.002 -0.002
[-0.006, 0.002] [-0.006, 0.002] [-0.006, 0.002] [-0.006, 0.001] [-0.006, 0.002]
Armed conflict -0.001 -0.001 -0.001 -0.001 -0.001
[-0.006, 0.003] [-0.006, 0.004] [-0.006, 0.004] [-0.006, 0.004] [-0.006, 0.004]
Intercept 0.016 0.015 0.015 0.014 0.014
[0.006, 0.026] [0.005, 0.024] [0.005, 0.026] [0.004, 0.023] [0.004, 0.023]
Num.Obs. 3609 3609 3609 3609 3629
R2 0.971 0.971 0.971 0.971 0.971
R2 Marg. 0.971 0.971 0.971 0.971 0.971
LOOIC -11663.0 -11660.8 -11655.1 -11658.7 -11767.1
LOOIC s.e. 427.7 428.3 428.7 427.9 422.9
WAIC -11663.8 -11663.8 -11658.8 -11660.0 -11767.9
Posterior means; 95% credible intervals in brackets
Table 7: Full results from Gaussian regression models predicting private civil liberties
Total barriers (t + 1) Barriers to advocacy (t + 1) Barriers to entry (t + 1) Barriers to funding (t + 1) Civil society repression (t + 1)
Total legal barriers -0.001
[-0.003, 0.002]
Total legal barriers (t - 1) -0.001
[-0.004, 0.001]
Barriers to advocacy -0.007
[-0.016, 0.003]
Barriers to advocacy (t - 1) 0.001
[-0.008, 0.011]
Barriers to entry 0.002
[-0.002, 0.008]
Barriers to entry (t - 1) -0.006
[-0.011, -0.001]
Barriers to funding -0.003
[-0.008, 0.002]
Barriers to funding (t - 1) 0.000
[-0.005, 0.005]
Civil society repression 0.021
[0.017, 0.025]
Civil society repression (t - 1) -0.019
[-0.023, -0.015]
Private civil liberties index (t) 0.968 0.967 0.969 0.969 0.967
[0.957, 0.979] [0.954, 0.978] [0.958, 0.981] [0.958, 0.979] [0.956, 0.978]
Polyarchy index 0.012 0.014 0.013 0.013 0.007
[0.002, 0.022] [0.005, 0.025] [0.003, 0.024] [0.004, 0.023] [-0.005, 0.018]
Log GDP per capita -0.001 -0.001 -0.001 -0.001 -0.001
[-0.002, 0.000] [-0.002, 0.000] [-0.002, 0.000] [-0.002, 0.000] [-0.002, 0.000]
Trade as % of GDP 0.001 0.001 0.001 0.001 0.001
[-0.002, 0.004] [-0.002, 0.004] [-0.002, 0.004] [-0.002, 0.004] [-0.002, 0.004]
Armed conflict 0.000 0.000 0.000 0.000 0.001
[-0.003, 0.003] [-0.004, 0.003] [-0.003, 0.003] [-0.004, 0.003] [-0.002, 0.004]
Intercept 0.030 0.029 0.030 0.028 0.026
[0.021, 0.039] [0.020, 0.039] [0.020, 0.039] [0.019, 0.036] [0.018, 0.036]
Num.Obs. 3609 3609 3609 3609 3629
R2 0.983 0.983 0.983 0.983 0.982
R2 Marg. 0.983 0.983 0.983 0.983 0.982
LOOIC -14213.9 -14209.9 -14208.4 -14205.6 -14101.5
LOOIC s.e. 449.7 449.6 451.9 449.6 488.9
WAIC -14214.5 -14210.5 -14211.4 -14208.1 -14103.5
Posterior means; 95% credible intervals in brackets
Table 8: Full results from Gaussian regression models predicting latent human rights
Total barriers (t + 1) Barriers to advocacy (t + 1) Barriers to entry (t + 1) Barriers to funding (t + 1) Civil society repression (t + 1)
Total legal barriers -0.013
[-0.032, 0.005]
Total legal barriers (t - 1) 0.007
[-0.011, 0.026]
Barriers to advocacy -0.024
[-0.093, 0.046]
Barriers to advocacy (t - 1) 0.008
[-0.061, 0.081]
Barriers to entry -0.027
[-0.063, 0.008]
Barriers to entry (t - 1) 0.019
[-0.017, 0.054]
Barriers to funding -0.016
[-0.053, 0.022]
Barriers to funding (t - 1) 0.006
[-0.035, 0.043]
Civil society repression 0.053
[0.024, 0.082]
Civil society repression (t - 1) -0.037
[-0.066, -0.008]
Latent human rights (t) 0.964 0.964 0.964 0.965 0.963
[0.951, 0.976] [0.952, 0.975] [0.950, 0.976] [0.952, 0.975] [0.951, 0.974]
