Does the presence of multiple political parties in the legislature facilitate the survival of authoritarian regimes, or is multipartism a sign of authoritarian weakness? This question lies at the core of major debates. Early studies of authoritarianism hypothesized that some degree of legislative pluralism helps dictators survive (Gandhi, Reference Gandhi2008), but others have argued that the existence of multiple parties undermines authoritarian stability (Svolik, Reference Svolik2012) and allows opposition forces to promote democratic change (Teorell and Wahman, Reference Teorell and Wahman2017). This question also informs urgent debates about opposition unity in authoritarian regimes.
This article shows that multipartism—even when designed purposefully by authoritarian incumbents—is an important driver of democratization. We make two contributions. First, we theorize the adoption of preemptive multipartism by authoritarian regimes. In doing so, we extend the literature on how nominally democratic institutions contribute to authoritarian survival (Gandhi and Przeworski, Reference Gandhi and Przeworski2006; Svolik, Reference Svolik2012; Pepinsky, Reference Pepinsky2014). Second, we show that this preemptive strategy ultimately leads to protracted top-down transitions. Authoritarian concessions deflect bottom-up democratization but cumulatively undermine the durability of dictatorship over time. In making this claim, we suggest that the literature has overestimated the importance of opposition fractionalization as an obstacle to regime change (Howard and Roessler, Reference Howard and Roessler2006; Arriola, Reference Arriola2013; Donno, Reference Donno2013; Wahman, Reference Wahman2013).
Extant studies invoke two opposite mechanisms to link multipartism and transitions to democracy. On the one hand, party competition allows for the possibility of a “stunning election” in which democratic forces surprisingly defeat authoritarian incumbents (Huntington, Reference Huntington1991). On the other hand, competition reassures authoritarian incumbents that they will be ready to survive a future democratic election (Wright and Escribà-Folch, Reference Wright and Escribà-Folch2012). Yet, multipartism may prompt democratization in the absence of a stunning defeat, and even when the incumbent does not expect to retain power after the transition.
We introduce a third possibility: concessions in the form of legislative multipartism prevent bottom-up democratization in the short run, but reduce the viability of dictatorship over the long run. The result is a protracted top-down transition, in which a sequence of locally rational decisions from the perspective of authoritarian incumbents ultimately creates conditions for regime change. This conclusion aligns with Treisman’s argument about a “slippery slope” of concessions that strengthen the opposition and helps explain intra-regime splits and the surprising delegation of power to regime insiders who pursue democratization (Treisman, Reference Treisman2020). We offer formal proof that protracted top-down transitions do not require miscalculation on the part of authoritarian rulers; rather, they reflect long-term processes that shift the equilibrium at the core of the authoritarian regime (Lindberg, Reference Lindberg and Lindberg2009).
The distinctive empirical implication of our argument is that authoritarian regimes with a greater number of legislative parties—reflecting a more fragmented opposition, but also a weaker ruling party—should experience an increase in the probability of a democratic transition. We test this hypothesis using event-history models, random forest classification trees to detect nonlinear effects, and two case studies. Our conclusion holds if we employ different regime classifications and after we instrument legislative multipartism and employ sensitivity analysis to overcome potential inference concerns. We present evidence for Latin America between 1945 and 2010, an extensive sample that allows us to compare a broad range in the effective number of parties (ENP) and different types of dictatorships, including hegemonic party rule in Mexico (1929–2000) and military rule in Brazil (1964–1985).
Our results have important implications for democratic party strategies under authoritarian rule. One important discussion centers on the role elections play in stabilizing autocracies. Isiksel and Pepinsky (Reference Isiksel and Pepinsky2025) highlight the normative justification for voting in authoritarian elections, while Knutsen et al. (Reference Knutsen, Nygård and Wig2017) emphasize that autocratic elections bring about regime instability just before or after the election but enhance survival in the long run. We complement these perspectives by highlighting the instrumental value of electoral concessions: while they may stabilize authoritarian rule in the short term, they also lay the groundwork for long-term democratization. Our findings align with Teorell and Wahman’s (Reference Teorell and Wahman2017) emphasis on multipartism as a driver of democratic change. However, whereas their account foregrounding incumbent responses to counter-elite mobilization blurs the distinction between top-down and bottom-up transitions, we theorize that authoritarian incumbents adopt multipartism to defuse bottom-up threats, but in so doing progressively erode the foundations of authoritarian rule, as the declining cost of conceding democratization prepares rulers to initiate top-down transitions. Evidence from two case studies further supports the preemptive top-down democratization process we theorize.
1. Multipartism and authoritarian survival
Scholars agree that dictators create nominally democratic institutions to secure and extend their power (Schedler, Reference Schedler2009; Pepinsky, Reference Pepinsky2014) by minimizing threats from the elites and from the masses (Magaloni and Kricheli, Reference Magaloni and Kricheli2010), as parties and legislatures may help contain internal splits and coopt the opposition (Magaloni, Reference Magaloni2006), mobilize support (Reuter, Reference Reuter2022), reduce monitoring problems (Svolik, Reference Svolik2012), and solve leadership disputes (Boix and Svolik, Reference Boix and Svolik2013). The literature, however, displays less consensus on the consequences of legislative multipartism. While some scholars state that multipartism undermines dictatorships (Howard and Roessler, Reference Howard and Roessler2006; Donno, Reference Donno2013), others claim that multiparty elections strengthen autocratic legitimacy (Magaloni, Reference Magaloni2008). Gandhi and Przeworski (Reference Gandhi and Przeworski2006, p. 1291) point out that multiparty legislatures “have an independent effect on the tenure of autocrats, stabilizing their rule.” Autocrats establish plural legislatures to attract international benefits (Miller, Reference Miller2020) and incorporate opposition forces, giving them a stake in the leader’s survival (Gandhi, Reference Gandhi2008).
These studies advance two contradictory arguments. The first one contends that multiparty elections make authoritarian regimes weaker by facilitating mobilization and creating counter-elites (Teorell and Wahman, Reference Teorell and Wahman2017). The second argument states that multipartism makes authoritarian parties stronger, preparing them to compete under democracy and thus facilitating liberalization (Wright and Escribà-Folch, Reference Wright and Escribà-Folch2012, p. 303). This contradiction reflects an implicit distinction between bottom-up and top-down democratization. Neither approach, however, explains why authoritarian elites would allow multiple parties in the first place.
