Good seed makes a good crop? The relationship between civil society and post-independence democracy levels

ABSTRACT The impact of civil society on democracy is contested. Some argue that Civil Society Organizations (CSOs) are important ‘schools of democracy’ and may foster democratic consolidation. Others emphasize that anti-democratic CSOs may undermine democracy. This debate is particularly relevant in the context of newly independent states. At that critical juncture, both democratic and authoritarian regime trajectories are possible. Societal preconditions – such as the state of civil society – can therefore be particularly relevant for the way forward. To what extent does the nature of CSOs prior to independence has an impact on the subsequent level of democracy? We argue that the existence of democratic CSOs prior to independence strengthens post-independence democracy whereas non-democratic CSOs have a detrimental effect. For the first time, this argument is empirically tested, using Varieties of Democracy (V-Dem) data on 91 cases of independence since 1905. The empirical results demonstrate that the presence of democratic CSOs prior to independence is positively correlated, whereas non-democratic CSOs are negatively correlated to democracy levels following independence.


Introduction
The impact of civil society on democracy has been discussed by scholars for many years (e.g., Putnam, 1995;Tocqueville, 1969). The bulk of empirical studies on the role of civil society and democratization have been case studies focusing on the states experiencing critical junctures such as, for example, revolutions (e.g., Stepan & Linz, 2013) or regime transitions (e.g., Bermeo, 1997). In particular, processes of independence are critical for changes in regime trajectory and provide a good opportunity for systematic, empirical studies. States sometimes venture down different paths following independence and form either a democratic or authoritarian new state even though they are similar in many aspects such as being located in the same region, similar time of independence, separating from the same sovereign state and even similar culture. This applies for instance to the Belarus in 1991 versus Estonia in 1992and Burma in 1948versus India in 1947 (Alamgir, 1997;Bennich-Björkman, 2007;Lenzi, 2002;Shani, 2016).
Nevertheless, the role of civil society in explaining the variation of regime outcomes following independence has so far only been studied through qualitative case studies (Bennich-Björkman, 2007;Shani, 2016). Therefore, this study aims to make a first contribution to the field by empirically exploring the relationship between pre-independence presence and nature of civil society organizations (CSOs), and post-independence democracy levels. 1 In particular, we focus on political CSOs, which can work towards making a country either more or less democratic. Democratic CSOs are willing to play by democratic rules and do not use violence as means to reach their goals. Conversely, non-democratic CSOs, also called 'uncivil society' (Kopecký & Mudde, 2003) or 'bad civil society' (Chambers & Kopstein, 2001), use violence as means to reach their goals, do not care for democratic rules and often hold undemocratic ideas (e.g., Chambers & Kopstein, 2001;Kopecký & Mudde, 2003).
As CSOs sometimes are not officially registered, their existence and the level of participation in them, are difficult to assess and quantify based on membership data, especially in years leading up to independence. The expert-coded data from Varieties of Democracy (Coppedge et al., 2020b) allow us tofor the first timequantify the participation in CSOs beyond formal memberships and during pre-independence periods.
The results of this study indicate a substantial positive relationship between the presence of democratic CSOs and a negative relationship between non-democratic CSOs before independence and democracy levels after independence. This finding supports arguments by several civil society scholars who stress the importance of differentiating between democratic and non-democratic CSOs (e.g., Chambers & Kopstein, 2001;Kopecký & Mudde, 2003).
The article proceeds as follows. First, we develop theoretical arguments based on the literature on civil society and democratization and present our hypotheses. Then we describe the data and methods used and provide a descriptive analysis of the data. We then move on to our results and provide a discussion on robustness and the results. Lastly, our conclusions along with an outline of future research areas on the subject.