Polyarchy index 0.075 0.081 0.082 0.076 0.017
[0.032, 0.118] [0.040, 0.123] [0.038, 0.123] [0.033, 0.119] [-0.057, 0.085]
Log GDP per capita 0.008 0.007 0.007 0.008 0.009
[0.000, 0.017] [0.000, 0.015] [0.000, 0.015] [0.001, 0.016] [0.001, 0.017]
Trade as % of GDP 0.024 0.024 0.024 0.023 0.022
[0.003, 0.043] [0.005, 0.046] [0.003, 0.045] [0.002, 0.043] [0.001, 0.041]
Armed conflict -0.005 -0.004 -0.004 -0.005 -0.002
[-0.030, 0.021] [-0.032, 0.019] [-0.029, 0.023] [-0.030, 0.020] [-0.028, 0.024]
Intercept -0.074 -0.076 -0.073 -0.076 -0.073
[-0.138, -0.009] [-0.132, -0.012] [-0.141, -0.011] [-0.138, -0.018] [-0.137, -0.015]
Num.Obs. 3609 3609 3609 3609 3629
R2 0.969 0.968 0.969 0.968 0.968
R2 Marg. 0.968 0.968 0.968 0.968 0.968
LOOIC 48.6 48.9 52.7 49.1 60.6
LOOIC s.e. 351.9 351.3 352.6 351.4 350.3
WAIC 46.7 47.7 51.3 48.4 59.7
Posterior means; 95% credible intervals in brackets


Abramowitz, Michael, and Nate Schenkkan. 2018. “How Illiberal Leaders Attack Civil Society.” Foreign Affairs, April 6, 2018. https://www.foreignaffairs.com/articles/central-europe/2018-04-06/how-illiberal-leaders-attack-civil-society.
Barry, Colin M., K. Chad Clay, and Michael E. Flynn. 2013. “Avoiding the Spotlight: Human Rights Shaming and Foreign Direct Investment.” International Studies Quarterly 57 (3): 532–44. https://doi.org/10.1111/isqu.12039.
Bell, Andrew, and Kelvyn Jones. 2015. “Explaining Fixed Effects: Random Effects Modelling of Time-Series Cross-Sectional and Panel Data.” Political Science Research and Methods 3 (1): 133–53. https://doi.org/10.1017/psrm.2014.7.
Bell, Sam R., Tavishi Bhasin, K. Chad Clay, and Amanda Murdie. 2014. “Taking the Fight to Them: Neighborhood Human Rights Organizations and Domestic Protest.” British Journal of Political Science 44 (4): 853–75. https://doi.org/10.1017/s0007123413000100.
Boulding, Carew. 2014. NGOs, Political Protest, and Civil Society. Cambridge: Cambridge University Press.
Brechenmacher, Saskia. 2017. Civil Society Under Assault: Repression and Response in Russia, Egypt, and Ethiopia. Washington, D. C.: Carnegie Endowment for International Peace. http://carnegieendowment.org/2017/05/18/civil-society-under-assault-repression-and-responses-in-russia-egypt-and-ethiopia-pub-69953.
Bürkner, Paul-Christian. 2017. brms: An R Package for Bayesian Multilevel Models Using Stan.” Journal of Statistical Software 80 (1): 1–28. https://doi.org/10.18637/jss.v080.i01.
Carothers, Thomas, and Saskia Brechenmacher. 2014. Closing Space: Democracy and Human Rights Support Under Fire. Washington, D. C.: Carnegie Endowment for International Peace.
Chaudhry, Suparna. 2016. “The Assault on Democracy Assistance: Explaining State Repression of NGOs.” PhD thesis, New Haven, Connecticut: Yale University.
Chaudhry, Suparna, and Andrew Heiss. 2021. “Closing Space and the Restructuring of Global Activism: Causes and Consequences of the Global Crackdown on NGOs.” In Beyond the Boomerang: New Patterns in Transcalar Advocacy, edited by Christopher L. Pallas and Elizabeth Bloodgood. Tuscaloosa, AL: University of Alabama Press.
CIVICUS. 2013. “Global Trends on Civil Society Restrictions: Mounting Restrictions on Civil Society: The Gap Between Rhetoric and Reality.”