2. Threats and opportunities for dictators
Why would multipartism facilitate a transition to democracy, and why would dictators admit multiple opposition parties in the legislature if that is the case? Both questions must be addressed together to account for a regime equilibrium. We explore some intuitive answers in this section, and address the most puzzling case of preemptive multipartism in the next section.
2.1. The threat of democratization
Existing studies suggest that multipartism ignites threats of democratization in at least two ways. First, multipartism spells regime change because it reflects the strength of opposition forces beyond the legislature. Opposition parties are frequently supported by outside groups—business associations, student organizations, or religious groups—that threaten the dictator’s hold on to power.
Central to this argument is the possibility of a “stunning election” (Huntington, Reference Huntington1991). Authoritarian regimes conduct unfair elections, but electoral manipulation is constrained by the need to achieve sufficient credibility. If they miscalculate the level of manipulation required to prevail, opposition forces may win and initiate a transition. However, to maximize their chances, opposition forces should converge around a single candidate. A very high number of legislative parties may indicate that democratic forces are too fragmented to prevail.
Second, multipartism may encourage rulers to reassess the costs of suppression. Political accommodation offers an alternative to a sheer crackdown, but it requires a mix of legislative concessions (which reduce the benefits of autocracy) with repression (which is still costly). While regime hard-liners prefer an equilibrium based on hegemony, with high benefits and high repression costs, soft-liners envision a new equilibrium based on controlled liberalization, with fewer benefits but lower repression costs (O’Donnell and Schmitter, Reference O’Donnell and Schmitter1986). In this context, opposition parties can use the legislature to extract further concessions.
2.2. Opportunities for dictators
If greater opposition presence in the legislature facilitates democratization, why do dictators tolerate party pluralism? The first explanation is exogenous: party systems are sticky (Lipset and Rokkan, Reference Lipset and Rokkan1967), and some authoritarian rulers simply inherit multiple parties. Although autocratic leaders have more room to maneuver than their democratic counterparts, altering entrenched legacies is costly. Some authoritarian interventions (e.g., in Brazil in the 1960s or Peru in the 1990s) temporarily reduce the number of opposition parties, but eventually prove unable to end multiparty politics. In such cases, multipartism is exogenous in the sense that it predates any electoral manipulation by the dictatorship. We leverage this exogenous source of variation for causal inference below.
To the extent that they can, however, autocrats manipulate the number of parties endogenously, either because they feel too strong or because they feel too weak. Dictators often liberalize competition when they feel confident about their own electoral prospects (Slater and Wong, Reference Slater and Wong2013). This is likely to occur when ruling parties possess large resources vis-à-vis the opposition, because incumbents can be competitive in a subsequent democracy (Wright and Escribà-Folch, Reference Wright and Escribà-Folch2012).
Dictators may also allow for preemptive multipartism when they feel weak. Mass mobilization increases the regime’s vulnerability by harming its “public image of invincibility” and encouraging internal defections (Magaloni, Reference Magaloni2006, p. 9). Legislatures with multiple parties facilitate “the identification of reliable bargaining partners, the revelation of information, and the avoidance of popular mobilization” (Gandhi, Reference Gandhi2008, p. 80). Electoral rules encouraging multipartism also help fragment opposition forces. Yet, this reason for political liberalization remains most puzzling, because concessions in the form of legislative pluralism also trigger the threats of democratization just documented above. We explain the strategic logic of this decision in the next section.
3. A theory of preemptive multipartism
A conventional explanation for the adoption of preemptive multipartism underscores authoritarian elites’ miscalculation. Dictators adopt partially free elections as a calculated gamble (Hyde, Reference Hyde2011). This strategy incentivizes multiple opposition contenders to run and reduces the risk of a unified opposition front. However, the opposition may ultimately remain united. If autocrats miscalculate, the attempt to create a fragmented opposition may lead instead to a stunning electoral defeat and a democratic transition.
We acknowledge this possibility, but note that the literature has neglected an alternative story. Our explanation of preemptive multipartism does not invoke miscalculation on the part of dictators and does not require the opposition to converge around a single party for a transition to occur. Quite often, “multiparty competitive elections are introduced before much in terms of real democratization has taken place” (Lindberg, Reference Lindberg and Lindberg2009, p. 318).
Conceptually, we treat authoritarian stability, liberalization, and democratization as analytically distinct outcomes. Multipartism may enhance regime durability in the short term, but it can also increase the likelihood of democratization over time. While democratization refers to a transition to competitive electoral rule, liberalization entails the relaxation of authoritarian controls without a full transition to democracy, often involving expanded civil liberties, freedom of expression, partial pluralism, or limited electoral competition (O’Donnell and Schmitter, Reference O’Donnell and Schmitter1986).
Our explanation builds on the distinction between internal and external regime change (Heine and Weiffen, Reference Heine and Weiffen2015). “Internal” regime transformations occur within the existing legal framework, while “external” change is imposed by regime opponents. The literature often describes internal change as top-down democratization, transitions through reform, or “pacted” transitions.Footnote 1 In contrast, the literature describes externally driven processes as bottom-up democratization, rupture, or revolution (Fishman, Reference Fishman2019).
To flesh out the logic of the argument, we rely on a formal model that purposefully assumes full information and pure strategies. Consider an autocracy with two actors: an authoritarian party A, and a democratic camp D composed of N parties, D = {D1, …, DN}. Parties in both camps seek to maximize their share of seats in the legislature, avoiding the costs of violence (repression or rebellion) if possible. Let s be the share of seats for the democratic bloc and 1 − s the share of seats controlled by the authoritarian party. The ENP in the legislature can range from 1 (if s = 0 under closed authoritarian rule) to N + 1 (when seats are evenly distributed across all parties).
Figure 1 presents the strategic setup. Authoritarian rulers can Repress democrats (at a cost r) and grant them an arbitrary share of seats s A as a unilateral concession. The effective number of legislative parties declines when concession s A is smaller, irrespective of how this bloc of seats is allocated among the N opposition parties. Alternatively, dictators can Tolerate a free election, allowing a top-down transition to democracy. Democrats, in turn, can Accept the concessions or Revolt and impose democratization from below at a cost m. If a transition to democracy (from above or from below) takes place, the electorate freely decides the size of the democratic camp in the legislature s E, and unveils the true level of public support for the authoritarian party, 1 − s E.
Figure 1. A model of preemptive multipartism.
