Political Institutions and Coups in Dictatorships

Does the creation of nominally democratic institutions help dictators stay in power by diminishing the risk of coups? We posit that the effectiveness of political institutions in deterring coups crucially depends on the types of plotters and their political goals. By providing a means to address the ruling coalition’s primary concerns about a dictator’s opportunism or incompetence, institutions reduce the necessity of reshuffling coups, in which the ruling coalition replaces an incumbent leader but keeps the regime intact. However, such institutions do not diminish the risk of regime-changing coups, because the plotters’ goals of overthrowing the entire regime and changing the group of ruling coalition are not achievable via activities within the institutions. Our empirical analysis provides strong empirical support for our expectations. Our findings highlight that the role of “democratic” institutions in deterring coups is rather limited as it only applies to less than 38% of coup attempts.

The burgeoning literature on comparative authoritarianism suggests that authoritarian rulers adopt nominally democratic institutions, such as political parties and legislatures, to consolidate their hold on political power (e.g., Boix & Svolik, 2013;Brownlee, 2007;Gandhi, 2008;Svolik, 2012). Particularly, given that the biggest threat to dictators' political survival comes from within rather than from outside the regime (Aksoy et al., 2015;Frantz & Stein, 2017;Svolik, 2012), many studies examine how authoritarian institutions mitigate threats of coup d'etats. For example, Boix and Svolik (2013) and Magaloni (2008) argue that authoritarian parties and legislatures help establish a system of credible power-sharing between a ruler and his ruling coalition, reducing the risk that the dictator will be overthrown by regime elites. As Woo and Conrad (2019) emphasize, authoritarian institutions coopt elites to reduce the risk of coups. In a different vein, Geddes (2005) argues that political parties decrease the likelihood of coups by giving party members a stake in the incumbent regime and making them more likely to oppose coups when they arise. The idea that autocratic rulers purposefully adopt authoritarian institutions to strengthen their rule and that these institutions promote autocratic stability is now well-accepted in the literature on comparative authoritarianism. 1 This "institutional turn" (Pepinsky, 2014) significantly contributes to the theoretical development in comparative authoritarianism. However, one shortcoming of this literature is that scholars have paid relatively little attention to the fact that unlike coups in democracies, coups in autocracies take two different forms: regime-changing coups and reshuffling coups. 2 Regime-changing coups entirely change both the ruling elites in power and the formal and informal rules for leadership selection and policies. Examples of regime-changing coups are the military coups in Egypt (1952), Iraq (1958), andLibya (1969) that ended dynastic rule by ousting the royal family from power (Geddes et al., 2018, p. 46). Reshuffling coups, on the other hand, preserve the regime and only replace the incumbent leader with another member of the same ruling elite. The coups in Argentina during the early 1970s that replaced only the junta leaders are an example of this type (Geddes et al., 2018, p. 46). Both types of coups described above contrast with coups against incumbent democracies, which only produce regime change by interrupting democratic processes.
Importantly, regime-changing and reshuffling coups qualitatively differ in terms of the perpetrators and their political goals. Reshuffling coups are typically conducted by members of the ruling coalition who have power to control leader selection and policy decisions in the regime. Members of the ruling coalition execute reshuffling coups to hold a dictator accountable without changing the membership of the ruling coalition or the rules for leader selection. These perpetrators seek to oust an opportunistic or incompetent incumbent leader while maintaining or enhancing their own power. On the other hand, regime-changing coups are typically conducted by regime members who are outside the ruling coalition. 3 When they are unhappy with the way leadership positions and political power are controlled by the existing group of ruling elites, these perpetrators seek to replace them and to change the rules for choosing leaders and policies. Given these significant differences, the failure to distinguish between the two types of coups significantly hinders efforts to understand the effects of authoritarian institutions on coups (Aksoy et al., 2015;Geddes et al., 2018).
