Civil Society’s Effects on Electoral Democracy
Theoretical Introduction
In Making Democracy Work, Putnam, Leonardi, and Nanetti (1993) attempt to explain the regional differences in institutional performance between the North and South of Italy using the proxy of civic community. Putnam, Leonardi, and Nanetti (1993) find that while socio-economic modernity, advanced by industrialization and economic development, can explain the high performance of Northern Italy, it cannot fully explain the regional differences in institutional performance across the country. Consequently, Putnam, Leonardi, and Nanetti (1993) deduce that civic community is a more powerful force driving effective governance. We know from Putnam, Leonardi, and Nanetti (1993) that the civic community conveys ideals of cooperation and solidarity, allowing for increased opportunities for collective action. We also know that the civic community is linked to political leaders being more egalitarian and less focused on hierarchical authority and clientilism. A non-civic community, fostering clientelism and hierarchical authority, might be more prone to corruption and the erosion of democratic institutions as leaders are more focused on transactional politics. Therefore we expect that as the civic community strengthens, a country is more likely to be categorized as a democracy. We can see this relationship in the Varieties of Democracy dataset referring to Figure 1 and Figure 2 . It is important to highlight that the y-axis in Figure 1 is the electoral democracy index. I find that electoral democracy is more practical in this context of analysis as this index focuses on core democratic principles like free and fair elections, voting rights, independent media, freedom of expression, and freedom of association, among others. This theoretical foundation led me to predict that civil society will be positively associated with electoral democracy, while I expect that state repression will moderate these positive effects.
Figure 1 plots the average V-Dem core civil society index (v2xcs_ccsi) against the average electoral democracy index (v2x_polyarchy) for all countries. Countries such as Nigeria and Liberia show relatively strong civil societies but weak electoral democracy, raising questions about the ability of CSOs to influence political outcomes in repressive or semi-democratic regimes.
In investigating what factors might undermine democracy within a strong civil society I came across Croke et al. (2016)’s examination of how education can decrease electoral participation in authoritarian regimes. They find that in Zimbabwe, more educated individuals were less likely to turnout to vote, attend political meetings, or contact government officials which the authors classify as deliberate disengagement. Croke et al. (2016) charactarize this phenomena by concluding that educated citizens are more politically aware and may then perceive participation as futile or as a way of avoiding risk in that engagement under an authoritarian regime may carry dangers like repression or surveillance. I then questioned, What institutional factors contribute to the coexistence of a strong civil society and weak electoral democracy?
One looking to answer such a complex question might look to Larry Diamond’s: Authoritarianism Goes Global: The Challenge to Democracy. Since the collapse of communism, Diamond outlines three trends that have altered the global political landscape being: the democratic surge, democratic backlash, and the authoritarian surge. In an attempt to gain a clearer picture of factors that might undermine democracy in a strong civil society I am most concerned with the authoritarian surge. Diamond illustrates that media initiatives by the “big five” authoritarian regimes like Russia’s RT, China’s CCTV, and Iran’s Press TV seek to undermine western ideals and control political narratives in said regimes. Authoritarian regimes also seek to undermine elections and silence opposition to their practices.
Building on these insights, I now turn my focus from broad institutional threats to democracy, to a more targeted analysis of the interaction between civil society repression and civil society strength and the subsequent influence on electoral democracy. In authoritarian regimes some civil society organizations are allowed to form / operate, while some states may restrict formation. Certain CSOs in authoritarian regimes may promote civic engagement in cultural and social domains leading to a stronger civil society, but political mobilization and pro-democracy coalitions are likely repressed undermining electoral democracy. In consideration of this, I looked to Civil Society and Political Change in Asia by Muthiah Alagappa, which is a comprehensive examination of the connection between civil society and political change in Asia. Alagappa (2004) investigates how CSOs interact with the state, political society, and economic factors across different regime types. Accordingly, the author makes three key propsitions: civil society is a distinct public sphere where non-state actors operate to influence state policy, CSOs play diverse roles ranging from opposition of authoritarian regimes to being regime-supported, and the strength of civil society depends on political structure and economic development. After considering an investigation of these factors and understanding civil society organizations [CSOs] as an integral piece to a strong civil society, I was led to the following question upon which I am basing my analysis: How does the interaction between CSO repression and CSO strength influence electoral democracy?