Theory: Civil Society, Independence and Democratization
A strong civil society has been claimed by many researches to increase the chances of a successful democratization in general, by creating 'constraints on the reproduction of privileged positions under authoritarianism' (Bernhard et al., 2017, p. 306), and working as a stabilizer and trigger of democracy via social trust (Diamond, 1999;Fukuyama, 1995;Putnam, 1995). Furthermore, they empower the people by monitoring officials and keeping citizens informed about their actions (Lindberg et al., 2017, p. 8), educate citizens in quality of policies and governance, and democratic rights, and train them in using democratic tools to influence their society (Gerring et al., 2005, pp. 330-331;Smith, 2000;Tocqueville, 1969, p. 522).
Whether CSO participation levels or the nature of civil society matters most for democracy outcomes has been widely discussed (Berman, 1997;Kopecký & Mudde, 2003;Tusalem, 2007). In the field of research arguing for the value of sheer participation, the classical work by Putnam (1995), Bowling Alone, connects a lowered trust for the government and other citizens to the decrease in CSO participation levels during the same time. Putnam (1995, p. 70ff) noticed how organized activity had decreased in the United States, and since members of civic associations are more likely to participate in politics, he concludes that there might be a link of social trust between low CSO participation and political disarray. Since Putnam, other scholars have argued mainly for the strength of civil society rather than its nature (e.g., Tusalem, 2007), and it is now widely assumed that a strong civil society 'contributes to the making and sustaining of democracy' (Grugel & Bishop, 2014, p. 136). Bernhard and Jung (2017) also find that post-Soviet states where political civil society was more engaged during the breakdown adopted more competitive political systems. Conversely, in countries where civil society was less engaged during the breakdown, the ruling elite created electoral authoritarian systems.
Others have argued that statements such as 'an active civil society is good for democracy' are 'invalid' and that it depends on which CSOs constitute the civil society (Kopecký & Mudde, 2003, p. 11). Way (2014) also lifts civil society nature rather than strength and claims that whilst CSOs direct and facilitate protests, civil society does not need to be strong to produce well-organized protests. Way (2014, pp. 38-39) specifically mentions the large-scale protests in Ukraine in 2004, Egypt in 2011and Cameroon in 1991 as examples of when weaker civil societies managed to organize very large protests. However, none of these protests resulted in a lasting democracy. Ukraine increased its level of democracy for five years before reverting to its original levels, Egypt one year, and the protests in Cameroon did not result in any democratization at all (see Figure 1). Berman (1997, p. 429) argues that activities such as organized bowling appear to have no effect on democratization if political institutions and political civil society are not revitalized, and we find examples supporting her argument in the Arab Spring. No matter how impressive the protests, Egypt did not manage to sustainably democratize, but Tunisia did (see Figure 1). In Tunisia, both secular liberals and Islamists were opposing the authoritarian government, and despite disagreeing in many areas, they held organized meetings where they agreed upon democracy rules as early as 80 years before transition (Stepan & Linz, 2013, p. 23). When the transition started, they were prepared for democratic rule. Egypt, on the other hand, had a large creative civil society but their political and democratic civil society was greatly underdeveloped. This demonstrates how democratic CSOs, as 'great free schools of democracy' (Tocqueville, 1969, p. 522), may increase the chances of a successful democratization process.
However, scholars have also raised concern that democratic and non-democratic CSOs most likely have opposite effects on democratization. For instance, Berman (1997) makes an example of how the civil society in the Weimar Republic was used by the non-democratic National Socialist German Workers' Party (NSDAP) for political gain. The high levels of socioeconomic cleavages resulted both in increased levels of CSO participation and a civil society largely consisting of dissatisfied citizens, which was used by the non-democratic NSDAP to gather support (Berman, 1997, p. 421). Chambers and Kopstein (2001) also warn about how non-democratic CSOs can nurture hatred and fear in its members. On July 4th, 1999, a member of the World Church of the Creator (a white supremacist organization) in the United States, went on a shooting rampage, targeting ethnic minorities. Chambers and Kopstein (2001, p. 837) mean that his hatred and fear of these ethnicities were nurtured by his membership in the association. Now, smaller CSOs such as the World Church of the Creator may not seem like a potent threat against a democracy the size of the United States. But Chambers and Kopstein (2001, p. 843) argue that there can be a spill-over effect from the extremist CSOs to the mainstream media and political discourse, normalizing its ideas. Examples of this can be found by the current president of the United States, Donald Trump, who (amongst countless other controversies) refuses to disavow support from David Duke and the KKK, 2 or condemn the white supremacist group Proud Boys, 3 without any serious consequences.
The warnings of the dangers constituted by non-democratic CSOs do not discredit theories of a positive relationship between CSO participation and democratization, but rather puts a light on the importance of civil society's nature. However, one should bear in mind that the classification of what constitutes democratic and non-democratic civil society is not always simple. CSOs often perform multiple tasks, change nature or means and their actions are often interpreted differently over time (Kopecký & Mudde, 2003).
Taken together, the current literature indicates that a combination of factors shapes the impact of civil society during critical junctures. Processes before and during the transition, such as cultural and socioeconomic features, shape the opportunities of civil society to facilitate democratization (Fishman, 2017). The Weimar Republic is a good example for this, as the non-democratic NSDAP benefited from high levels of socioeconomic inequality to gather popular support (Berman, 1997). Furthermore, the previous regime influences both which elites are available after transition, and which actors are seen as legitimate post-transition Fishman, 2017).