In this paper, we posit that the effects of authoritarian institutions on the likelihood of coups crucially depend on the type of coups under consideration. We argue that the presence of political parties and legislatures in authoritarian regimes reduces the likelihood of reshuffling coups, while authoritarian institutions have no meaningful impact on the likelihood of regime-change coups. The primary goal of ruling coalition members is to make a dictator commit to satisfactory power-sharing agreements or to punish the dictator for policy failures. Political institutions enable them to address their concerns over a dictator's opportunism and incompetence by facilitating collective actions among ruling elites and improving the transparency. Therefore, well-institutionalized political parties and legislatures diminish the need for punishing opportunistic leaders via violent and risky means.
However, the same logic does not apply to regime-changing coups. Those plotting this type of coup aim to change the ruling elites and the rules for leader selection and policy decisions. These goals are incompatible with the existing institutions' values and rules and, thus, are not attainable via activities within the institutions. Without providing a peaceful means of achieving regime change, existing political institutions will not reduce plotters' willingness to use violent means to pursue their goals. Further, while political institutions mobilize mass support toward the regime and would deter regime-change coups, this deterrent effect of institutions will be offset by the institutions' effects in increasing the feelings of relative deprivation and grievance among regime members outside the ruling coalition. Consequently, we anticipate that political institutions do not affect the likelihood of regime-changing coups.
Using time-series cross-sectional data comprising the 118 authoritarian countries over the years 1946 to 2010, we present supporting evidence for our argument. The existence of political parties and legislatures in authoritarian regimes is associated with a decreased likelihood of reshuffling coup attempts. Contrarily, authoritarian political institutions do not significantly affect the likelihood of regime-change coup attempts. These results are robust to employing a matching algorithm to construct a balanced sample, exploring institutional variations, estimating various model specifications, and using alternative estimators.
Our findings have important implications for the literature on comparative authoritarianism. First, our study highlights the importance of carefully disaggregating political actors with distinct interests when evaluating the role of political institutions in authoritarian regimes. Existing theories that argue for authoritarian institutions' regime-preserving effects tend to exclusively focus on how institutions mediate conflicts between dictators and their ruling coalition (e.g., Boix & Svolik, 2013;Magaloni, 2008). However, more than 62% of coup attempts are regime-change coups that would replace the ruling coalition in power with a new group of ruling elite (Chin et al., 2020). While existing frameworks help understand how institutions mitigate the risk of reshuffling coups led by members of the ruling coalition, they cannot explain whether and how institutions affect regime-changing coups. As the literature on coups highlights, there are important variations among coups regarding the types of plotters and their interests (e.g., Aksoy et al., 2015;Albrecht & Eibl, 2018;De Bruin, 2019;Singh, 2014;Sudduth, 2021). 4 Lumping together these actors with distinct interests within the state apparatus prevents us from fully understanding how institutions affect authoritarian regime survival. More generally, while the literature on comparative authoritarianism focuses on how authoritarian institutions protect regimes from threats from within and outside the regime (e.g., Svolik, 2012), this dichotomization of threats toward authoritarian regimes would obscure significant variations among potential challengers within the regime (or within the society) and, thus, fail to capture the divergent effects of institutions on these heterogeneous actors.
Second and relatedly, our finding that political institutions do not have meaningful effects on regime-changing coups indicates that institutions are not as effective in helping authoritarian regimes survive as the literature previously claims. Scholars argue that political institutions play a critical role in managing intra-elite conflict and have coup-inhibiting effects (e.g., Boix & Svolik, 2013;Magaloni, 2008;Reuter & Remington, 2009). However, among all the 310 coup attempts from 1946 to 2010, 193 coup attempts are regime-change coups, while 117 are reshuffling coups (Chin et al., 2020). Authoritarian political institutions do not affect regime-changing coups that account for more than 62% of coup attempts. As the principal threat most autocratic leaders face stems from within the regime in the form of coups, political institutions' role in maintaining and promoting authoritarian regime survival might be much more limited than is typically assumed in the literature.