Let us now turn to more concrete literature on the history of CSO repression and how the repression of CSOs may impact democracy. In the mid 1990s, Rutzen (2015) asserts that CSOs enjoyed a positive reputation within the international community stemming from their contributions to health, education, culture, etc. Rutzen (2015) reinforces that there was a shift toward declining enthusiasm for supporting CSOs especially in countries that had undergone political change in the 1980s and 1990s. Over time, said governments felt the need to strengthen state institutions and consolidate power. Most of this focus came from within “hybrid” or “semi-authoritarian regimes”, in that these regime types held elections but showed little commitment to democratic values like human rights and the rule of law, principles that CSOs are so keen in supporting. Consequently, Rutzen (2015) finds that governments began tightening control on CSOs. Leaders within autocratic regimes adopted ideas similar to Putin’s concept of “managed democracy” which quickly devolved into “managed civil society”. Rutzen (2015) finds two key patterns that emerged: CSOs were allowed to function as long as they remained uninvolved in politics, while some states tried to co-opt CSOs while repressing those that resisted government control. As a result, civic space quickly contracted and Rutzen (2015) invokes data from the International Center for Not-for-Profit-Law to show that between 2004 and 2010, more than fifty countries considered or did enact measures to repress civil society. According to Rutzen (2015) some of the many legal measures used by governments are: requiring government approval for international funding for CSOs, stigmatizing internationally funded CSOs, capping the amount of international funding a CSO could receive, and using defamation and treason to bring criminal charges against CSOs receiving international funding. Important to highlight are the justifications that governments used for CSO repression being, protecting state sovereignty, promoting transparency and accountability in civil society and enhancing aid effectiveness. An example of repression tactics highlighted by Rutzen (2015) is that in Ethiopia, CSOs are prohibited from receiving more than 10% of their funding from international sources if said CSOs look to promote democracy.
I next turn to Bernhard et al.’s Civil Society and Democracy in an Era of Inequality to understand how CSOs may impact electoral democracy across different regime types. Bernhard, Fernandes, and Branco (2017) outlines the Gramscian view of CSOs as a way to resist authoritarianism. The Gramscian view holds that CSO mobilization can trigger democratic transitions through challenging authority and building networks of support in opposition to the regime. CSO repression, of course, works against this possibility of challenging authoritarian regimes. The Tocquevillian view holds that in democratic regimes, CSOs promote political engagement, participation, and responsiveness. However, Bernhard, Fernandes, and Branco (2017) also note that authoritarian elements in civil society, when linked to the state and allied with political actors, can impede equality and democratic outcomes in newly established democracies. Bernhard, Fernandes, and Branco (2017) conclude that civil society is decisive in the emergence of democracy and to the depth of democracy, however, Bernhard, Fernandes, and Branco (2017) importantly highlight that civil society may only be decisive at critical moments of major historical change such as a regime transition. It may therefore be that during these regime changes, CSOs are repressed the most, and therefore this contribution will be critical in robustness tests of my findings.
I will measure civil society strength by the proxies of CSO structure, CSO entry / exit into public life and CSO participatory environment. CSO structure measures the degree to which smaller versus larger CSOs dominant the political space. Low fragmentation of CSOs signals competition between smaller and larger CSOs to influence policy and mobilize support and political engagement, while high CSO fragmentation signals large CSOs dominate and are accorded special weight by policy makers. A CSO space with more competition for policy influence is a stronger one that is less prone to government influence. CSO entry and exit into public life measures the extent to which the governments controls CSOs and the political activities they are engaged in. Under monopolistic control, the only CSOs allowed to engage in political activity are government sponsored. While, when CSO entry and exit into public life in unconstrained, the government does not impede CSO formation and operation. Finally, CSO participatory enviorment measures the degree to which pepole are involved in civil society organizations.
Finally, why might CSO repression moderate the effect of civil society strength on electoral democracy? Critically, the presence of CSOs suggests a strong civil society in certain contexts. However, their ability to promote electoral democracy, foster political engage, and express interest depends on the political climate in which the CSO operates. When state repression is high, even well-organized and broadly supported CSOs may face barriers to mobilization and engagement such as cuts to international funding and legal constraints. Accordingly, even if CSO formation and operation within public life is possible, repression can undermine CSOs’ ability to influence democratic outcomes.