Declaration of Independence as a Critical Juncture
Perhaps contrary to common intuition, countries gaining independence is not a thing of the past. There have been several waves of independence throughout the twentieth century, and new cases are still occurring, with the latest state to have gained independence being South Sudan (2011). Regions such as Northern Ireland (United Kingdom), Catalonia (Spain) and Kurdistan (Iraq, Syria, Turkey, Iran and Armenia) may also eventually gain independence and would China lose some of its provinces, new independent states such as Tibet, Xinjiang and Hong Kong may form. When a state gains independence, either from a foreign sovereign state or by state dissolution or unification, it by default creates a new state. This is a seminal event that shapes the future political, social and cultural future of the new country. To understand why states have such varying odds at democratizing independence, the research on the mechanisms and dynamics of those struggles is of fundamental importance (e.g., Bennich-Björkman, 2007;Bernhard et al., 2004;Oldenburg, 2010).
In many cases, independence struggles involve some sort of democratization, that is, to form a new state that provides the citizens with more political influence and power than the current polity does. However, conclusions from research on the role of CSOs in democratization processes in already independent states may not necessarily always be applicable to states gaining independence. An independence process is motivated most fundamentally by the desire to form a new state and is not necessarily driven by the desire for democracy. In some countries, such as Estonia, the independence struggles were closely related to democratization (Bennich-Björkman, 2007;Surzhko-Harned, 2010). However, independence processes may also be led by anti-democratic forces or forces neutral to democratization. Many colonies gained independence, not by a desire to democratize (although that may have existed in the countries), but by the decision by the colonial powers to liquidate their colonies, which in turn was mainly motivated by factors such as international relationships rather than pro-democratic ideals (Hargreaves, 2014). Some semi-independent countries, such as Bhutan, had an independence process that was mainly driven by a quest for self-determination, rather than pro-democratic forces, and hence they did not experience much change in their level of democracy following independence (Nga et al., 2019). In other countries, motivations behind independence struggles are based on territory, cultural belonging, religion or ethnicity. For instance, the independence process in Eritrea was primarily driven by territorialism and the right to uphold the old colonial borders as an independent polity, rather than a democratization struggle (Clapham, 1996). Therefore, it is important not to equate independence processes with democratization struggles in already independent states.
Whatever the motivation for independence, it is a critical juncture that brings fundamental changes to the political landscape and new institutions and rules for governance are formed. Critical junctures can be defined as 'relatively short periods of time during which there is a substantially heightened probability that agents' choices will affect the outcome of interest' (Capoccia & Kelemen, 2007, p. 348). A declaration of independence fits this definition, as: (a) it is a relatively short period of time compared to the process it instigates (a new state); (b) there is a high probability that the choices of the involved agents will affect the outcome, in this case, the level of democracy after independence. Nevertheless, no large-N study has so far addressed the potential of newly independent states to democratize. Rather, the field has focused on democratization in relation to other critical junctures in already independent states (e.g., Acemoglu & Robinson, 2000;Teorell, 2010;Tilly, 2000), and on qualitative regional or single case studies of states gaining independence (e.g., Goldman, 1999;Shani, 2016;Surzhko-Harned, 2010).

Civil Society and Independence
Some scholars argue that the role and impact of civil society before the general regime transition is not easily predicted by observing the patterns of organization and mobilization in the authoritarian regime (e.g., Fishman, 2017). However, case studies of countries gaining independence suggest that the chances of forming a democratic new state at independence may be affected by their levels of civil society participation, and the nature of that civil society before independence (e.g., Bennich-Björkman, 2007;Stepan & Linz, 2013, p. 23;Surzhko-Harned, 2010). Democratic CSOs are believed by some to prepare society and actors for democracy by outlining the rules of the democracy they want to institute, as well as planning the process of democratization, facilitating a quick and thoughtful process (e.g., Bennich-Björkman, 2007, p. 340;Stepan & Linz, 2013, p. 23). Heading towards independence then is a new state led by a civil society with a plan for their democratization process, and a ready set of rules and outline of the new way of governance. This also generates minimal time for takeover by non-democratic actors and minimal space for setbacks and delays in the forms of disagreements and irresolution. 4 These governance-preparing and setback-preventing aspects of democratic CSOs insinuate a democracy-facilitating effect.
Some argue that civil society organizations cannot exist within authoritarian states before a process of democratization has begun and they are at least partially allowed. O' Donnell and Schmitter (1986, p. 48), for example, write that '[b]y trivializing citizenship and repressing political identities, authoritarian rule destroys selforganized and autonomously defined political spaces'. However, CSOs do not necessarily need to be legal or visible to exist and have an impact. As Howell (2012, pp. 1-2) puts it: 'In many countries, [CSOs] operate in highly repressive political contexts with restrictive legal and regulatory environments, and face constant harassment in their day-to day work'. We see examples of this also in countries gaining independence, such as Estonia (Bennich-Björkman, 2007) and India (Jayal, 2007), where a grassroot civil society had existed, and formed the base for democratization, well before independence.
In India, political CSO participation can be traced back as far as to the 1800s, and in 1885, the national umbrella organization, the Indian National Congress (INC) emerged and later came to represent a successful pro-democracy political opposition (Jayal, 2007, p. 144). In Estonia, the underground movement-, reading-and discussion groups served as platforms for preparation of democracy and produced parties, congress members and even the first Prime Minister of independent Estonia in 1992 (Bennich-Björkman, 2007, pp. 325-334;Surzhko-Harned, 2010, pp. 635-637). Both countries successfully and rapidly democratized, much thanks to the work by pre-independence democratic CSOs.
Based on this discussion, we formulate two hypotheses regarding the role of civil society before independence for democracy levels after independence: H1: There is a positive relationship between the presence of democratic CSOs before independence and post-independence democracy levels.
H2: There is a negative relationship between the presence of non-democratic CSOs before independence and post-independence democracy levels.
If our findings support H1, this would suggest that the presence of democratic CSOs before independence can be used to predict the levels of democracy after independence, and underline the importance of facilitating for those organizations leading up to independence.
Would H2 be supported, this would affirm the warnings of non-democratic civil society put forward by scholars, and demonstrate the importance of limiting the strength and role of these CSOs during times of state separation and independence.