———. 2021. “CIVICUS Monitor.” https://monitor.civicus.org/.
Conrad, Courtenay R., and Emily Hencken Ritter. 2019. Contentious Compliance: Dissent and Repression Under International Human Rights Law. Oxford: Oxford University Press. https://doi.org/10.1093/oso/9780190910976.001.0001.
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, et al. 2020. “V-Dem [Country-Year/Country-Date] Dataset V10.” Varieties of Democracy (V-Dem) Project. https://doi.org/10.23696/vdemds20.
Cramer, Maria. 2020. “Judge Reverses Convictions of Activists Who Left Water for Migrants.” The New York Times, February 5, 2020. https://www.nytimes.com/2020/02/05/us/tucson-border-activists-conviction-reversed.html.
Davenport, Christian. 2007a. State Repression and the Domestic Democratic Peace. New York: Cambridge University Press. https://doi.org/10.1017/cbo9780511510021.
———. 2007b. “State Repression and Political Order.” Annual Review of Political Science 10 (June): 1–23. https://doi.org/10.1146/annurev.polisci.10.101405.143216.
Davenport, Christian, and David A. Armstrong. 2004. “Democracy and the Violation of Human Rights: A Statistical Analysis from 1976–1996.” American Journal of Political Science 48 (3): 538–54. https://doi.org/10.1111/j.0092-5853.2004.00086.x.
Dietrich, Simone, and Amanda Murdie. 2017. “Human Rights Shaming Through INGOs and Foreign Aid Delivery.” Review of International Organizations 12 (1): 95–120. https://doi.org/10.1007/s11558-015-9242-8.
Dupuy, Kendra, and Aseem Prakash. 2018. “Do Donors Reduce Bilateral Aid to Countries with Restrictive NGO Laws? A Panel Study, 1993–2012.” Nonprofit and Voluntary Sector Quarterly 47 (1): 89–106. https://doi.org/10.1177/0899764017737384.
Dupuy, Kendra, James Ron, and Aseem Prakash. 2016. “Hands Off My Regime! Governments’ Restrictions on Foreign Aid to Non-Governmental Organizations in Poor and Middle-Income Countries.” World Development 84 (August): 299–311. https://doi.org/10.1016/j.worlddev.2016.02.001.
Fariss, Christopher J. 2014. “Respect for Human Rights Has Improved over Time: Modeling the Changing Standard of Accountability in Human Rights Documents.” American Political Science Review 108 (2): 297–318. https://doi.org/10.1017/s0003055414000070.
Fariss, Christopher J., Michael R. Kenwick, and Kevin Reuning. 2020. “Estimating One-Sided Killings from a Robust Measurement Model of Human Rights.” Journal of Peace Research 57 (6): 801–14. https://doi.org/10.1177/0022343320965670.
Gibney, Mark, Linda Cornett, Reed Wood, Peter Haschke, Daniel Arnon, Attilio Pisanò, and Gray Barrett. 2019. “The Political Terror Scale 1976–2018.” http://politicalterrorscale.org/.
Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Håvard Strand. 2002. “Armed Conflict 1946–2001: A New Dataset.” Journal of Peace Research 39 (5): 615–37. https://doi.org/10.1177/0022343302039005007.
Gohdes, Anita R., and Sabine C. Carey. 2017. “Canaries in a Coal-Mine? What the Killings of Journalists Tell Us about Future Repression.” Journal of Peace Research 54 (2): 157–74. https://doi.org/10.1177/0022343316680859.
Goldstein, Robert Justin. 1978. Political Repression in Modern America: From 1870 to the Present. Boston: G. K. Hall.
Heiss, Andrew. 2017. “Amicable Contempt: The Strategic Balance Between Dictators and International NGOs.” PhD thesis, Durham, NC: Duke University.
———. 2019. “Taking Control of Regulations: How International Advocacy NGOs Shape the Regulatory Environments of Their Target Countries.” Interest Groups and Advocacy 8 (3): 356–75. https://doi.org/10.1057/s41309-019-00061-0.
International Center for Not-For-Profit Law. 2021. “The Civic Space Initiative.” https://www.icnl.org/our-work/global-programs/the-civic-space-initiative.
Keck, Margaret E., and Kathryn Sikkink. 1998. Activists Beyond Borders: Advocacy Networks in International Politics. Ithaca: Cornell University Press.
