Mind the Gap: Bridging the Theoretical and Empirical Divide of Party System Institutionalization and Authoritarian Successor Parties

Transcription

1 Mind the Gap: Bridging the Theoretical and Empirical Divide of Party System Institutionalization and Authoritarian Successor Parties Allen Hicken University of Michigan Darin Self Cornell University March 11, 2016 Abstract Authoritarian governments bequeath a variety of legacies to their democratic successors. This paper investigates the extent to which the nature of the authoritarian regimes, specifically choices over whether to build a ruling party and whether to allow (circumscribed) multiparty competition, shapes the type of party system that emerges after democratization. Using a new dataset on party and party system characteristics we test the argument that institutionalized party systems in new democracies are more likely to emerge when prior authoritarian regimes have invested in party building 1. 1 Paper prepared for the MPSA annual meeting - April Do not cite without author permission

2 1 Introduction Beginning with Samuel Huntington s seminal 1968 (Huntington, 1968) work on party institutionalization a large literature has developed with a focus on party and party system institutionalization. The focus from parties to the party system in the literature was largely spurred by Mainwaring et al. (1995). In their seminal piece on party systems in Latin America Mainwairing and Scully extend Huntington s idea of party institutionalization to that of the system as a whole. For party systems to be institutionalized, they argue, party systems must meet four criteria; stability in the rules and nature of interparty competition, parties with strong roots in society, legitimacy of the electoral process, and cohesive, disciplined, and autonomous parties (Mainwaring et al., 1995, pg. 5-6). Institutionalized party systems exhibit a pattern of behavioral regularity (Calvert, 1995; Hicken and Martinez Kuhonta, 2011) in the electoral process which serves to provide stability and strength to the democratic system. An institutionalized party system is a critical component of democratic development and success. Without institutionalized parties and party systems societal preferences and demands do not have an institutionalized channel to government. With political parties decoupled from society political parties have weakened incentive to provide the public goods that serve broad societal interests and enhance the likelihood of democratic consolidation and stability. The lack of an institutionalized party system can expose the democratic system to charismatic, patronage, or clientelistic elites who, in furthering their own interests, do little do advance the public interest (Mainwaring et al., 1995; Hicken and Martinez Kuhonta, 2011). 1

3 Because of the importance of party system institutionalization for democratic consolidation, theorizing about the determinants of party system institutionalization has tended to focus on the party system within a broader democratic context. 2 In fact Mainwaring et al. (1995) almost conflates party system institutionalization with democracy. While we agree that an institutional party system is likely necessary for robust democratic rule, we proceed from the premise that at least some of the key determinants of institutionalization are located prior to a transition to democracy itself. Building on Hicken and Martinez Kuhonta (2011) we explore the role of authoritarian parties or party systems on party system institutionalization in democracies. Hicken and Martinez Kuhonta (2011) hypothesize that party systems in democratic systems might be influenced by their authoritarian predecessors. While they present some evidence in support of their argument, they do not systematically test their hypothesis. The primary purpose of this piece is to build on Hicken and Martinez Kuhonta (2011) and analyze the extent to which certain factors of authoritarian rule explain variance in party system institutionalization in current democracies. Specifically, we focus on how variance in the types of authoritarian rule (Geddes, 1999; Hadenius and Teorell, 2007) and the degree to which autocrats invest in institutionalize parties and party systems shapes the type of party system that emerges post-transition. Our findings speak to both the literature on party system institutionalization and the literature on democratic consolidation. 2 On the determinants of party system institutionalization see (Roberts and Wibbels, 1999; Reich, 2001, 2004; Tavits, 2005; Mainwaring and Torcal, 2006; Tavits and Annus, 2006; Mainwaring and Zoco, 2007; Lupu and Stokes, 2010). 2

4 Democracies are not de novo entities. Most emerged from authoritarian regimes, albeit in different ways and at different paces. Likewise, democratic parties and party systems are often not de novo institutions. In some democracies, political parties are echoes or even direct descendants of parties that competed in earlier episodes of democracy. In addition, many of the new third wave democracies emerged from authoritarian systems with some semblance of electoral and party competition. Indeed, Magaloni and Kricheli (2010) and Gandhi and Reuter (2013) find that of autocratic regimes which existed from over 50% boasted single or hegemonic party systems. Many political parties that dominated authoritarian rule continue to survive and some have even thrived after democratic transition. The existence and survival of parties that held power during authoritarian rule (henceforth known as Authoritarian Successor Parties (Roberts, 2013a; Loxton, 2015a)) within democratic systems is somewhat perplexing given their role in overseeing human rights abuses and economic or political suppression (Loxton, 2015a). But, while this may be perplexing to some the survival and success of authoritarian successor parties is undeniable. Preliminary work by James Loxton shows that of 68 countries that underwent a democratic transition from 1974 to produced a prominent authoritarian successor party. Of these 48 countries there are 38 where the authoritarian successor party returned to power. The empirical reality of authoritarian successor parties surviving and thriving demonstrates that in order to understand the party system institutionalization within democracies our theories must account for the role of parties and party systems developed prior to democratization. In other words theories of party system institutionalization must account for the historical 3