Conceptualizations and Operationalizations
Many of the conceptualizations and data in this study originate from Varieties of Democracy (V-Dem). V-Dem is run by an international network of political scientists and researchers and is based at the Department of Political Science, University of Gothenburg, Sweden. The data cover 202 countries and 758 variables regarding democracy. It spans from 1789 to the present and is collected by contributions from over 3000 local country experts around the world (Coppedge et al., 2020a, pp. 29-40;Coppedge et al., 2020b).

Units of Analysis: Cases of Independence
By the definition of independence used in this study, there must be at least ten years of non-independence before the independence date for a country to be regarded as having gained independence. 5 This means that, for example, European countries that were occupied by the Nazis are not seen as having lost and gained independence, as the occupations did not last over a decade. For conceptual clarity, this study also applies the decade rule to the time after independence. This means that a country must be independent for a minimum of ten years after its independence date to be considered successfully having gained independence. Independence is therefore operationalized by using the V-Dem dataset variable 'Independent states', which categorizes country-years as either 0 = not independent, or 1 = independent (Coppedge et al., 2020a, pp. 179-180). The first country-year that is coded as '1', preceded by at least 10 '0''s, and followed by at least 10 '1''s, is then considered to be the year of independence for a given country.

Dependent Variable: Post-Independence Democracy Levels
The dependent variable of this study is the level of democracy ten years after independence. Democracy is in this study conceptualized by the more specific term liberal democracy. Liberal democracy is based on the more fundamental concept of electoral democracy, which is a governance characterized by regular electoral competition for support from an electorate based on extensive suffrage (Munck & Verkuilen, 2002;Schumpeter, 1950). The elections affect the composition of the executive and are not marked by fraud or systematic irregularities. Between elections, citizens enjoy the freedom of expression and freedom of association, and independent media presenting alternative views on political matters exist (Coppedge et al., 2020a, p. 42). Liberal democracy is defined as electoral democracy, with the addition of a strong focus on rule of law, checks and balances, and protection of individuals' rights and liberties from being infringed by an overly powerful government and the tyranny of the majority (Coppedge et al., 2020a, p. 43;Dahl, 1989;Zakaria, 1997). This study conceptualizes 'democracy' as liberal democracy rather than the narrower electoral democracy, as it is more complete due to its focus on constitutionalism, and is frequently used to measure democracy levels in recent research (e.g., Lührmann et al., 2018;Mainwaring & Bizzarro, 2019;Trein, 2020). Liberal democracy is operationalized in this study by the variable 'Liberal Democracy Index' (LDI) in the V-Dem dataset (Coppedge et al., 2020a, p. 43), which is an interval level running from 0 (=liberal democracy is not achieved at all) to 1 (=liberal democracy is fully achieved). The value of this variable ten years after independence in a given country is our dependent variable.

Independent Variable: Presence of Democratic and Non-Democratic Anti-System CSOs Before Independence
This study recognizes that there are different types of CSOs with different attitudes towards democracy, which may have different effects on democratization. Therefore, it distinguishes these groups based on their nature, and relationship towards democratic rules. The two main independent variables of this study are therefore the presence of active, anti-system, democratic and non-democratic CSOs. 6 The V-Dem variables on anti-system CSOs used in this study that constitute the CSO block in analysis are: Democratic anti-system CSOs: The V-Dem variable 'CSO anti-system movement character' (v2csanmvch_4) is used to capture activity in those CSOs that are 'perceived by most disinterested observers as willing to play by the rules of the democratic game, willing to respect constitutional provisions or electoral outcomes, and willing to relinquish power (under democratic auspices)'. This variable is converted from bimodal to interval scale by the V-Dem measurement model, running from 0 to 1 (Coppedge et al., 2020a, p. 185). 7 Non-democratic anti-system CSOs: The V-Dem variable 'CSO anti-system movement character' (v2csanmvch_5) is used to capture activity in those CSOs that are 'perceived by most disinterested observers as unwilling to play by the rules of the democratic game, not willing to respect constitutional provisions or electoral outcomes, and/or not willing to relinquish power (under democratic auspices)'. This variable is also converted from bimodal to interval scale by the V-Dem measurement model, running from 0 to 1 (Coppedge et al., 2020a, p. 185) (see note 7).

