Civil society organizations (CSOs) are key for democratization (O’Donnell and Schmitter 1986; Bernhard and Karakoç 2007) and democratic consolidation (Dahl 1971; Bratton 1989; Diamond 1994; Mechkova et al. 2019; Bernhard et al. 2020).Footnote 1 However, governments increasingly restrict independent CSOs, e.g., through prohibiting foreign funding, launching smear campaigns, and arresting civil society activists (Christensen and Weinstein 2013; Glasius 2018; Glasius et al. 2020; Smidt et al. 2021).Footnote 2 We investigate whether mounting restrictions on CSOs affect the quality of horizontal accountability institutions like parliaments and courts.

Horizontal accountability institutions like parliaments and courts are core democratic institutions. They can both monitor (e.g., through access to classified information) and punish government misbehavior (e.g., votes of no confidence or threats of impeachment) on a day-to-day basis (Mechkova, Lührmann and Lindberg 2019, 41). While their quality is partially endogenous to governments’ decisions (Bermeo 2016; Levitsky and Ziblatt 2018; Svolik 2019), horizontal accountability institutions also have agency on their own and can resist against government attacks on their independence. Thus, the relation between government-imposed restrictions on CSOs and a change in the quality of horizontal accountability is neither pre-determined by nor mechanically following from government-initiated moves toward authoritarian rule, but an empirical puzzle. Inquiry into this puzzle helps us understand pathways of executive aggrandizement.

Existing research on civil society only recently started to investigate the consequences of restrictions on CSOs, suggesting that they hamper international shaming campaigns (Smidt et al. 2021), reduce public health (Heinzel and Koenig-Archibugi 2022), or worsen respect for human rights (Chaudhry and Heiss 2022). Yet, how mounting restrictions affect the quality of horizontal checks and balances has not been explored. Moreover, most existing research on autocratizationFootnote 3 just mentions governments’ interference with CSOs and conceptualizes it as symptomatic of authoritarian governance rather than as having independent explanatory power in the process of democratic decline (Levitsky and Ziblatt 2018; Mounk 2018).

To our best knowledge, only two studies offer informed hypotheses regarding a directional relationship between restrictions on CSOs and horizontal accountability. While Mechkova et al. (2017, 166) claim that restrictions follow improvements in horizontal accountability, helping incumbents to off-set heightened constraints, Bernhard (2020, 356) suggests that restrictions on CSOs follow deteriorations in horizontal accountability because authoritarian leaders can now bring down CSOs as the “last defense” of democracy.

Not only are these claims contradictory, but existing studies also discard the possibility that restrictions imposed on CSOs temporarily precede and facilitate a deterioration of horizontal accountability. This manuscript theoretically develops this possibility and offers the first systematic test of the relationship between restrictions on CSOs and the change in the quality of horizontal accountability institutions in a global sample over a long time period.

We argue that restrictions on CSOs facilitate the erosion of horizontal accountability. First, CSOs monitor government behavior and they raise the alarm when governments violate democratic norms, for example, through mobilizing opposition voters or reaching out to international allies. Restrictions likely hamper such activities, implying fewer constraints on governments’ attempts to dismantle horizontal checks and balances. Second, CSOs provide crucial services to horizontal accountability institutions to control government behavior, such as expertise, monitoring, and independent testimony. If restrictions interfere with CSO activity, then horizontal oversight institutions may find it harder to effectively check on the government. Our main expectation is that if restrictions grow more severe, the quality of horizontal accountability institutions like courts and parliaments subsequently deteriorates.

Using global data for the period 1992–2018, our statistical analyses support this expectation. Increases in restrictions correlate with subsequently lower quality of horizontal accountability institutions. We interpret this correlation as testifying to a dynamic whereby restrictions on CSOs enable the deterioration of horizontal oversight. To empirically validate this interpretation, we examine two alternative explanations: (i) government-imposed restrictions on CSOs are made possible by and thus follow the dismantling of horizontal checks on governments (reverse causation) and (ii) mounting restrictions and declining horizontal accountability are both pre-determined by authoritarian incumbents and anti-democratic structures (omitted variable bias). As expected, we find little evidence that these alternative explanations are the only drivers of the correlation between increases of restrictions and lower horizontal accountability. Finally, we use in-depth analyses of two “typical cases”—Kenya and Turkey in the 2013–2021 period—to illustrate the plausibility of our causal argument.

Our findings make important contributions to two literatures: First, they inform research on the nature of regime transformation processes that disaggregate different types of accountability institutions and actors. Specifically, we show that inference with CSOs is one pathway to executive aggrandizement and authoritarian rule (Bernhard et al. 2020; Lührmann et al. 2020; Lührmann and Lindberg 2019; Mechkova et al. 2017; Mechkova et al. 2019). Second, our findings add to the literature on the consequences of shrinking civic spaces, providing first cross-national evidence that restrictions on CSOs can harm core democratic institutions (Chaudhry and Heiss 2022; Heinzel and Koenig-Archibugi 2022; Smidt et al. 2021). Our results are also informative for practioners and policy-makers interested in protecting civic space. Government leaders often justify restrictions on CSOs with reference to democracy, claiming that CSOs are self-appointed, elitist, and accountable to external constituencies (Carothers and Brechenmacher 2018). We help debunk such myths, showing that restrictions on CSOs weaken core institutions of democracy. Overall, our findings imply that when international organizations and activists protect CSOs (as in the case of Kenya), they also protect horizontal accountability institutions such as parliaments and courts against authoritarian power grabs. But overlooking the often-subtle restrictions on CSOs plays into the hands of authoritarian governments.

What We Know About Horizontal Accountability Institutions and Restrictions on CSOs

Democracies seldom suddenly collapse but tend to gradually erode (Bermeo 2016). Increasing restrictions on CSOs and deteriorating quality of horizontal accountability institutions are both mentioned as symptoms of gradual erosion of democracy (e.g., Levitsky and Ziblatt 2018). Yet, their precise relationship remains underexplored in both the literature on autocratization and the literature on civil society.