5 legacy of authoritarianism. Because most democracies emerge from authoritarian regimes it is critical to understand whether and how characteristics of authoritarian rule influence the democratic party system. While Hicken and Martinez Kuhonta (2011) argue that authoritarian legacies play a role in the institutionalization of party systems they do not explore whether or how differences in authoritarian regime structure shape variance in party system institutionalization after democratic transitions. This is the scope of this study. Starting with Geddes (1999) a burgeoning literature has developed that attempts to catalog the types of authoritarian regimes and assess which types are more durable. Some of this work has even explored how autocrats use political parties to ensure their survival (Geddes, 2005; Schedler, 2006; Gandhi and Lust-Okar, 2009; Svolik, 2009, 2012; Wright and Escriba-Folch, 2012; Boix and Svolik, 2013). In this piece we will explore how variation in party structures within authoritarian regimes influences the institutionalization of party systems after democratic transitions. If the institutionalization of party systems is partially dependent upon authoritarian legacies the type of authoritarian rule should have a significant impact on how institutionalization takes place after a democratic transition. However, while some authoritarian regimes embed elections and parties into their political system, by definition these systems are not entirely open and fair (Levitsky and Way, 2002, 2010). Indeed, Mainwaring et al. (1995) explicitly remove Mexico and Paraguay from their analysis because the institutions at the time did not allow for fair and open competition within the party system. This introduces a few challenges for developing a theoretical framework to explore the relationship between authoritarian legacies and 4

6 party system institutionalization. The first challenge in developing a theory of the legacy of authoritarianism lies in the fact that the rules of the game under authoritarian rule effectively prevents the full institutionalization of party systems. As Mainwaring notes, (Mainwaring et al., 1995) in autocracies in which only the ruling party is allowed it makes no sense to talk about a party system, though we may useful examine the degree to which the party itself is institutionalized. Even where autocrats allow for opposition parties (e.g. Suharto s Indonesia or Singapore) electoral competition is usually a façade as autocrats game the system to prevent opposition parties from competing on the same terms as the ruling party. This places a limit on developing a theoretical framework that links party system institutionalization under authoritarian rule directly to that under democracy. Furthermore, the stability of the party system during authoritarian rule is largely artificial, which obviates the typical indicators of institutionalization like electoral volatility or party age (Mainwaring et al., 2015). Thus, a study of institutionalization under autocracy should focus more on parties than on the interaction between them, and because opposition parties are significantly disadvantaged under authoritarian rule much of the focus should be on how individual ruling parties institutionalization. A second challenge is the fact that most authoritarian parties are guaranteed electoral success and survival. Compared to their democratic counterparts this may limit their incentives and ability to fully institutionalize. Thus a theory about the role of authoritarian successor parties must account for how authoritarian successor parties that do not fully institutionalize under authoritarian rule can still affect the 5

7 way in which the democratic party system develops. Indeed, the theory developed herein argues that authoritarian successor parties acquire some antecedent strengths (Slater and Wong, 2013) which better equip them to withstand the growing pains of a democratic transition and open electoral competition. Finally, even if authoritarian parties are able to rely on antecedent strengths to succeed under democracy our question is not just about individual parties but about the party system as a whole. A theory of the historical legacy of authoritarian parties must connect the institutionalization of individual parties with the institutionalization of the party system. This is not necessarily a straightforward process party system institutionalization does not necessarily follow from the institutionalization of individual parties (Randall and Svåsand, 2002). In order to address these issues we develop a theory of the authoritarian shadow. In its simplest form our argument is that high levels of party system institutionalization under authoritarian regimes should be associated with higher levels of party system institution post transition, ceteris paribus. In addition, we explore whether the type of authoritarian regime in place affects the incentives for political elites to invest in or institutionalize political parties under autocracy, and hence the degree of institutionalization after transition. Where the authoritarian regime is structured in such a way as to induce political elites to invest in parties we should see higher levels of institutionalization. By contrast, where parties within these systems do not exist or function as a minimalist or peripheral organizations then we should see lower levels of institutionalization. Following this reasoning we expect to observe that as authoritarian regimes use political parties as a means to govern the legislative pro- 6