Control Variables
In regression, we control a number of variables related to influential theories that may complement the explanatory power of the presence of democratic CSOs. Below, we present a brief summary of each theory along with its operationalization in this article.
Civil society environment. To acknowledge theories that argue that civil society participation acts as 'schools of democracy' and increases support for democracy by allowing citizens to 'practice' democracy (e.g., Tocqueville, 1969, p. 522), and that an environment with a high level of civil society participation and well-organized parties is a good breeding-ground for democratization, we include the following V-Dem variables: . 'CSO participatory environment', which accounts for the number of active CSOs, and popular involvement in them (Coppedge et al., 2020a, pp. 183-184). . The V-Dem 'party organizations' variable, which accounts for how many national political parties have permanent organizations (Coppedge et al., 2020a, p. 92). This is to ensure that we do not overestimate the results of 'democratic CSOs', as democratic CSOs are likely more common in countries with higher levels of civil political organization.
Modernization theory. The most basic assumption of the modernization theory is that economic and social development, or socioeconomic development, has a positive relationship to democracy. Lipset (1959), for example, concluded that development helps create a democratic culture, and found that countries with high levels of socioeconomic development more frequently are democratic than those with a lower score. Przeworski et al. (2000) found that development acts as a stabilizer of democracy levels, rather than as a trigger for democratization. In an effort to capture the essence of the modernization theory, we control for the following variables: . GDP per capita is a central part of the concept of socioeconomic development (e.g., Inglehart & Welzel, 2010, p. 551). The V-Dem variable used is 'GDP per capita, logged, base 10' (Coppedge et al., 2020a, p. 339). . Equal distribution of resources index. An equal distribution of resources creates a larger middle class and reduces the number of poor people, one of the most theoretically important aspects of socioeconomic development (Teorell, 2010 pp. 24-28, and the references therein). The V-Dem variable used is 'Equal distribution of resources index' (Coppedge et al., 2020a, pp. 54-55). . Equal political power. Having an equal distribution of resources is toothless if only the richest individuals have political influence. The V-Dem variable used to capture the distribution of power over socioeconomic groups is 'Power distributed by socioeconomic position' (Coppedge et al., 2020a, p. 192).
Democratic experience. Previous experiences of the democratic rule may also affect the prospects for democratization. All types of political regimes are attempting to socialize their citizens into supporting their political system (Pop-Eleches & Tucker, 2017, p. 3). Citizens who have experienced life under democracy have also learned what constitutes 'good policy' and governance (Gerring et al., 2005, pp. 330-331). This means that if a population has positive previous experiences of democracy before its transition, the country is more likely to successfully democratize, due to the larger support and knowledge of democracy amongst its citizens.
. To account for the influence from previous democratic experiences, we include a variable for the level of democracy before independence. This reflects the level 13 years prior to the post-independence democracy level, which is our dependent variable.
Neighbourhood theory. According to the neighbourhood theory, societies are more likely to support democratization if they are located in the vicinity of already democratic states or states that support democracy (e.g., Berg-Schlosser, 2008;Teorell, 2010, pp. 86-89). Vice versa, they are less likely to democratize if they are located in the vicinity of states that are non-democratic or oppose democracy.
. To account for the levels of democracy in the different politico-geographical regions, we generate a variable for the average level of democracy in each country's region at the time of the country's independence using the V-Dem variable 'Region (politicogeographic 6-category)', which divides the world's countries into six regions (Coppedge et al., 2020a, pp. 337-338) (see note 7), the year of independence and the LDI (Coppedge et al., 2020a, p. 43).
Democratization trends. We can identify three 'waves' of democratization in history, in which many countries democratized during a short period of time (Huntington, 1991). Thus, countries that gain independence during one of these waves might have higher chance of democratization at independence due to democratizing international influences.
. To address such temporal trends, we include 11 decade dummies in our analyses, starting at 1900-1909, up to 2000-2009. Lastly, as all variables produce VIF values below 10, multicollinearity is not a concern for our models.

Descriptive Analysis
The distribution of cases over the six regions and their respective democracy levels 10 years after independence are illustrated in Figure 2. On a global median, countries have reached a LDI score of 0.14 (on a 0-1 scale) ten years after independence. Western Europe and North America stand out from the rest, clearly being more democratic post-independence than other regions, with a median score of 0.645. Otherwise, few new states manage to create a substantial democracy, failing to reach a median over 0.4 on the LDI ten years after independence, and Sub-Saharan Africa is the only region with a median below the global score (0.109 on a 0-1 scale). As many Sub-Saharan states were colonies of west European states before their independence, this is sometimes explained by colonialism, whereby some scholars claim that the colonization by European states has a negative impact on future democratization attempts (e.g., Bernhard et al., 2004). However, our data suggest that all statuses prior to independence (colony, occupied, part of other state or semi-independent) have similar relationships to democracy levels after independence, 8 concurring instead with empirical conclusions by, for example, Teorell (2010, p. 142), that 'Colonial origin … were not robustly related to democratization during the third wave [of democratization]'. 9 The scatterplots in Figure 3 illustrate the relationship between pre-independence democratic and non-democratic CSOs and post-independence democracy levels. It shows that although the cases with a high occurrence of democratic CSOs before  independence (≥0.5) are relatively few (n = 15), they generally reach higher levels of democracy (≥0.5) after independence, whilst the cases with low levels of democratic CSOs (<0.5) mostly have lower levels of democracy (<0.5) after independence, although the cases with low levels of democratic CSOs have a wider spread. We also see that whilst there are relatively few cases with non-democratic CSOs (>0, n = 17) all fall short of reaching even mid-high (≥0.4) levels of post-independence democracy levels. 10