The literature on autocratization can be divided into studies describing how regimes become more autocratic and studies explaining why this autocratization process happens. Regarding the first strand, the most prevalent autocratization mode in the past 30 years has been described as “executive aggrandizement”: elected incumbents use legal and oftentimes democratic mechanisms to weaken checks and balances (Lührmann and Lindberg 2019; Bermeo 2016).Footnote 4 Only very few studies explicitly theorize the role of restrictions on CSOs in executive aggrandizement. A notable exception is Mechkova et al. (2017) who argue that threats to CSOs “open to a gradual and often hidden ‘tightening of the screws’ that can quietly negate high-profile advances in the electoral arena.” If restrictions on CSOs, however, would compensate for democratic improvements in other areas, then increases in restrictions would, contrary to our expectation, positively correlate with horizontal accountability. By contrast, Bernhard et al. (2020) sees CSOs in Eastern Europe as “the last layer of accountability,” suggesting that restrictions on CSOs occur only after substantive decline in horizontal and vertical accountability. If his argument generalizes, increases in restrictions would not correlate with subsequently lower horizontal accountability. As these contradictory findings suggest, exactly how restrictions on CSOs relate to the quality of horizontal accountability institutions requires further research.

Regarding the second strand of autocratization literature, i.e., on why regimes become gradually more autocratic, several causes have been identified: the emergence of (ethno-)populism (Stroschein 2019; Vachudova 2020), the collusion of conservative parties with populist anti-democrats (Levitsky and Ziblatt 2018), the erosion of democratic preferences in the population (Mounk 2018), or the political polarization that prevents voters from allying against authoritarian incumbents (Svolik 2019; cf. also Singer 2018). Recent events also facilitated autocratization, for instance, the global financial crisis in 2008 and the fear about immigration in 2015 (Bernhard 2020). What has remained largely underexplored is how decline in the quality of one element of democracy (e.g., shrinking civic spaces) may help explain decline in the quality of another element of democracy (e.g., horizontal checks on executives).Footnote 5 Of course, the decline in any element of democracy is likely partly endogenous to the behavior of (authoritarian-minded) incumbents (Bermeo 2016; Levitsky and Ziblatt 2018; Svolik 2019). Nevertheless, a crackdown on CSOs may also independently impact the subsequent ability of parliaments, courts, and other oversight institutions to control the executive. Whether this is the case is the puzzle that this manuscript investigates.

The literature on civil society has discussed restrictions on CSOs as a government strategy to minimize domestic and international accountability (Bakke et al. 2019; Christensen and Weinstein 2013; Glasius 2018; Glasius et al. 2020; Smidt et al. 2021). For example, Smidt et al. (2021, 1273) view government-imposed restrictions as a means of governments to win the “struggle for the command of information,” e.g., to hide their misbehavior from scrutiny and avoid punishment. Similarly, Glasius (2018) conceptualizes restrictions as one authoritarian practice for “sabotaging accountability” by disabling access to information and voice.

Beyond theoretical claims, there is also evidence that severe restrictions on CSOs lead to reductions in domestic protest and international shaming as well as to government abuse and mismanagement (Smidt et al. 2021; Heinzel and Koenig-Archibugi 2022). Yet, how interference with CSOs impacts the quality of horizontal accountability institutions like courts and parliaments has not been theorized. To our knowledge, there is also no large-N cross-national empirical study on this relationship. We fill these gaps.

Argument: The Impact of Restrictions Against Civil Society

We argue that restricting CSOs diminishes obstacles to governments’ attacks on horizontal accountability institutions, such as independent monitoring of government behavior, high election turnout by opposition voters, popular pro-democracy protests, and international shaming campaigns.Footnote 6 Moreover, restricting CSOs weakens collaboration between CSOs and horizontal accountability institutions, thereby reducing the latter’s ability to control the executive. The next two subsections elaborate each claim, respectively.

Removing Obstacles to Anti-democratic Reform

Existing research shows that CSOs help preserve democracy (e.g., Bernhard et al. 2020). One reason for this is that CSOs monitor government policies and practices and disseminate the information gleaned through monitoring via their own reports, exchanges with media, and events like protest marches. According to Diamond (1994, 10), “a free press is only one vehicle for providing the public with a wealth of news and alternative perspectives. Independent [civil society] organizations may also give citizens [and other actors] hard-won information about government activities that does not depend on what the government says it is doing.”

Another reason for why CSOs preserve democracy is that CSOs create domestic networks where citizens can interact, build trust, and share information (Boulding 2014). CSOs also build networks with international allies, including foreign NGOs and international organizations (Keck and Sikkink 1998). CSOs use these networks as well as their knowledge and material mobilization resources (such as finance and offices) to bring critical information to the attention of voters, mass publics, and international allies and mobilize these actors to punish governments’ anti-democratic behavior (McCarthy and Zald 1977).

However, when CSOs face restrictions, their capacity to monitor and inform and to network and mobilize shrinks. First, restrictions on CSOs imply that governments face less scrutiny. Movement and visitation restrictions on CSOs directly prevent their monitoring activity. Yet, other types of restrictions also divert attention away from monitoring politically sensitive issues. For instance, when the Ethiopian government passed a law prohibiting foreign-funded NGOs from working on democracy and human rights, most foreign-funded CSOs closed operations. The few CSOs that survived “rebranded” or “restructured” their activities to less sensitive topics (Dupuy et al. 2015). Taken together, laws and practices that constrain CSO activity likely reduce the flow of information on government attacks against horizontal oversight institutions.

Furthermore, restrictions reduce CSOs’ capacity to make use of mobilization networks and resources. When governments curtail funding opportunities, CSOs may lack the means to produce outreach materials for citizens or to travel abroad to meet with international activists. But other physical and legal restrictions also incentivize CSOs to shift away from public activity and toward self-preservation. For instance, restrictions like surveillance or arbitrary arrests can deter CSO activists from mobilizing others to avoid further repression. Finally, defamation campaigns may discredit CSOs’ public image, diminishing their capacity to mobilize citizens.

The negative repercussions of restrictions on CSOs’ functions—i.e., information provision and mobilization—remove obstacles to governments’ attempts to dismantle horizontal accountability institutions. First, when CSOs cannot provide critical information, it is harder for citizens to punish governments’ violations of democratic norms at the ballot box. Restrictions thus diminish opposition voter turnout (e.g., Ginsburg and Huq 2018, for a similar argument). As the case of Turkey demonstrates, repression of the right to protest was pre-emptively used in 2017 to de-mobilize civil society before an anti-democratic constitutional referendum that removed horizontal checks on executive power (Arslanalp and Deniz Erkmen 2020; ICNL 2019b).