8 cess or electoral market we will observe greater institutionalization of political parties under both authoritarian and democratic rule. 2 Mind the Gap 2.1 Determinates of Institutionalization The work of Mainwaring et al. (1995) provided the foundation for future research into the institutionalization of party systems. In this piece Mainwaring and Scully built on the work of Huntington (1968) to demonstrate how party systems as a whole institutionalize. Huntington (1968) argued that institutionalized parties are adaptable, complex, autonomous and coherent. Adaptability refers to parties ability to respond to changing circumstances. Organizationally complex parties rely more on the organization of the party itself than on individual elites. Autonomous parties operate independent of other social groupings and primarily exist to win elections and legislate. And finally, coherence requires that individual parties hold together and are not undermined by intra-party factionalism. While this framework has been very influential in the study of parties, the ability for political scientists to measure institutionalization of individual parties in a comparative manner was and still is supremely difficult. In part because of these difficulties we have seen a shift in comparative political science to the study of party systems beginning with Dix (1992). Building on both Dix (1992) and Huntington (1968) Mainwaring et al. (1995) extended and refined these ideas to produce a new definition party system institutionalization a definition that has become the point of departure for most subsequent literature. 7

9 In their definition of party system institutionalization Mainwaring and Scully outline four components for party system institutionalization: 1) stability in the rules and nature of interparty competition, 2) parties with stable roots in society, 3) parties and the electoral process viewed as legitimate, and 4) parties that are cohesive, disciplined and autonomous. The literature on the causes of party system institutionalization is large 3 and explores wide range of possible explanatory variables. These include: the development of programmatic or ideological linkages (Tavits, 2005; Mainwaring and Torcal, 2006), strong economic performance (Lewis- Beck and Stegmaier, 2000), institutional constraints (Roberts and Wibbels, 1999; Tavits, 2005), the strength or salience of social cleavages (Lipset and Rokkan, 1967; Tavits, 2005; Birnir, 2006), or the time or timing of democratization (Mainwaring and Zoco, 2007; Mainwaring and Olivares, 2015). A related literature looks at the factors that contribute to de-institutionaliation, including the dilution of party brands (Lupu, 2014), significant breaks from party brands (Roberts, 2013a), fluctuations in economic performance (Roberts and Wibbels, 1999; Tavits, 2005), and shocks to the electoral marketplace via the introduction of new cleavages or grievances (Bartolini and Mair, 1990; Madrid, 2005). Almost without exception these studies focus on explanatory variables solely within a democratic context. But to what extent do experiences prior to transition shape democratic party systems? Is the type of party system that emerges somehow contingent on the organization of political life under authoritarian auspices? 3 Most of these studies focus on electoral volatility as either a key component of institutionalization, or a natural consequent (Mainwaring and Olivares, 2015) with volatility associate with lower levels of institutionalization. 8

10 2.2 Institutionalization of Authoritarian Parties As discussed earlier, we are sympathetic to the argument that party system institutionalization in autocratic contexts means something different than it does under democracy (Mainwaring and Olivares, 2015) (Hicken and Martinez Kuhonta, 2011, chpt. 14). After all, it makes little sense to talk about the stability of inter-party competition where only one party is allowed to compete and where the electoral outcomes are designed to be stable over time. Yet, while the party system institutionalization under authoritarian rule may be incompatible under the strict framework of Mainwaring et al. (1995) many authoritarian regimes do in fact contain parties and we know that parties are not solely instruments of democracy. Those parties exhibit variation in the degree to which they are cohesive, autonomous organizations with strong societal roots, and there is also variation in the extent to which those parties, and elections themselves, are seen as legitimate parts of the political environment. This suggests two questions. First, if the structure of authoritarian rule restricts the full institutionalization of party systems does it still make sense to talk about the degree of individual party institutionalization? Secondly, if individual parties can institutionalize does the degree of institutionalization affect party system institutionalize at a later date? One obvious way the characteristics of autocratic systems might shape party competition post-transition is through the existence of authoritarian successor parties. We hypothesize that where authoritarian successor parties are able to adapt and survive under democracy they serve as an anchor to democratic possibility. Their existence brings added stability to the fledgling party system and may provide in- 9

11 centives for competitors to develop stronger, more institutionalized parties in order to compete. But what factors shape whether authoritarian parties are able to successfully transition to democracy? The factors discussed in the existing literature sound very much like the characteristics of institutionalized parties discussed above. For example, Grzymala-Busse (2002) argues that parties which are highly professionalized during authoritarian rule are better able to remain cohesive and adapt to democratization without becoming fatally tied to the ideological preferences of past elites. In addition to professionalization, another factor that contributes to the success of authoritarian successor parties is their linkages to society during authoritarian rule. During authoritarian rule voters and elites have incentives to form alliances and remain loyal to the ruling party, but ruling parties vary in the extent to which they invest in thick versus thin organizations. In other words, some authoritarian leaders choose to invest in institutionalized parties, while others do not. Ruling parties are in a good position to make the kinds on investments that can help the party compete after democratization (Geddes et al., 2014). For example, ruling parties can draw on state and economic ties to create large, robust support networks (Kitschelt and Singer, 2015), including financial support networks (Cheng, 2006). Additionally, these parties have the ability to build local bases of support (e.g. local branches) opportunities that may be denied the opposition due to legal restrictions (Slater and Wong, 2013) or lack of resources. Overall recent work in the field demonstrates that successful authoritarian successor parties tend to be those that are cohesive, have territorial organization, boast clientelistic networks, and rely on developed economic 10