Method
The aim of this study is to evaluate the general role of civil society nature at the time of independence, focusing on finding a general relationship, rather than exact explanations for certain cases. Therefore, a large-N analysis is conducted, using OLS regressions, including variables related to complementary theories as controls. However, including all available cases of newly independent countries in the data, the total N only reaches 91, and in models including GDP levels, the N is only 60, due to missing GDP data. 11 Thus, we have a relatively small number of observations, which in social sciences are quite common (Coppedge, 2012, p. 262). Accordingly, null findings should be interpreted with caution as statistical significance might have been achieved with a larger number of cases.
The dependent variable, level of liberal democracy after independence, is included at ten years after independence (t + 10), and all independent variables are included at three years before independence (t − 3), since one year before independence (t − 1) could be considered during an ongoing independence process, as independence is not something gained overnight. Also, if using values from a long time before independence, the link to post-independence democracy levels is weaker, and there is a higher risk that existing relationships go unnoticed. 12

Results
Although 115 countries gained independence within the studied time frame, only 91 cases are included in the analysis. This is because some countries lack data on variables before independence. This concern, for example, countries that previously were part of Yugoslavia or the Soviet Union. The results from regression analysis with independent variables at three years before independence are presented in Table 1.

Democratic and Non-Democratic CSOs
The first model in the main regression table (Table 1) demonstrates that the presence of democratic CSOs prior to independence is positively related to levels of democracy after independence (ß = .375) at a statistically significant level. Correspondingly, the presence of non-democratic CSOs prior to independence is negatively related to post-independence democracy levels (ß = −.366) at a statistically significant level.
Having established the expected directions of the main relationships, we move on to include our control variables. Model 2 includes all control variables, save GDP/Capita, and the results hold with statistical significance. In model 3, we include the GDP variable, which reduces the total number of observations from 91 to 60, as the GDP variable contains 31 missing values. Including the GDP variable in model 3 also reduces the coefficient (ß = .0854), and increases the calculated probability value to p > .1. Yet, model 3 shows a sustained negative effect of non-democratic CSOs prior to independence (ß = −.215, p < .1), despite the inclusion of the GDP variable and reduced total number of observations. Model 2 and model 3 are identical apart from the inclusion or exclusion of the GDP variable. Model 2, which does not contain the GDP variable. The exclusion of the GDP variable yields an increase in the total number of observations, as well as the coefficient and significance level for democratic CSOs (ß = .142, p < .05). The coefficient and significance level for non-democratic CSOs is also higher (ß = −.258, p < .05). This indicates that models with a greater number of observations tend to achieve higher levels of statistical significance (all else equal).
To ensure that the difference in results between models 2 and 3 are not due to the inclusion and exclusion of the GDP/Capita variable, we ran the same regression as in model 2, but with the smaller sample of model 3 (model 4). The results are more similar to those of model 3 than model 2, indicating that the differences in results between models 2 and 3 are mainly derived from the difference in sample size, and not from the inclusion or exclusion of the GDP/Capita variable.
The model fit is also relatively high for all models including controls.

Robustness and Discussion
First, we investigated whether the lower number of observations in model 3 due to missing GDP data is significantly correlated to any of the control variables or with any type of statehood previous to independence. The results are presented in Table A2 in the appendix and show that previously semi-independent countries, and those that gained independence between the 1920s and the 1940s, are positively correlated to being excluded from model 3 due to missing GDP data, and thus make up a smaller proportion of the total cases in this model. Conversely, previous colonies and countries that gained independence in the 1950s and 1960s, are proportionally more in model 3 compared to other models. These contextual factors have however not affected the effects of the main independent variables or control variables, which are not significantly correlated to missing GDP data. Figure 2 demonstrates that countries in the region 'Western Europe and North America' democratize to a greater extent than any other region, following independence. If there is something besides the presence of democratic CSOs that makes this region stand out in this fashion, they could be biasing the results somewhat. It is also possible that the large proportion of Sub-Saharan African countries (49% of included observations) skews the results. With this in mind, we ran a series of regressions specifically investigating the effects of region using dummies (Table A3, appendix). They show that being located in the 'Asia/Pacific' region (model 6), or the 'Western Europe/ North America' region (model 10) is positively related to democracy levels following independence. Oppositely, being located in the 'Sub-Saharan Africa' region is negatively related to post-independence democracy levels (model 11). Nevertheless, the results indicate enduring and significant results for both democratic and non-democratic CSOs throughout all models testing for regional effects.
As the protection of civil rights logically facilitates a strong and pluralistic civil society, the additional focus on rule-of-law and civil liberties in the LDI may raise concerns that it is creating an 'overlap by construction' with civil society-related variables. To inspect whether our results hold when using the level of the narrower Electoral Democracy Index (EDI) after independence rather than the LDI, we performed the same regressions as in Table 1 but with EDI as the dependent variable, and EDI to calculate the regional average democracy level. 13 To account for democratic experience whilst ensuring that there is no overlap with the civil society variables, we used the Clean Elections Index instead of the LDI (see Table A4 in the appendix). Even though the EDI does not include the same focus on civil liberties and rule-of-law, the results are very similar to those of the original regressions in Table 1. Although the EDI regressions lose all significance in model 3 (due to the low number of observations), the models produce slightly higher coefficients in models 1 and 3 than the corresponding models in the main regression table (Table 1). This shows that the results hold for analysis using a narrower model of democracy, as well as the more conventional model that this study uses.
Tables A5 and A6 in the appendix display the regression results when using independent variables at five and ten years before independence. From these tables, we can see that both the effect of democratic and non-democratic CSOs hold when moving further away from the independence date. The presence of democratic CSOs at five years before independence keeps a slightly lower yet positive and significant effect, as presented in model 18, Table A5 (ß = .120, p < .1). The presence of non-democratic CSOs keeps a negative and significant effect (ß = −.213, p < .1). When moving to ten years before independence, the coefficients increase somewhat again. Yet, when using the independent variables at ten years before independence and keeping a higher N (N = 87), the presence of democratic CSOs still has a positive and significant effect, as shown in Table  A6, model 22: (ß = .128, p < .1), and the presence of non-democratic CSOs keeps a negative and significant effect (ß = −.204, p < .1). This shows that although the effect is stronger closer to the date of independence, the assigned nature of the CSOs in this study is not a product of some initial stages of the independence process. They reflect the CSO environment in the country, which even well before independence affects the post-independence democracy levels.
Finally, we also ran the main regressions from Table 1 but with a 20-year lag on the dependent variable (LDI) instead of the original ten-year lag. The results show that the direction of the correlations is the same as in the regressions using a ten-year lag in Table  1, and the values hold significance (see Table A7 in the appendix). This means that the presence of democratic and non-democratic CSOs prior to independence appears to produce effects that last some years past the independence stage, although they might weaken slightly over time. Thus, these additional models show that findings are also somewhat robust to a different lag structure.
In sum, our analyses provide robust support for both of our hypotheses. There seems to be a positive relationship between the presence of democratic CSOs before independence and post-independence democracy levels. At the same time, there is a negative relationship between the presence of non-democratic CSOs before independence and post-independence democracy levels.
The analysis results indicate that the nature of anti-system CSOs prior to independence contributes to an explanation of the variation in democracy outcomes in states gaining independence. In line with what Chambers and Kopstein (2001), Kopecký and Mudde (2003) and Berman (1997) claim, whether there are democratic or non-democratic CSOs active when a country is moving towards independence is crucial for its level of democracy post-independence.