Furthermore, restrictions diminish CSO-led mobilization efforts against governments’ anti-democratic reform. For example, in Hong Kong, shortly after a restrictive security law was passed, one of the most prominent pro-democracy activists (Joshua Wong) withdrew from the civil society group Demosisto and later, the group was disbanded (Amnesty International 2020). In subsequent months, pro-democracy mobilization dwindled significantly (New York Times 2020). In Russia, survey evidence corroborates that Russian citizens became more suspicious of CSOs’ activities as a result the government-imposed designation “foreign agent” for CSOs, that obtain foreign funding, making it difficult for these CSOs to reach the wider public and, presumably, hindering mobilization for pro-democracy protest (Brechenmacher 2017, 23-24).

Finally, when restrictions prevent domestic CSOs from mobilizing against governments’ anti-democratic behavior, authoritarian moves may slip under the international community’s radar (Smidt et al. 2021). But even if foreign governments do not solely rely on CSOs for detecting authoritarian behavior, local CSO-led mobilization may be necessary to  legitimate international criticism (Bob 2005). As such, when restrictions diminish the flow of information and mobilization of CSOs, international “naming and shaming” that has the potential to deter executive aggrandizements becomes less likely or less effective.

Figure 1 summarizes our first causal mechanism underpinning the hypothesized relationship between mounting restrictions on CSOs and lower-quality horizontal oversight. Restrictions diminish CSOs’ capacity and motivation to collect critical information and use mobilization networks and resources. In turn, more limited flows of information and mobilization efforts diminish the ability of other actors—specifically, voters, mass publics, and international allies—to constrain government-sponsored reforms that weaken horizontal accountability institutions.

Fig. 1
figure 1

Visualization of argument part I

Weakening Control Functions of Horizontal Accountability Institutions

Restrictions on CSOs can also directly lower the power of horizontal accountability institutions to constrain the executive. Existing research shows that while almost any regime has formally independent courts, parliaments, and other representative institutions, independent CSOs are needed for “breathing democratic life into the bones of formal representative institutions” (Schedler 2010, 78).

For starters, CSOs make horizontal accountability institutions effective in controlling the executive by providing information. Specifically, parliaments often rely on “fire-alarm oversight” to control the executive, which implies that parliamentarians rely on independent information provided by CSOs to punish and deter government misbehavior (McCubbins and Schwartz 1984; Scherer et al. 2008). CSOs broker critical information to parliamentarians, who can then exercise oversight, for example, vote against anti-democratic bills or inquire into government misbehavior.

Similarly, constitutional courts depend on CSOs to acquire information to rule against executive power grabs (Engstrom and Low 2019). That is, CSO activists often function as witnesses or provide expert information to court judges and prosecutors. In Kenya, for example, CSOs supported a legal complaint against government fraud in the 2017 elections by providing evidence for alleged electoral manipulation (Oloo 2018).

Government-imposed restrictions on CSOs can thus impede the functioning of horizontal accountability institutions by diminishing access to timely and high-quality information. When CSOs must deal with mounting restrictions, they have fewer resources to advise members of parliaments and to provide testimony in committees of inquiry or courts. Moreover, government-sponsored defamation against CSOs—i.e., portrayals of CSOs as elitist, foreign-sponsored, and unaccountable—not only discredits CSOs in the broader public but also convinces parliamentarians and court judges that it is ineffective or even counter-productive to rely on CSO-provided information. Such defamation may also make parliamentarians and court judges reluctant to engage with CSOs out of fear of becoming targets of government repression too (Bullain and Rutzen 2018).

Finally, as alluded to above, CSO mobilization efforts can add both weight and legitimacy to anti-government positions taken by parliaments, courts, and other oversight institutions. But restrictions reduce CSOs' mobilization efforts, weakening the pressure that horizontal oversight institutions can exert on the executive. For example, research shows that when high court judges do not observe mass mobilization against authoritarian moves, then they are less likely to issue decisions against governments and more readily align with a perceived pro-incumbent majority preference (Chilton and Versteeg 2018).

Figure 2 summarizes our second causal mechanism. Restrictions diminish the collaboration between CSO activists and horizontal accountability institutions. Horizontal oversight institutions subsequently lack CSO-provided information and CSO mobilization-based signals of mass support, which diminishes their ability to constrain the government. Taken together, we derive the following observable implication: Mounting restrictions on CSOs negatively correlate with the quality of horizontal accountability institutions.

Fig. 2
figure 2

Visualization of argument part II

The Independent Agency of CSOs

CSOs retain agency in dealing with restrictions on their activity (e.g., Fransen et al. 2021; Smidt et al. 2021). While restrictions like foreign funding prohibitions—in the most drastic cases—force the closure of independent organizations (Dupuy et al. 2015), CSOs can sometimes adapt to a more restrictive environment and even organize counter-mobilization.

The ability to adapt may partly hinge on the severity of restrictions (e.g., the number of diverse types of restrictions and the number of CSOs affected by them). Civil society activists may be able circumvent sporadic interference with a few selected organizations. By contrast, bundles of restrictions on several CSOs make it more difficult to adapt and resist (Smidt et al. 2021).Footnote 7

When CSOs do not succumb to restrictions but circumvent them, they can continue to defend horizontal oversights institutions from government attacks. However, if CSOs close shop or go into hiding, the decline in the quality of horizontal accountability institutions may be longer-lasting and more drastic. We probe this expectation in the case studies.

Research Design

We evaluate the main hypothesis using statistical analyses across a global set of countries in the period 1992–2018. We choose this period, because CSOs have significantly grown in number and influence since the end of the Cold War, especially in the Global South and former communist countries. Thus, the period of analysis ensures that all countries in the sample have a non-zero probability of exhibiting government strategies to counter CSO activity. Our analyses end in 2018 due to data availability, when we started this project. The unit of analysis is country-year.

We additionally examine our argument in two cases: Kenya and Turkey in the 2013–2017 period. An increase in restrictions on CSOs in Kenya in 2013 led to a temporary decline in the quality of horizontal accountability institutions. Restrictions on CSO activity in 2013 in Turkey helped erode horizontal checks on executive power. Since both cases support the main observable implication of our argument and are thus “typical cases” (Seawright and Gerring 2008), we can use these cases to trace the two proposed causal mechanisms (Lieberman 2005).