12 ties (Loxton, 2015a,b). 2.3 Bridging the Gap If it is the case that ruling parties that institutionalization under authoritarian auspices are more likely to survive the transition to democracy, can these authoritarian successor parties then help the new democratic party system itself institutionalize? As Randall and Svåsand (2002) notes, the institutionalization of a single party does not necessarily lead to the institutionalization of the entire system. Nonetheless, there are a number of reasons why the institutionalization of authoritarian successor parties may actually help the party system as a whole institutionalize. Countering Randall and Svåsand (2002), Mainwaring and Olivares (2015) argue that individual parties with a strong organization should increase overall party system institutionalization. This happens in three ways. First, individual parties that institutionalize may boost the institutionalization of the entire system by shaping the balance of power between rival forces (Roberts, 2013a). In order to compete against an institutionalized party opposition forces face incentives to coalesce (LeBas, 2011; Riedl, 2014). In effect, the existence of an authoritarian party increases the environmental hostility (Tavits, 2013, p. 159) which then induces opposition parties to become more institutionalized if they hope to compete and win elections against the authoritarian successor parties. Second, the increased professionalization developed under authoritarian rule may prove beneficial if party members defect from the successor party to form new parties and bring with them organizational skills and experience. Finally, authoritarian successor parties 11

13 also may carry a strong party brand with them into the democratic period, provided of course they choose not run from their past. The continuation of this brand should help reduce the costs to voters of identifying the party positions (Lupu, 2014) across the parties competing in the new electoral marketplace. To summarize, we argue, first, that ruling parties protected under authoritarian rule do vary in their degree of institutionalization. Some parties choose to develop linkages to society and core constituencies, establish a professionalized organization, develop strong party brands, and cultivate financial resources. Second, those parties that do invest in institutionalization are more likely to survive the transition to democracy. Third, these institutionalized successor parties then serve to improve overall system institutionalization by encouraging the opposition to coalesce, heightening environmental hostility, and promoting distinct party brands. We hypothesize that the incentives to engage in these forms of behavior will vary depending upon the type of authoritarian rule. Authoritarian regimes that use parties and party systems as instruments of rule 4 may open the door to greater levels of party system institutionalization prior to democratization. We now turn to our hypothesese. 3 Hypotheses There are a number of empirical implications that follow from the argument we just outlined. The first, and most simple, is that authoritarian systems with more institutionalized parties/party systems should be associated with more institutionalized 4 As opposed to authoritarian regimes simply using parties as a façade to appear legitimate to domestic or international opposition 12

14 party systems post-transition. Hypothesis 1: The levels of institutionalization under autocracy should be positively associated with the level of party system institutionalization under democracy. Beyond this simple hypothesis we posited a mechanism by which institutionalization under autocracy is transmitted into the democratic party system namely, through the existence and success of authoritarian successor parties. Two empirical implications of this argument are as follows. Hypothesis 2: Authoritarian successor parties are more likely where the level of institutionalization of the ruling party under autocracy was high. Hypothesis 3: Where authoritarian successor parties are competitive the level of party system institutionalization should be higher. Finally, we also argued that autocracies that choose to invest in parties will have higher levels of institutionalization than those that do not. Hypothesis 4: Authoritarian regimes that use parties will have higher levels of party system institutionalization under autocracy. We are still collecting data on authoritarian successor parties so in this initial analysis 13

15 we test only hypotheses 1 and 4. 4 Research Design and Data We argue that the structure and design of authoritarian regimes will have a lasting impact on the institutionalization of party systems after countries transition to democracy. To test this assertion we employ a large-n global analysis of party system institutionalization for countries that have experienced some authoritarian interlude between 1946 and Previous attempts to determine causes of party system institutionalization have largely been limited to individual regions such as Latin America or Eastern Europe. With a global sample we are able to explore the determinants of party system institutionalization independent of geographic regions. Huntington s work initially focused on the institutionalization of individual parties. The difficulty of measuring individual party institutionalization, however, led scholars to shift their focus to the institutionalization of party systems as a whole (Mainwaring et al., 1995). In these studies party system institutionalization is typically measured in one of two ways: using electoral volatility or average party age as proxies for party system institutionalization. These two proxies, however, only capture one aspect of party system institutionalization as conceived by Mainwaring et al. (1995) - the regularity of intra-party competition, in the case volatility, and organizational strength in the case of party age. 5 Our approach is to avoid these fairly crude measures in favor of newly available data on party systems across the world. In 5 What s more, volatility itself might more usefully be thought of as a symptom of poor institutionalization, rather than a constituent part Mainwaring and Olivares (2015). 14

