DEPARTMENT OF POLITICAL SCIENCE PLAYING MUSICAL CHAIRS: THE ROLE OF ELITE SUPPORT DURING DEMOCRATIC TRANSITION Samuele Arioli Master’s thesis: 30 credits Programme: Master’s Programme in Political Science Date: 25/05/2025 Supervisor: Fabio Angiolillo Words: 15835 Abstract What role does the support of elite groups play in sustaining democratic transition? While the literature on elite defections has emphasized internal ruptures within regime coalitions as triggers for democratization, less attention has been paid to how elite dynamics shape the success of democratic transitions once they begin. This thesis argues that elite support during transition is both dynamic and interest-based. Drawing from a novel panel dataset covering 135 episodes of democratic transition between 1900 and 2019, I conduct logistic and time-to-event analyses to investigate the relationship between elite group support and democratic transition, focusing on three such groups. The results show that military support decreases the probability of transition by 9.2%, while business elite support has the opposite result, increasing it by 8.9%. Party elite support, instead, proves crucial as transition is ongoing: At any given point, countries where the party elites support the regime are approximately 58% more likely to complete their transition successfully compared to those without party elite support. By disaggregating elite groups and tracking their changing support through transition episodes, this study adds nuance to the current scholarship on democratization and points at the role of elite support realignment as a potential driving factor of successful democratic transitions. Keywords: democratic transition; elite support; military; party elites; business elites; time-to- event analysis List of Contents Introduction ................................................................................................................................ 1 Literature Review ....................................................................................................................... 3 Democratization ..................................................................................................................... 3 Elites’ Political Agency ......................................................................................................... 6 Determinants of Elite-Led Democratization .......................................................................... 8 Theoretical Framework ........................................................................................................... 10 Military ................................................................................................................................ 12 Political Elite ........................................................................................................................ 14 Economic Elite ..................................................................................................................... 16 Research Design ...................................................................................................................... 18 Logistic Regression Model .................................................................................................. 18 Cox Proportional Hazards Model ........................................................................................ 18 Dependent Variable ............................................................................................................. 19 Independent Variables ......................................................................................................... 21 Covariates ............................................................................................................................ 22 Descriptive Statistics ............................................................................................................ 23 Results ...................................................................................................................................... 26 Elite Group Support and Overall Likelihood of Democratic Transition ............................. 26 Successful Democratic Transition over Time by Elite Group Support ............................... 28 Regression Diagnostics ........................................................................................................ 32 Robustness Tests .................................................................................................................. 33 Discussion and Limitations .................................................................................................. 38 Conclusion ............................................................................................................................... 39 Bibliography ............................................................................................................................ 41 Appendix .................................................................................................................................. 48 List of Tables Table 1. Elite group support and likelihood of democratic transition. .................................... 27 Table 2. Elite group support and hazard of successful democratic transition. ........................ 31 Table 3. Elite support group (50%) and likelihood of democratic transition. ......................... 34 Table 4. Elite support group (continuous) and likelihood of democratic transition. ............... 35 Table 5. Elite group support (50%) and hazard of successful democratic transition. ............. 36 Table 6. Elite group support (continuous) and hazard of successful democratic transition. ... 37 Table 7. Descriptive summary of the variables ....................................................................... 48 Table 8. Episodes of democratic transition by country (1900–2024). ..................................... 51 Table 9. Cox proportionality assumption for a model including business support. ................ 55 Table 10. Multicollinearity. ..................................................................................................... 56 Table 11. Cox proportionality assumption. .............................................................................. 57 List of Figures Figure 1. Trajectory of regime support after transition onset across three elite groups. ......... 12 Figure 2. Average regime support during episodes of democratic transition. ......................... 23 Figure 3. Trends of regime support within five years since democratic transition onset. ....... 24 Figure 4. Elite support and EDI across three cases of democratic transition. ......................... 25 Figure 5. Average marginal effects of elite group support with 95% confidence intervals. ... 28 Figure 6. Cumulative probabilities of democratic transition by elite group with 95% confidence intervals. ................................................................................................................ 29 Figure 7. Adjusted cumulative probabilities of democratic transition by elite group with 95% confidence intervals. ................................................................................................................ 32 Figure 8. Distribution of regime support across all social groups (1789–2024). .................... 49 Figure 9. Distribution of transition lengths. ............................................................................. 50 Figure 10. Scaled Schoenfeld residuals. .................................................................................. 58 Introduction At dawn on New Year’s Day 1958, an Air Force mutiny shook Caracas. The fighter planes were directed to Miraflores Palace, the residence of Venezuelan dictator Marcos Pérez Jiménez. Although the coup ultimately failed, general strikes were called for January 21, prompting Jiménez to order a violent crackdown. The military, however, refused to comply and sided with the protesters, forcing Jiménez to flee the country two days later (Chin, Carter and Wright, 2021). Theories on elite defection would explain the fall of the dictatorship and the country’s subsequent transition to democracy as originating from ruptures among regime insiders (O’Donnell and Schmitter, 1986; Svolik, 2012; Haggard and Kaufman, 2016; Reuter and Szakonyi, 2019; del Río and Higashijima, 2024). Yet, a closer look at the unfolding events reveals a far more complex landscape of actors. Following Jiménez’s departure, a group of Army officers led by Admiral Wolfgang Larrazábal established a provisional government. The transitional junta remained in power until democratic elections were held in December that year, after which it committed to returning to the barracks. Meanwhile, the leaders of three major opposition parties convened at Puntofijo Residence to sign an agreement whereby they would respect the electoral results, prevent single-party hegemony, and create a government of national unity. As the new system began to take shape, these parties sought the support of key industrialists and business enterprises for their project of institutional and economic relaunch, promising state subsidies and guarantees against future expropriation in return. In 1961, a new Constitution was signed, marking the successful end of Venezuela’s transition to democracy. This puzzle challenges current knowledge about the role of elite defections for democratic transition, leading to a central question: What role does the support of elite groups play in sustaining democratic transition? To investigate this research problem, I develop a theory on the patterns of regime support for the military, political, and economic elites. These groups are theoretically prioritized in virtue of their leverage over force, representation, and capital which positions them as unique social actors able to continuously and substantially impact national governance. Since these groups share the same goal of self-perpetuation, their support to the emerging regime follows a strategic calculation of the anticipated costs and benefits their members will face once democracy becomes consolidated. I adopt a novel empirical approach based on the Episodes of Regime Transformations (ERT) framework (Maerz et al., 2024), operationalizing democratic transitions as episodes involving 1 a change of regime from autocracy to democracy. This implies the introduction of multiparty, free, and fair elections, and the fulfilment of Dahl’s minimal institutional guarantees for de facto accountability. Moreover, one deriving benefit of measuring regime change as an episode is the ability to unpack the transition process over time and track the duration of regime support by each elite group. To measure elite regime support during transition, I use the data collected by Knutsen et al. (2025) and included in the V-Dem dataset v15 (Coppedge, Gerring, Knutsen, Lindberg, Teorell, Altman, et al., 2025). For my empirical analysis, I merge the ERT and V-Dem datasets to create a panel dataset covering 135 episodes of democratic transition between 1900 and 2019. I conduct a logistic regression analysis to estimate the overall likelihood of democratic transition. The findings are robust and show that support by the military decreases the probability of transition by 9.2%, while support by the business elites has the opposite result, increasing it by 8.9%. In a second step, I conduct time-to-event analysis to assess the impact of regime support in ongoing transition episodes. The results indicate that, at any given point during the transition, countries where party elite support the regime are approximately 58% more likely to complete their transition successfully compared to those without party support. This study is a direct contribution to the literature on regime support coalitions, by offering a large-N quantitative analysis of how three elite groups influence the prospects for democratic transition. It also extends elite agency theories by examining these groups separately and identifying generalizable causal patterns during transitional periods. In doing so, the study shows that ruling elites are not a monolithic bloc but consist of distinct groups with divergent incentives to support democratic reform. Finally, this thesis introduces a tension with the literature on elite schisms by raising the possibility that, although elite groups are guided by different interests, they may converge in their support for the emerging regime as transition unfolds – pointing to a dynamic of elite realignment, particularly between party and business elites. The rest of this thesis is divided into five sections. Section two reviews previous literature on democratization, analyzing its principal conceptualizations as well as the empirical contributions of two major strands of research. It also elaborates on the role of elites during regime change and concludes by identifying the key determinants of elite-led democratization. Section three develops the theoretical framework and lists the main hypotheses. Section four outlines the empirical models, details variable operationalization, and provides a descriptive overview of the data. Section five presents the empirical results, along with diagnostics and 2 robustness tests. It also discusses the results and outlines the study’s limitations. Finally, section six summarizes the main findings and concludes by suggesting directions for future research. Literature Review Democratization In its most basic formulation, democratization designates the process through which a country becomes more democratic. This implies a movement toward a political system in which the government is continuously responsive to the preferences of its citizens, considered as political equals (Dahl, 1971, p. 1). In Dahl’s framework, such sustained accountability is guaranteed through six political institutions: elected officials, free and fair elections, freedom of expression, alternative sources of information, associational autonomy, and inclusive citizenship (Dahl, 1998, p. 85). This conceptualization offers a more substantive threshold than other minimalist definitions centered on the de jure presence of multiparty elections (see, for example, Schumpeter, 1942, p. 269), proving helpful to guard against cases where national elections – albeit existent – are executed under conditions that compromise electoral freedom and fairness (Diamond, 2002). Therefore, democratization is understood as a macro-dynamic of regime transformation toward better democratic institutions and practices. The literature identifies three types of democratization: political liberalization, democratic transition, and democratic consolidation. Although analytically different, these