Moreover, while the cases are similar in several respects (e.g., a regional neighborhood with semi-democratic countries and the availability of large-scale external democracy support), there are also crucial differences (e.g., the geopolitical status, the level of socio-economic development, and the legacy of military dictatorship). As such, evidencing the proposed causal pathways in both Kenya and Turkey supports the generalizability of these pathways.

Finally, the cases differ in the severity of restrictions (i.e., restrictions in Turkey were more severe) and the historical strength of CSOs at the outset of the rise in restrictions (i.e., Kenya had a slightly stronger civil society sector). Therefore, we can explore the proposed nuances, i.e., that more severe restrictions on CSOs lead to longer and more drastic declines in horizontal accountability institutions.

Dependent Variables

Our dependent variable captures the quality of horizontal accountability institutions using an index designed by Lührmann et al. (2020). The index is available in the V-Dem dataset version 10 (Coppedge et al. 2020; Pemstein et al. 2020). It is continuous and approximately normally distributed ranging from -2.212 to 2.207 in the full sample. The index is constructed using a hierarchical latent variable model. The model inputs are three groups of variables measuring horizontal accountability.

The first group includes variables measuring judicial independence and activity, i.e., the degree of lower court independence, the degree of high court independence, the degree of government compliance with lower courts, the degree of government compliance with high court, and the degree of government respect of the constitution. The second group includes variables capturing legislative independence and activity, i.e., whether an independent legislature exists, the degree of independent investigation of government wrongdoing by the legislature, and the degree of independent questioning by the legislature. The third group consists of one variable measuring the independence and activity of other oversight bodies, i.e., the degree of independent questioning by the comptroller general, general prosecutor, or ombudsman.

Each of the variables within the three groups is based on multiple expert judgments (on the ordinal scales implied by the variables), which are aggregated using latent variable models. More information on coding and aggregation is found in Pemstein et al. (2020).

Independent Variables

To capture the imposition of restrictions on CSOs, we rely on two datasets and measures. First, we use a dataset by Bakke et al. (2019) that records the types of restrictions against CSO activity in a given country and year. Their source data are US State Department reports in the period 1995–2016. We create an annual count of 12 restriction types: banning specific CSOs, curtailing travel, restricting their visits to government sites, limiting their domestic funding sources, limiting their international funding sources, creating difficulties in obtaining visas or denying visas, making it difficult to register, censoring their publications, harassing civil society activists, surveilling activists, arresting activists, and killing activists.

Second, we employ a variable from the V-Dem dataset version 10 that captures the degree to which the government restricts CSOs (v2csreprss) in the full period of analyses, 1992–2018 (Bernhard et al. 2015; Pemstein et al. 2018). The variable aggregates judgments made by multiple V-Dem expert coders regarding the question of whether the government attempt to repress CSOs. The experts indicate their answer on a five-point scale from no repression to severe repression. Their judgments are then aggregated using a latent variable model that returns a continuous, approximately normally distributed variable ranging from − 3.503 to 2.879. We reverse the original scale of this variable and transform it into purely positive values. At the high end of the scale, governments violently pursue all CSOs. At the low end, CSOs are free to organize without fear of sanctions.

While higher values on the Bakke et al. measure indicate more forms of restrictive measures imposed on CSOs, higher values on the V-Dem measure indicate that restrictions target a greater number of organizations and activists and that restrictions are more violent. Despite these differences in conceptualizing restrictions on CSOs, the two measures are positively and significantly correlated (rho = 0.5892, p-value = 0.000). In Appendix G, we visually compare the V-Dem and Bakke et al. measures.

Both measures are not without limitations. On the one hand, the count of restriction types by Bakke et al. implicitly assumes that each restriction type is equally severe. This assumption could be violated, for example, if arrests of activists hamper CSO activity more than visa difficulties. We therefore re-estimate our models with the V-Dem measure using expert judgments of the overall severity of restrictions. On the other hand, the measure of restrictions from the V-Dem project may suffer from consistency bias. Specifically, some of the V-Dem expert coders that code restrictions on CSOs may also be responsible for coding our V-Dem dataset-based dependent variable, i.e., the degree of horizontal accountability. Since human coders strive for consistency in their judgments, a correlation between restrictions and horizontal accountability might be an artifact. Yet, the pool of expert coders rating horizontal accountability and the pool of expert coders evaluating restrictions on CSOs are not identical. Complementing the analyses with the restriction measure from Bakke et al. further alleviates concerns that results are driven by consistency bias.

We operationalize the year-to-year changes in restrictions by calculating the difference between contemporaneous restrictions on CSOs and restrictions one year previously. Overtime changes rather than levels are the correct measure because we theorize about how newly imposed restrictions can make it harder for CSOs to prevent executive aggrandizement. But our results hold for levels of restrictions, too.

Identification Strategy

We use linear regression analyses to test whether increases in restrictions on CSOs are associated with a lower level of horizontal accountability. Yet, even if restrictions and horizontal accountability are negatively associated, interpreting these correlations as supportive of our argument—that restrictions remove obstacles to attacks on horizontal accountability institutions and hamper their control functions—requires us to examine two alternative explanations.

The first alternative explanation is reverse causation: An incumbent may have already tightened executive control over horizontal accountability institutions when she decides to impose restrictions on CSOs and then just continues dismantling horizontal accountability institutions. To exclude that reverse causation is driving our results, we control for two lags of the dependent variable, i.e., the previous quality of horizontal accountability institutions. In other words, we hold horizontal accountability in the two previous years constant and test whether the imposition of more restrictions on CSOs still correlates with lower-quality horizontal accountability in the contemporaneous year.

In addition, we conduct a descriptive sequencing analysis. If the onset of substantive restrictions against CSOs temporarily precedes the onset of a decline in horizontal accountability at least in some countries, then this should alleviate concerns about reverse causation. Thus, we select countries where we observe a substantive decline in horizontal accountability over a two-year period (i.e., a decline by at least one-quarter of a standard deviation) and sort these countries into two groups:

  • The first group of countries exhibits “ex ante restriction episodes”: Horizontal accountability remains stable (i.e., only fluctuates within one standard deviation) within five years before the substantive decline in horizontal accountability, while restrictions increase substantively (i.e., by at least one-quarter of a standard deviation) in this period.