Conclusion
We have performed the first-ever cross-regional large-N empirical study of countries gaining independence, whilst separating the level of participation from the democratic nature of CSOs, and our analyses indicate that the nature of active CSOs is pivotal for states gaining independence. Results show that the presence of democratic CSOs prior to independence is positively related to democracy levels after independence, and nondemocratic CSOs are negatively related to post-independence democracy levels, whilst accounting for other influential factors such as politico-geographical neighbourhood, previous levels of democracy-related variables and modernization factors.
The outcomes of this study concur with the arguments by for example Chambers and Kopstein (2001), Kopecký and Mudde (2003) and Berman (1997), who claim that civil society nature is vital for democracy outcomes, as we find that the role of civil society during independence struggle is highly relevant for the democratic outcome. Our conclusions are contrast with those by . They find that the strength of civil society during critical junctures is more crucial to the regime outcome than the strength of civil society prior to the critical juncture (Bernhard et al., 2017, p. 300). Contrary, we find that the presence of CSOs that are willing to play by democratic rules prior to the critical juncture of independence, is positively related to the level of democracy after independence. Vice versa, the presence of CSOs not willing to play by democratic rules prior to independence is negatively related to post-independence democracy levels. The discrepancy in our conclusions may stem from the difference in types of critical junctures under study. Whilst  focused on junctures such as economic crisis, regime transition, party system realignment and social revolution; we studied the obtainment of independence. This encourages continued research on the effects of civil society on the outcomes of various critical junctures.
We find that the role of civil society during independence struggles is highly relevant for the democratic outcome. Our results also diverge from studies of democratization of independent states that claim that high participation levels, in general, consolidate democracy (e.g., Putnam, 1995). When it comes to civil society and independence, participation alone is not of most weight when forming a new state, but rather whether CSOs active in the country are willing to play by democratic rules or not. We need to pay attention to the potential detrimental impact of active presence by non-democratic CSOs in countries moving towards independence. When non-democratic CSOs are present, a new state has low chances of building a liberal democracy. However, the presence of democratic CSOs before independence increases its chances of democratization.
Independence is a critical juncture and a seminal event in the history of a state and deserves more attention from empirical research. In the context of independence, the nature of civil society is clearly an important factor in determining whether the new state develops in a democratic or authoritarian direction. Future studies need to investigate how the international community can help facilitate the emergence and survival of these pro-democratic CSOs during processes and events leading up to a new potentially independent state. We also recognize arguments by Fishman (2017), those similar processes are influenced by complex relationships between cultural, socioeconomic and political institutions. Therefore, we also encourage future studies to shed light on the mechanisms by which different factors influence democracy outcomes after independence, recognizing these processes as pivotal critical junctures. Notes 1. CSOs are here specified as voluntary organizations, separated from the state as well as from private life, consisting of citizens with common goals and interests, formed to discuss and develop ideas, and working towards realizing those ideas and aspirations. They can take the form of 'interest groups, labor unions, religiously inspired organizations (if they are engaged in civic or political activities), social movements, professional associations and classic nongovernmental organizations (NGOs), but not businesses, political parties, government agencies or religious organizations that are primarily focused on spiritual practices' (Coppedge et al., 2020a, p. 181). 2. http://www.bbc.com/news/world-us-canada-40915569; https://www.independent.co.uk/ news/world/americas/kkk-trump-david-duke-tucker-carlson-election-2020-a9609491.html. 3. https://www.bbc.com/news/av/election-us-2020-54351472; https://abcnews.go.com/Politics/ extremist-group-relishes-trumps-mention-debate/story?id=73341826. 4. One example of this occurring can be found in Egypt following the Arab Spring. The two main groups that opposed the ousted president Mubarak had still not held a single meeting discussing a way forward, four months after his ouster. Thus followed a growing sense of distrust in the Egyptian political atmosphere, and the military was left in control for over a year (Stepan & Linz, 2013, p. 23).