Political Institutions and Coups in Dictatorships
The vast majority of authoritarian regimes function with the help of political institutions traditionally associated with democratic regimes, such as political parties and legislatures. A large body of work has argued that dictators establish nominally democratic institutions to help them address various threats to their power and to improve their chances of political survival (e.g., Boix & Svolik, 2013;Gandhi, 2008;Gandhi & Lust-Okar, 2009;Magaloni, 2008;Svolik, 2012). Specifically, political institutions in authoritarian regimes are thought to address two types of threats to political leaders' power. One type of threat comes from groups and individuals outside the regime (i.e., civil war, revolutions, and mass protest), and the other from challenges posed by those within the regime (i.e., coups) (Svolik, 2012).
On the former, scholars have argued that authoritarian institutions are effective in broadening the basis of support for incumbent regimes and thus help dictators counteract external threats (Gandhi, 2008;Lust-Okar, 2005;Malesky & Schuler, 2010). Dictators use parties and legislatures to distribute rents and privileges, which enables them to co-opt citizens and potential regime opponents (Bueno de Mesquita et al., 2003). Political parties provide their members with jobs, political privileges, and other economic benefits, reducing their incentives to challenge the regime (Geddes, 2005;Svolik, 2012). Authoritarian legislatures are also useful in co-opting the opposition. By giving opposition groups a formal say in the policymaking process and policy compromises, legislatures ensure that potential opponents have a stake in the regime. Authoritarian institutions, particularly parties, can also build and mobilize mass support for the regime because they are effective in operating a patronage system (Geddes, 2005;Magaloni, 2006;Magaloni & Kricheli, 2010).
Though the role of institutions in addressing threats from outside the regime is important, the threat of coup d'etats coming from within the regime is considered most crucial in authoritarian regimes. The literature highlights that an overwhelming majority of authoritarian leaders lose power through coups rather than through regular leadership transition, popular uprisings, or civil wars (e.g., Svolik, 2012). About one-third of leader exits in dictatorships are removal via coup (Frantz & Stein, 2017, p. 942), and they would suffer severe post-tenure punishment following their removal via coup. Coup attempts would also result in more general violence with a significant number of deaths (De Bruin, 2019), and would have long-term effects on a country's political stability and economy, even in the case of failed coups (e.g., Powell & Chacha, 2016). These points strongly suggest that to maintain political stability and power, dictators need to adopt effective strategies to reduce the likelihood of coup attempts.
Many researchers claim that political institutions reduce the risk of coups (e.g., Boix & Svolik, 2013;Bove & Rivera, 2015;Geddes, 2005;Magaloni, 2008;Miller, 2020;Olar, 2019;Svolik, 2012;Woo & Conrad, 2019). Political institutions, for example, can stabilize power-sharing arrangements between a dictator and regime elites by facilitating collective action among ruling elites (Boix & Svolik, 2013;Gehlbach & Keefer, 2012). Political institutions increase regime elites' capacity to coordinate actions against dictators by increasing information flows and providing an institutional forum for elites to regularly meet (Albertus & Menaldo, 2018;Gehlbach & Keefer, 2012). By making credible the coup threat posed by ruling elites, political institutions deter leaders from opportunistic behaviors and make dictators commit to power-sharing deals (Magaloni, 2008). Political institutions also reduce coup risk by enabling regime elites to monitor the behavior of dictators by institutionalizing regular interactions between dictators and regime elites (Boix & Svolik, 2013;Svolik, 2012). Enhanced transparency, in turn, facilitates the dictator's compliance with the power-sharing rules and reduces regime elites' skepticism about the dictator making good on those promises. In short, a scholarly consensus is that nominally democratic institutions, such as parties and legislatures, reduce the probability of coup attempts.