  • The other group exhibits “ex post restriction episodes”: Restrictions remain stable (i.e., only fluctuated within one standard deviation) within five years before the substantive decline in horizontal accountability, but horizontal accountability declines substantively (i.e., by at least one-quarter of a standard deviation) in this period.

If some countries exhibit ex ante restrictions, then this is further evidence in favor of our argument that restrictions on CSOs independently facilitate the erosion of horizontal accountability institutions.

The second alternative explanation is that unobserved reasons (omitted variables) explain the co-occurrence of government-initiated restrictions on CSOs and the erosion of horizontal accountability constraints. One class of unobserved reasons might be time-invariant structural difference between countries; for example, authoritarian legacies may both increase restrictions on CSOs and reduce horizontal accountability (Bernhard et al. 2017). Thus, we include country fixed effects.Footnote 8 Another class of reasons that may explain an association between restrictions and horizontal accountability varies over time. As detailed below, we control for several time-varying potentially confounding factors. In Appendix A (Table A4), we also show our main models with year fixed effects to control for global temporal shocks.

However, the thorniest challenge to identification is that the government restrictions on CSOs and the deterioration of horizontal accountability institutions are partly endogenous to the decisions of incumbent governments. To counter the possibility that both CSO restrictions and declining horizontal accountability are explained by the presence of authoritarian-minded incumbents, we take two routes.

First, we control for incumbent leader’ attributes with incumbency period fixed effects (i.e., intercepts). For example, in Kenya, the period 1992–2001 ruled by Moi has its own intercept, the period 2002–2012 ruled by Kibaki has its own intercept, and the period 2013–2018 ruled by Kenyatta has its own intercept. If authoritarian-minded government leaders would explain both restrictions on CSOs and the quality of horizontal oversight, then this control would substantively reduce the explanatory power of restrictions. We show that it does not.

Second, we examine whether restrictions on CSOs also correlate with auxiliary outcomes that are related to horizontal accountability but not directly in the purview of governments. Specifically, we examine whether increases in restrictions also correlate with the level of citizens’ engagement in public affairs. Citizens’ engagement is part of our causal mechanisms explaining decline in horizontal oversight but not completely controlled by governments (i.e., citizens have agency). If mounting restrictions on CSOs reduce citizens’ engagement, then this supports our argument that restrictions remove obstacles to the erosion of horizontal check-and-balance institutions. Then, we examine how restrictions relate to aspects of horizontal accountability, which are not directly in the purview of governments: the degree to which members of the legislature investigate executive decisions and the degree to which they question the executive in practice. If restrictions on CSOs indeed reduce this independent activity of the parliament, then this lends credence to our argument.

Of course, none of these additional tests are perfect. However, together with the two illustrative case studies, we hope that they buttress the empirical validity of our argument.

Control Variables

Neighboring democracies could help prevent a deterioration of the de facto power of horizontal oversight institutions and protect CSOs. Thus, we control for the mean regime type in a geographic region and year (not including the country under observation) using data on regime type from V-Dem (Coppedge et al. 2020; Pemstein et al. 2020).Footnote 9 Using V-Dem data, we also construct a variable for the legacy of military dictatorship and whether the current system is presidential. Both factors can lower the chances of consolidated horizontal check-and-balance institutions and potentially weaken civil society (Svolik 2008). Moreover, we include the current strength of independent civil society, which has been shown to influence democratization and democratic stability and plausibly affects the de-mobilizing impact of government-sponsored restrictions (Bernhard et al. 2017; Bernhard et al. 2020).

We also control for anti-government protest events using data from Clark and Regan (2016) and internal armed conflict using data from UCDP-PRIO Armed Conflict Dataset (Gleditsch et al. 2002; Pettersson et al. 2019) because contention can trigger restrictions and precede attacks on democratic institutions. Moreover, we control for economic development with GDP per capita and GDP growth, which are associated with authoritarian reversal and can also affect the CSO sector (Svolik 2008). We also include oil wealth that may inhibit democracy. Data for GDP per capita, growth, and oil rents are taken from the World Bank (World Bank 2019). All control variables are lagged by 3 years to avoid post-treatment bias.

Cross-National Results

The regression models presented in Table 1 support our expectation: Increases in restrictions on CSOs correlate with a lower level of horizontal accountability institutions (for full results, see Appendix A).

Table 1 Arellano-Bond models of quality of horizontal accountability institutions

Figures 3 and 4 illustrate this result. The magnitude of the negative correlation varies across the measures of restrictions. A standard deviation increase in restrictions measured with the Bakke et al. data is associated with a significant decrease in horizontal accountability by 0.01 units (see Fig. 3). A standard deviation increase in restrictions measured with the V-Dem data is associated with a significant decrease in horizontal oversight by 0.04 units (see Fig. 4).

Fig. 3
figure 3

Marginal effect of mounting restrictions (Bakke et al.)

Fig. 4
figure 4

Marginal effect of mounting restrictions (V-Dem)

A substantive explanation for the stronger effect coefficient on the V-Dem measure might be that it captures physically coercive restrictions against CSOs, while the Bakke et al. measure captures the variety of legal, regulatory, and practical restriction types on CSOs. Mounting coercive restrictions may be more difficult to circumvent than legal impediments, explaining the greater decline in horizontal accountability associated with the former. Yet, consistency bias among V-Dem coders may also account for the stronger finding. Therefore, the negative marginal effect of the Bakke et al. measure of restrictions lends credence to our argument.

While we naturally refrain from interpreting the coefficients on the control variables as causal effects, we note that only GDP growth and the strength of CSOs positively and consistently correlate with the quality of horizontal accountability institutions. In line with previous research, economically productive countries with a stronger civil society thus seem more insulated form democratic decline in horizontal accountability institutions (see Appendix A). Overall, the results provide evidence that increases in government-imposed restrictions serve as a warning sign of subsequent erosion of checks and balances.

Testing Alternative Explanations

As discussed above, by controlling for two lags of the dependent variable, we reduce the likelihood that our results are driven by autocratization trends, i.e., a prevailing decline in horizontal accountability.Footnote 10 In addition, our descriptive sequencing analyses show that nine countries see a significant crackdown on organized civil society (ex ante restrictions) before the quality of horizontal oversight institutions deteriorates. Five cases show the opposite pattern of ex post restrictions, i.e., a significant deterioration in horizontal accountability followed by restrictions imposed on CSOs and further decline in horizontal accountability.