In this game, democrats revolt when s E − m > s A, but they never revolt if the authoritarian party is willing to concede top-down democratization. In turn, authoritarian elites prefer a controlled process of democratization to bottom-up revolt. Rulers can prevent a revolt by granting a minimal level of concessions, s A = s E − m. Given this choice, authoritarian elites must assess whether the cost of toleration is lower than the cost of sustaining the regime through repression and concessions, i.e., s E < r + s A.
The game has two equilibrium outcomes: (1) authoritarian rulers repress opponents and allow some degree of multipartism to deflect mobilization, up to the point at which (2) r > m; that is, the required repression (or the concessions to avoid it) are so costly that incumbents tolerate democratization. Recurrent renegotiation of the first equilibrium (authoritarian stability), given by shifts in r or m, may ultimately lead to the second outcome (a top-down transition). Rulers prioritize political survival, making it rational to grant concessions in the short run even if they can anticipate the downstream effect of concessions over the long run.
Receiving concessions amplifies democratic forces’ organizational capacity for electoral mobilization and strengthens their ability to establish coalitions with societal actors to engage in anti-regime protests (Trejo, Reference Trejo2014; Woo and Conrad, Reference Woo and Conrad2019), but the enhanced chances of democratization also increase their willingness to negotiate a transition. The key insight of our model remains that while authoritarian rulers liberalize to prevent democratization, increasingly granting concessions lowers the cost of democracy for incumbents.
The two panels in Figure 2 illustrate the equilibrium outcomes as a sequence. The horizontal line reflects the level of concessions, ranging from s A = 0 (a one-party legislature) to s A = s E (a top-down transition allowing for an unconstrained multiparty system). In Figure 2.1, r ≤ m and authoritarian rulers can offer limited pluralism to deactivate bottom-up mobilization. However, concessions must grow as the opposition’s cost of mobilization declines. In Figure 2.2, r > m, and authoritarian rulers are better off by allowing an internal transition.
Figure 2. Equilibrium outcomes at different levels of multipartism.
Given that the ENP is increasing on s A, Figure 2 encapsulates the authoritarian quandary: under the right conditions, an optimal level of multipartism deflects the risk of bottom-up democratization, but as concessions accumulate, they encourage top-down democratization because elites have fewer incentives to defend the regime. The consequences of this process are clear if we consider two historical moments in the life of a regime (i.e., two independent rounds of the game). An initial level of concessions s 1 may prevent revolt in the first period, but sets elite benefits at a level that makes top-down democratization likely if further concessions s 2 > s 1 are required in the next round. Even assuming perfect information—i.e., without invoking miscalculation on the part of dictators—this setup illuminates why rulers permit multiple parties and why soft-liners may eventually prefer to democratize.
3.1. Four empirical implications
Although economic crises or social unrest may alter the costs of mobilization—and thus increase the level of concessions required to appease the opposition—legislative multipartism is not a proxy for those exogenous shocks. Even in the absence of such shocks (e.g., when multipartism is inherited), authoritarian regimes with multiparty systems should be prone to democratize, because access to fewer legislative seats diminishes elites’ incentives to defend the regime. Thus, irrespective of whether authoritarian regimes inherit multipartism or adopt it as a preemptive measure, the theory implies that a greater number of legislative parties will increase the probability of a top-down transition to democracy.
This general hypothesis entails four more nuanced observable implications. First, an increase in the number of parties should facilitate democratization, but not other forms of authoritarian turnover. Second, multiparty autocracies should be more likely to democratize, even when they have inherited multipartism from past regimes. Third, the probability of a democratic transition will rise abruptly once the number of parties expands beyond a threshold (given by s A = s E—r), presumably near the point at which the ruling party loses control of the legislature. Fourth, preemptive multipartism should operate through protracted top-down democratization.
4. Data and methods
We primarily test these implications using quantitative evidence, and further use process tracing to capture the nuances of the pathway leading to top-down democratization. For the quantitative analysis, we take advantage of a dataset covering 20 Latin American countries, which allows us to measure the ENP (our indicator for multipartism) across all dictatorships between 1945 and 2010.Footnote 2 Our units of analysis are authoritarian country-years (N = 584). The dataset includes alternative measures of regime change, as well as institutional and economic indicators as control variables.
Latin America is an attractive region to test our theory for at least four reasons. First, it has hosted a wide range of authoritarian regimes. Almost three-quarters of the 584 authoritarian regime-years in our sample had legislatures, and almost 50% of them had more than one party. According to Cheibub et al.’s (Reference Cheibub, Gandhi and Vreeland2010) classification, 55% of country-years in our sample were led by the military and 45% by civilians. Such diversity contrasts with Eastern Europe, where single-party autocracies dominated, and fewer than 3% were military regimes.
Second, Latin America experienced regime changes throughout the whole period, a pattern that contrasts with other regions. Most Western European countries democratized after World War II. Many Asian and African countries, in turn, introduced multiparty systems in the wake of the Cold War. Latin America undertook multiple waves of democratization between 1945 and 2010.
Third, we need a continuous measure of the ENP to analyze nonlinear effects. We must distinguish whether the presence (Wright and Escribà-Folch, Reference Wright and Escribà-Folch2012) or the number of opposition parties makes a difference. While worldwide comparative data on legislative parties only recently became available through the V-Party project (Lührmann et al., Reference Lührmann, Düpont, Higashijima, Kavasogly, Marquardt, Bernhard, Döring, Hicken, Laebens and Lindberg2020), legislative data have long been available for Latin America. A focus on this region allows us to overcome the use of coarse measures (e.g., no legislature, one party, multiple parties) characteristic of previous studies. Fourth, the literature on Latin America has debated extensively whether multiparty systems are a source of instability for presidential democracies (Cheibub, Reference Cheibub2007), without exploring an equivalent question in the context of dictatorships.
4.1. Regime change
The dependent variables in our models capture whether regime change takes place in any given year, the direction of the change, and the type of democratization. We analyze changes in regimes, not changes in government.
We employ three alternative measures of regime change. The first measure is dichotomous and simply captures the end of an authoritarian regime. The second indicator is trichotomous, capturing whether the regime survives through the end of the year, is overthrown by a new dictatorship, or experiences a transition to democracy. The trichotomous measure allows for a placebo test, because our hypothesis implies that multipartism should induce democratization, not just the demise of the authoritarian regime. Our third measure is binary and contrasts regime survival with top-down democratization—the specific mode of transition our argument predicts.