democratizing pathways can be sequential in principle, but are not immune to crises of instability and, ultimately, to reversals (Linz and Stepan, 1978). First, political liberalization refers to transformations within an authoritarian regime, usually through reforms that limit oppression and censorship, strengthen individual and collective rights, and open up the political arena to opponents (del Río and Higashijima, 2024, p. 3). Autocratic leaders tend to implement such reforms as a strategy to deflate public pressures or coopt the opposition (Linz, 1990, p. 148). However, although liberalization increases the costs of state repression and the benefits associated with individual and social action, it cannot by itself guarantee a shift to democracy, since power is still heavily concentrated in the hands of the autocratic rulers (O’Donnell and Schmitter, 1986, p. 6). This capacity to define unilaterally rules and procedures proves evident in the context of elections, where incumbents manipulate electoral rules to tip the playing field in favor of the regime’s own candidates (Lührmann, 3 Tannenberg and Lindberg, 2018, p. 63). Therefore, in these “liberalized authoritarian systems” or “dictablandas” (O’Donnell and Schmitter, 1986, p. 9), ruler accountability remains weak despite improvements in citizen equality. Second, transitions extend beyond episodes of liberalizing autocracy by eliciting a change from one regime to another. Regime change involves fundamental variations in the “sets of formal and informal rules and procedures for selecting national leaders and policies” (Geddes, 1999, p. 116). A country undergoes a democratic transition when it departs from autocracy and turns into a type of democracy that satisfies the minimal requirements of competition and participation to guarantee de facto accountability. Far from being a smooth and linear process, democratic transitions are characterized by uncertainty and discontinuity, since the very structure of authority and the rules for achieving political power become subject of heavy contestation (O’Donnell and Schmitter, 1986, p. 6). Indeed, under the propulsion of liberalizing reforms, various political actors and groups set to rewrite the rules of the game until consensus in favor of democracy is reached. Consequently, drafting constitutional contracts is seen as a crucial step in any transition (Linz, 1990, p. 157). A transition to democracy is considered complete when there is sufficient agreement about the procedures to produce a government which is chosen through free and fair elections, enjoys effective policy-making ability, and whose (executive, legislative, and judicial) branches do not share power with other non-elected bodies (Linz and Stepan, 1996, p. 3). Finally, democratic consolidation represents the third democratizing type or phase. Born as a label for the study of new democracies, the term “consolidation” has been subject to much debate due to its teleological connotation (O’Donnell, 1996, p. 38) and conceptual ambiguity (Schedler, 1998, p. 102). While some accounts equate consolidation with a process of enriching the social and distributional qualities congruent with democracy (democratic progress), others are more focused on the safeguards achieved by democratic rule against authoritarian creep (democratic survival). Schedler observes how consolidation is deeply connected with “expectations of regime continuity” (Schedler, 1998, p. 103). For democracy to be stable and thrive, however, there needs to occur a strategic and normative shift whereby citizens and politicians alike agree that democracy has become “the only game in town” (Linz and Stepan, 1996; Diamond, 1997). In newly established democracies, this means the emergence of a certain degree of loyalty to democratic procedures as the sole means for regulating competition for office and resolving conflicts. According to Diamond (1997, pp. 18–19), consolidation unfolds along three mutually constitutive dimensions. First, accountable and representative 4 institutions are extended to every level of governance (democratic deepening). Second, political actors internalize the rules of competition and agree to constrain the exercise of power within clear constitutional boundaries (political institutionalization). Finally, over time, the regime learns to fulfill citizens’ expectations for economic stability, public order and freedom (regime performance). These three democratizing pathways are generally conceptualized and measured according to two prominent approaches: transitologist and incrementalist. The transitologist school focuses on exploring the conditions surrounding democratic transitions and breakdowns as key moments of regime change, and gained prominence during the world’s third wave of democratization (Huntington, 1991). This analytical preference assumes boundaries between categorical opposites and underscores transitions as “interval[s] between one political regime and another” (O’Donnell and Schmitter, 1986, p. 6). This implies a binary understanding of democracy and non-democracy (i.e., autocracy), whereby the crossing of some nominal measure or threshold on a continuous indicator is required for a transition to be identified (Collier and Adcock, 1999, p. 547). Boix, Miller and Rosato (BMR) compiled a dataset covering political regimes from 1800 to 2007 that offers a prominent example of comprehensive dichotomous measure of democracy (Boix, Miller and Rosato, 2013, p. 1530). A country is coded democratic if it satisfies a set of criteria related to contestation (i.e., the chief executive is elected and accountable, and the legislature is elected in free and fair elections) and to participation (i.e., a majority of adult men is enfranchised). The incrementalist school casts aside conceptual binarism in favor of a more graded approach that seeks to explain temporal changes in levels of democracy. Incrementalist scholars position real-world regimes along an autocracy-democracy continuum to fully appreciate the continuous nature of democracy (Bollen and Jackman, 1989, p. 612), where full democracy represents the ideal type (see Dahl, 1971, 1998). Empirically, Coppedge and collaborators widen the scope of analysis to the broader phenomenon of democratization and test a set of determinants on four dependent variables: the level and annual change of democracy, as well as annual upturns and downturns (Coppedge et al., 2022, p. 17). However, the validity of such approaches is constrained by a set of limitations. On the one hand, dichotomies drive transitologists to neglect nuances among observations that fall in the same regime category. Therefore, autocratic countries without elections or with no de facto competition (e.g., North Korea) are often mentioned in the same breath with those where 5 executive offices and legislatures ritually face elections, albeit under highly unequal conditions (e.g., Cameroon) (Nord et al., 2024, p. 17). This has two implications. First, all regimes within the same categories have equal probabilities to undergo substantial changes. Second, regime transformations either succeed or do not, thus overlooking cases where democratization kicks off but fails to stabilize or others where democratic breakdown is preempted. On the other hand, incrementalists tend to assume that equidistant changes in either direction on the scale amount to the same phenomenon and share similar causes (except for Coppedge et al., 2022). This fails to appreciate that equal amounts of change can have very different implications for countries situated at the lower end of the continuum (e.g., Syria) compared to those near the top (e.g., Costa Rica) (Nord et al., 2024, pp. 62–63). Moreover, the focus on comparing democracy levels neglects temporary but substantial changes that might occur between the two selected data points. For instance, measuring Tunisia’s democracy levels between 2011 and 2022 through this lens might point to stability, while failing to account for the short-lived but transformative impact of its revolution. Additionally, both approaches prioritize short-term effects over long-term processes, and treat democratization and autocratization either as two different research agendas (as in the case of transitologists) or as meaningful equivalents (the incrementalists). Elites’ Political Agency In the vast literature on regime change, there exists widespread consensus among scholars that the outcome of democratization is influenced by a combination of structural conditions and leaders’ choices. While structural factors – such as economic development (Lipset, 1959; Boix and Stokes, 2003), inequality (Acemoglu and Robinson, 2005), natural resources (Barro, 1999), and culture (Fish, 2002) – offer the permissive condition for a country to become democratic, political agency has been shown to serve as the generative condition, acting as short-term catalyst of change (Treisman, 2020). Among political actors, incumbent leaders and elites are in a favorable position to initiate change. Broadly defined, elites are “persons with power individually, regularly, and seriously to affect political outcomes at the macro level of organized societies” (Higley, Field and Groholt, 1976, p. 17). The literature on democratization presents incumbent elites and leaders either as rational actors who continuously engage in risk assessment to maximize their interests, or as agents prone to take suboptimal decisions due to imperfect information or inaccurate calculation. These opposite conceptualizations are reflected in two broad strands of research. 6 A first stream of literature approaches democratization as “a deliberate choice on the part of political leaders” (Rustow, 1970, p. 355). According to one account, richer elites would concede liberalizing reforms as a way to coopt poorer segments of society who have mobilized to expropriate them (Acemoglu and Robinson, 2005). Alternatively, rulers may want to increment the democratic credentials of the regime to legitimize policies that increase their revenues along with economic growth (North and Weingast, 1989). Looking at authoritarian elites in party-based autocracies, Slater and Wong distinguish between democratization through “weakness” and democratization through “strength” (Slater and Wong, 2013, p. 718). Democratization through “weakness” takes place when authoritarian elites concede democracy as a last resort, given an imminent risk of violent overthrow. According to Geddes (1999), dominant party regimes – which tend to be the most robust and durable among regime types (Geddes, Wright and Frantz, 2014, p. 319) – act in this way to prevent the emergence of some other authoritarian regime that will persecute them. As an example, in 2011 the military and civilian elites that had supported Ben Ali’s regime in Tunisia actively negotiated a transition to democracy in face of mounting social turmoil (Geddes, Wright and Frantz, 2014, p. 313). Opposed to this, in democratization through “strength” authoritarian elites decide to liberalize when the risks associated with democratization are lower than the potential deriving benefits. This is called a “conceding-to-thrive scenario” (Slater and Wong, 2013, p. 718): High incumbent capacity provides the dominant party elite which faces declining popularity with enough confidence that it can liberalize without sparking major socio-economic instability and losing its relevance as a political actor in the long term. Notably, although authoritarian legacies might prove difficult to eradicate (Albertus and Menaldo, 2018), this type of incumbent-led democratization does not pose an inherent risk of future regression as most authoritarian successor elites have been able to reinvent themselves as committed democratic actors who play according to the rules of the game (Grzymala-Busse, 2002, 2020; Riedl et al., 2020). The role of conservative elites in shaping the rise of democracy during the first wave of democratization in Western Europe illustrates this point (Ziblatt, 2017). A second stream of literature takes a more critical stance toward politicians’ intentions and behaviors. Given that democratization often implies high uncertainty as institutions and norms undergo fundamental changes, some researchers shifted their analytical focus from negotiated transitions to examining the potential causal role of incumbent mistakes on regime outcomes. Treisman’s (2020) study finds that over two thirds of the cases of democratic transition that have occurred since 1800 can be attributed to dictators who weakened their own grip on power 7 by committing serious blunders in their attempt to block democracy. Such errors are categorized according to whether incumbents (i.e., the leader and/or the elite members) took suboptimal choices in relation to domestic outsiders, regime insiders, or international actors. Recurrent mistakes involve incumbents over-/underestimating their own or their opposition’s popularity and resources. In the first case, the incumbent can call for an election or plebiscite to buttress their rule but end up underperforming, thus galvanizing the opposition and eliciting elite defections (Treisman, 2020, p. 796). In 1988 Pinochet called for a referendum to continue its presidential mandate but failed to gain enough support, prompting him to leave office one year and a half later and allow for Chile’s democratic transition to commence. In the second case, instead, the incumbent might neglect the growing power of its opposition and fail to use reforms as a tool to demobilize and coopt rival factions (Treisman, 2020, p. 793). In 1848, at the outbreak of the February Revolution, King Louis-Philippe I refused all concessions before being violently overthrown. Although analytically different, these two approaches might illuminate different phases of a country’s democratization history, where deliberate choice and incumbent mistakes intersect. The Spanish democratic transition exemplifies this possibility: Franco and his associates mistakenly handed power over to King Juan Carlos and Adolfo Suárez who eventually turned out to be detrimental to the authoritarian state. These figures played an essential role in spearheading the transition to democracy, which was confirmed by the adoption of a new constitution in 1978 (Treisman, 2020, p. 805). Determinants of Elite-Led Democratization What might induce ruling elites to allow democratic change? This dilemma has been central to much agency-centered literature on democratization. Stemming from this, it is possible to outline a typology of pressures that might lead elites in power to take decisions that – whether intentionally or not – end up improving a country’s democratic institutions. These pressures can be grouped into three broad