16 the following sections we outline how we measure party system institutionalization and discuss the measurement of our key explanatory variables. 4.1 Measuring Party System Institutionalization Our primary measure of party system institutionalization, which we will refer to as PSI, is a new measure developed by the Varieties of Democracy Project (Hereafter V-Dem). 6. The construction of PSI, which is a index of five party system-related components, was a two stage process. The first stage is the aggregation of ordinal ratings provided by multiple country experts (five or more) for the five individual components (party organization, branches, linkages, distinct party platforms, and legislative party cohesion 7. In the second stage, the outputs of these first stage analyses are aggregated into the PSI index using Bayesian factor analysis techniques. 8 6 We merged a number of datasets to V-Dem V.5 for this study. A significant issue when merging datasets is the variation in coding country names. To address this problem we used an R package developed by Darin Self to uniformly code countrynames across time. This package can be downloaded at 7 The specific questions used to measure these components can be found at the V-Dem website. v-dem_codebook_v43.pdf See section 2.16 Party institutionalization index 8 We draw on Bernhard et. al. (2015) for this summary of process: The individual components of the index are aggregated from the responses that experts provided to the relevant questions on the V-Dem survey. These multiple ordinal rating were then aggregated into a unified, continuous and reliable variable using Bayesian item response theory (IRT) models. Because individual raters might vary with regard to the way they interpret the questions, and in terms of reliability and consistency, these models are useful because they incorporate the information encoded in the variation in raters perceptions, and in reliability levels across and within coders into the estimation process (Bollen and Paxton 2000, Jackman 2004). IRT models assume that the variable being measured is latent, that it cannot be measured directly and that each coders response includes a degree of error. The model uses patterns of disagreement across ratings to estimate coder level errors, and down weighs the ratings provided by coders who are deemed less informative. The Bayesian framework allows us to estimate country-year level parameters that capture the latent variable of interest, along with thresholds differentiating the levels of the variable. For a more detailed discussion see Pemstein et al. (2015). 15

17 For more details on the characteristics of the PSI index (its strengths and weaknesses) see Bizzarro et. al. (2016). The V-Dem data includes observations for 193 countries with fairly regular coverage from 1900 to As can be viewed in the summary table below the PSI index ranges from 0-1. While V-Dem boasts an impressive scope of data across space and time, though the sample employed in this study is far more limited do the availability of supporting data. 4.2 Explanatory Variable Our key explanatory variable is the institutionalization of the parties/party system under autocracy. We take three approaches to operationalizing this concept. Our primary measure is the average PSI index during authoritarian period (PriorPSI ). Figure 1 9 compares the level of PSI in autocracies to the level in democracies. A couple of things are apparent. First, there is a good deal of variation in PSI within both democracies and autocracies. Second, on average, the level of PSI is higher in democracies than in autocracies, consistent with Mainwaring (2015). However, we are interested in whether the level of PSI under autocracies helps predict the level PSI under democracy. 9 We use a very relaxed definition of democracy for this comparison. Countries were coded as being democratic if their Polity score was 0 16

18 Figure 1: Distribution of PSI by Regime Type To supplement our primary analysis we also we employ two additional sets of variables as indicators of the party system under the antecedent regime. First, we use the data created by (Wahman et al., 2013) (Hereafter ARD). ARD builds and improves upon work initially done by Geddes (1999) to categorizing authoritarian regime types. Specifically, they divide autocracies into the following 6 categories 10 : Military, Monarchy, Multi-party, No-party, One-party, and Other. 11 A challenge with this and all such classifications is the reality of hybrid regimes for example, military regimes that form a political party. Thus, we expect these estimates to be noisier than our direct measure of PriorPSI. Recall that we argue that institutionalization of parties under autocracy should affect the institutionalization post-transition. Thus, where parties don t exist, we expect the level of post-transition PSI to be lower. This 10 Unfortunately there are not enough observation for us to test the effect of Monarchy and No-party 11 By contrast, Geddes (1999) lists five regime types: Military, Personal, Party, Oligarchy, and Monarchy. Geddes categorization has been criticized for being over simple, and for neglecting the distinction between single an hegemonic party regimes. 17