To verify the robustness of our findings, we code alternative versions of the dependent variables using three different data sources: Geddes et al. (Reference Geddes, Wright and Frantz2014), henceforth GWF; Cheibub et al. (Reference Cheibub, Gandhi and Vreeland2010), henceforth CGV; and Svolik (Reference Svolik2012). Regime change is a rare event, representing roughly 7% of all observations. While all three data sets allow us to distinguish the replacement of a dictatorship by a new autocracy or by a democracy, only the GWF data set allows us to distinguish top-down from bottom-up transitions.Footnote 3 Accordingly, we rely primarily on GWF for our main analyses and further reconstruct modes of transition using qualitative case studies later in the paper.
4.2. Multipartism
Our main predictor is the degree of legislative multipartism, measured through the ENP in the lower (or only) House of Congress (Laakso and Taagepera, Reference Laakso and Taagepera1979). Data for this variable originates from multiple historical sources compiled for this project. The sample also includes dictatorships without legislatures, as discussed in the next section. We employ ENP as our main independent variable because the democratization literature has debated the effects of multipartism, but explore alternative measures for robustness. The ENP accounts not only for the dictator’s share of seats but also for the opposition’s fractionalization. We measure ENP at t − 1 to avoid endogeneity; our measure never captures ENP after the transition.
4.3. Confounders
The analysis controls for the presence of personalist regimes, since those often constrain the number of parties and are less likely to democratize (source is GWF). Additional controls capture the proportion of other countries in the region that are democratic (Mainwaring and Pérez-Liñán, Reference Mainwaring and Pérez-Liñán2014), as well as dichotomous measures of internal and international armed conflict (data from Clio Infra). The models also include per capita GDP (in thousands of 2000 US dollars) and annual economic growth (data from World Development Indicators and Penn World Tables) to account for modernization. To address the resource curse (Ross, Reference Ross2001), a dichotomous variable takes a value of 1 when oil and mineral exports represent more than 10% of the gross national income. Finally, we control for the size of the country’s population (in millions) and levels of ethnic and religious fractionalization.
As further robustness checks, additional models include two post-treatment variables. As incumbents may profit from a fragmented opposition (a side-benefit of multipartism), we include a dichotomous indicator coded as 1 when the ENP in the opposition bloc exceeds 2.0. In addition, because concessions are often intended to ease social protest, the models also include unrest—the sum of riots, strikes, and anti-government demonstrations from Banks and Wilson (Reference Banks and Wilson2017).
4.4. Estimation
To model regime survival, we employ discrete-time survival models to estimate how regime duration affects the probability of breakdown (Box-Steffensmeier and Jones, Reference Box-Steffensmeier and Jones2004). We account for duration dependence using a flexible polynomial of regime age, and we meet all conditions under which logit is a useful estimator for discrete-time survival models: the data include only years when regimes are at risk or end, and the baseline hazard is explicitly modeled. We complement this approach with a random forest classifier (Breiman, Reference Breiman2001), which makes fewer functional form assumptions and can capture complex, nonlinear relationships. While this method enhances robustness, it does not provide easily interpretable coefficients; therefore, we use it primarily to estimate the optimal number of parties that predict democratization.
We seek to estimate the effect of multipartism, which presumably facilitates a transition to democracy, aside from the countervailing effect of having a legislature, which stabilizes the authoritarian regime (Wright and Escribà-Folch, Reference Wright and Escribà-Folch2012). The sample naturally presents missing values for ENP in regimes without legislatures. Prior studies have solved this problem using dichotomous (Teorell and Wahman, Reference Teorell and Wahman2017) or trichotomous (no legislature, one party, many parties) measures (Gandhi, Reference Gandhi2008). Unfortunately, this approach truncates the variance of our main predictor. Building on Wright and Escribà-Folch (Reference Wright and Escribà-Folch2012), we adopt a simple solution for this problem. The ENP is coded as 0 in the absence of a legislature, and a dichotomous variable indicating the presence of a legislature is included in all models. We validate this strategy in Table A2 of the appendix.
5. First implication: multipartism prompts democratic transitions
Table 1 presents fully specified models, using the GWF classification to measure regime change. Model 1.1 assesses the probability of autocratic breakdown, Model 1.2 estimates competing risks of an autocracy becoming a democracy or a new autocracy, and Model 1.3 investigates the likelihood of top-down democratization.
Table 1. Regime termination (Geddes, Wright & Frantz)
Note: Entries are logistic estimates (standard errors). Regime survival is always the baseline category. In Model 1.1, the outcome is the end of regime according to Geddes, Wright & Frantz. Model 1.2 estimates competing risks of transition from autocracy to autocracy or democracy. Model 1.3 limits the outcome to top-down democratization.
* p < 0.05, ** p < 0.01.
The results in Table 1 align with our theoretical expectations. Model 1.1 indicates that, as the number of parties increases, authoritarian regimes are more likely to collapse. The effect is statistically and substantively significant: each one-unit increase in the ENP more than doubles the odds of regime breakdown (odds ratio = 2.06). We estimate that a one-standard-deviation increase in the ENP increases the probability of authoritarian regime breakdown by over 9 percentage points, holding other covariates constant. Contrary to claims that multipartism signals opposition weakness, it more often reflects the declining strength of autocrats. Indeed, the effect of opposition fragmentation is not statistically significant.
Model 1.2 more directly tests the first observable implication by estimating competing risks of a dictatorship being replaced by a subsequent autocracy or by a competitive regime. The test shows that a fragmented party system does not trigger all types of regime instability. In line with our theoretical model, multipartism prompts democratization (A-to-D) but not autocratic turnover (A-to-A). Substantively, a one-standard-deviation increase in ENP raises the predicted probability of democratization by nearly 10 percentage points.
Lastly, Model 1.3 further investigates the key insight of our argument—that multipartism paves the way for top-down democratization. Consistent with our model, we find that party pluralism significantly predicts the likelihood of top-down democratization. Substantively, a one-standard-deviation increase in ENP augments the predicted probability of top-down democratization by over 7 points, and each additional “effective” party is associated with a 176% increase in the odds of a top-down transition (relative to autocratic survival), holding other variables constant (odds ratio = 2.76). Conversely, we find that the ENP is inversely (albeit insignificantly) correlated with bottom-up democratization (see Table A2 in the appendix). The results suggest that bottom-up democratization may not be deterred when multipartism is low, whereas allowing for multipartism creates conditions conducive to top-down democratization.