This introduction began with the foundational insights of Putnam et al, which emphasize that civic community is a core driver of institutional performance and democratic survival. Therefore, I began examining factors that may undermine democracy even within a strong civil society. I then refined my investigation to state repression of CSOs. We know from existing literature that CSOs play a critical role in upholding democracy and state repression of CSOs may weaken democracy. This contribution, building on the existing literature, attempts to understand CSO strength through the lenses of entry and exit into public life, CSO fragmentation, and participatory environment.
I proceed as follows: I construct an index of civil society strength by combining three V-Dem variables — v2cseeorgs, which measures how freely civil society organizations can enter and exit public life, and v2csstruc, which captures the degree of fragmentation and competition among CSOs, and v2csprtcpt, which captures the involvement of people in CSOs. Together, these variables represent the degree to which civil society is both institutionally open and structurally competitive and effective in engaging citizens, forming a more comprehensive measure of CSO strength. I then interact this index with civil society repression, measured by v2csrprss. This interaction allows me to test how repression conditions the effect of civil society strength on electoral democracy, measured by v2x_polyarchy. The goal is to assess whether repression undermines not just the formation and structure of CSOs, but also their capacity to translate into democratic outcomes. This modeling approach tests my core hypothesis: the effect of civil society strength on electoral democracy weakens as state repression increases.
Hypothesis
I hypothesize that civil society strength has a positive effect on electoral democracy. However, this expected positive relationship is conditional on the level of CSO repression. Specifically as repression increases, the democratic benefits of a strong civil society weaken. Simply, my logic follows the assumption that in regimes where repression is high, CSOs are less likely to form and operate freely and engage in advocacy and political mobilization contributing to lower levels of electoral democracy.
Civil society strength is measured by a composite index of CSO entry and exit (v2cseeorgs), CSO structure (v2csstruc), and CSO participatory environment (v2csprtcpt). While CSO repression is measured by a VDem variable: v2csrprss.
To test this hypothesis, I create a civil society strength index by combining v2cseeorgs and v2csstruc_2, and v2csprtcpt and interact this index with v2csrprss in my model. I chose to use v2csstruc_2 for my index of civil society strength because it holds that neither large nor small CSOs dominate and each type of CSO contends to have their voices heard by policy makers. This ensures uniform fragmentation and limited dominance of one type of CSO in the civic space contributing to more opportunities to gain support within the state and represent interests on policy making decisions.
Data-Set and Specific Variables to Consider
| Summary Statistics of Key V-Dem Variables | ||||
|---|---|---|---|---|
| Mean | Median | Min | Max | |
| State Control of Entry and Exit of CSOs into Public Life | −0.19 | −0.34 | −3.31 | 3.69 |
| CSO Participatory Environment | −0.20 | −0.32 | −3.40 | 3.17 |
| CSO Structure - Fragmentation | 0.34 | 0.29 | 0.00 | 1.00 |
| Civil Society Strength Index | 0.43 | 0.41 | 0.01 | 0.93 |
| Civil Society Organization Repression | 0.10 | 0.28 | −3.36 | 3.73 |
| Electoral Democracy Index | 0.26 | 0.17 | 0.01 | 0.92 |
The main variables I am considering in my analysis are CSO repression (v2csreprss) (which I reversed so that higher figures indicate more repression), an index of civil society strength (civil_society_strength), and an electoral democracy index (v2x_polyarchy). These variables come from the v-dem dataset where the unit of analysis is a country and the sample is most sovereign and some non-sovereign territories.
The Varieties of Democracy dataset is not based on government self-reports or documents and instead relies on country experts who code conditions based on specific prompts. Since expert-coders can disagree, V-Dem applies a Bayesian Item Response Theory to estimate true scores for each country and year. Every variable in the dataset comes with an entry in the V-Dem codebook and all coder level data is publicly available.
The CSO repression variable indicates to what extent the government represses CSOs and varies in terms of monopolistic control to severely repressive at the highest level and not repressive at the lowest. I created the index of civil society strength through merging CSO fragmentation, CSO entry and exit into public life, and CSO participatory environment and standardizing each variable along a 0 to 1 scale with equal weight. Figures 2, 3, and 4 illustrate the relationships between the variables while the summary table provides the mean, median, and range of each variable of interest.