19 includes the Military, and Other regime types, where we expect to see a negative association with post-transition PSI. By contrast, we expect post-transition PSI to be highest in systems where multi-party competition was allowed, ceteris paribus. Finally, our expectations about the level of institutionalization under single party rule, and therefore the level of PSI under democracy, are unclear. We anticipate that the level of institutionalization under single party rule will be less than in multi-party regimes, but more than under other regime types. However, we are uncertain what to expect in terms of post-transition institutionalization. On the one hand, single ruling parties may be institutionalized and thus in a good position to become an authoritarian successor party. On the other hand, it is possible the monopoly on political organization in single party regimes stunts the development of new parties post-transition and therefore hampers institutionalization. Under single-party rule no other parties are allowed to compete in elections and in many cases opposition parties are completely banned. The existence of a singleparty rule deprives opposition parties of the ability to openly establish linkages with society, build a party brand, develop professional cadre of staff and candidates, and raise funds. This may also deincentivize the ruling party from investing in organizational capacity and building linkages to society. In essence single-party rule prevents the opposition from developing any antecedent strengths that may be carried into democracy while also making. Thus our expectations about the effect of single party rule are mixed. If the authoritarian successor party effect dominates we expect a positive association, but if the monopoly effect dominates we expect a negative relationship. 18

20 We use ARD to determine authoritarian regime type and transitions to democracy, and include in our final sample only those cases where the the policy score in the last year in the data set is 2. ARD limit their coding of regimes to cases after the year Thus, our analysis using ARD limits us to cases where a country had some year of autocracy in 1972 or later. If a country was coded as democratic in 1972 but had some autocratic experience prior to 1972 we are unable to capture the previous authoritarian regime type for those cases 12. Finally, in addition to PSI and ARD we also use Svolik (2012) (hereafter PAR) data on the politics of authoritarian rule to model PSI as a function of the antecedent regime 13. Unlike Wahman et al. (2013) and Geddes et al. (2014), Svolik (2012) breaks autocratic regimes into their component parts rather than creating an overarching typology for all them. These component parts include which type of Executive and Legislature are present as well the type of Military and Party System structure Within each type we code the sub-types as binary variables. As an example, there are three categories for types of the military: Civilian, Corporate, and Personal. We create a single dummy variable for each of these types of military rule and use each individual type as a treatment. 12 The majority of these cases are 1 st or 2 nd wave democracies such as the United States, Canada, and Western European states. Though we are forced to exclude those cases we are able to capture the majority of countries in the 3rd wave of democracy 13 Svolik (2012) codes authoritarian regimes dating back to This significantly expands the time series from Wahman et al. (2013) allowing us to capture most of the 2 nd and 3 rd wave democracies but still excluding 1 st wave democracies 14 Because Svolik (2012) codes changes in these component parts prior to a full regime change we also code an alternate to his Party System to account of the party system that was the system during the majority of time of the prior regime 15 Svolik (2012) codes each regime as a cross section of countries and year ranges rather than a full panel. Using an R package designed by Darin Self we expand the date to a full country-year panel. This package can be found at 19

21 4.3 Control and Balance Variables We incorporate a number of variables in addition to the core explanatory variables discussed above as controls as well as covariates to balance the data. Previous studies that model determinants of party system institutionalization include some well-known structural and institutional factors. We incorporate additional controls for GDP from the IMF as well the level of ethno-linguistic fractionalization ELF. Data from both of these variables come from the Quality of Governance Standard Dataset. We also included variables measuring the effective number of electoral parties (ENEP), proportional representation (PR), and whether the democratic regime is presidential (Presidentialism), all from Selway and Self (2016). We also calculate how many transitions a country has experienced within the time-frame of the sample (Regime Number) as well as the duration Duration (in years) of these regimes. Finally, a significant hypothesis within the party system institutionalization literature is that party systems become more institutionalized over time and through repeated elections. To account for this hypothesis we use data drawn from the V-Dem dataset to capture the number of electoral Rounds under democracy. 4.4 Methodology To estimate the causal effects of the antecedent regime structure on subsequent party system institutionalization we employ both matching and regression techniques Because our variation in our primary explanatory variable (PSI ) is relatively homogeneous over time we condense our sample into a cross-section in order to fit a linear model. 17 For the matching approach we pool our data instead of condensing it into a cross-section 20

22 Table 1: Descriptive Statistics Statistic N Mean St. Dev. Min Max PSI 2, PriorPSI 1, ENEP 2, PR 2, Presidentialism 2, ELF 2, GDP 2, Number 2, Duration 2, USSRSatellite 2, We specify our linear model as follows: y i = β + x 1i β 1i + x 2i β 2i + ε i where x 1 i is a binary variable coded for each sub-type of the antecedent regime structure we wish to effectively use as our treatment, x 2 i is a set of control variables previously discussed, and ε i is the error term. To estimate the effect of sanctions on regime type we also employ a matching technique using the GenMatch package in R. GenMatch uses genetic matching to balance data which can then be used to estimate the average treatment effect on the treatment. Genetic Matching is a balancing method which employs an evolutionary search algorithm using the Mahalanobis distance to balance the data to allow multivariate matching(diamond and Sekhon, 2013). When the Mahalanobis distance is not optimal for achieving balance GenMatch searches over the space of distance 21