5.1. Robustness tests
Table 1 presents estimates using measures of regime change that rely on Geddes et al. (Reference Geddes, Wright and Frantz2014). We replicated the analysis using alternative measures based on Cheibub et al.’s (Reference Cheibub, Gandhi and Vreeland2010) and Svolik’s (Reference Svolik2012) classifications of political regimes. The results, presented in Tables A3 and A4 and Figure A1 in the appendix, confirm the findings in Table 1. Likewise, Figure A2 illustrates how survival rates decline over time, showing that regimes with a greater number of parties are less likely to survive due to their greater susceptibility to top-down (rather than bottom-up) democratization. Although the three datasets differ in their timing of several events, our conclusions remain unchanged: Multipartism destabilizes authoritarian regimes by promoting democratization.
An important concern is that the ENP may capture both opposition strength and party system fragmentation. We address this measurement issue in two ways. First, we show that our results hold when controlling for lagged levels of democracy using V-DEM’s Electoral Democracy Index, which helps account for prior regime characteristics that may jointly influence both ENP and the likelihood of democratization—such as regimes where multipartism emerges as a consequence rather than a cause of democratization (see Table A5 in the appendix).
Second, we estimated our models using the ruling party’s share of seats as the primary independent variable. While the literature has emphasized the consequences of multipartism and opposition unity, the main factor in our theoretical model is the size of concessions (or conversely, the share of seats retained by authoritarian rulers). The results, shown in Tables A6–A8 of the appendix, reinforce our claim: as concessions expand (and thus the dictator’s share of seats diminishes), the likelihood of democratization increases—but not the probability of autocratic turnover.
Lastly, to ensure that the results are not driven by a small subset of countries or by temporally clustered transitions, we show that our results are robust to the inclusion of country and year fixed effects (Tables A12 and A13 of the appendix).
6. Second implication: inherited multipartism leads to democratization
Some unobserved variables could drive an increase in the number of parties while also making a transition more likely. For example, in response to exogenous shocks, reformers could liberalize the party system as a conscious step toward democratization. This creates a likely problem of endogeneity by inducing a positive correlation between changes in the number of parties and regime change, even though the number of parties may not cause regime change.
We address this concern by leveraging exogenous (inherited) sources of multipartism. As explained above, inherited multipartism should be equally relevant to our argument, because it also imposes fixed costs on authoritarian elites. We estimate the probability of democratization employing dummy variables as coded by GWF, CGV, and Svolik, using a two-stage duration estimator with a probit link that treats the ENP as an endogenous covariate. We instrument multipartism after 1945 using a dichotomous measure of proportional representation (PR) in the electoral system before 1945, the average district magnitude in the last election prior to 1945, and an interaction between these two variables. Electoral institutions before 1945 presumably shaped the number of parties over the long run (Amorim Neto and Cox, Reference Amorim Neto and Cox1997), but they predate any unobserved shocks that encouraged liberalization after World War II. We further discuss the estimation in the appendix (see “Endogeneity” section).
Table 2 summarizes the results for the second-stage model (full results presented in the appendix, Tables A9-1 and A9-2). For measures of the dependent variable based on GWF and CGV, the effect of multipartism is positive, statistically significant, and greater in magnitude than in naïve models. Although Svolik’s dataset generates weaker results (p = 0.19), there are few reasons to believe that our results are driven by endogeneity.
Table 2. Endogeneity models: second stage—democratization
Note: Probit estimates with a continuous endogenous covariate (ENP). Controls omitted to save space (specification equivalent to Model 1.1). Instruments are PR and Average District Magnitude before 1945 (plus interaction). For full results, see Tables A9-1 and A9-2 in the appendix. GWF Geddes, Wright & Frantz; CGV Cheibub, Gandhi & Vreeland.
* p < 0.05, ** p < 0.01.
Table 2 also reports F-tests for variables excluded in the second stage, indicating that PR and district magnitude are relevant instruments for multipartism. We also report the Sargan test to evaluate the validity of the overidentifying restrictions. The null hypothesis is that all instruments are uncorrelated with the contemporaneous error term. We are unable to reject the null (p > 0.9 for GWF, p > 0.4 for CGV, p > 0.5 for Svolik), which suggests that all instruments are valid.
6.1. Assessing exclusion restrictions
Table 2 enables us to dismiss the most significant source of endogeneity—the possibility that short-term exogenous shocks drive both party liberalization and democratization simultaneously—because it captures the effect of inherited multipartism. However, we must weigh potential violations of the exclusion restriction. Such violations could happen as a result of direct effects or non-ignorability. Instruments could have a direct effect on the outcome through a causal path other than the exposure variable. This violation seems implausible, as it is unlikely that proportionality and district magnitude will undermine a regime operating decades later other than through the party system. Still, because autocrats strategically design electoral systems and preserve them when advantageous (Gandhi et al., Reference Gandhi, Heller and Reuter2022; Higashijima, Reference Higashijima2022), contemporary electoral institutions may serve as a key pathway through which the exclusion restriction is violated. However, we find that pre-1945 electoral systems are only modestly correlated with the electoral formula in the corresponding postwar country-year (r = 0.24), show no correlation with regime change outcomes (r = 0.03 using GWF coding), and that controlling for current formulas does not meaningfully alter the results of our instrumental variable approach (see the appendix, Tables A10-1 and A10-2). This suggests that any direct influence of historical electoral systems on regime outcomes through alternative pathways such as contemporary institutions is limited.
Alternatively, in the absence of instrument randomization, an unobserved confounder could explain both the adoption of proportionality in the past and regime change at the present. This violation is more plausible, as two possibilities come to mind. The first is that the lasting social cleavages that prompted the adoption of PR in the past continue to pose a challenge for authoritarian rule years later. The second possibility is that elites introduced proportionality in response to past revolutionary threats, and those threats persist. While ignorability is ultimately a non-testable assumption, we address both possibilities next.
6.2. Alternative explanations as threats to causal inference
Because the two stories that challenge the ignorability assumption could also confound our results in Table 1, we address them separately in this section and devote special attention to them in the appendix. Both stories involve stable predictors (social cleavages and latent regime instability) that potentially affect the outcome variable (regime change) and the instrument (the introduction of PR in the distant past), but also the treatment (the number of parties at the present).