categories: vertical, horizontal, and intra-ruling elites. Vertical pressures involve bottom-up forms of community engagement that directly challenge ruling elites. Protests, demonstrations, and boycotts can serve as a useful tool to expose the regime’s weaknesses and create instability within the ruling coalition. Non-violent popular mobilization can drive liberalization in autocracies as elections increase the costs of repression for the regime and serve as “focal points” for activists to mobilize in case of frauds and irregularities (Teorell and Wahman, 2018, p. 82). Thus, research on competitive 8 authoritarianism has emphasized how multiparty elections are able to introduce “structural vulnerabilities” in the form of regime challengers (Schedler, 2013, p. 147; cf. Morgenbesser and Pepinsky, 2019). Horizontal pressures stem from opposition elites. As long as they are sufficiently organized and present an electorally viable alternative (Donno, 2013, p. 706), opposition parties and candidates pose a credible threat to authoritarian stability by forging splits within the ruling coalition (especially between hard-liners and soft-liners) and exploiting popular discontent as a source of legitimacy for its own actions. In particular, once contentious collective action emerges, opposition actors can articulate popular protests into coherent political demands and a broader critique of the incumbent’s right to stay in power (Slater and Wong, 2013, p. 720; Sato and Wahman, 2019). As an example, in Ghana the opposition led by the National Patriotic Party (NPP) leveraged public discontent against the ruling National Democratic Congress (NDG) right before the 1996 elections, managing to obtain important reforms in electoral quality (Sato and Wahman, 2019, p. 1431). Intra-elite pressures emerge from fractures within the circle of regime’s supporters. The transitions school has long emphasized that “there is no transition whose beginning is not the consequence – direct or indirect – of important divisions within the authoritarian regime itself” (O’Donnell and Schmitter, 1986, p. 19). There is overall empirical consensus that elite schisms represent a major threat to authoritarian stability, since they are capable of fundamentally rearranging the power configuration within the regime (Geddes, 1999; Svolik, 2012; Djuve, Knutsen and Wig, 2020). Understood as “instances in which regime-affiliated elites voluntarily abandon the ruling coalition in order to challenge the regime” (Reuter and Szakonyi, 2019, p. 554), defections imply a re-shuffling of important skills, resources, and organization capabilities from leading to oppositional forces. Such critical events usually follow disrupting shocks like economic crises, which are likely to erode the regime’s ability to maintain its clientelist relations with its supporters and effectively co-opt the opposition (Djuve, Knutsen and Wig, 2020). Member defection typically responds to a strategic logic of costs and benefits aimed at maximizing the candidate’s personal revenue (Reuter and Szakonyi, 2019). To sum up, the reviewed literature on political elite’s incentives and democratization presents causal arguments that emphasize either power asymmetries between actors or the unintended effects of authoritarian mistakes. Although the shift from socio-structural to political determinants testifies to the advancement of classical elite theory within regime studies, this 9 research area remains limited in assessing the ways different elite groups navigate the internal dynamics of regime transformation and its implications on the prospects for democracy. Indeed, we still know little about the behavioral patterns and strategic decisions of distinct elite groups during democratization. This underscores the importance of disaggregating elites from the institutional character of the regime (cf. Linz and Stepan, 1996; Geddes, 1999). Elites might differ as to the specific logics of survival, which, in turn, shape both the extent of engagement with the democratizing process and the resources and skills they bring to the emerging regime. Therefore, this thesis contributes to ongoing research on the social origins of regime change by developing a testable theory on the patterns of regime support of three elite groups common to virtually every modern state and examining their implications for a successful democratization. Theoretical Framework Of the three democratizing pathways, democratic transitions involve the most profound regime transformation, whereby arbitrary power becomes constrained by a set of rules that institutionalize uncertainty over political outcomes (Przeworski, 1991, p. 14). Transitions are not instantaneous events but rather lengthy processes that typically start with the enactment of some liberalizing reforms and culminate in a founding democratic election. Over the course of several years, major shifts often occur in the power relations between societal actors, as well as in how these actors pursue their interests. This underlines the dynamic nature of regime support coalitions, as institutional change goes in hand with a reconfiguration of those groups upon which the regime depends for legitimacy. It is, therefore, crucial to investigate the role of elite groups during episodes of democratic transitions. By unpacking the concepts of elite and regime support, and by zooming in the transition process itself, it is possible to identify broader trends in elite support and assess its effects on the prospects for democracy. Regime support constitutes a vital source of energy for any political system. Along with demands, support serves as input that the system is called upon to process in order to develop policies and other authoritative decisions – i.e., the output (Easton, 1957, p. 387). In this study, I focus specifically on elite support for those in power, which is particularly consequential in autocratic and transitional regimes, where the distinction between regime and government remains often blurred (Svolik, 2012, p. 4; Knutsen et al., 2025, p. 5). Indeed, although in democracies the regime (i.e., the rules of the game) can be analytically separated from the 10 government (those who play by them), such separation cannot be presumed until a transition is complete. Because the principal goal of elites is self-perpetuation (Field, Higley and Burton, 1990, p. 151), their support for the ruling authorities is mainly interest-based. During periods of political discontinuity, such as transitions, elites supporting democratic reforms act in the way they do to protect their rights and interests (Albertus and Menaldo, 2014, p. 578). Therefore, as regime change unfolds, elites’ alignment with emerging democratic institutions can be explained as a strategic calculation of the anticipated costs and benefits their members will face once democracy becomes consolidated. Three such groups stand out the most as they have potentially vital consequences for the functioning of any political regime: the military, the political, and the economic elites. As the exclusive holders of “autonomous national power” (Mills, 1956, p. 6), they occupy key positions within the national security apparatus, state institutions, and major corporations. Their influence stems from the strategic command posts they control in the fundamental structures of modern society and the valued experiences they extract therefrom – i.e., prestige, power, and wealth. Consequently, they continuously and substantially impact national governance. Although analytically distinct, these three domains of national power do not work in isolation; rather, they are able to intersect, develop mutual dependencies, and grow networks of communication and influence that reinforce their everyday dominance (Mills, 1956, p. 8). Figure 1 visually illustrates the proposed conceptualization. During democratic transitions, regime support varies across elite groups as a response to opposite interests. When autocratic survival is low, the military concedes democracy as a self-extrication strategy aimed at protecting its corporate interests. Once returned to the barracks, the military’s capacity to credibly affect the regime declines with the expansion of civilian control over the security apparatus. In this early phase of transition, political parties (the political elite) begin to institutionalize and emerge as stable supporters of democracy, since electoral competition becomes the only way of maintaining or accessing power. Over time, as the political elite signals credible commitment to the emerging democratic institutions and future prospects of investment increase, economically powerful actors follow such cues by gradually aligning with the political elite to vouchsafe their rights and ensure continued influence within the new institutional order. 11 Figure 1. Trajectory of regime support after transition onset across three elite groups. Military During the height of the Cold War the military represented a pivotal actor in the management of domestic politics of most countries. Brazil (1964–1985), Nicaragua (1936–1979), South Korea (1961–1982), and Spain (1939–1975) are typical examples where sustained military presence in the state apparatus marked a distinct version of autocratic rule. Indeed, these “military regimes” constituted the second most common regime type in the 1970s, only after dominant-party regimes (Geddes, Wright and Frantz, 2014, p. 316). Since the 1980s onwards, however, military-led governments have declined dramatically and, in a number of countries, allowed for the emergence of lasting democratic institutions (for example, Chile after 1988). Military officers are more likely to be the initiators of a transition to democracy when they perceive that the regime is about to tumble (Geddes, Wright and Frantz, 2018, p. 207). As military rule begins to deteriorate following disrupting events such as economic crises, unsuccessful wars, and popular mobilizations, the officer corps decides to extricate itself from the direct responsibilities of governance and attempts a prompt return to the barracks. In such circumstances, the overarching purpose of maintaining the corporate interests of the military- as-institution trumps any consideration of holding onto public office (Linz and Stepan, 1996, p. 67). Consequently, democracy appears as the “safest bet” for a military that desires to protect 12 its autonomy, unity, and permanent functions of defense rather than being forced to ouster by a rival military faction, insurgents, or foreign intervention. The probability of a negotiated transition increases when the military is able to ensure immunity guarantees for itself. Military leaders usually do so by striking bargains with civilians. In Venezuela, for instance, the military under Admiral Larrazábal agreed to relinquish power in exchange for amnesty for human rights violations committed under the Jiménez dictatorship (Karl, 1990, p. 11). Such military-initiated deals often include informal alliances with political parties expected to succeed in the future democratic politics, providing officers a shield against prosecution and revanchism (Wright and Escribà-Folch, 2012, p. 293). A case in point is Guatemala, where the military leadership – implicated in forced disappearances during the civil war – backed the Christian Democracy and its presidential candidate, Cerezo Arévalo, to minimize the risk of post-transition accountability (Jonas, 2000, p. 11; Wright and Escribà-Folch, 2012, p. 293). As transition unfolds and founding democratic elections determine public offices, the active support of the military for the regime gradually becomes less needed. Once the military officers withdraw to their barracks, the state can successfully usher in a process of demilitarization, consolidating civilian control over the security apparatus. While this shift does not necessitate an immediate ideological commitment to the democratic principles by the military (Agüero, 1995, p. 22), it marks a fundamental transformation in military-civilian relations, ensuring that all areas of government policy remain in the hands of elected civilians and the head of state assumes the role of commander-in-chief (Agüero, 1995, p. 16; Grewal, 2023, p. 35). In this way, the military becomes professionalized along meritocratic principles of recruitment and promotion, and its messianic self-image is replaced by a focus on external defense instead of domestic political interventionism (O’Donnell and Schmitter, 1986, p. 31). This needs not imply a shrunk coercive capacity; rather, civilian rulers may keep improving the military capabilities of the state to ensure a monopoly of violence in society and limit the likelihood of future coups by providing efficiency wages to its military corps (Besley and Robinson, 2010, p. 661). The pace of demilitarization, however, varies according to the extent of military entrenchment in the political institutions of the state prior to transition. In Brazil, for example, the military managed to retain a number of entry points and veto powers well into the transition process, partly due to former members of the military-backed National Renewal Alliance (ARENA) integrating in the newly founded political parties (Hagopian, 1990, p. 155). Similarly, in post- 13 1983 Turkey, the military remained a latent threat to government stability and exerted policy influence through the National Security Council (Aydınlı, 2012, p. 101). Conversely, in Argentina, where the military publicly carried the blame for the Malvinas/Falklands war debacle, President Alfonsín moved swiftly to restructure the armed forces. The reforms included abolishing clientelist ties to the state, re-adjusting the military budget, and prosecuting nine top officers directly responsible for state terrorism under the first three juntas (Nassif, 2020, p. 9). Attaining civilian supremacy is, therefore, considered essential, especially in countries that where the military was highly empowered during the previous autocratic rule. Failure to establish such control can lead to crises of reversibility, as unchecked and highly politicized militaries may still be able to come out of the barracks and obtain their preferred outcome. This was the case in post-Marcos Philippines, where the military engaged in six coup attempts between 1986 and 1990. A partial settlement was eventually reached, under which the military recognized the Aquino administration in exchange for protection of its veto powers over important areas of national security (Bello and Gershman, 1990, p. 44). In sum, the military is more likely to allow for a democratic transition to take place when regime survival appears untenable and imposes risks to the corporate interests of the military- as-institution. Thus, the