23 metrics to find something better and iterates through this process until it can no longer improve the balance. GenMatch generalizes the Mahalanonbis distance by including an additional weight matrix: d(x i, X j ) = { (X i X j ) T (S 1/2 ) T W S 1/2 (X i X j ) } 1 2 where W is a k k positive definite weight matrix and S 1/2 is the Cholesky decomposition of S which is the variance-covariance matrix of X (Diamond and Sekhon, 2013, pg. 6). The balance matrix X supplied for analysis is: X = {GDP, P riorp SI, ENP, ELF, P olity, Duration, P residential, Number} 5 Findings We argue that PriorPSI will have a significant causal effect on the level of PSI. In addition to this we also hypothesize that autocrats will vary the extent to which they invest in party system institutionalization dependent upon the structure of the authoritarian regime. We begin our analysis by specifying a fixed effects model of PSI on PriorPSI with the results reported in Table 2. Our initial model finds a statistically significant and strong positive relationship between PSI and PriorPSI. To test the strength of this relationship we add a number of other variables that could possibly mitigate the strength of this relationship. As previously discussed, the passage of time after 22

24 Table 2: Estimated Effect of PriorPSI on PSI Dependent variable: PSI (1) (2) (3) (4) PriorPSI (0.055) (0.054) (0.040) (0.040) Electoral Rounds (0.004) (0.005) Control Variables No No Yes Yes Observations Adjusted R Note: p<0.1; p<0.05; p<0.01 transitioning to democracy may be another cause of institutionalization. To operationalize this we include a continuous measure of the number of elections at time t. In the second model we estimate a beta coefficient with a positive slope for both PriorPSI and Electoral Rounds. However, introducing Electoral Rounds reduced the size of the beta coefficient for PriorPSI suggesting that the initial model suffered from omitted variable bias. Because of the possibility of more omitted variable bias we specified two more models that re-estimate Models 1 and 2 in Table 2 with controls 18. The introduction of these controls nearly halves the slope of PriorPSI. This coefficient, however, remains fairly stable when controlling for Electoral Rounds Controls: ENEP, GDP, ELF, PR, Presidential, USSR Satellite, Number, and Duration 19 We note here the reduction in the number of observations and potential bias in our estimators as the presence of missing data is likely the function of factors which may also influence PSI 23

25 We argue that these models identify 20 a positive causal relationship between PriorPSI and PSI. We find that party system institutionalization does indeed continue on into democracy after the end of authoritarianism and that as PriorPSI increases there is a substantial increase in the subsequent PSI after democratization. This suggests that democratization does not produce a tabula rasa for party systems and that democratization should not be considered a starting point for party systems. These findings affirm our initial beliefs; that there is an authoritarian legacy that influences how party systems take root in democracies. Having identified a positive relationship between the institutionalization of party systems under authoritarian rule and the institutionalization of party systems after democratization we turn our focus to understanding how differences in authoritarian regime structure effects the institutionalization of party systems post-transition. As previously stated, we hypothesize that authoritarian regimes that promote incentives to invest in party rule will have an overall positive impact on post-transition PSI. To test this we specify both bivariate and multivariate regressions 21 to estimate the effect of the antecedent regime structure on the expected level of PSI. Our first models employ the Wahman et al. (2013) data. Wahman et al. (2013) divide authoritarian regimes into six types: Military, Monarchy, Multi-Party, No-party, and One-party. Because of a data availability we only estimate the effect of Military, Multi-Party, and One-party 22. We estimate two models, a bivariate and multivariate model, for 20 In subsequent papers we will improve our identification strategy to reduce error in our identification strategy. 21 All observations are drawn from a sample where polity2 2 unless otherwise stated. 22 We code each regime structure as a dummy with a score of 1 if regime type matches the desired regime type and 0 for all other types. 24

26 each type of authoritarian regime and present the results in Table 3 and Table 4. The coefficients for Military, Multi-party and Single-party regime types are significant in the bivariate specifications. As expected the coefficient for the Military is negative while Multi-party is positive. Finally, the coefficient for Single-party is negative. This would suggest that the monopoly effect of single party regimes swamps the successor party effect. These models would estimate the causal effect of each regime type if the regime type is independent of other factors that also explain levels of party system institutionalization. We do not believe this to be the case and include a number of controls in subsequent models. These subsequent models alter the coefficients for all regime types. We also graph these estimates below with coefficient plots. With these models we find that only the regime type Military has a significant influence on subsequent PSI. We estimate that authoritarian regimes that have some authoritarian Military experience are expected to have a slightly higher (0.06) PSI than other regimes types. This result runs counter to the hypotheses laid out previously where we believed that authoritarian regimes with some party structure would have a positive effect on PSI after transitions to democracy. However, note that these models include Prior-PSI as a control, which, if, hypothesis 4 is correct, should be highly correlated with autocratic regime type. Thus, it is not surprising that when we both PriorPSI and the regime type in the model that the effects of the latter are washed away. More on this below. In addition to Table 3 and Table 4 we graph the coefficients of the models with their standard errors in Figure 2 below. These figures add greater clarity to the 25