Kim, Hunjoon, and Kathryn Sikkink. 2010. “Explaining the Deterrence Effect of Human Rights Prosecutions for Transitional Countries.” International Studies Quarterly 54 (4): 939–63. https://doi.org/10.1111/j.1468-2478.2010.00621.x.
Kozenko, Andrey. 2015. “Pure Pragmatism—Nothing Personal.” Meduza, May 21, 2015. https://meduza.io/en/feature/2015/05/21/pure-pragmatism-nothing-personal.
Krain, Matthew. 1997. “State-Sponsored Mass Murder: A Study of the Onset and Severity of Genocides and Politicides.” Journal of Conflict Resolution 41 (3): 331–60. https://doi.org/10.1177/0022002797041003001.
Lebovic, James H., and Erik Voeten. 2009. “The Cost of Shame: International Organizations and Foreign Aid in the Punishing of Human Rights Violators.” Journal of Peace Research 46 (1): 79–97. https://doi.org/10.1177/0022343308098405.
Murdie, Amanda, and Tavishi Bhasin. 2011. “Aiding and Abetting: Human Rights INGOs and Domestic Protest.” The Journal of Conflict Resolution 55 (2): 163–91. https://doi.org/10.1177/0022002710374715.
Nielsen, Richard A. 2013. “Rewarding Human Rights? Selective Aid Sanctions Against Repressive States.” International Studies Quarterly 57 (4): 791–803. https://doi.org/10.1111/isqu.12049.
Poe, Steven C., and C. Neal Tate. 1994. “Repression of Human Rights to Personal Integrity in the 1980s: A Global Analysis.” American Political Science Review 88 (4): 853–72. https://doi.org/10.2307/2082712.
Press Trust of India. 2015. “Foreign Funding: Government Cancels Licences of Nearly 9,000 NGOs.” The Indian Express, April 28, 2015. http://indianexpress.com/article/india/govt-cancels-licences-of-nearly-9000-ngos/.
———. 2019. “14,500 NGOs Banned by Government from Receiving Foreign Funds.” The Times of India, December 4, 2019. https://timesofindia.indiatimes.com/india/14500-ngos-banned-by-govt-from-receiving-foreign-funds/articleshow/72366997.cms.
Putnam, Robert D., Robert Leonardi, and Raffaella Y. Nanetti. 1994. Making Democracy Work: Civic Traditions in Modern Italy. Princeton: Princeton University Press.
R Core Team. 2021. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. https://www.r-project.org/.
Richards, David L. 1999. “Perilous Proxy: Human Rights and the Presence of National Elections.” Social Science Quarterly 80 (4): 648–65.
Ritter, Emily Hencken, and Courtenay R. Conrad. 2016. “Preventing and Responding to Dissent: The Observational Challenges of Explaining Strategic Repression.” American Political Science Review 110 (1): 85–99. https://doi.org/10.1017/S0003055415000623.
Schnakenberg, Keith E., and Christopher J. Fariss. 2014. “Dynamic Patterns of Human Rights Practices.” Political Science Research and Methods 2 (1): 1–31. https://doi.org/10.1017/psrm.2013.15.
Sikkink, Kathryn. 2011. The Justice Cascade: How Human Rights Prosecutions Are Changing World Politics. New York: W. W. Norton & Company.
Stan Development Team. 2021. Stan Modeling Language Users Guide and Reference Manual. http://mc-stan.org.
Stephan, Maria J., and Erica Chenoweth. 2008. “Why Civil Resistance Works: The Strategic Logic of Nonviolent Conflict.” International Security 33 (1, Summer): 7–44. https://doi.org/10.1162/isec.2008.33.1.7.
UCDP/PRIO. 2020. UCDP/PRIO Armed Conflict Dataset Codebook.” https://ucdp.uu.se/.
Ward, Michael D., and Andreas Beger. 2017. “Lessons from Near Real-Time Forecasting of Irregular Leadership Changes.” Journal of Peace Research 54 (2): 141–56. https://doi.org/10.1177/0022343316680858.
Weiss, Meredith L. 2009. “Edging Toward a New Politics in Malaysia: Civil Society at the Gate?” Asian Survey 49 (5): 741–58. https://doi.org/10.1525/as.2009.49.5.741.

  1. We use Stan (Stan Development Team 2021) through R (R Core Team 2021) and brms (Bürkner 2017) to estimate our models. We generate 4 MCMC chains for each model with 2,000 iterations in each chain, 1,000 of which are used for warmup. All chains converge; we assess convergence with visual inspection.↩︎