military views democratization as a strategy to extricate itself from the direct responsibilities of governance by returning to the barracks. As the transition unfolds and civilian control is extended onto the security apparatus, the military’s role as key regime supporter declines and its role is confined to matters of external defense. Accordingly, the probability of a successful transition is expected to increase as military support declines during transition. I then formalize the first two hypotheses as: Hypothesis 1: Regime support by the military is negatively associated with the occurrence of a democratic transition. Hypothesis 2: During democratic transition, lower levels of regime support by the military are associated with higher likelihood of successful democratic transition. Political Elite In democracies, the political elite primarily consists of leaders and members of political parties in the legislature, especially those holding executive power. Political parties are an endemic feature of any democracy (Stokes, 1999, p. 245), as they serve as a linkage between citizens 14 and the state, ensuring that governments remain responsive to public preferences and translate constituents’ demands into policy decisions (Dahl, 1971, p. 1). The main interest driving party elite’s behavior during transition is the desire to maintain or acquire power. To enhance their chances of remaining politically relevant and reduce the risk of post-exit retribution, many autocratic successor parties in Central and Eastern Europe have successfully managed to regenerate themselves as legitimate political contenders after the collapse of communist rule by fundamentally changing their symbols, organizational bases, and agendas (Grzymala-Busse, 2002). In developmental Asia, other ruling parties pre-empted direct confrontation to their rule and solved lapses in popularity by spearheading political reform toward greater competition. Confident of their developmental records, party decision- makers of Taiwan’s Kuomintang allowed opposition forces to compete in the 1986 general elections. Since then, the party has maintained high levels of electoral success, ensuring unabated control over the legislature (Slater and Wong, 2013, pp. 723–724). Other prominent examples include South Korea’s Democratic and Justice Party under Roh Tae-Woo and Indonesia’s Golkar under Habibie (Slater and Wong, 2013, p. 722). Beyond authoritarian successor parties, most party organizations emerge and become key sources of democratic support as basic freedoms of association and political competition improve. Thus, in the early phases of transition, when public expectations for democracy rise, these newly founded parties must find enough resources to be able to articulate popular demands and effectively mobilize voters once elections are called. This entails developing strong societal roots, stabilizing inter-party competition patterns, securing a stable membership base, maintaining organizational autonomy, and expanding throughout the national territory (Mainwaring, 1998, pp. 68–69). Party-system institutionalization greatly influences the prospects of democracy, as it strengthens accountability mechanisms both vertically with civil society and horizontally across parties (Bernhard et al., 2020, p. 5). In turn, this provides incentives to both government and opposition parties to support democratic reforms, increasing the likelihood of a successful transition as democratic stances become embedded across the party system at large (Angiolillo, Wiebrecht and Lindberg, 2025, p. 6). During the 1970s in Mexico, the decades-ruling Institutional Revolutionary Party (PRI) implemented a series of electoral reforms to boost the competitive façade of the system in response to opposition boycotts. However, growing electoral success of two major rival parties, combined with rising popular discontent following the 1982 debt crisis, forced the PRI to open negotiations with the opposing Party of National Action (PAN) and allow for substantial 15 reforms to take place, including the establishment of a Federal Electoral Institute to oversee elections (Nelson, 2014, p. 110; Lewis, no date). Over time, the PRI’s popularity declined drastically and its hegemony over Mexican politics ended with the 2000 elections, when PAN- backed Vicente Fox became president. In sum, political parties are the main pillars of democracy, serving as a linkage between citizens and state. While some ruling elites may concede democracy from a position of strength, most parties emerge as fundamental freedoms are established, enabling them to build societal support, organizational capacity, and national reach. During this period, support for democracy spreads across both governing and opposition parties as electoral competition becomes the only way through which they can maintain or compete for office. For these reasons, the probability of a successful transition is expected to increase in parallel with growing support by the political elite as the transition progresses. I then formalize the third and fourth hypotheses as: Hypothesis 3: Regime support by the political elite is positively associated with the occurrence of a democratic transition. Hypothesis 4: During democratic transition, higher levels of regime support by the political elite are associated with higher likelihood of successful democratic transition. Economic Elite The economic elite consists of influential actors of a country’s economy, such as industrialists, corporate executives, and private investors. Their substantial rents and tax revenues position them as powerful pressure group within political regimes and during regime changes. Similar to other elite types, their support for democratic reforms reflects a strategic decision to meet long-term interests. Despite the flow of rents and special privileges they may enjoy at a point in time, economic elites are not inherently safer under autocracy. Indeed, the absence of strong accountability mechanisms can limit their ability to effectively monitor incumbents, making their property rights vulnerable to state interference (Albertus and Menaldo, 2014, p. 579). In fact, autocratic rulers may use sanctions and expropriations as instruments of control against rival elites, redistributing resources to more loyal supporters (Albertus and Menaldo, 2010, pp. 4–8). In Zimbabwe, President Mugabe’s extensive program of land redistribution was aimed to eliminate white-owned farms, expand the patronage system, and reward the regime’s main support base – the black peasantry (Shaw, 2003, p. 76; del Río and Higashijima, 2024, p. 9). 16 Similarly, in post-1929 Mexico, the PRI leadership pursued a series of large-scale land reforms as a means to consolidate control and increase its chances of political survival (Díaz-Cayeros et al., 2015, p. 7). Thus, when autocracy breaks down and conditions are in place for the emergence of a democratic alternative, economic elites are expected to support regime change as democratization increases future prospects for investment, economic freedom, and secured property rights (Doucouliagos and Ulubaşoğlu, 2008; Acemoglu et al., 2019). As the new political leadership manifests support for institutional change, this signals credible commitment to economic actors – both beneficiaries and outsiders under the previous regime –, who are incentivized to engage with the democratizing coalition to vouchsafe their rights and interests and ensure the continuation of de facto power once the transition is complete. However, the role of the economic elite during transitions may take different forms, depending on its relative bargaining power vis-à-vis other societal forces. Thus, while class compromise represents the best course of action in contests marked by social upheaval and high labor mobilization (Przeworski and Wallerstein, 1982, p. 216), economic elites may, in other cases, secure greater influence in the constitutional-making process and craft institutional arrangements to preserve rents at the expense of future redistribution, perpetuating a pattern of elite-biased democracy (Rajan, 2009, p. 180). A case in point is Ecuador, where the Supreme Court and Congress during the 1970s, acting in defense of economic interests, successfully blocked a series of policies intended to increase redistribution through corporate taxes and land reforms (Albertus and Menaldo, 2014, p. 582). In sum, while the economic elite may benefit from rents and special privileges under autocracy, their long-term interests are often better served under democratic rule. Thus, as the transition unfolds and the political leadership signals strong commitment to institutional change, economic elites are likely to follow such cues and engage with the emerging democratic coalition in order to make sure their interests are protected in the future. Their support can provide important resources and legitimacy to the emerging democratic regime. For these reasons, the economic elite’s support for the regime is expected to increase gradually over the course of the transition, as the perceived benefits of the new regime become more certain. I then formalize the last two hypotheses as: Hypothesis 5: Regime support by the economic elite is positively associated with the occurrence of a democratic transition. 17 Hypothesis 6: During democratic transition, higher levels of regime support by the economic elite are associated with higher likelihood of successful democratic transition. Research Design In this section I discuss the empirical models for testing the hypotheses, the sources of the data used, and the operationalization of key variables and covariates. I also present descriptive statistics and conduct country-specific checks to assess face validity. Logistic Regression Model To investigate the direct relationship between elite group support and the occurrence of a democratic transition – as formulated in Hypotheses 1, 3, and 5 –, I employ logistic regression analysis. This empirical method estimates the probability of a binary dependent variable 𝑌 to be equal to 1, given a set of predictors 𝑋. Formally, this probability is expressed as: 1 𝑃(𝑌 = 1 | 𝑋 ) = 1 + 𝑒!(#!$ #"&"$ ⋯ $ ##& (1) #) Here, 𝑃(𝑌 = 1 | 𝑋 ) denotes the conditional probability of the outcome 𝑌 = 1 given the values of the independent variables 𝑋), … , 𝑋*. 𝛽+ is the intercept, while 𝛽) to 𝛽* are the coefficients associated with each predictor. The constant 𝑒 refers to the base of the natural logarithm. ) $(&!' &")"' ⋯ ' & ) ) is the logistic function used to estimate a predicted probability between )$ , # # 0 and 1 of observing the outcome (Harrell , 2015, p. 220). Since democratic transitions occur in only 10% of the sample observations, they qualify as rare events in the context of logistic regression analysis. Therefore, I adopt Kosmidis and Firth’s (2009) bias-reduction method to reduce estimation bias due to the low number of transition cases. I discuss the operationalization of the dependent variable further below. Cox Proportional Hazards Model Testing Hypotheses 2, 4, and 6 requires a statistical model that accounts for time since transition onset. For this reason, I employ David Cox’s proportional hazards (PH) model. Of widespread use in medical research for the analysis of survival time data, this model has been adopted by quantitative social sciences to address questions focusing on conditions affecting time to specific events, such as revolutions, coups, or, as in this case, successful democratic transitions. 18 The Cox PH model is “a survival analysis regression model, which describes the relation between the event incidence, as expressed by the hazard function and a set of covariates” (Bradburn et al., 2003, p. 431). The term hazard refers to the probability that a unit experiences the event at a given point in time. Formally, this probability is expressed as: ℎ(𝑡 | 𝑋) = ℎ+(𝑡) × exp(𝛽)𝑋) +⋯+ 𝛽*𝑋*) (2) Here, ℎ(𝑡 | 𝑋) denotes the instantaneous risk of the event occurring at time 𝑡 given the values of the independent variables 𝑋), … , 𝑋*. The term ℎ+(𝑡) is the baseline hazard, which measures the (time-varying) hazard if all the independent variables are equal to 0. The impact of the predictors is measured by the size of their respective coefficients 𝛽), …, 𝛽*. exp (𝛽)𝑋) +⋯+ 𝛽*𝑋*) is the relative risk or hazard ratio (HR) of each covariate – that is, the effect of each independent variable on the probability of an event occurring at any given time. The Cox PH model is semi-parametric. So, while it does not assume any specific statistical distribution for the baseline hazard, it does require the predictors to have a proportional and multiplicative effect on the hazard over time. Therefore, this proportionality assumption implies that hazard functions for any two units are proportional, and their HRs remain constant over time (Bradburn et al., 2003, p. 432). Dependent Variable I draw data on democratization from the ERT dataset v15 (Maerz et al., 2024). This dataset adopts a comprehensive methodological approach that treats political regimes as the same set of phenomena varying in their proximity to the ideal type of democracy and in their fulfilment of Dahl’s six institutional guarantees for de facto accountability (Dahl, 1998, p. 85; Maerz et al., 2024, p. 970). Its authors define regime transformations as “periods when a country undergoes sustained and substantial changes along the democracy-autocracy continuum” (Maerz et al., 2024, p. 970). They operationalize these variations by means of V-Dem’s Electoral Democracy Index (EDI), which ranges from 0 (not democratic) to 1 (ideal democracy) (Teorell et al., 2019). To be classified as an episode of regime transformation, the ERT requires a starting annual change of at least ±0.01 and a subsequent cumulative shift of at least ±0.10 over the course of the episode. An episode is considered ongoing if the EDI score (1) registers at least one change within five years, (2) does not experience a single-year reversal ³0.03, and (3) does not undergo a cumulative reversal of 0.10 over a five-year period. The episode is coded as complete in the year when a cumulative change of ±0.10 has occurred since 19 episode onset and immediately precedes any of the previous three conditions for termination (Maerz et al., 2024, p. 972). Compared to other datasets, the ERT framework applies a directional definition of regime transformation that captures subtle variations in the political structure of a country, without requiring it to shift from one regime to the other (i.e., regime change). Moreover, it also distinguishes among successful and failed episodes, enabling research on the factors leading to divergent outcomes. In this way, the dataset identifies a total number of 671 episodes from 1900 to 2024. As the primary interest is in the relationship between regime support by elite group and democratic transition, I rely on the ERT’s “democratization outcome” variable (dem_ep_outcome) (Edgell et al., 2025a). This variable captures detailed outcomes of the transformation episodes, dividing them into six categories.1 Outcomes are coded on the basis of the Regimes of the World (RoW) classification by Lührmann, Tannenberg and Lindberg (2018, p. 63) which distinguishes among four regime types along the continuum: closed and electoral autocracies; electoral and liberal democracies. In this case, I focus on democratic transition, which is coded as 1.2 To be classified as democratic transition, an episode must involve a change of regime from (closed or electoral) autocracy to (electoral or liberal) democracy. This implies the introduction of multiparty, free, and fair elections, and the fulfilment of Dahl’s minimal institutional guarantees (Maerz et al., 2024, p. 972). Accordingly, I construct a dichotomous variable, “democratic transition,” for use in Equation 1. The variable takes value 1 if the country-year is undergoing an episode of democratic transition and 0 if it is a stable autocracy – that is, not experiencing an ERT and ranked as (closed or electoral) autocracy on the RoW scale. Since the Cox PH model (Equation 2) is based on periods between starting and ending points, I construct a second variable called “event.” The variable equals 1 if the given country-year is the final year of transition, and 0 if the transition is ongoing or if no transition occurs (i.e., the country-year is a stable autocracy). To measure time to event – also known as the survival object in Cox’s framework –, I construct the variable “transition years,” which captures the number of years a country has been in 1 These are: no democratization episode; democratic transition; preempted democratic transition; stabilized electoral autocracy; reverted liberalization; deepened democracy; uncertain outcome (Edgell et al., 2025b). 