Figure 2 shows the strong and positive relationship between my civil society strength index and electoral democracy, a relationship that is justified and theoretically sound being based on a plethora of past literature and analysis. Figure 3 shows the relationship between civil society strength and repression. Finally, for figure 4, I created a time lag plot of repression and civil society strength to uncover if time is a factor in causing states to be repressive towards CSOs. Even in time t+5, state repression is still high when civil society is strong. This shows that repression lags behind CSO strength, suggesting repression is likely reactive to CSO strength, but further testing and theorizing will be required. Finally, these figures provide foundational insight into confirming my hypothesis that civil society strength is positively associated with electoral democracy. After I run my regressions, I will be able to construct further visualization of the interaction between repression and civil society strength and its effects on electoral democracy.
Foundation of Empirical Extensions
One empirical extension I will conduct to bring me closer to a causal inference is to restrict my analysis to countries with a mid-to-high level of GDP per capita. This restriction will allow me to compare countries that are relatively similar in development, which is known to influence not only civil society engagement and strength, but also democratic ideals and transitions. If my results are consistent with this restriction, I will be able to rule out development as a confounding variable.
I am treating my interaction between repression and civil society strength as another empirical extension. Through this interaction, I gain the ability to show that the effect of civil society strength on electoral democracy is not universal - it may further depend on regime repressiveness towards CSOs. The observable implications of this interaction are as follows: In high repression contexts, I might expect a weak relationship between civil society strength and electoral democracy. Accordingly, CSOs within said contexts may be forced to avoid politics, or are co-opted by the regime in power to only support the regime itself. CSOs may also exist but mainly operate in non-political environments like health or education, while any CSO engaging in political mobilization is restricted or dissolved. It is important to acknowledge that I also might see in my regression model that the interaction between repression and civil society is not significant, meaning repression does not moderate the relationship between civil society strength and electoral democracy.
I am treating the CSO fragmentation variable in my civil society strength index as an empirical extension in itself. While CSO entry and exit and participatory environment measure the accessibility and engagement opportunities in civil society, fragmentation captures the structural conditions of civil society. A civic space where neither small nor large CSOs dominate (v2csstruc_2) suggests diversity in representation and more opportunities for democratic responsiveness. Without accounting for this dynamic, the index might overesimate civil society strength where CSOs are monopolized or co-opted by government. Therefore, including this measure in my civil society strength index reflects quality of competition and CSOs’ opportunity to shape political outcomes.
Using a lagged repression variable as an empirical extension in my regression models is important as it acknowledges civil society’s and repression’s impact on electoral democracy is unlikey to be instantaneous. By introducing a lag structure in repression, I can better capture cause and effect sequence. This approach also helps address concerns over reverse causality or reverse inference where for example, high civil society strength could be causing repression.
Finally, regressing with a linear time trend variable accounts for unobserved global or temporal dynamics that could confound the relationship between civil society strength, repression, and electoral democracy. Over time, factors such as war or technological change could influence every country in the dataset. By regressing with a linear time trend variable, I control for these unaccounted for shocks or gradual shifts that could introduce bias otherwise. This also reinforces the internal validity of my analysis in isolating the variation in electoral democracy that is due to country-specific factors rather than time dependent shifts.
Research Design
Main Regression Model
\[ \text{Electoral Democracy}_{it} = \beta_0 + \beta_1 \cdot \text{Civil Society Strength}_{it} + \beta_2 \cdot \text{Repression}_{it} + \beta_3 \cdot (\text{Civil Society Strength}_{it} \times \text{Repression}_{it}) + \epsilon_{it} \]
Regression Model With Covariates
\[ \text{Electoral Democracy}_{it} = \beta_0 + \beta_1 \cdot \text{Civil Society Strength}_{it} + \beta_2 \cdot \text{Repression}_{it} + \beta_3 \cdot (\text{Civil Society Strength}_{it} \times \text{Repression}_{it}) + \beta_4 \cdot \text{Education}_{it} + \beta_5 \cdot \text{GDP per capita}_{it} + \beta_6 \cdot \text{Social Media Censorship}_{it} + \epsilon_{it} \]
The covariates I decided to include are education level (e_peaveduc), GDP per capita (e_gdppc) [hoping to rule out as confounder] and government social media censorship (v2smgovsmcenprc). My decision to include these covariates follows the simple intuition based on past literature that more educated individuals are more likely to participate in civic life and absorb political information. Also, social media censorship reduces the amount of political information available to citizens, making it less likely for citizens to be able to engage in civic life.