Figure 5 illustrates the overtime sequences of these countries using red bars for periods of declining horizontal accountability after ex ante restrictions and grey bars for periods of declining horizontal accountability after ex post restrictions. We note three results: First, among the countries that experience drastic increases in restrictions and subsequently drastic declines in horizontal oversight, we find countries from all world regions. Both the sequence and its occurrence in all geographic regions lend support to our argument that mounting restrictions facilitate the erosion of parliaments, courts, and other oversight bodies.

Fig. 5
figure 5

Sequencing analyses

Second, most countries however experience both a rise in restriction and a decline in horizontal accountability at the same time and the changes in restrictions and horizontal accountability happen in a more gradual fashion. As such, the countries in Fig. 5 with a clear sequence and substantive changes from one year to the next are the exception. Third, the ex post restriction cases do not necessarily counter our argument because even ex post restrictions can facilitate further erosion of horizontal oversight bodies, as we see in Nicaragua (2010–2012) or Zambia (2012–2018).

Furthermore, we want to ensure that the correlation between mounting restrictions and lower horizontal oversight is not driven by the prevalence of authoritarian government leaders alone. Therefore, we re-estimate our main model controlling for incumbency period fixed effects (i.e., time-invariant attributes of incumbent leaders during their period in office). The size and significance of the coefficient on restrictions on CSOs remain virtually the same regardless (see Appendix C). We conclude that leaders’ authoritarian predisposition cannot explain the association between mounting restrictions and declining horizontal oversight.

Moreover, we explore whether a crackdown on CSOs also relates to auxiliary outcomes that usually help determine the quality of horizontal oversight but are less likely to be in the purview of governments. As Fig. 6 illustrates, our additional regression analyses show that interference with CSOs also negatively correlates with citizens’ public engagement. This result aligns with our argument that restrictions lower obstacles (here citizens’ engagement) to governments’ executive aggrandizement. Moreover, restrictions on CSOs negatively correlate with the degree to which parliaments investigate and question the executive in practice. This finding buttresses the argument that restrictions diminish the power of horizontal oversight institutions (see Appendix D and E for full tables).

Fig. 6
figure 6

Correlations between restrictions and mechanism-related outcomes

While none of these identification strategies are flawless, together, we think they provide convincing evidence that restrictions imposed on CSOs contribute independently to the erosion of horizontal check-and-balance institutions.

Finally, another explanation for the negative correlation between increasing restrictions and deteriorating horizontal accountability is that restrictions target anti-system CSOs, which then still reduce the quality of horizontal accountability institutions. While we think that this explanation is less plausible considering the mounting evidence of shrinking civic spaces for democracy-promoting CSOs, Appendix F shows that when we exclude countries with a high prevalence of anti-system CSOs, our findings remain the same.

Tracing the Hypothesized Pathway

Using evidence from a variety of secondary sources, including news articles, reports, and academic literature, we examine the hypothesized pathways of our argument in two cases: Kenya and Turkey from 2013 to 2017. We have chosen these cases because their pattern of mounting restrictions and declining horizontal accountability aligns with our argument, allowing us to inquire whether our proposed causal mechanisms are responsible for this pattern. As Fig. 7 illustrates, in Kenya, a minor increase in restrictions between 2012 and 2013 goes hand in hand with a temporary decline in horizontal accountability.Footnote 11 In Turkey, a positive trend in government-imposed restrictions since 2006 and a steep increase in restrictions from 2013 onward co-occur with a drastic and long-lasting decline in horizontal oversight.Footnote 12

Fig. 7
figure 7

Illustration of patterns of restrictions and horizontal accountability

In both cases, we show that restrictions reduced CSO mobilization efforts and, thereby, diminished obstacles (citizen protest and opposition voting) to government-sponsored attacks on horizontal oversight institutions. We also find evidence in Turkey that restrictions reduced the capacity of CSOs to provide inputs for horizontal oversight institutions, diminishing the latter’s effective control over the executive. In addition, the case of Kenya also illustrates that minor increase in restrictions against an initially strong civil society may temporarily hamper CSOs but then spark counter-mobilization, thereby helping to preserve horizontal accountability in the long run.

Kenya: Increasing Restrictions on CSOs and a Temporary Bump in Horizontal Oversight

CSOs played a crucial role in the democratization struggles in Kenya. Civil society confronted colonial governments in the first half of the twentieth century and forced authoritarian president Moi to hold elections and accept defeat at the beginning of the twenty-first century (Mati 2020). When Uhuru Kenyatta took office in 2013, history had taught him that CSOs pose challenges to government power. The Kenyatta administration immediately tried to reign in independent civil society through defamation campaigns and legal harassment. Restrictions had some crippling effects on CSO mobilization and initially helped weaken horizontal oversight institutions. However, the vibrant CSO scene in Kenya quickly adapted to the relatively moderate restrictions and even pushed back against anti-civil society legislation, which in turn helped protect horizontal accountability institutions.

Just after inauguration in 2013, Kenyatta’s government targeted CSOs with restrictive measures. It refused to implement the progressive civil society legislation (ICNL 2019a), the Public Benefits Organizations (PBO) Act, using the outdated NGO Coordination Act to deregister critical organizations (The Observatory for the Protection of Human Rights Defenders 2018, 4–5). Moreover, the Kenyatta administration proposed several restrictive amendments to the PBO Act, for example, capping foreign funding for domestic CSOs to a maximum of 15% of their income (Goitom 2013). In parallel to these (attempted) legal restrictions, harassment and defamation intensified (Bakke et al. 2019). CSOs faced smear campaigns led by high-ranking government officials, freezing of bank accounts and assets, and arbitrary arrests (CIVICUS 2015).

At least initially, restrictions on CSOs had crippling effects. In the period 2014–2019, there were increasing rates of CSO deregistration, an unusual decrease in CSO budgets, and much slower CSO-sector growth compared to previous years (NGOs Co-ordination Board 2019). Deregistration affected some critical and powerful CSOs, such as the African Centre for Open Governance (AfriCOG). Kenyan BBC Africa Business Editor warned: “Kenya is sliding into a dictatorship” and this trajectory “began with the vilification of civil society as an ‘evil society’ by senior aides to the president” (Madowo 2018).