2 The code is applied to all the country-years of the episode, regardless of whether a democratic deepening also occurred (Edgell et al., 2025b). 20 transition since onset, up to and including the present year. Specifically, the variable is computed as the difference between the observation year and the first year of democratization plus 1, for the subset of country-years undergoing democratic transition. Values equal to 0 indicate that the country-year is a stable autocracy. Independent Variables Regime support is operationalized through V-Dem’s “regime support groups” variable (v2regsupgroups). This variable captures the coalition of social groups the regime relies upon to maintain power. Similar to Geddes (1999, p. 116), a regime is defined by “the set of (formal and informal) rules that are essential for selecting political leaders and maintaining them in power” (Djuve, Knutsen and Wig, 2020, p. 924). In other words, a regime determines whom selects policies and how these policies are selected. Regimes tend to differ in the number of social groups they rely upon for power and legitimacy. A supporting group, thus, comprises individuals who, first, prefer the current regime over alternative forms, and, secondly, offer crucial financial resources, organizational capacity, and expertise without which the regime they back would be seriously undermined (Knutsen et al., 2025, p. 5). It is important to note that, while regime and government are distinct concepts in democracies, the two tend to overlap in autocracies and, likely, in transitional regimes as well (Knutsen et al., 2025, p. 5). This is due to the fact that, although a democracy requires a separation of the rules for selecting leaders and the leaders per se, such separation cannot be presumed for a country until a democratic transition is fully complete. This point is central to the research problem at hand, which concerns how elite group support for those in power shapes the prospects for democratization. The variable captures 13 social groups,3 and it is based on expert coders’ ratings aggregated into interval scale (0–1) to define whether a given social group is supporting the regime (1) or not (0). Because V-Dem’s collection of expert-coded data relies on a minimum quota of five coders per country (Coppedge, Gerring, Knutsen, Lindberg, Teorell, Marquardt, et al., 2025, p. 16), each score reflects the level of coder agreement on whether a certain social group supports the regime for a given year. Therefore, higher scores (closer to 1) resonate with high 3 These are: the aristocracy; agrarian elites; party elites; business elites; the state bureaucracy; the military; an ethnic or racial group(s); a religious group(s); local elites; urban working classes; urban middle classes; rural working classes; rural middle classes; a foreign government or colonial power (Coppedge, Gerring, Knutsen, Lindberg, Teorell, Altman, et al., 2025). 21 consensus across coders that the given group is supporting the regime, while lower values (closer to 0) reflect high agreement that the group does not. Hence, each social group’s variable is measured continuously (0–1), although it aims to represent whether the group is supportive of the regime. I solve this issue through a conservative approach to turn scores with high likelihood of support into supporting groups (1) and leave mid-to-low likelihood of support as non-supporting groups (0). To do so, I dichotomize group variables by assigning any score ³0.75 (i.e., at least 75% of coder agreement that the group is supporting the regime) as indicating group support (1), and all others indicating lack of support (0). Figure 8 in the Appendix shows the distribution of regime support across all 13 social groups from 1900 to 2024 following this approach. To inspect the role of military, political and economic elites during democratic transitions, I focus on three such groups: party elites (of the party or parties holding executive power) (v2regsupgroups_2), business elites (v2regsupgroups_3), and the military (v2regsupgroups_5). Although the variable includes several other elite groups that may influence regime change, these three groups are theoretically and empirically prioritized in view of their leverage over representation (parties), capital (business), and force (military), which positions them as unique actors with the ability to seriously influence regime dynamics. Covariates I employ a set of control variables for economic, political, and social dimensions. First, to account for the correlation between economic development and democracy (Knutsen et al., 2019), I include (logged) GDP per capita and GDP growth as % of GDP (Fariss et al., 2022). Second, since popular protests may constitute important threats to authoritarian stability, especially when they coordinate with political parties (Sato and Wahman, 2019, p. 1423), I use a variable measuring the frequency of mass mobilization events such as demonstrations, strikes and sit-ins (Coppedge, Gerring, Knutsen, Lindberg, Teorell, Altman, et al., 2025). Third, I include a variable capturing electoral quality via electoral management body capacity to account for the role of elections in regime dynamics (Donno, 2013; Coppedge, Gerring, Knutsen, Lindberg, Teorell, Altman, et al., 2025). Finally, to control for the effects of political inclusion on broad-based regime support and democracy (Angiolillo, Wiebrecht and Lindberg, 2025, p. 11), I include a variable measuring the distribution of power by social group 22 (Coppedge, Gerring, Knutsen, Lindberg, Teorell, Altman, et al., 2025). Descriptive summary of the variables is reported in Table 7. Descriptive Statistics The ERT dataset captures a total of 147 successful democratic transitions over the past 124 years (see Table 8). These episodes reflect shifts on the RoW scale from autocracy to democracy, followed by a founding democratic election. Most of these cases include further democratic deepening as well. Between 1900 and 2024, 111 countries had at least one transition. Figure 2 displays the frequency of regime support across party elites, business elites, and the military during episodes of democratic transition between 1900 and 2024. Panel A shows the distribution based on country-years in transition. Party elites are the most prominent, showing regime support for 47.5% of all country-years. They are followed by the business elites (19.5%) and, lastly, by the military (11.2%). Panel B accounts for the potential skew caused by longer transitions by focusing on transition episodes. It captures the percentage of transition episodes in which each group supported the regime for at least one year. Party elites remain the first (62.6%), followed by the business elites (27.9%) and, lastly, by the military (21.8%). Together, the panels underscore the role of party elites as the most common support base across all countries in transition, resonating with the findings of Geddes, Wright and Frantz (2014, p. 318). Figure 2. Average regime support during episodes of democratic transition. 23 Figure 3 presents aggregate trends in regime support across the focal elite groups during the first five years of transition – a timeframe corresponding to the median transition duration in the sample (see Figure 9). Support from party elites experiences a steep rise in the first two years of transition and continues to increase until the fourth year, reaching nearly half of all the country-years in transition. Despite a small dip in year five, regime support from political parties remains at relatively high levels. Business elites show a more gradual increase in regime support over time, with an evident uptick in year three, reaching less than 25% of the country- years in the last two years of transition. In contrast, regime support from the military declines sharply around transition onset and remains relatively low – around 10% of the country-years –, before experiencing a small rise in year five. Figure 3. Trends of regime support within five years since democratic transition onset. Next, I conduct country-specific checks to assess face validity of the independent variables. Figure 4 compares the variation in elite support coalitions with the EDI trend for three countries in transition. Each colored bar corresponds to a specific elite group, while the black line traces the development of the EDI through the transition episode. Venezuela experienced a democratic transition between 1958 and 1961. It stands out as a crucial case of elite support realignment that sequentially involved all three focal groups. After Jiménez’s deposition from power in 1958, the military agreed to leave public office in exchange for an amnesty for past abuses, becoming an apolitical body focused on external defense. This prompted a resurgence of party politics as old and new political forces stepped in to fill the 24 institutional vacuum. Parties’ commitment to the emerging regime was signaled through a foundational agreement signed between the representatives of the three major parties – the Democratic Action (AD), the Social Christian Party (COPEI), and the Democratic Republican Union (URD) –, with the aim of upholding democratic electoral rules and normalizing inter- party competition (Karl, 1990, p. 11). This is reflected in a rapid increase in the EDI score from 0.26 in 1958 to 0.6 in 1959. The business elites appear to show support for the new regime only in the final year of transition (EDI = 0.66), following negotiations with parties and labor unions that resulted in substantial state subsidies and guarantees against property socialization, as codified in the 1961 Constitution (Karl, 1987, p. 84). Figure 4. Elite support and EDI across three cases of democratic transition. Hungary transitioned to democracy between 1988 and 1991, marking the end of state communism. The graph indicates that regime support by party elites is constant over the entire episode, proving consistent with the post-communist Hungarian Socialist Party’s (MSzP) reinvention as a committed democratic competitor (Grzymala-Busse, 2002). This is accompanied by a quick rise in the EDI score, equal to 0.8 in 1990. Business elites shored up the emerging democratic regime in the second half of the transition, corresponding to the beginning of a series of government-sponsored stabilization policies aimed to liberalize the economy and increase privatization (Bartlett, 1996, p. 48). 25 Lastly, democratic transition in Bangladesh spanned over 12 years (1986–1997), representing a typical case of protracted transition characterized by the gradual demilitarization of the state after military rule. Following a small rise in the EDI in 1987, the country’s democracy level stabilized around 0.3 until a new positive change of +0.2 in the sixth year (EDI = 0.49). Indeed, until 1990 General Ershad managed to combine military rule with piecemeal reforms to coopt the opposition – including holding partially free elections in 1986. However, mounting protests throughout the country and growing discontent among high-ranking military officers led Ershad to resign and allow for elections in 1991 (Maniruzzaman, 1992, p. 208). Although the end of martial law enabled parties to re-enter the political arena, civilian control over state power was a gradual process. The risk of military intervention persisted, as exemplified by the 1996 failed attempt by General Nasim to mobilize troops against President Rahman Biswas (Imtiaz Ahmed, 2006, p. 290). Results The empirical analysis – based on a time series spanning 1900 to 2019 – consists of four parts. First, I test the direct relationship between elite group support and the overall likelihood of democratic transition through logistic regression. Second, I move to time-to-event analysis to inspect whether the probability of a successful democratic transition over time varies according to elite group support. Third, I discuss post-regression diagnostics. Finally, I re-estimate the models for two alternative operationalizations of the predictor to test the robustness of the results. Elite Group Support and Overall Likelihood of Democratic Transition Table 1 presents the results of the logistic regression analysis. Although logistic coefficients are notably difficult to interpret – as they reflect changes in the log-odds of the outcome for a one-unit increase in the predictor –, it is still possible to determine their direction and level of statistical significance. The first three models display the output for the bivariate relationship between each support group and democratic transition. Models 1 and 2 seem to confirm a positive correlation between party (p < 0.01) and business support (p < 0.001) and the likelihood of democratic transition. In Model 3, military support is negatively associated with the occurrence of a democratic transition (p < 0.001). 