While I analyze the relationship between civil society strength and electoral democracy, I am not fully committed to making a causal claim, but will implement many analysis tools to approach causality. This approach remains subject to standard limitations of observation analysis and research. Unobserved factors such as political culture, international pressure, or shifts in regime strategy may influence both civil society and democratic institutions.
Findings (Main Regression Model)
| (1) | |
|---|---|
| + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | |
| Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001 | |
| Intercept | 0.020*** |
| (0.003) | |
| Civil Society Strength | 0.426*** |
| (0.006) | |
| Repression | 0.028*** |
| (0.001) | |
| Civil Society Strength × Repression | -0.272*** |
| (0.002) | |
| Num.Obs. | 26172 |
| R2 | 0.790 |
| R2 Adj. | 0.790 |
When civil society strength and repression are both zero, the expected level of electoral democracy is .020 on the V-Dem polyarchy scale. In contexts where repression is 0, a full 1 unit increase in civil society strength is associated with a .426 increase in electoral democracy. When civil society strength is 0, a 1 unit increase in repression is associated with a .028 increase in electoral democracy. Finally, the negative coefficient on the interaction terms indicates that the effect of civil society strength on electoral democracy is .272 points weaker for a 1 unit increase in repression. As repression increases, the positive effect of civil society strength on electoral democracy declines. This effect is strong, however, as I progress with my analysis the effect may diminish.
Findings (Regressing with Country Level Fixed Effects)
| (1) | |
|---|---|
| + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | |
| Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001. Country fixed effects included but not displayed. | |
| Intercept | -0.041*** |
| (0.007) | |
| Civil Society Strength | 0.464*** |
| (0.006) | |
| Repression | 0.027*** |
| (0.002) | |
| Civil Society Strength × Repression | -0.255*** |
| (0.003) | |
| Num.Obs. | 26172 |
| R2 | 0.850 |
| R2 Adj. | 0.849 |
After controlling for country level fixed effects, when civil society strength and repression are both 0, the expected electoral democracy level is -.041. Within countries over time, when repression is held constant, a 1 unit increase in civil society strength is associated with a .464 increase in the electoral democracy index. When civil society strength is held constant, a 1 unit increase in CSO repression is associated with a .027 point increase in electoral democracy. Similar to the main regression model, the marginal effect of CSO strength on electoral democracy declines as repression increases. For each one unit increase in repression, the effect of civil society strength on electoral democracy decreases by .255, a strong effect.
Findings (Regressing with Covariates and Country Level Fixed Effects)
| (1) | |
|---|---|
| + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | |
| Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001. Country fixed effects included but not displayed. | |
| Intercept | -0.025 |
| (0.018) | |
| Civil Society Strength | 0.150*** |
| (0.018) | |
| Repression | -0.036*** |
| (0.005) | |
| Civil Society Strength × Repression | -0.042*** |
| (0.008) | |
| Education Level | 0.050*** |
| (0.004) | |
| GDP per Capita | -0.000 |
| (0.000) | |
| Gov't Social Media Censorship | 0.040*** |
| (0.003) | |
| Num.Obs. | 2620 |
| R2 | 0.972 |
| R2 Adj. | 0.971 |
When all predictors are 0, the expected electoral democracy index is 0.025. Holding education level, GDP per capita, and social media censorship constant, a 1 unit increase in civil society strength is associated with a .150 increase in electoral democracy within countries over time. Holding the same factors constant, within countries over time, a 1 unit increase in repression is associated with a .036 decrease in electoral democracy. The interaction term suggests that within countries over time, holding the same factors constant, for each 1 unit increase in repression, the positive effect of civil society strength is expected to decrease by .042. This interaction effect is weaker than earlier models, but is still statistically significant and consistent with my hypothesis. This effect being smaller likely indicates that the covariates or country-level effects can explain the variance in my interaction term.