The first possibility is that social cleavages that prompted the adoption of PR in the past still account for multipartism and also pose challenges for the survival of authoritarianism. Lipset and Rokkan (Reference Lipset and Rokkan1967) famously argued that cleavages shape the number of parties over the long run. Plural societies are likely to adopt PR, and our estimates would be biased if unobserved social heterogeneity affected not only the number of parties but also political instability. In Tables 1 and 2 (see the appendix, Table A9-1), we already deal with this possibility by controlling for ethnic and religious fractionalization. Our results hold after adjusting for those stable cleavages. It is still possible that some unobserved cleavages are not properly captured by fractionalization measures. But in this case, confounding should induce a positive association between ENP and all forms of authoritarian instability, not only democratic transitions. Our placebo test (Model 1.2) dismisses this possibility.
A second possibility is that elites offered PR in the past to deflect revolutionary threats that still endanger authoritarian rule (Gjerløw and Rasmussen, Reference Gjerløw and Rasmussen2022). Recurrent threats could activate two mechanisms: (1) they could induce new concessions to opposition parties, potentially leading to a top-down transition, or (2) they could increase the risk of revolution, leading to a bottom-up transition or a new authoritarian regime. The first mechanism would simply confirm our argument. The second mechanism, however, would confound the effect of multipartism (resulting from the early introduction of PR). As with the case of cleavages, we address this possibility directly through regression adjustment and indirectly through placebo tests and sensitivity analysis. We measure enduring regime threats by computing the total number of prior regime changes since 1945. Our results, reported in Table A11 in the appendix, show that multipartism remains a significant predictor of regime change after this adjustment. Additionally, all models control for internal conflict and social unrest as proxies for potential revolutionary threats. It is still possible that some enduring threats are not properly captured by past regime instability or social unrest. Even in that case, there is no reason to believe that revolutionary threats should only promote transitions to democracy. This path (from past threats, to PR, to multipartism) should create a positive correlation between ENP and overall authoritarian instability. However, as our placebo test shows (Model 1.2), multipartism only affects democratic transitions, not authoritarian turnover.
To further assess the risk of unobserved confounding, we conducted a sensitivity analysis that estimates adjusted coefficients based on a hypothetical omitted variable that is k-times stronger a confounder than a given covariate (Cinelli and Hazlett, Reference Cinelli and Hazlett2020). Following our discussion on bottom-up threats and findings from Table 1, we investigate how strong an unobserved confounder would have to be, relative to social unrest, to fully explain away the effect of multipartism. The results presented in Figure A3 from the appendix suggest that any hypothetical confounder would need to be more than 5–10 times more predictive of ENP and regime change than the observed effect of social unrest. Given the theoretical and empirical reasons indicating that social revolts facilitate regime breakdowns, it becomes highly unlikely that the effects of multipartism derive from bias in unobserved confounding.
7. Third implication: optimal number of parties
Our theory implies that authoritarian rulers should permit an expansion in the number of parties up to the point at which they are willing to concede democratization. This suggests that the function connecting the ENP and democratization is not linear (or smoothly monotonic), but rather anticipates a steep change when authoritarian parties lose legislative majorities. An alternative function is suggested by the “stunning elections” hypothesis, in which the opposition coordinates on a single candidate to win. This explanation anticipates that the likelihood of a transition will increase as the opposition bloc matches the strength of the authoritarian bloc, but will decline if the opposition fragments. Here, the function linking the number of parties and the probability of transitions is also nonlinear, but looks like an inverted U.
In this section, we relax the functional form by employing machine-learning techniques. We use nonparametric multivariate classification trees (CART). CART models offer several benefits, including the ability to detect nonlinearities in the risk function without overfitting and to identify interactions between covariates. We employ random forests, an ensemble of several randomized trees that reduce the inherent instability present in single decision trees and eschew a partitioning process biased toward continuous predictors (Breiman, Reference Breiman2001).Footnote 4 Each tree is grown employing random sampling with replacement. Random forests introduce an additional source of randomness, as only a subset of variables is selected to build each tree. This process leads to disparate trees. Finally, all trees are combined to determine fitted values. We construct an ensemble of 1,500. Random forests usually yield more precise predictions than classic models.
Results for alternative measures of democratization, presented in Figure 3, show that the probability of transitions increases with multipartism. The risk of democratization is low when the regime incorporates some opposition but retains tight control of the legislature. The optimal number of parties for authoritarian rulers is somewhere between one and two. The prospects of autocratic collapse increase as rulers offer further concessions, with major shifts observed around 2 and 3.5 legislative parties, depending on the data. This evidence does not disprove the importance of coordination among opposition parties, but suggests that transitions also occur when opposition forces are fragmented. (For additional details on CART models, see the appendix, Figures A4 and A5.)
Figure 3. Probability of democratization (random forests).
8. Fourth implication: protracted top-down democratization
We now turn to case studies, which allow us to explore the protracted sequence that leads to top-down democratization. We document causal mechanisms by analyzing the sequences leading to democratization in two cases, Mexico and Brazil. The two countries experienced different types of dictatorships—party rule in Mexico (1929–2000) and military rule in Brazil (1964–1985)—and were unusually enduring. The presence of legislatures, which accommodated factions of the ruling party and opposition, combined with rules for executive alternation, contributed to their survival. In response to social pressures, the two regimes allowed for multipartism. The opposition fragmented but gained spaces in the national legislature and subnational governments, and liberalization evolved into a process of top-down democratization.
The cases illustrate the two analytic moments conveyed by Figure 2. In the first stage, authoritarian rulers granted limited concessions to stabilize the regime, deactivating bottom-up pressures. In the second moment, additional reforms—championed by leaders who worried about further discontent—allowed the opposition to make significant inroads into institutional spaces. The ruling elites lost significant power, but did not escalate repression to preserve the regime, allowing a top-down transition instead.
8.1 Mexico: the end of party hegemony
The Institutional Revolutionary Party (PRI) ruled a hegemonic regime between 1929 and 2000, longer than any non-communist party in modern history. The PRI won every governorship until 1989 and retained its majority in Congress until 1997. However, fear of social insurrection encouraged the PRI to provide opposition parties with increasing space. Together with other forms of patronage, legislative seats initially helped presidents coopt their opponents (Magaloni, Reference Magaloni2010). The road toward multipartism involved several electoral reforms, motivated by social pressures and by rebel threats along ethno-territorial lines (Trejo, Reference Trejo2012). A protracted transition ultimately occurred within the existing legal framework (Loaeza, Reference Loaeza2000), fueled by multiple agreements (Eisenstadt, Reference Eisenstadt2004). In 2000, almost four decades after the first multiparty reforms, the PRI peacefully relinquished power to the presidential candidate of the National Action Party (PAN).