26 Model 4 represents the fully specified model accounting for potential confounders. Here, the coefficient for party support turns negative but insignificant. Conversely, the coefficient for business support remains significant at p < 0.001 and increases in magnitude. Military support, instead, slightly decreases but maintains its significance level at p < 0.001. Among the controls, only Electoral Management Body capacity and power distributed by social group are statistically significant (p < 0.001) and positively associated with democratic transition. Table 1. Elite group support and likelihood of democratic transition. Model 1 Model 2 Model 3 Model 4 Outcome: Democratic Transition Party support 0.638** -0.438 (0.238) (0.280) Business support 1.195*** 1.395*** (0.282) (0.350) Military support -1.509*** -1.450*** (0.290) (0.370) Log GDP per capita 0.144 (0.212) GDP growth 0.013 (0.017) Mobilization 0.143 (0.112) EMB capacity 0.492*** (0.142) Power distributed by 1.034*** social group (0.167) Year FEs Ö Ö Ö Ö AIC 5640.1 5554.7 5425.1 3988.2 RMSE 0.29 0.29 0.29 0.24 N 9317 9317 9317 9317 Note: Standard errors clustered by country in parentheses. P-value: p < 0.001***, p < 0.01**, p < 0.05*, p < 0.1+. 27 To better interpret the regression coefficients, Figure 5 displays the average marginal effect of the three key support groups on the likelihood of democratic transition. Marginal effects are used to reflect the change in the predicted probability of the outcome as the predictor increases by one unit, all else equal. The results indicate that support by the military decreases the likelihood of transition by 9.2% (p < 0.001), while support by the business has the opposite effect, increasing the probability associated with democratic transition by 8.9% (p < 0.001). Conversely, party support has an insignificant marginal effect (p = 0.15). Figure 5. Average marginal effects of elite group support with 95% confidence intervals. The lack of significance for party support presents a puzzling result. One possible reason lies in what the variable measures, as it captures regime support by elites belonging to government parties alone. Thus, while it is true that some transition cases have witnessed a strong government spearheading a process of transition to democracy from the very start (Riedl et al., 2020), these preliminary results suggest that elites of government parties are not always prone to democratize and may even represent barriers to transition. This begs the question of whether party elites support exerts a different effect once the transition is already underway. I investigate this point further in the next section. Successful Democratic Transition over Time by Elite Group Support To examine how elite group support influences the likelihood of a successful democratic transition over time, I begin with Kaplan-Meier product-limit estimator, which measures 28 survival probabilities following transition onset. Here, survival is conceptualized as the amount of time a country unit remains in transition – that is, the period between autocracy and the establishment of democracy. Given my interest in successful transitions (i.e., the event occurring), I construct cumulative distribution functions to visualize the probability of a successfully completed transition over time. Figure 6 contrasts cumulative distribution curves by the support status (support versus no support) of party elites and the military. Each curve shows an upward-sloping trend, with each step representing the moment in time when a given country unit completes its transition to democracy. Panel A shows that, once a transition begins, the probability of a successful outcome is higher when party elites support the regime than when they do not. Indeed, after ten years since onset, the cumulative probability of success reaches 23% with party support, compared to 15% without support. Panel B reveals the opposite for military support. When the military remains to back the regime after onset, countries are approximately half as likely to successfully complete their transition after 10 years (11%) compared to cases without military support (20%). Figure 6. Cumulative probabilities of democratic transition by elite group with 95% confidence intervals. Next, I turn to the Cox PH model to test the relationship between elite group support and a successful transition. This semi-parametric model proves well-suited for time-to-event analysis, as it allows to examine both the occurrence of the event and its timing, accounting for possible confounders. Table 2 displays results from two bivariate models and a final multivariate model with control variables. It is important to note that, due to a violation of the proportionality assumption, I exclude the variable for business support from the models. As a result, the corresponding hypothesis cannot be tested within this empirical framework, since including the variable would risk 29 compromising the validity of the model estimates, as confirmed by Table 9 in the Appendix. In addition, I restrict the sample to transition episodes lasting less than 25 years, which corresponds to the empirical breakpoint observed in the cumulative distribution plots. Coefficients are exponentiated to display hazard ratios. HRs enable easy comparison across models as they express whether a variable increases or decreases the likelihood of a successful transition and quantify the effect size. Ratios higher than 1 increase the likelihood of the event occurring at any given point in time, while ratios lower than 1 decrease it. Party support seems to increase the likelihood of a successful transition, with an HR equal to 1.714 in Model 1 and an HR equal 1.578 in Model 3. At any given point in time during the transition, countries where party elites support the regime are about 58% more likely to complete their transition than those without party support. Statistical significance, however, shifts from p < 0.01 in Model 1 to p < 0.05 in Model 4. Regarding military support, instead, Model 2 shows an HR equal to 0.341 (p < 0.05). This implies that, at any given point in time during the transition, countries where the military support the regime are about 66% less likely to complete their transition than those without military support. However, the predictor of military support loses significance when other factors are accounted for in Model 3. Finally, variables related to the economy – GDP per capita (log) and GDP growth – attain significance but their HRs suggest a negative effect on the likelihood of successful transition. Conversely, the capacity of the Electoral Management Body seems to have the opposite effect. 30 Table 2. Elite group support and hazard of successful democratic transition. Model 1 Model 2 Model 3 Outcome: Timing of Successful Democratic Transition Party Support 1.714** 1.578* (0.305) (0.302) Military Support 0.341* 0.539 (0.156) (0.259) Log GDP per capita 0.566*** (0.081) GDP growth 0.926** (0.022) Mobilization 1.062 (0.074) EMB capacity 1.456** (0.190) Power distributed by social 1.142 group (0.138) AIC 1560.3 1561.8 1536.3 RMSE 0.12 0.12 0.12 N 9312 9312 9312 Note: Robust standard errors in parentheses. P-value: p < 0.001***, p < 0.01**, p < 0.05*, p < 0.1+. Figure 7 plots the adjusted cumulative distribution curves from Model 3. Panels A and B reinforce Figure 6 by taking into account the control variables. Although the confidence intervals partially overlap, Panel A and Panel B show that, all else equal, the effects of party and military support on the likelihood of a successful transition have opposite directions as time since onset progresses. 31 Figure 7. Adjusted cumulative probabilities of democratic transition by elite group with 95% confidence intervals. Altogether, logistic and Cox models provide a comprehensive picture of how elite group support influences episodes of democratic transition. Regime support by the military and business elites seem associated with lower and higher overall likelihoods of transition, respectively (see Table 1). This is not true for party support, suggesting a spurious relationship with the overall occurrence of a transition. Yet, once transition begins, time-to-event analysis reveals that party support significantly accelerates the process (see Table 2). This implies that, despite not being positively correlated with the overall occurrence of a transition, regime support by party elites does play an important role once the regime change is underway. Indeed, as the political arena opens up and the quality of elections improves, alternation in government can bring new actors to the forefront who are more committed to democratic reforms. Their support to the changing regime increases the probability of a successful transition within a shorter time period, compared to scenarios lacking party support. Regression Diagnostics As a goodness-of-fit measure of Table 1, I compare Akaike’s information criterion (AIC) across models to see which one best approximates the data. A lower AIC is indicative of better model quality (Harrell , 2015, p. 172). Model 4 displays a much lower AIC than the other bivariate models, suggesting higher predictive power. For Table 2 as well, the fully specified Model 3 has a lower AIC value compared to Models 1 and 2, suggesting that goodness of fit improves once control variables are accounted for. I also test for multicollinearity to check whether some variables are mutually correlated, potentially causing unstable coefficients and inflated standard errors. I compute generalized variance inflation factors (GVIF) on Model 4 of Table 1. Table 10 in the Appendix displays 32 the results. All predictors have GVIF-adjusted values below 1.3, indicating no signs of multicollinearity. Finally, to assess the validity of the Cox regressions, I use the function cox.zph() on Model 3 of Table 2 to test for independence between residuals and time (proportionality assumption). Table 11 in the Appendix confirms insignificant results for each covariate and for the model as a whole, suggesting no signs of violation . As an additional diagnostic check, I produce graphs of the scaled Schoenfeld residuals against time (see Figure 10). Each graph displays a smoothing spline fit with dashed lines representing a +/- 2 standard-error interval around the fit. There is no evidence of systematic departures from the spline fit, implying that the b estimates for each variable are constant over time. Robustness Tests To assess the robustness of the findings, I re-estimate the models using two different operationalizations of the elite group support variables. The first follows Knutsen et al.’s (2025) approach and lowers the threshold for determining support from ³ 0.75 to ³ 0.50. By including middling scores around 0.5 (or 50%), this strategy aims to account for those groups which appear to exert “an intermediate degree of political influence” (Knutsen et al., 2025, p. 12) at a given point in time. The second, instead, avoids dichotomization altogether and maintains the original continuous scale of the variable, as produced by V-Dem’s cross-coder mean aggregation (0–1). Rather than imposing a binary distinction of support versus non- support, this alternative strategy allows for a more granular understanding of degree of group support as well as of inter-coder agreement. Table 3 presents the results of the logistic regressions using 50% as a threshold to determine group support. The coefficients for the three focal groups do not seem to substantially vary from Table 1, although the effect magnitude for business and military support in Model 4 is slightly lower than in the respective fully specified model of Table 1. EMB capacity and power by social group remain significant and positively correlated with democratic transition. 33 Table 3. Elite support group (50%) and likelihood of democratic transition. Model 1 Model 2 Model 3 Model 4 Outcome: Democratic Transition Party support 0.842** -0.306 (0.262) (0.362) Business support 1.216*** 1.066*** (0.218) (0.291) Military support -1.732*** -1.366*** (0.233) (0.290) Log GDP per capita 0.039 (0.230) GDP growth 0.004 (0.017) Mobilization 0.133 (0.115) EMB capacity 0.538*** (0.147) Power distributed by 0.905*** social group (0.166) Year FEs Ö Ö Ö Ö AIC 5593.8 5424.5 5190.1 3927.3 RMSE 0.29 0.29 0.28 0.24 N 9317 9317 9317 9317 Note: Standard errors clustered by country in parentheses. P-value: p < 0.001***, p < 0.01**, p < 0.05*, p < 0.1+. Table 4 re-estimates the models using the continuous version of the variable. In Model 4 we can see that, while party support remains insignificant, the coefficients for business and military 34 support increase in size (p < 0.001). Standard errors, however, increase as well from Table 3, indicating lower levels of precision in the regression estimates. Table 4. Elite support group (continuous) and likelihood of democratic transition. Model 1 Model 2 Model 3 Model 4 Outcome: Democratic Transition Party support 1.657*** -0.369 (0.389) (0.578) Business support 2.519*** 2.748*** (0.457) (0.592) Military support -3.479*** -3.378*** (0.440) (0.512) Log GDP per capita -0.026 (0.241) GDP growth 0.008 (0.019) Mobilization 0.175 (0.121) EMB capacity 0.523** (0.163) Power distributed by 0.954*** social group (0.172) Year FEs Ö Ö Ö Ö AIC 5513.4 5334.5 5034-1 3636.9 RMSE 0.29 0.28 0.27 0.23 N 9317 9317 9317 9317 Note: Standard errors clustered by country in parentheses. P-value: p < 0.001***, p < 0.01**, p < 0.05*, p < 0.1+. 35 Turning to time-to-event analysis, Table 5 shows the output of Cox regressions using the 50% threshold of support. Compared to Model 3 of Table 2, party support shows a higher HR (74%) at p < 0.05, while military support remains insignificant. GDP growth still suggests a significant negative effect on the likelihood of successful democratic transition. Power by social group also achieves significance (p < 0.05) and is shown to increase the likelihood of success at any point in time during the transition. The covariate for GDP per capita (log) is excluded due to a violation of the proportionality assumption. Table 5. Elite group support (50%) and hazard of successful democratic transition. Model 1 Model 2 Model 3 Outcome: Timing of Successful Democratic Transition Party support 2.239** 1.743* (0.571) (0.466) Military support 0.765 0.838 (0.168) (0.189) GDP growth 0.937** (0.022) Mobilization 1.005 (0.072) EMB capacity 1.095 (0.121) Power distributed by social group 1.292* (0.144) AIC 1557.6 1568.1 1550.9 RMSE 0.12 0.12 0.12 N 9312 9312 9312 Note: Standard errors clustered by country in parentheses. P-value: p < 0.001***, p < 0.01**, p < 0.05*, p < 0.1+. 