Initial Discussion
The results from my fully specified regression model including country-level fixed effects and covariates (education level, GDP per-capita, and government social media censorship) provide important insight into the relationship between CSO repression, civil society, and electoral democracy. First, civil society strength is strongly and positively correlated with electoral democracy. The coefficient on civil society strength (β = .150; p < .001) suggests that within countries over time, an increase in civil society strength, while holding covariates constant, is associated with a substantial increase in the electoral democracy index. Looking at civil society repression, the association between repression and electoral democracy is negative, but modest (β = −.036; p < .001), which aligns with expectations that more repressive environments tend to suppress democratic outcomes.
Critically, the interaction term between civil society strength and repression is negative and significant (β = −.042; p < .001). This finding is particularly important for my theoretical framework; it suggests that the effect of civil society strength is indeed conditional on the level of repression. As repression increases, the positive effect of civil society strength declines. This finding supports the notion that while civil society plays an important role in promoting electoral democracy, its ability to do so is constrained under more repressive conditions. In highly repressive environments, CSOs may face legal, financial, or organizational constraints that hinder their ability to build democratic accountability or support political participation.
I hypothesized that civil society strength is positively associated with electoral democracy, but that the effect weakens as repression increases. I found a positive and statistically significant relationship between civil society strength and electoral democracy, supporting my main hypothesis. Additionally, my subsequent hypothesis on the conditioning of repression on civil society’s effect is supported both in terms of the conditioning mechanism and direction. The interaction term between civil society strength and repression (β = −.042; p < .001) suggests that repression does condition civil society’s effect on electoral democracy — and the effect becomes weaker, not stronger, under repressive conditions. I now move to implement these extensions to further approach causal inference.
Empirical Extension (Lagged Repression)
| (1) | |
|---|---|
| + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | |
| Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001. Country fixed effects included but not displayed. | |
| Intercept | -0.006 |
| (0.018) | |
| Civil Society Strength | 0.156*** |
| (0.018) | |
| Lagged Repression | -0.027*** |
| (0.005) | |
| Civil Society Strength × Lagged Repression | -0.056*** |
| (0.008) | |
| Education Level | 0.045*** |
| (0.004) | |
| GDP per Capita | 0.000 |
| (0.000) | |
| Gov't Social Media Censorship | 0.047*** |
| (0.003) | |
| Num.Obs. | 2620 |
| R2 | 0.973 |
| R2 Adj. | 0.972 |
This lag structure captures the intuition that the effects of civil society and repression on electoral democracy are unlikely to be immediate. By interacting current civil society strength with lagged repression, I can assess whether past government repression conditions the positive effect of electoral democracy. The negative and significant interaction between civil society strength and lagged repression (β = .056, p < 0.001) indicates that civil society’s positive impact on electoral democracy, when repression occurred in the previous year, is weaker than assessing repression in the same year. This is important for my causal inference ensuring that the potential cause in repression precedes the observed changes in electoral democracy.
Empirical Extension (Restricting to Above Median of GDP per Capita)
| (1) | |
|---|---|
| + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | |
| Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001. Country fixed effects included but not displayed. | |
| Intercept | 0.046 |
| (0.029) | |
| Civil Society Strength | 0.154*** |
| (0.018) | |
| Lagged Repression | -0.036*** |
| (0.005) | |
| Civil Society Strength × Lagged Repression | -0.042*** |
| (0.008) | |
| Education Level | 0.038*** |
| (0.004) | |
| Gov't Social Media Censorship | 0.047*** |
| (0.003) | |
| Num.Obs. | 2107 |
| R2 | 0.980 |
| R2 Adj. | 0.978 |
Within countries over time that are above the median of GDP per capita, the negative and significant interaction term of .042 (p < .001) reinforces the condtioning of the effects of repression on the effects of civil society strength on electoral democracy. Specifically, for each 1 unit increase in repression one year ago, the effect of civil society strength on electoral democracy decreases by .042 within countries over time while holding covariates (education level, government social media censorship) constant. The observation of this effect within richer countries allows me to rule out economic development as a confounding variable.