8.1.1 Concessions prevent external democratization
The moment depicted in Figure 2.1 took place in Mexico between the 1960s and the 1980s, when the PRI granted concessions in response to adverse circumstances. Many reforms enacted at this stage sought to deactivate social conflict, as a cycle of protest mobilized peasants, workers, and unions, and led to the emergence of guerrillas that threatened the stability of the regime (Del Campo and López Leyva, Reference Del Campo and López Leyva2004).
Amid increasing conflicts, President Adolfo López Mateos (1958–1964) initiated a first wave of reforms in 1963, replacing the majoritarian system that prevented opposition parties from gaining seats in the lower house. By introducing “party deputies”—15% of the seats reserved for minority parties—the PRI sought to project an image of pluralism (Valdés Zurita, Reference Valdés Zurita2018). Faced with “the threat of a nation risen to arms” after the 1968 protests (Johnson, Reference Johnson1978, p. 161) and the possibility that the student movement would foster broader opposition, President Luis Echeverría (1970–1976) adopted a policy of apertura democrática that included press liberalization and a progressive rhetoric (Riding, Reference Riding1976). In 1978, a new reform doubled the size of the Chamber of Deputies and allocated a quarter of the seats to minority parties through PR. The PRI’s goal was threefold: to reward itself disproportionately, to divide its opponents, and to coopt smaller parties, discouraging violent mobilization (Magaloni, Reference Magaloni2006). Although the electoral system hindered coordination among opposition parties, it fostered multipartism.
Debates on how to handle increasing political competition and economic crises agitated the PRI, broadening the split between dinosaurios (dinosaurs) and técnicos (technocrats) in the 1980s. Dinosaurios were the “ward bosses”: they managed the distribution of patronage, mobilized support for PRI candidates, and advocated state-led development (Gill, Reference Gill, Kopstein and Lichbach2005). The técnicos represented a reformist faction and promoted laissez-faire economics. Despite tactical differences, both PRI factions sought to preserve the regime. In 1986, a reform expanded the Chamber and increased PR, while ensuring PRI control through a governability clause that granted the winning party a majority of seats.
8.1.2 Further concessions encourage internal democratization
After an economic crisis in 1982 and a poorly managed earthquake in 1985, the costs of mobilization for the rightist PAN and the leftist Party of the Democratic Revolution (PRD) declined. Opposition parties expanded by taking advantage of voters’ dissatisfaction and the new electoral system, and the regime was no longer able to create an image of invincibility (Magaloni, Reference Magaloni2006).
As a result, the PRI rigged the 1988 election. To regain legitimacy, a new wave of electoral reforms took place in the 1990s, weakening the PRI’s control. In 1990, Mexico created the Electoral Federal Institute (IFE), an autonomous body. A 1993 reform further reduced the maximum number of seats that any single party could hold in the lower house. Such an expansion of multipartism was remarkable because it implied that the PRI would henceforth require the collaboration of opposition parties to achieve the two-thirds majority necessary for any constitutional amendment.
In this context, the uprising of the Zapatista National Liberation Army (EZLN) in 1994 prompted additional concessions. Three months after the Zapatista rebellion, Congress passed a constitutional reform increasing citizen control over electoral councils. This reform sought to avert a revolutionary coalition of rural rebels, urban left parties (PRD), and social movements (Trejo, Reference Trejo2012, p. 235).
Afraid of insurgents, the government negotiated new reforms. In 1996, Congress approved a “definitive electoral reform,” which enhanced the independence of the electoral tribunal and granted opposition parties unprecedented access to public and private media. While the 1977 reform allowed for the first multiparty legislature, the 1996 reform led to the establishment of the first divided government. By 1997, negotiating legislative coalitions was easier for the ruling technocrats than dealing with the hard-liners of their party.
In line with our argument, Mexican elites progressively relinquished their ability to fix the number of parties. Before the first electoral reform, the ruling party controlled all seats in Congress. After 1963, the PRI’s legislative share in the lower house decreased from 97% to 83%. With the 1978 reform, PRI lost an additional 10% while PAN and PRD extended their presence in the legislature. Following the new reforms of the 1990s, the ruling party’s share of seats decreased to approximately 60%. By the time of the transition in 2000, the PRI barely held 48% of the seats.
Mexico’s history illustrates how multipartism helped stabilize the regime, and how it eventually facilitated democratization. Bottom-up pressures prompted repeated reforms, and opposition parties negotiated spaces for participation (Eisenstadt, Reference Eisenstadt2004). Despite facing coordination problems due to ideological disagreement, the two parties challenging the PRI cooperated on electoral reforms and facilitated a transition to democracy.
8.2 Brazil: the end of military rule
The Brazilian dictatorship of 1964–1985 represented one of the most stable bureaucratic-authoritarian regimes of the twentieth century (O’Donnell and Schmitter, Reference O’Donnell and Schmitter1986). However, the regime’s candidate lost indirect elections for the presidency in 1985, and the military relinquished power. This outcome followed “the most protracted and controlled case on record of the liberalization of an authoritarian regime from above” (von Mettenheim, Reference von Mettenheim1995).
8.2.1 Concessions prevent external democratization
The Brazilian military planned for a long-term intervention and tolerated legal opposition (Stepan, Reference Stepan1971). While military juntas in Argentina, Chile, and Uruguay closed their legislatures, in Brazil congressional elections were held every four years.
Brazilian rulers permitted political parties, although with strong restrictions. Between 1965 and 1978, government regulations imposed a two-party system in which the pro-regime Aliança Renovadora Nacional (ARENA) competed against a single opposition force, the Movimiento Democrático Brasileiro (MDB; Desposato, Reference Desposato2001). The MDB struggled to serve as an autonomous voice during 1969–1974, when repression and torture reached their heights. Repression undermined the party and ARENA overwhelmingly won the 1970 congressional elections. However, the MDB offered a channel for opposition demands, and it gained relevance after 1974.
Despite the opposition’s weakness, the dictatorship did not enjoy enough popular support to govern without repression and manipulation of electoral laws. Social pressures led military officers to believe that some liberalization was necessary, and the MDB’s weakness suggested that an opening carried no significant risks. The “slow road to democratization” began with General Geisel’s triumph over hard-liners in 1973 (Skidmore, Reference Skidmore and Stepan1989). Liberalization carried out by soft-liners sought to split the opposition and coopt social and labor movements. As Mainwaring (Reference Mainwaring1986) notes, the abertura enjoyed widespread support among powerful political and economic actors, as well as among an opposition that was unable to overthrow the regime.