5. This study considers a state as independent if it (a) has a relatively autonomous administration over some territory, (b) is considered a distinct entity by local actors or the state it is dependent on. Polities excluded from the list are: states that have some form of limited autonomy […]; are alleged to be independent but are contiguous to the dominant entity […]; de facto independent polities but recognized by at most one other state […]. Occupations or foreign rule are considered to be an actual loss of statehood when they extend beyond a decade. (Coppedge et al., 2020a, pp. 179-180) We consider a country consolidated as independent 10 years after its independence date. This rule is largely built on the logic used in Huntington's definition of a consolidated democracy. According to Huntington (1991, pp. 266-267), a democracy is considered consolidated after two successful power transitions following democratic elections. This usually takes eight to ten years (depending on the length of the electoral cycle). We allow the same amount of time to pass before we measure the level of democracy in the new state, which then has had enough time to reach an eventual substantial level of democracy. 6. We use the V-Dem definition of an anti-system CSO as being … organized in opposition to the current political system. That is, it aims to change the polity in fundamental ways, e.g., from democratic to autocratic (or vice-versa), from capitalist to communist (or vice-versa), from secular to fundamentalist (or vice-versa) … it must also have a 'movement' character, which is to say a mass base and an existence separate from normal electoral competition. (Coppedge et al., 2020a, pp. 184-185) 7. Which countries that are included in each regional dummy variable can be seen in Table A1 in the appendix. 8. See Figure A1 in the appendix, which illustrates the relationships between different preindependence statuses and post-independence democracy levels. 9. For a complete list of which countries that are included in this study and their respective status prior to independence, see Table A1 in the appendix. 10. For standardized residuals histograms, see Figure A2 in the appendix. 11. Unfortunately, the GDP variable from the Maddison Project Database, which is included in the V-Dem dataset, only spans back to 1950 for certain countries. This means that several countries whose GDP data is missing before 1950 and that gained independence before 1953 are excluded from analyses that include GDP as a control. Some countries also lack GDP data for other reasons. We have used all available GDP data included in the V-Dem dataset and imputed the data when it was missing only for a few years. This concerns Indonesia (used GDP value for one year prior), the Philippines (GDP value from three years prior) and Taiwan (GDP value from three years prior). A complete account of which countries were excluded from analysis for this reason can be found in Table A1 in the appendix. 12. For a comparison of the results to regressions using the main independent variable five and ten years before independence, see Tables A5 and A6 in the appendix. 13. The EDI consists of a freedom of association index (v2x_frassoc_thick), clean elections index (v2xel_frefair), freedom of expression index (v2x_freexp_altinf), elected officials index (v2x_elecoff) and a suffrage variable (v2x_suffr). The LDI is aggregated using the EDI, a rule of law index (v2xcl_rol), judiciary constraints on the executive index (v2x_jucon), legislative constraints on the executive index (v2xlg_legcon) and a bicameral legislature variable (v2lgbicam) (Coppedge et al., 2020a, pp. 42, 43, 48).
Consolidator under Grant 724191, PI: Staffan I. Lindberg, V-Dem Institute, University of Gothenburg, Sweden; as well as by internal grants from the Vice-Chancellor's office, the Dean of the College of Social Sciences, and the Department of Political Science at University of Gothenburg.

Disclosure Statement
No potential conflict of interest was reported by the author(s).

Data Availability Statement
The data that support the findings of this study are openly available in the V-Dem Dataset Version 10, at https://doi.org/10.23696/vdemds20.            Table A5 is smaller than in the corresponding models in Table 1 because there is missing data on the GDP variable for Cambodia, Laos and the Democratic Republic of Vietnam, and on all variables for Slovenia and North Korea, earlier than three years before independence. Data source: Coppedge et al. (2020b). + p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001. Notes: Standard errors in parentheses. 11 decade dummy variables are included in models 21, 22, and 23, but not reported. The number of observations (N ) is lower in model 22 than in model 20 because of missing data in the equal distribution of resources variable for Norway and the democratic experience variable for Timor-Leste, ten years before independence. Data source: Coppedge et al. (2020b). + p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001.  Table A7 than in Table 1, as Montenegro and Timor-Leste gained independence <20 years ago, meaning that there is no data on their level of democracy 20 years after independence. Data source: Coppedge et al. (2020b). + p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001.