36 Table 6 re-estimates the Cox models for continuous support. Model 3 shows that party support increases its magnitude (114%), however the significance level passes from p < 0.05 to p < 0.1. For the first time, the fully specified model displays for military support an HR equal to 0.464 and significant at p < 0.05. This implies that, at any given point in time during the transition, countries where the military supports the regime are about 54% less likely to complete their transition than those without military support. GDP growth and power by social group remain significant. Table 6. Elite group support (continuous) and hazard of successful democratic transition. Model 1 Model 2 Model 3 Outcome: Timing of Successful Democratic Transition Party support 3.532*** 2.144+ (1.373) (0.892) Military support 0.494* 0.464* (0.174) (0.180) GDP growth 0.939** (0.022) Mobilization 1.054 (0.076) EMB capacity 1.121 (0.122) Power distributed by social group 1.233+ (0.144) AIC 1558.4 1565.5 1549.0 RMSE 0.12 0.12 0.12 N 9312 9312 9312 Note: Standard errors clustered by country in parentheses. P-value: p < 0.001***, p < 0.01**, p < 0.05*, p < 0.1+. 37 Discussion and Limitations The logistic regression analysis provides robust evidence for Hypothesis 1: “Regime support by the military is negatively associated with the occurrence of a democratic transition.” Specifically, it finds that military support decreases the probability of a democratic transition by 9.2%. The analysis also shows robust support for Hypothesis 5: “Regime support by the economic elite is positively associated with the occurrence of a democratic transition,” as business elite support increases the probability of democratic transition by 8.9%. However, the association between party elite support and the probability of transition results statistically insignificant, failing to confirm Hypothesis 3: “Regime support by the political elite is positively associated with the occurrence of a democratic transition.” Turning to the Cox analysis, I find robust evidence to confirm Hypothesis 4: “During democratic transition, higher levels of regime support by the political elite are associated with higher likelihood of successful democratic transition.” In particular, at any given point during the transition, countries where party elites support the regime are approximately 58% more likely to complete their transition successfully compared to those without party support. This suggests that, while party support may not trigger democratization, it does sustain transitions once they are ongoing. There is mixed evidence, however, on Hypothesis 2: “During democratic transition, lower levels of regime support by the military are associated with higher likelihood of successful democratic transition.” Indeed, alternative model specifications provide some grounds for expecting a negative association (see Table 6), but this relationship lacks consistent significance. Finally, due to a violation of the proportionality assumption, I am unable to test Hypothesis 6: “During democratic transition, higher levels of regime support by the economic elite are associated with higher likelihood of successful democratic transition.” Next, I acknowledge three key limitations. First, at a general level, this study relies on a theory which explains democratic transition as an elite-driven affair. Although this limits its applicability to other types of transition scenarios, the study does not discount the importance of other dynamics initiated by non-elite actors – such as civil society organizations – or the potentially relevant interactions between domestic elite groups and foreign actors (see Tolstrup, 2013). Second, causal inferences must be made with extreme caution, as this study does not rely on an experimental design. Moreover, issues of reverse causality must be considered, as it is plausible that democratization itself may influence elite decisions to support the regime. 38 Finally, the inability to test the role of business elite support during transition episodes calls for alternative statistical models capable of capturing time-varying effects more accurately. Conclusion This thesis aims to answer the question “What role does the support of elite groups play in sustaining democratic transition?” Therefore, I develop a testable theory on the patterns of regime support for the military, political, and economic elites. Empirically, I use a two-pronged strategy using logistic regressions and time-to-event analysis on a dataset covering 135 episodes of democratic transition between 1900 and 2019. I find that regime support by the military decreases the probability of transition by 9.2%, while support by the business elites has the opposite effect, increasing it by 8.9%. The association between party elite support and overall likelihood of transition is not significant. However, time-to-event analysis shows that party elite support does sustain transition once the process is underway: At any given point during a transition episode, countries where party elites support the regime are approximately 58% more likely to complete their transition successfully compared to those without party support. While I find mixed evidence for military support, I am unable to test the effect of business elite support due to a violation of model assumptions. This study provides new insights into the phenomenon of democratic transition. In particular, it shows that elites differ in how they impact democratization, suggesting that they function less as a unified bloc and more as a coalition of actors with divergent interests. While military support appears to obstruct democratic transitions, business elites may favor the establishment of democratic institutions, and party elites emerge as crucial actors sustaining democratization once transitions are ongoing. These findings add more nuance to the actor-centered literature on democratization by raising the possibility that – in addition to elite schisms – elite realignment, particularly between party and business elites, can also play a role in advancing democratization. At the same time, this study acknowledges (but does not empirically test) that some patterns of elite regime support may be vulnerable to crises of reversibility. The military, for instance, may initially withdraw from politics but later reassert its influence in a coup-like fashion unless civilian supremacy is successfully achieved over all areas of policymaking. In conclusion, this thesis opens two main avenues for future research. The first is empirical: Future studies should more closely examine the dynamics of business elite support and its effects on democratic transition. The second is broader and calls for a deeper exploration of the 39 possible interactions between elite actors and other social forces. Expanding the analytical scope to include new actors – such as state bureaucracies, middle and working classes, or ethnic groups – could further enhance our understanding of one of the most transformative processes in political life. 40 Bibliography Acemoglu, D. et al. (2019) ‘Democracy Does Cause Growth’, Journal of Political Economy, 127(1), pp. 47–100. Available at: https://doi.org/10.1086/700936. Acemoglu, D. and Robinson, J.A. (2005) Economic Origins of Dictatorship and Democracy. Cambridge: Cambridge University Press. Available at: https://doi.org/10.1017/CBO9780511510809. Agüero, F. (1995) Soldiers, Civilians, and Democracy: Post-Franco Spain in Comparative Perspective. Johns Hopkins University Press. Albertus, M. and Menaldo, V. (2014) ‘Gaming Democracy: Elite Dominance during Transition and the Prospects for Redistribution’, British Journal of Political Science, 44(3), pp. 575–603. Available at: https://doi.org/10.1017/S0007123413000124. Albertus, M. and Menaldo, V. (2018) Authoritarianism and the Elite Origins of Democracy. Cambridge: Cambridge University Press. Available at: https://doi.org/10.1017/9781108185950. Albertus, M. and Menaldo, V.A. (2010) ‘If You’re Against Them You’re with Us: The Effect of Expropriation on Autocratic Survival’. Rochester, NY: Social Science Research Network. Available at: https://papers.ssrn.com/abstract=1601138 (Accessed: 21 March 2025). Angiolillo, F., Wiebrecht, F. and Lindberg, S.I. (2025) ‘Party Systems, Democratic Positions, and Regime Changes: Introducing the Party-System Democracy Index’, British Journal of Political Science, 55, p. e42. Available at: https://doi.org/10.1017/S0007123424000887. Aydınlı, E. (2012) ‘Turkey Under the AKP: Civil-Military Relations Transformed’, Journal of Democracy, 23(1), pp. 100–108. Barro, R.J. (1999) ‘Determinants of Democracy’, Journal of Political Economy, 107(S6), pp. S158–S183. Available at: https://doi.org/10.1086/250107. Bartlett, D.L. (1996) ‘Democracy, Institutional Change, and Stabilisation Policy in Hungary’, Europe-Asia Studies, 48(1), pp. 47–83. Bello, W. and Gershman, J. (1990) ‘Democratization and Stabilization in the Philippines’, Critical Sociology, 17(1), pp. 35–56. Available at: https://doi.org/10.1177/089692059001700102. Bernhard, M. et al. (2020) ‘Parties, Civil Society, and the Deterrence of Democratic Defection’, Studies in Comparative International Development, 55(1), pp. 1–26. Available at: https://doi.org/10.1007/s12116-019-09295-0. Besley, T. and Robinson, J.A. (2010) ‘Quis Custodiet Ipsos Custodes? Civilian Control over the Military’, Journal of the European Economic Association, 8(2–3), pp. 655–663. Available at: https://doi.org/10.1111/j.1542-4774.2010.tb00535.x. 41 Boix, C., Miller, M. and Rosato, S. (2013) ‘A Complete Data Set of Political Regimes, 1800– 2007’, Comparative Political Studies, 46(12), pp. 1523–1554. Available at: https://doi.org/10.1177/0010414012463905. Boix, C. and Stokes, S.C. (2003) ‘Endogenous Democratization’, World Politics, 55(4), pp. 517–549. Available at: https://doi.org/10.1353/wp.2003.0019. Bollen, K.A. and Jackman, R.W. (1989) ‘Democracy, Stability, and Dichotomies’, American Sociological Review, 54(4), pp. 612–621. Available at: https://doi.org/10.2307/2095882. Bradburn, M.J. et al. (2003) ‘Survival Analysis Part II: Multivariate data analysis – an introduction to concepts and methods’, British Journal of Cancer, 89(3), pp. 431–436. Available at: https://doi.org/10.1038/sj.bjc.6601119. Chin, J.J., Carter, D.B. and Wright, J.G. (2021) ‘The Varieties of Coups D’état: Introducing the Colpus Dataset’, International Studies Quarterly, 65(4), pp. 1040–1051. Available at: https://doi.org/10.1093/isq/sqab058. Collier, D. and Adcock, R. (1999) ‘DEMOCRACY AND DICHOTOMIES: A Pragmatic Approach to Choices about Concepts’, Annual Review of Political Science, 2(Volume 2, 1999), pp. 537–565. Available at: https://doi.org/10.1146/annurev.polisci.2.1.537. Coppedge, M. et al. (eds) (2022) Why Democracies Develop and Decline. Cambridge: Cambridge University Press. Available at: https://doi.org/10.1017/9781009086974. Coppedge, M., Gerring, J., Knutsen, C.H., Lindberg, S.I., Teorell, J., Altman, D., et al. (2025) ‘V-Dem Codebook v15’, Varieties of Democracy (V-Dem) Project [Preprint]. Coppedge, M., Gerring, J., Knutsen, C.H., Lindberg, S.I., Teorell, J., Marquardt, K.L., et al. (2025) ‘V-Dem Methodology v15’, Varieties of Democracy (V-Dem) Project [Preprint]. Dahl, R.A. (1971) Polyarchy; participation and opposition. New Haven : Yale University Press. Available at: http://archive.org/details/polyarchypartici0000dahl (Accessed: 21 February 2025). Dahl, R.A. (1998) On democracy. New Haven: Yale University Press. Diamond, L. (1997) ‘Consolidating Democracy in the Americas’, The Annals of the American Academy of Political and Social Science, 550, pp. 12–41. Diamond, L. (2002) ‘Elections Without Democracy: Thinking About Hybrid Regimes’, Journal of Democracy, 13(2), pp. 21–35. Díaz-Cayeros, A. et al. (2015) ‘Authoritarian Survival and Poverty Traps: Land Reform in Mexico [Feb. 2015]’, CDDRL Working Papers [Preprint]. Available at: http://cddrl.fsi.stanford.edu/publication/authoritarian-survival-and-poverty-traps-land- reform-mexico (Accessed: 22 March 2025). 42 Djuve, V.L., Knutsen, C.H. and Wig, T. (2020) ‘Patterns of Regime Breakdown Since the French Revolution’, Comparative Political Studies, 53(6), pp. 923–958. Available at: https://doi.org/10.1177/0010414019879953. Donno, D. (2013) ‘Elections and Democratization in Authoritarian Regimes’, American Journal of Political Science, 57(3), pp. 703–716. Available at: https://doi.org/10.1111/ajps.12013. Doucouliagos, H. and Ulubaşoğlu, M.A. (2008) ‘Democracy and Economic Growth: A Meta- Analysis’, American Journal of Political Science, 52(1), pp. 61–83. Available at: https://doi.org/10.1111/j.1540-5907.2007.00299.x. Easton, D. (1957) ‘An Approach to the Analysis of Political Systems’, World Politics, 9(3), pp. 383–400. Available at: https://doi.org/10.2307/2008920. Edgell, A.B. et al. (2025a) ‘Episodes of Regime Transformation Dataset (v15.0)’. Varieties of Democracy (V-Dem) Project. Available at: www.github.com/vdeminstitute/ert. Edgell, A.B. et al. (2025b) ‘Episodes of Regime Transformation Dataset (v15.0) Codebook’. Varieties of Democracy (V-Dem) Project. Available at: www.github.com/vdeminstitute/ert. Fariss, C.J. et al. (2022) ‘New Estimates of Over 500 Years of Historic GDP and Population Data’, Journal of Conflict Resolution, 66(3), pp. 553–591. Available at: https://doi.org/10.1177/00220027211054432. Field, G.L., Higley, J. and Burton, M.G. (1990) ‘A New Elite Framework for Political Sociology’, Revue européenne des sciences sociales, 28(88), pp. 149–182. Fish, M.S. (2002) ‘Islam and Authoritarianism’, World Politics, 55(1), pp. 4–37. Available at: https://doi.org/10.1353/wp.2003.0004. Geddes, B. (1999) ‘What Do We Know About Democratization After Twenty Years?’, Annual Review of Political Science, 2(Volume 2, 1999), pp. 115–144. Available at: https://doi.org/10.1146/annurev.polisci.2.1.115. Geddes, B., Wright, J. and Frantz, E. (2014) ‘Autocratic Breakdown and Regime Transitions: A New Data Set’, Perspectives on Politics, 12(2), pp. 313–331. Available at: https://doi.org/10.1017/S1537592714000851. Geddes, B., Wright, J. and Frantz, E. (2018) How Dictatorships Work: Power, Personalization, and Collapse. 1st edn. Cambridge University Press. Available at: https://doi.org/10.1017/9781316336182. Grewal, S. (2023) Soldiers of Democracy? Military Legacies and the Arab Spring | Oxford Academic. Online edition. Oxford University Press. Available at: https://academic.oup.com/book/46755 (Accessed: 12 March 2025). Grzymala-Busse, A. (2020) ‘Consequences of Authoritarian Party Exit and Reinvention for Democratic Competition’, Comparative Political Studies, 53(10–11), pp. 1704–1737. Available at: https://doi.org/10.1177/0010414019897683. 