The decision to liberalize allowed the regime to reduce repression, but it unleashed a new dynamic. Anticipating a triumph under the effects of the “economic miracle” (Bacha and Bonelli, Reference Bacha and Bonelli1999), the regime let MDB run a relatively free campaign in 1974. However, the opposition claimed key victories, winning 16 of the 22 contested Senate seats. ARENA’s success in 1970 had been achieved under extreme conditions—proscription and intimidation of opponents, military operations against guerrilla movements, and a forced recess of Congress. The MDB’s success in 1974 cast doubts on the regime’s ability to control the electoral process (von Mettenheim, Reference von Mettenheim1995).
In 1977, President Geisel issued a set of electoral and legislative reforms aimed at protecting ARENA. The reforms saved ARENA’s legislative majority in the 1978 election, but the MDB performed even better than in 1974. Military rulers concluded that the two-party system was “no longer working to their advantage” (Power, Reference Power2000, p. 63).
8.2.2 Further concessions encourage internal democratization
In 1979, the military exploited a long-overdue party reform to liberalize the two-party system, with the goal of dividing the opposition. The share of seats for the MDB (renamed PMDB) declined, but the new rules created further opportunities. Opponents formed new parties like the Partido dos Trabalhadores (PT) and revived old labels such as the Partido Democrático Trabalhista (PDT). The effective number of opposition parties increased from 1.0 in 1978 to 2.6 in 1980, while the government’s party (renamed PDS) progressively shrank. The regime responded in 1981 with a new set of defensive reforms—including the adoption of a straight-ticket ballot to leverage coattail effects—to no avail. Before the abertura began, the government controlled 72% of the Chamber of Deputies and 86% of the Senate. In 1982, when direct elections for governors as well as for federal and local representatives were reintroduced, the regime lost its majority in the lower house but maintained a 67% share in the Senate.
Emboldened, the opposition defied autocracy in Congress and in the streets, demanding political rights and new concessions. Meanwhile, social movements challenged the dictatorship (Humphrey, Reference Humphrey1982). Party leaders and civil society groups operated independently, although they eventually cooperated to organize mass demonstrations calling for direct presidential elections in 1983–1984.
The abertura reached a tipping point in 1983, as the regime lost its ability to control the presidential succession. As internal struggles intensified, mobilizations demanding direct elections for the presidency emerged (Mainwaring, Reference Mainwaring1986). Massive demonstrations led some PDS officials (including the vice-president) to support direct elections, despite military threats. The campaign convinced many party leaders that the regime had to bargain with the opposition as a way out of the political crisis.
The PMDB’s presidential candidate ultimately prevailed in the 1985 indirect election. By then, more than five parties had representation in the Chamber of Deputies, and the party system would soon become one of the most fragmented in the world. Multipartism posed coordination problems, yet opposition parties served to channel popular demands, taking advantage of the top-down liberalization process and pushing Brazil toward democracy.
9. Leveraging multipartism
Students of authoritarian regimes have long debated whether multipartism helps dictators survive (Gandhi and Przeworski, Reference Gandhi and Przeworski2006; Gandhi, Reference Gandhi2008) or facilitates democratization (Donno, Reference Donno2013). Our study contributes to this debate by demonstrating that authoritarian concessions reduce the risk of external regime change while also encouraging internal democratization. In contrast to social diversity (Gerring et al., Reference Gerring, Hoffman and Zarecki2018), party diversity facilitates regime change.
Our analysis helps clarify why multipartism promotes transitions. Past studies either argued that multipartism makes authoritarian parties weaker, creating counter-elites (Teorell and Wahman, Reference Teorell and Wahman2017), or makes them stronger, preparing them to compete under democracy (Wright and Escribà-Folch, Reference Wright and Escribà-Folch2012). The theory in this paper reconciles these views to account for puzzling instances of preemptive multipartism. Incumbents tolerate counter-elites to contain bottom-up revolt, but concessions make regime soft-liners reassess the costs of suppressing free elections. This creates conditions for a protracted top-down democratization. Democratic transitions driven by this mechanism occur over a long period of party fragmentation, rather than as the result of a stunning election. Quantitative empirical evidence from 20 Latin American countries indicates that dictatorships become unstable when the effective number of legislative parties exceeds a threshold of two. We further documented the causal mechanism by reconstructing how authoritarian rulers granted concessions to deflect bottom-up revolts, ultimately leading to protracted top-down transitions in Mexico (1929–2000) and Brazil (1964–1985).
Our findings have important implications for opposition forces in authoritarian regimes. Although a unified opposition bloc is presumably best for confronting authoritarian rulers in elections (Howard and Roessler, Reference Howard and Roessler2006; Arriola, Reference Arriola2013; Donno, Reference Donno2013; Wahman, Reference Wahman2013), opposition parties may fail to unify (Jiménez, Reference Jiménez2021), or voters may fail to support the coalition (Gandhi and Ong, Reference Gandhi and Ong2019). This article shows that opposition forces should not relinquish electoral spaces even if they are unable to unite. Authoritarian rulers grant electoral concessions to quell unrest and divide the opposition in the short term, but their concessions ultimately create conditions for party pluralism in the next round of negotiations.
These conclusions resonate with historical studies showing that parties were decisive players in democratization in Europe (Capoccia and Ziblatt, Reference Capoccia and Ziblatt2010), Latin America (Madrid, Reference Madrid2019), and elsewhere (Grzymala-Busse, Reference Grzymala-Busse2006). In an age of competitive authoritarian regimes, it is essential to understand when and how opposition forces can build momentum for democratization. If dictators tolerate institutions that potentially destabilize their regimes, democrats need to leverage this potential.
Supplementary material
The supplementary material for this article can be found at 10.1017/psrm.2025.10062.
To obtain replication material for this article, please visit https://doi.org/10.7910/DVN/5ZAWXQ.
Acknowledgements
We are grateful to Ana Arjona, Juan Albarracín, Daniel Brinks, Michael Coppedge, Kent Eaton, Candelaria Garay, Alisha Holland, Juan Pablo Luna, Gerardo Munck, Tom Mustillo, Emilia Simison, George Tsebelis, Amy Erica Smith, and Guillermo Trejo for their valuable feedback on earlier versions of this work. We also thank participants at the 2017 Latin American Political Science Association (ALACIP) conference and the 2019 Annual Democratization Cluster Workshop at the Kellogg Institute, University of Notre Dame.
Open access