Indeed, the Kenyatta government actively worked to dismantle the elaborate system of horizontal check-and-balance institutions that had been put in place by the 2010 constitution. From 2013 onward, “the executive forced through a number of bills that diminished the power of the legislature” (Ebole and Odhiambo 2017, 3). These included the Kenya Defense Forces Bill, 2015; the Kenya Information and Communication Act; and the Security Laws Act. Moreover, “the executive sought to amend the Judicial Service Act … to increase the powers of the president in appointing the chief justice” (Ebole and Odhiambo 2017, 4). Beyond executive aggrandizement via legislation, the government also just ignored norms and rules that previously guaranteed judicial independence. In 2015, the government appointed two supporters of the ruling coalitions as members of the Judicial Service Commission (Mr. Arap Korir Bett and Ms. Waceke Guchu), an important state body that appoints judges, including those of the Supreme Court (Sunday Nation Team 2015). Finally, “the executive also routinely disregarded binding court orders, revealing a lack of respect for the role and powers of the judiciary” (Ebole and Odhiambo 2017, 4).

Supporting the first proposed mechanism, government-imposed restrictions soaked up CSOs’ resources for monitoring government attacks on horizontal oversight institutions and mobilizing mass publics in their defense. For example, while Kenyan newspapers reported the appointment of government supporters to the Judicial Service Commission as a tactic to dismantle horizontal checks on the executive (Sunday Nation Team 2015), neither newspaper archives nor any of the major databases of protest events—the Armed Conflict and Events Database (ACLED) and the Social Conflict in Africa Database (SCAD)—record any public protest event against this decision in the period 2013–2018 (Raleigh et al. 2010; Salehyan et al. 2012).Footnote 13 When President Kenyatta ridiculed the nullification of 2017 presidential election results by stating “six people have decided they will go against the will of the people” (Burke 2017), only a few residents of Bonyamatutu village in Nyamira County, the home of Chief Justice Maraga, protested against Kenyatta’s verbal attacks on the judiciary (Raleigh et al. 2010). There was no mass mobilization against government-sponsored attacks on horizontal oversight institutions. Restrictions and rhetorical vilification of CSOs likely helped the government to relieve pro-democracy pressures.

We find no evidence for the second causal mechanism of our argument. Instead, nuancing our argument, CSOs banded together with horizontal oversight institutions to fend off restrictions on their activity (Berger-Kern et al. 2021; Cheeseman and Dodsworth 2023). CSOs persuaded parliamentarians of the negative developmental consequences of government-sponsored restrictive amendments to the above-mentioned progressive civil society legislation (the PBO Act). In December 2013, a majority in parliament voted against adding foreign funding caps and other restrictive amendments. In 2014 and 2015, CSOs maintained the momentum and thwarted government proposals for further restrictive legislation. While the government still refuses to implement the progressive PBO Act, CSOs could effectively defend its space from deteriorating further during the 2013–2017 period.

CSO activists used their alliance with members of parliaments and court judges to thwart further executive aggrandizement. For instance, in 2017, the pro-government majority in the National Assembly failed to remove the independent auditor general. Responding to a case filed by a Kenyan human rights activist, the High Court stopped the National Assembly from discussing a petition against the auditor general (Ogemba 2017). In 2018, prominent CSOs helped halt a constitutional reform project (the Building Bridges Initiative), which would have enlarged the executive and enhanced its control over the judiciary. Specifically, CSOs supported a High Court judgment that deemed the reform project unconstitutional (Kenya Human Rights Commission 2021; Progressive International 2021; Transparency International Kenya 2021).

Overall, resilience of the Kenyan civil society contributed to the temporary nature of the decline in horizontal oversight. Existing research suggests numerous explanations for CSOs’ resistance, e.g., broad-based CSO coalitions, pre-existing networks, civil society sector strength (Berger-Kern et al. 2021), and international support (Cheeseman and Dodsworth 2023, 627–28). We add that the relatively moderate nature of restrictions in Kenya (i.e., administrative burdens and defamation rather than outright prohibition and large-scale physical repression) likely also contributed to CSOs’ ability to circumvent government interference and mount successful resistance.

Turkey: Severe Restrictions on CSOs and Substantial Dismantling of Horizontal Oversight

Since the onset of the EU accession process from 1999 onward, CSOs emerged as “important and influential actors” in Turkey (Eslen Özerkan and Mutlu 2008, 20). Yet, after the slowdown of EU accession negotiations in 2006 and after the brutal repression of anti-government protests in 2013, government-imposed restrictions truncated their new-won power. The de-mobilization of the civil society sector created propitious conditions for the authoritarian ambitions of populist leader Recep Tayyip Erdoğan. With the 2017 constitutional referendum, Turkey transitioned from a parliamentary democracy to a presidential system devoid of independent horizontal oversight.

Between May and September 2013, the so-called Gezi Park protests mobilized nearly 3.5 million protestors (Anisin 2016, 411). The small-scale environmental action to save the Gezi Park in Istanbul evolved into a mass mobilization event against the incumbent AK party and then-prime minister Recep Tayyip Erdoğan. The government’s reaction was harsh. The repression of protesters led to several deaths, over 8000 injured, and 4900 detainees (Amnesty International 2013, 56–57). Tellingly, this wave of repression was accompanied by anti-civil society statements from Erdoğan and his close supporters (Taş 2015, 780).

After the protests, the surge in restrictions against independent CSOs was encompassing and severe.Footnote 14 First, government authorities oppressed critical voices based on ambiguous legal frameworks. Extensive auditing, frequent fiscal penalties, and the arbitrary closure of CSOs became frequent (Yabanci 2019, 291). Second, the government actively co-opted civil society by purging anti-government actors in existing CSOs and the founding of new pro-government CSOs. A few months after the Gezi protests, the government launched an initiative to create a government-friendly network of CSOs (Yabanci 2019, 294ff).

Restrictions effectively de-mobilized many independent CSOs. In June 2016, Erdoğan announced that “civil society organizations working against the state have largely been destroyed” (Anadolu Agency 2016).Footnote 15 Indeed, independent researchers observed that “[t]he result [of the crackdown] is restricted engagement of autonomous civil society. Local and issue-based demands cannot be amplified and conveyed to a broader audience and [they are] definitely excluded from policymaking” (Yabanci 2019, 294). The government effectively “tamed” Turkish civil society “rendering it politicized, segregated and disabled” (Yabanci 2019, 293).