Further Approaching Causal Inference (Regressing with a Linear Time Trend)
| (1) | |
|---|---|
| + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 | |
| Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001. Country fixed effects included but not displayed. | |
| Intercept | 0.080* |
| (0.035) | |
| Civil Society Strength | 0.151*** |
| (0.018) | |
| Lagged Repression | -0.036*** |
| (0.005) | |
| Civil Society Strength × Lagged Repression | -0.042*** |
| (0.008) | |
| Education Level | 0.030*** |
| (0.006) | |
| Gov't Social Media Censorship | 0.048*** |
| (0.003) | |
| Num.Obs. | 2107 |
| R2 | 0.980 |
| R2 Adj. | 0.978 |
Holding repression, time, and covariates constant, within the same above gdp per-capita median country, and using a lagged repression variable, a one-unit increase in civil society strength is associated with a .151 increase in electoral democracy. It is important to highlight that the interaction term between CSO remains negative and significant. This indicates that the positive relationship between civil society strength and electoral democracy is weakened when accounting for repression that occurred last year. For each 1 unit increase in lagged repression, the marginal effect of civil society strength on electoral democracy decreases by .042. This model accounts for all time invariant differences across countries meaning the interaction term accounts for effects within countries over time. The inclusion of the linear time trend in this model now accounts for secular changes over time like global shifts in civic norms to ensure that the interaction term is not capturing long-term movement in repression and civil society.
For the final piece of my analysis, I constructed a visualization of predicted values of electoral democracy as a function of civil society strength and lagged repression (Figure 5). The darker colors represent lower levels of repression, while the yellow and orange represents high levels of repression (using lagged repression variable). This plot indicates that civil society strength has a strong positive effect on electoral democracy; however, the color gradient suggests that this effect is moderated as repression increases.
Discussion and Policy Implications
Across all models, civil society strength shows a strong, positive, and statistically significant relationship with electoral democracy. This finding is robust across all empirical extensions confirming my expectation that civil society is essential for promoting democracy. The interaction term between civil society strength and repression consistently shows negative and significant coefficients, although somewhat weak after including covariates and other empirical extensions. This supports my hypothesis that the positive impact of civil society strength on electoral democracy is moderated and weakened by CSO repression. While my analysis does not confirm a causal relationship, it offers strong and conclusive evidence that approaches causality. By incorporating country-level fixed effects, I control for unobserved country level factors that could bias estimates. I also rule out GDP-per capita as a confounder by restricting analysis to countries above the median of gdp per-capita. I also include covariates such as education and social media censorship, factors that are known to affect democratic outcomes and civil society strength. By introducing a lag structure, I capture the effect of past repression on current electoral democracy, establishing a sequence of cause and effect, and address concerns on reverse causality. Finally, by regressing with a linear time trend, I can account for global changes over time such as war or democratization waves that could influence countries. Together, these choices strengthen the argument that the relationships I observed are meaningful political dynamics.
My results show that civil society strength is consistently associated with higher levels of electoral democracy. This effect, however, is moderated under regimes repressing CSOs. Governments and international actors should therefore prioritize legal protection for CSOs including rights to funding access and the opportunity to engage in civic life without fear of persecution. Since civil society strength is a strong indicator of electoral democracy, institutions like Freedom House or the World Bank should integrate CSO based metrics to assess democracy. My findings also have clear implications for EU foreign policy given Poland and Hungary’s attacks on civil society in recent years. The EU’s ability to enforce democratic ideals through the Conditionality Regulation or Article 7 procedures should also depend on member states enacting legislation to repress civil society organizations.
Civil society is the foundation of democracy, but its ability to enhance and uphold democratic norms is dependent on the political space it operates in. When repression rises, this foundation is systematically eroded making the defense of the civic space essential to the survival of democracy.
References
Alagappa, Muthiah, ed. 2004. Civil Society and Political Change in Asia: Expanding and Contracting Democratic Space. Stanford, CA: Stanford University Press.
Bernhard, Michael, Tiago Fernandes, and Rui Branco. 2017. “Introduction: Civil Society and Democracy in an Era of Inequality.” Comparative Politics 49 (3): 297–309. https://www.jstor.org/stable/26330959.
Croke, Kevin, Guy Grossman, Horacio A. Larreguy, and John Marshall. 2016. “Deliberate Disengagement: How Education Can Decrease Political Participation in Electoral Authoritarian Regimes.” American Political Science Review 110 (3): 579–600.
Putnam, Robert D., Robert Leonardi, and Raffaella Y. Nanetti. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton, NJ: Princeton University Press.