27 findings presented in Tables 3 and 4. We see that regime types Military, Multi-party, and Other each have the expected negative impact on the PSI of the subsequent democratic regime, while multi-party has a a positive impact. 23 This supports our reasoning that authoritarian regimes that allow for some party competition leads to a higher expected level of PSI. This is especially notable when compared to the expected level of PSI for Single-party authoritarian regimes. It appears that eliminating space for opposition parties to form or operate in the authoritarian context causes party systems to have lower expected levels of PSI after transitions to democracy. This suggests that some party competition, even if uneven competition, leads to better outcomes for party system institutionalization than completely dominate party competition by a single party. Using these models we find some support for our authoritarian regime structure hypotheses. However, we find a significant change in the coefficient for PriorPSI upon adding the regime type into the model. Recall that we opened our empirical tests by modeling the effect of PriorPSI on PSI with results presented in 2. Using the same controls that we do for the models presented in Tables 3 and 4 the coefficient for PriorPSI increases substantially while at the same time washing out the effects of the regime type. 23 When we add controls, specifically priorpsi, the the signs of these variables flip with the error term for Multi-party and Other barely touching or crossing zero. 26

28 Table 3: Estimated Effect of Antecedent Regime Structure Military (0.030) (0.040) Dependent variable: PSI (1) (2) (3) (4) Multi-Party (0.028) (0.036) PriorPSI (0.083) (0.079) Electoral Rounds (0.003) (0.003) Intercept (0.015) (0.079) (0.015) (0.077) Control Variables No Yes No Yes Observations Adjusted R Note: p<0.1; p<0.05; p<

29 Table 4: Estimated Effect of Antecedent Regime Structure Single-Party (0.037) (0.050) Dependent variable: PSI (1) (2) (3) (4) Other (0.041) (0.061) PriorPSI (0.079) (0.080) Electoral Rounds (0.003) (0.003) Intercept (0.014) (0.076) (0.014) (0.076) Control Variables No Yes No Yes Observations Adjusted R Note: p<0.1; p<0.05; p<

30 Figure 2: Coefficient Plots for Tables 3 and 4. In each of these initial models we found that PriorPSI explains a significant amount of variation in PSI for democracies with an authoritarian past while the authoritarian regime type only explains a limited portion of the variation. However, noting the increase of the coefficient for PriorPSI when adding the regime type to the model we used a system of equations to identify the effect of the different authoritarian regime types on PSI. To do this we regressed PriorPSI on each of the four regime types and extracted the residuals. We then regressed PSI on each of these residuals only to identify the expected outcomes of PSI as a function of PriorPSI and regime type and present the results in Table 6. Using a system of equations we find that countries with the antecedent regime type Military have a much higher expected PSI than those that had the antecedent regime type of Multi-party, Single-party, or Other. We are especially interested in the change in party system institutionalization countries undergo after democratization. Table 5 gives the average level of PSI prior to democratization. Unlike the initial OLS models this the result in Table 6 identify a higher expected 29

31 Table 5: Average Level of PSI During Authoritarianism PriorPSI Type Military Multi-party Single-party Other level of PSI for Single-party and Other antecedent authoritarian regime types than Multi-party 24. This is especially startling when considering the average level of PSI which we report in Table 5. In Table 5 we see that Multi-party regimes have the highest level of PSI of all authoritarian regime types. It appears, however, that these countries are not expected to experience any increase in PSI after a democratic transition the expected level of PSI found in the models reported in Table 6 and the average PSI prior to democratization are almost the same. This appears to show that party systems emerging from Multi-party do not undergo any radical shifts. This is not the case, however, for Military, Single party, and Other regime types, which experience radical upward movement of PSI after democratization. This dramatic shift in PSI is shown in Figure 3 for each type of authoritarian rule. Not that all party systems drastically increase their level of PSI after democratization with the exception of Multi-party regimes. This is an intuitive finding as these countries need to undergo less radical changes to adopt democratic procedures. 24 To draw a modern example the expected level of PSI of a country with the antecedent regime of Military would compare with current day Ghana with the estimated level of PSI for Single-party and Multi-party would compare to current Egypt and Colombia respectively. 30

32 Table 6: Regression on Residuals of Structure and Prior PSI Military Multi-Party Single-Party Other Dependent variable: PSI (1) (2) (3) (4) (0.082) (0.079) (0.080) (0.076) Electoral Rounds (0.003) (0.003) (0.003) (0.003) Intercept (0.058) (0.060) (0.059) (0.058) Control Variables Yes Yes Yes Yes Observations Adjusted R Note: p<0.1; p<0.05; p<