43 Grzymala-Busse, A.M. (2002) Redeeming the Communist Past: The Regeneration of Communist Parties in East Central Europe. Cambridge: Cambridge University Press (Cambridge Studies in Comparative Politics). Available at: https://doi.org/10.1017/CBO9780511613388. Haggard, S. and Kaufman, R.R. (2016) Dictators and Democrats: Masses, Elites, and Regime Change. Princeton University Press. Available at: https://press.princeton.edu/books/hardcover/9780691172149/dictators-and-democrats (Accessed: 22 May 2025). Hagopian, F. (1990) ‘“Democracy by Undemocratic Means”?: Elites, Political Pacts, and Regime Transition in Brazil’, Comparative Political Studies, 23(2), pp. 147–170. Available at: https://doi.org/10.1177/0010414090023002001. Harrell , F.E. (2015) Regression Modeling Strategies: With Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis. Cham: Springer International Publishing (Springer Series in Statistics). Available at: https://doi.org/10.1007/978-3- 319-19425-7. Higley, J., Field, G.L. and Groholt, K. (1976) Elite Structure and Ideology: A Theory with Applications to Norway. New York: Columbia University Press. Huntington, S.P. (1991) ‘Democracy’s Third Wave’, Journal of Democracy, 2(2), pp. 12–34. Imtiaz Ahmed, S. (2006) ‘Civilian supremacy in democracies with “fault lines”: The role of the parliamentary standing committee on defence in Bangladesh’, Democratization, 13(2), pp. 283–302. Available at: https://doi.org/10.1080/13510340500523994. Jonas, S. (2000) ‘Democratization through Peace: The Difficult Case of Guatemala’, Journal of Interamerican Studies and World Affairs, 42(4), pp. 9–38. Available at: https://doi.org/10.2307/166340. Joseph A Schumpeter (1942) Capitalism Socialism And Democracy. Available at: http://archive.org/details/in.ernet.dli.2015.190072 (Accessed: 21 February 2025). Karl, T.L. (1987) ‘Petroleum and Political Pacts: The Transition to Democracy in Venezuela’, Latin American Research Review, 22(1), pp. 63–94. Karl, T.L. (1990) ‘Dilemmas of Democratization in Latin America’, Comparative Politics, 23(1), p. 1. Available at: https://doi.org/10.2307/422302. Knutsen, C.H. et al. (2019) ‘Economic development and democracy: An electoral connection’, European Journal of Political Research, 58(1), pp. 292–314. Available at: https://doi.org/10.1111/1475-6765.12282. Knutsen, C.H. et al. (2025) ‘Behind the Throne: Regime support coalitions around the world, 1789-2020’, British Journal of Political Science, 55, p. e37. Available at: https://doi.org/10.1017/S0007123424000656. Kosmidis, I. and Firth, D. (2009) ‘Bias reduction in exponential family nonlinear models’, Biometrika, 96(4), pp. 793–804. 44 Lewis, S.E. (no date) ‘Mexico’s Democratic Transition and Consolidation since 1977’, in W. Beezley (ed.) The Oxford Handbook of Mexican History. Oxford University Press, p. 0. Available at: https://doi.org/10.1093/oxfordhb/9780190699192.013.35. Linz, J.J. (1990) ‘Transitions to Democracy’, The Washington Quarterly, 13(3), pp. 143–164. Available at: https://doi.org/10.1080/01636609009445400. Linz, J.J. and Stepan, A. (1978) The Breakdown of Democratic Regimes. Johns Hopkins University Press. Available at: https://doi.org/10.56021/9780801820090. Linz, J.J. and Stepan, A. (1996) Problems of Democratic Transition and Consolidation. Johns Hopkins University Press. Available at: https://doi.org/10.56021/9780801851575. Lipset, S.M. (1959) ‘Some Social Requisites of Democracy: Economic Development and Political Legitimacy’, American Political Science Review, 53(1), pp. 69–105. Available at: https://doi.org/10.2307/1951731. Lührmann, A., Tannenberg, M. and Lindberg, S.I. (2018) ‘Regimes of the World (RoW): Opening New Avenues for the Comparative Study of Political Regimes’, Politics and Governance, 6(1), pp. 60–77. Maerz, S.F. et al. (2024) ‘Episodes of regime transformation’, Journal of Peace Research, 61(6), pp. 967–984. Available at: https://doi.org/10.1177/00223433231168192. Mainwaring, S. (1998) ‘Party Systems in the Third Wave’, Journal of Democracy, 9(3), pp. 67–81. Maniruzzaman, T. (1992) ‘The Fall of the Military Dictator: 1991 Elections and the Prospect of Civilian Rule in Bangladesh’, Pacific Affairs, 65(2), pp. 203–224. Available at: https://doi.org/10.2307/2760169. Mills, C.W. (1956) The Power Elite. Oxford University Press, USA. Morgenbesser, L. and Pepinsky, T.B. (2019) ‘Elections as Causes of Democratization: Southeast Asia in Comparative Perspective’, Comparative Political Studies, 52(1), pp. 3–35. Available at: https://doi.org/10.1177/0010414018758763. Nassif, H.B. (2020) ‘Military Politics and Democratic Transition: Combining Rationality, Culture, and Structure’, in Oxford Research Encyclopedia of Politics. Available at: https://doi.org/10.1093/acrefore/9780190228637.013.1934. Nelson, J.M. (2014) ‘Will Malaysia Follow the Path of Taiwan and Mexico?’, Journal of Democracy, 25(3), pp. 105–119. Nord, M. et al. (2024) ‘Democracy Winning and Losing at the Ballot: Democracy Report 2024’. Rochester, NY: Social Science Research Network. Available at: https://doi.org/10.2139/ssrn.4774409. North, D.C. and Weingast, B.R. (1989) ‘Constitutions and Commitment: The Evolution of Institutions Governing Public Choice in Seventeenth-Century England’, The Journal of Economic History, 49(4), pp. 803–832. Available at: https://doi.org/10.1017/S0022050700009451. 45 O’Donnell, G.A. (1996) ‘Illusions About Consolidation’, Journal of Democracy, 7(2), pp. 34–51. O’Donnell, G.A. and Schmitter, P.C. (1986) Transitions from Authoritarian Rule: Tentative conclusions about uncertain democracies. Johns Hopkins University Press. Available at: https://cadmus.eui.eu/handle/1814/22062 (Accessed: 4 February 2025). Przeworski, A. (1991) Democracy and the Market: Political and Economic Reforms in Eastern Europe and Latin America. Cambridge: Cambridge University Press (Studies in Rationality and Social Change). Available at: https://doi.org/10.1017/CBO9781139172493. Przeworski, A. and Wallerstein, M. (1982) ‘The Structure of Class Conflict in Democratic Capitalist Societies’, American Political Science Review, 76(2), pp. 215–238. Available at: https://doi.org/10.2307/1961105. Rajan, R.G. (2009) ‘Rent Preservation and the Persistence of Underdevelopment’, American Economic Journal: Macroeconomics, 1(1), pp. 178–218. Available at: https://doi.org/10.1257/mac.1.1.178. Reuter, O.J. and Szakonyi, D. (2019) ‘Elite Defection under Autocracy: Evidence from Russia’, American Political Science Review, 113(2), pp. 552–568. Available at: https://doi.org/10.1017/S0003055419000030. Riedl, R.B. et al. (2020) ‘Authoritarian-Led Democratization’, Annual Review of Political Science, 23(Volume 23, 2020), pp. 315–332. Available at: https://doi.org/10.1146/annurev-polisci-052318-025732. del Río, A. and Higashijima, M. (2024) ‘Democratic Reforms in Dictatorships: Elite Divisions, Party Origins, and the Prospects of Political Liberalization’, Comparative Political Studies, p. 00104140241302772. Available at: https://doi.org/10.1177/00104140241302772. Rustow, D.A. (1970) ‘Transitions to Democracy: Toward a Dynamic Model’, Comparative Politics, 2(3), pp. 337–363. Available at: https://doi.org/10.2307/421307. Sato, Y. and Wahman, M. (2019) ‘Elite coordination and popular protest: the joint effect on democratic change’, Democratization [Preprint]. Available at: https://www.tandfonline.com/doi/abs/10.1080/13510347.2019.1645127 (Accessed: 6 February 2025). Schedler, A. (1998) ‘What is Democratic Consolidation?’, Journal of Democracy, 9(2), pp. 91–107. Schedler, A. (2013) ‘The Politics of Uncertainty: Sustaining and Subverting Electoral Authoritarianism’, in A. Schedler (ed.). Oxford University Press, p. 0. Available at: https://doi.org/10.1093/acprof:oso/9780199680320.003.0006. Shaw, W.H. (2003) ‘“They Stole Our Land”: debating the expropriation of white farms in Zimbabwe’, The Journal of Modern African Studies, 41(1), pp. 75–89. Available at: https://doi.org/10.1017/S0022278X02004159. 46 Slater, D. and Wong, J. (2013) ‘The Strength to Concede: Ruling Parties and Democratization in Developmental Asia’, Perspectives on Politics, 11(3), pp. 717– 733. Stokes, S.C. (1999) ‘Political Parties and Democracy’, Annual Review of Political Science, 2(Volume 2, 1999), pp. 243–267. Available at: https://doi.org/10.1146/annurev.polisci.2.1.243. Svolik, M.W. (2012) The Politics of Authoritarian Rule. Cambridge: Cambridge University Press (Cambridge Studies in Comparative Politics). Available at: https://doi.org/10.1017/CBO9781139176040. Teorell, J. (2010) Determinants of Democratization: Explaining Regime Change in the World, 1972–2006. Cambridge: Cambridge University Press. Available at: https://doi.org/10.1017/CBO9780511762727. Teorell, J. et al. (2019) ‘Measuring Polyarchy Across the Globe, 1900–2017’, Studies in Comparative International Development, 54(1), pp. 71–95. Available at: https://doi.org/10.1007/s12116-018-9268-z. Teorell, J. and Wahman, M. (2018) ‘Institutional stepping stones for democracy: how and why multipartyism enhances democratic change’, Democratization, 25(1), pp. 78–97. Available at: https://doi.org/10.1080/13510347.2017.1318279. Tolstrup, J. (2013) ‘When can external actors influence democratization? Leverage, linkages, and gatekeeper elites’, Democratization, 20(4), pp. 716–742. Available at: https://doi.org/10.1080/13510347.2012.666066. Treisman, D. (2020) ‘Democracy by Mistake: How the Errors of Autocrats Trigger Transitions to Freer Government’, American Political Science Review, 114(3), pp. 792–810. Available at: https://doi.org/10.1017/S0003055420000180. Wright, J. and Escribà-Folch, A. (2012) ‘Authoritarian Institutions and Regime Survival: Transitions to Democracy and Subsequent Autocracy’, British Journal of Political Science, 42(2), pp. 283–309. Available at: https://doi.org/10.1017/S0007123411000317. Ziblatt, D. (2017) Conservative Parties and the Birth of Democracy. Cambridge: Cambridge University Press (Cambridge Studies in Comparative Politics). Available at: https://doi.org/10.1017/9781139030335. 47 Appendix Table 7. Descriptive summary of the variables Mean SD Min Max N Democratic transition 0.08 0.27 0 1 13182 Event 0.01 0.11 0 1 13182 Electoral democracy index 0.26 0.26 0.01 0.92 26775 Transition year 0.25 1.65 0 27 27913 Party support 0.26 0.44 0 1 26574 Business support 0.18 0.38 0 1 26574 Military support 0.28 0.45 0 1 26574 Log GDP per capita 1.91 1.10 -0.42 5.57 22191 GDP growth 2.83 3.80 -34.50 61.60 21981 Mobilization -0.53 1.34 -3.57 4.06 19115 EMB capacity -0.28 1.63 -3.19 3.32 25225 Power by social group -0.18 1.43 -2.79 3.33 27709 48 Figure 8. Distribution of regime support across all social groups (1789–2024). 49 Figure 9. Distribution of transition lengths. 50 Table 8. Episodes of democratic transition by country (1900–2024). Transition Transition Country Country Period Period Albania 1998–2006 Canada 1919–1922 Argentina 1963–1964 Cape Verde 1981–1996 Argentina 1983–1984 Chile 1958–1964 Armenia 2010–2019 Chile 1988–1994 Austria 1918–1921 Colombia 1990–1992 Bangladesh 1986–1997 Costa Rica 1950–1954 Barbados 1944–1969 Croatia 1992–2010 Belarus 1991–1992 Cyprus 1970–1991 Belgium 1918–1921 Czechia 1920–1921 Belgium 1944–1950 Czechia 1990–1991 Benin 1990–1995 Denmark 1901–1902 Bhutan 2007–2009 Denmark 1945–1946 Bolivia 1982–1995 Dominican Republic 1978–1983 Bosnia and Dominican Republic 1995–2000 1996–2001 Herzegovina Ecuador 1978–1980 Botswana 1959–1970 El Salvador 1992–2000 Brazil 1975–1992 Estonia 1919–1927 Bulgaria 1990–2002 Estonia 1993–1993 Burkina Faso 1991–2000 Fiji 1963–1973 Burkina Faso 2016–2016 Fiji 1992–1997 51 Transition Transition Country Country Period Period Finland 1917–1920 Jamaica 1953–1956 France 1945–1948 Jamaica 1984–2003 Georgia 2004–2004 Japan 1947–1955 German Democratic Kenya 2010–2014 1990–1990 Republic Kosovo 2000–2003 Germany 1919–1921 Latvia 1922–1923 Ghana 1992–2017 Lesotho 1990–1993 Greece 1974–1982 Lesotho 2002–2003 Guatemala 1984–2008 Liberia 2005–2007 Guinea-Bissau 2014–2019 Lithuania 1919–1921 Guyana 1986–2003 Luxembourg 1918–1920 Honduras 1990–1995 Luxembourg 1945–1947 Hungary 1988–1991 Madagascar 1990–1994 Iceland 1904–1905 Malawi 1992–1995 India 1950–1953 Malawi 2009–2017 India 1977–1978 Malawi 2020–2021 Indonesia 1998–2005 Maldives 2005–2009 Ireland 1921–1928 Maldives 2018–2024 Israel 1949–1950 Mali 1992–1993 Italy 1944–1960 Malta 1947–1948 Ivory Coast 2010–2017 Malta 1962–1964 52 Transition Transition Country Country Period Period Mauritius 1968–1969 Panama 1979–1992 Mexico 1988–2004 Papua New Guinea 1970–1975 Moldova 1991–1995 Paraguay 1989–1993 Moldova 2009–2011 Peru 1978–1981 Mongolia 1990–1993 Peru 1995–2002 Montenegro 2020–2024 Philippines 1986–1990 Namibia 1980–1995 Poland 1919–1920 Nepal 2014–2016 Poland 1989–1992 Netherlands 1918–1920 Portugal 1970–1983 Netherlands 1945–1949 Romania 1990–1997 Nicaragua 1984–1991 Sao Tome and 1987–1992 Principe Niger 1993–1994 Senegal 1990–1994 Niger 2000–2005 Serbia 2000–2003 Niger 2011–2012 Seychelles 2013–2024 Nigeria 2010–2015 Sierra Leone 1999–2003 North Macedonia 1994–1998 Slovenia 1990–1992 North Macedonia 2002–2003 Solomon Islands 1960–1982 North Macedonia 2017–2019 Solomon Islands 2007–2024 Norway 1906–1906 South Africa 1990–2005 Norway 1945–1946 South Korea 1982–2000 Palestine/West Bank 1994–2005 53 Transition Transition Country Country Period Period Spain 1931–1932 Tunisia 2011–2012 Spain 1976–1982 Türkiye 1962–1967 Sri Lanka 1947–1950 Türkiye 1983–1992 Sri Lanka 1994–1995 Ukraine 1991–1994 Sri Lanka 2010–2018 Ukraine 2005–2007 Suriname 1949–1955 Ukraine 2016–2020 Suriname 1992–2000 United Kingdom 1919–1923 Sweden 1917–1922 United States of 1919–1923 America Taiwan 1986–2001 Uruguay 1911–1925 Thailand 1992–2001 Uruguay 1934–1947 The Gambia 2017–2018 Uruguay 1980–1995 Timor-Leste 1998–2024 Vanuatu 1970–1980 Trinidad and Tobago 1956–1967 Venezuela 1958–1961 54 Table 9. Cox proportionality assumption for a model including business support. Variable Chi-squared df p-value party_support 0.0442 1 0.8336 business_support 8.8526 1 0.0029 military_support 0.5133 1 0.4737 log_gdppc 3.5714 1 0.0588 gdp_growth 0.2515 1 0.6160 v2cagenmob 0.0898 1 0.7644 v2elembcap 0.8194 1 0.3654 v2pepwrsoc 0.2230 1 0.6368 GLOBAL 22.1087 1 0.0047 55 Table 10. Multicollinearity. Variable GVIF df GVIF^(1/(2*df)) party_support 1.149993 1 1.072377 business_support 1.150604 1 1.072662 military_support 1.114465 1 1.055682 log_gdppc 1.565768 1 1.251306 gdp_growth 1.201382 1 1.096076 v2cagenmob 1.121770 1 1.059136 v2elembcap 1.512092 1 1.229672 v2pepwrsoc 1.184012 1 1.088123 factor(year) 1.551694 119 1.001848 56 Table 11. Cox proportionality assumption. Variable Chi-squared df p-value party_support 0.0442 1 0.833 military_support 0.5071 1 0.476 log_gdppc 3.5707 1 0.059 gdp_growth 0.2527 1 0.615 v2cagenmob 0.0905 1 0.764 v2elembcap 0.8208 1 0.365 v2pepwrsoc 0.2193 1 0.640 GLOBAL 9.8232 7 0.199 57 Figure 10. Scaled Schoenfeld residuals. 58