At the same time, the quality of horizontal check-and-balance institutions drastically declined. In the aftermath of the 2016 coup attempt, the government ruled without parliamentary or judicial oversight under a prolonged state of emergency, inter alia, dismissing thousands of court judges and prosecutors from their functions (UN Human Rights Office 2016; Mariniello 2019; Yildiz 2019). The attacks on horizontal accountability culminated in the 2017 constitutional reform (still under state of emergency). The position of prime minister was abolished, and the president became the head of the executive. The president now had the power to appoint ministers, prepare the budget, choose a share of the Supreme Court judges, declare a state of emergency, and dismiss parliament. In addition, the parliament lost its right to scrutinize ministers or propose an inquiry (Lowen 2017). The constitutional reform ultimately brought about “a populist, one-man system that jeopardizes legislative and judicial independence” (Ekim and Kirişci 2017).

Supporting the first proposed mechanism, restrictions on CSOs reduced the ability of electorate to hinder the dismantling of horizontal oversight bodies. Turkish civil society lent limited support to election observation efforts during the referendum, because many organizations feared reprisals for such an engagement (OSCE Office for Democratic Institutions and Human Righ 2017, 1–3). More importantly, defamation campaigns equating CSOs supporting the “no” campaign with terrorist supporters diminished the effectiveness of their voter outreach and information-sharing. As the OSCE Limited Referendum Observation Mission remarked, the constitutional referendum “took place on an unlevel playing field and the two sides of the campaign did not have equal opportunities” (OSCE Office for Democratic Institutions and Human Righ 2017, 1).

Furthermore, government-imposed restrictions on CSOs lowered the prospects of mass mobilization as an obstacle to executive aggrandizement. Preceding the 2017 referendum, the OSCE Limited Referendum Observation Mission noted the obstruction of CSOs which supported the “NO” campaign against the constitutional change (OSCE Office for Democratic Institutions and Human Righ 2017, 1–3). While CSOs helped mobilize for some protest activity in 2017 (especially after the election board’s decision to approve some 1.5 million unstamped referendum ballots), their mobilization attracted far fewer citizens than the 2013 Gezi demonstrations and were much more short-lived and sporadic (Zihnioglu 2019). Footnote 16

Supporting the second proposed mechanism, the Turkish case illustrates how restrictions diminished the parliament’s control function. On 19 October 2011, the Turkish parliament established a 12-member “conciliation committee” to discuss revisions to the existing constitution. International observers like the Council of Europe positively remarked that the conciliation committee initiated “consultations with of all the driving forces in Turkish society” (Council of Europe 2013). Indeed, by April 2012, about 68,000 written comments and testimonies had been received and 160 hearings had been held, including with civil society groups (Keyman and Kanci 2013). Moreover, “a plethora of Turkish CSOs assembled and established platforms to influence the possible contents of a new constitution.” Through these platforms, CSOs generated a public debate, wrote reports, and organized conferences to convey their demands and proposals (Kuşku Sönmez 2019, 972).

After the crackdown on CSOs in 2013, none of the civil society groups that were engaged with the conciliation committee survived. The development of 2017’s constitution project did not involve any CSOs. As the European Commission notes about the preparation of the 2017 constitutional reform, “there was no genuine opportunity for open discussion with all political forces nor did it involve civil society” (European Commission 2018). In January 2017, parliament approved all 18 constitutional amendments of 2017. The missing input from CSOs contributed to ineffectiveness of the Turkish parliament in constraining executive aggrandizement through constitutional reform.

Finally, the comparison between Turkey and Kenya suggests that the severity of restrictions initiated by the AK party government together with the relative weakness (or rather newness) of the Turkish civil society sector helps explain the limited resilience. The large-scale crackdown on CSOs diminished civic activism legacies of the 2013 Gezi protests and marginalized civil society groups in political affairs (Zihnioglu 2019), thereby contributing to the steep decline in the independent power of horizontal accountability institutions.

Conclusion

This demonstrates that there is a strong correlation between mounting restrictions on CSOs and lower quality of horizontal accountability institutions. Our analyses further suggest that this correlation is unlikely to be explained by reverse causation (i.e., a prevailing trend of deteriorating horizontal accountability) or omitted variable bias (e.g., the authoritarian mindset of incumbent leaders and persistent anti-democratic structures) alone. Instead, the evidence leads us to conclude that restrictions on CSOs are likely one of the drivers of accelerated erosion of horizontal accountability institutions.

Qualitative evidence from Kenya and Turkey further supports our causal story. In Turkey, severe restrictions de-mobilized CSOs, thereby diminishing mass protests and hampering electoral opposition to executive aggrandizement. Restrictions also reduced the collaboration between CSOs and parliament, weakening the latter’s ability to constrain constitutional amendments. In Kenya, moderate restrictions reduced CSOs’ ability to mobilize citizens for mass protests, potentially facilitating a temporary weakening of horizontal oversight institutions. However, CSOs’ counter-mobilization helped forestall further executive aggrandizement. From a comparative perspective, the less severe level of restrictions in Kenya is one plausible reason for CSOs’ resilience and effective resistance.

The manuscript connects and contributes to research on shrinking civic space and the literature on democratic decline. While others have explored how government interference with CSOs affect international shaming campaigns (Bakke et al. 2019), health service provision (Heinzel and Koenig-Archibugi 2022), and the survival of CSOs (Dupuy et al. 2015), we show that severe restrictions also have repercussions on core democratic institutions like parliaments and courts. Our findings thus suggest that restrictions on CSOs are not only symptoms of authoritarian governance (e.g., Diamond 2015; Foa and Mounk 2016, 2017; Levitsky and Ziblatt 2018). Governments that go after civil society can more effectively dismantle horizontal oversight institutions. For policy, the study offers evidence that concerns about “shrinking civic spaces” are warranted. The protection of CSOs does not only support civil society's important work but also hinders backsliding in the quality of horizontal accountability institutions like parliaments and courts.

Restrictions on CSO activity are not the only authoritarian tool that may facilitate democratic erosion. Future research should systematically investigate the institutional consequences of other government strategies to avoid accountability, i.e., the killing of journalists, interference in academic organizations, and restrictions on Internet freedom. What our analyses reveal is that restrictive measures may sometimes appear less severe than other methods of repression, such as torture and extrajudicial killings, but they can have significant detrimental consequences for the quality of democracy.