Theory
Though these studies improve our understanding of the roles of political institutions in reducing the probability of coup d'etats, they overlook fundamental differences in coup type. We claim that the effectiveness of political institutions in preventing coup attempts crucially depends on both the type of plotters and their political goals. When plotters are members of the ruling coalition who have access to the regime's decision-making and political power, their primary goals are to make the dictator commit to power-sharing agreements. By providing the ruling coalition with the institutional means of achieving this goal, political institutions reduce the would-be plotters' incentives to punish the dictator for opportunistic behavior. On the other hand, when regime members outside of the ruling coalition aim to change the entire group of ruling elites who hold power and the way policy decisions are made, political institutions have no influence on coup attempts. Their goals to overthrow the regime are incompatible with the values and rules underlying the existing institutions and thus are not achievable within the current political system. Without providing the means of achieving plotters' goals, political institutions do not reduce the necessity that these individuals take violent steps to achieve their goals.
For the discussion below, we first define central actors. We call individuals and groups that have positions within the state apparatus (e.g., the government and the security apparatus) regime members. As the term coup d'etats refers to an attempt by individuals within the state apparatus to remove the sitting head of government using unconstitutional means (Powell & Thyne, 2011), potential coup plotters are by definition regime members (e.g., Svolik, 2012). We then further disaggregate regime members into (a) coalition insiders and (b) coalition outsiders. Coalition insiders are those who hold key positions with power to influence policy and receive political perks inside the ruling coalition. They are similar to the "inner circle," the "leadership group," and the "ruling group" in other studies. Members of the ruling body of military juntas, such as the Supreme Council of the Armed Forces in Egypt, or members of the politburo central committee in communist countries, are examples of coalition insiders. On the other hand, coalition outsiders are regime members, but are outside the ruling coalition. Even though they have positions within the executive branches, the government or the repressive apparatus, they are sidelined within the regime and excluded from key positions with power to influence policy and enjoy political privileges.
Drawing on Aksoy et al. (2015), we distinguish between two different types of coups: (a) reshuffling coups that replace an incumbent leader while preserving the power of the incumbent ruling coalition and (b) regimechange coups that overthrow the entire autocratic regime by ousting the leader and the leader's primary support coalition. Here a regime is defined as "a set of formal and informal rules that identify the group from which leaders can come and determine who influences leadership choice and policy" (Geddes et al., 2014, p. 314). Regime-change coups thus intend to replace the group of elites who have the capacity to control leadership selection and policy choices with another group of elite, while reshuffling coups aim to preserve the regime and replace a leader with another who belongs to the ruling coalition in power. From 1946 to 2010, reshuffling coups account for approximately 38% of all coup attempts (117 times), while regime-change coups account for 62% (193 times).
Since reshuffling coups and regime-change coups are fundamentally different in terms of the characteristics of potential plotters and political goals they aim to achieve, whether and how political institutions affect the likelihood of a coup depends on the type of coups under consideration. In this section, we will first elaborate our theoretical logics of (1) why authoritarian institutions will diminish the probability of reshuffling coups, and then explicate the logics of (2) why authoritarian institutions do not reduce the likelihood of regime-change coups.

Why Authoritarian Institutions Reduce the Probability of Reshuffling Coups
Reshuffling coups are often executed by coalition insiders who want to replace the incumbent leader but preserve the existing regime and their power within it (Aksoy et al., 2015, p. 429). The same group of ruling elites will continue to rule the country, and the formal and informal rules for choosing leaders and deciding policies will remain intact after reshuffling coups. The ruling coalition consists of both civilian and military elites that hold key positions in the regime. They enjoy privileges, rents, and policy influence in the regime in exchange for tacit or manifest support for the dictator and the regime.
The primary goals that members of the ruling coalition seek to achieve through reshuffling coups are to make a dictator commit to share enough power and resources with them and/or to punish the dictator for any violation of power-sharing agreements or policy failures. Though the dictator must sign on to a power-sharing agreement to win the support of the ruling coalition, such agreements are plagued by a commitment problem due to the lack of an independent authority to enforce the deal (Svolik, 2012). In order to deter the dictator from reneging on the power-sharing deal, the ruling coalition needs to make credible threats of punishment. As Boix and Svolik (2013) nicely put it, "[p]ower-sharing in authoritarian regimes is . . . ultimately sustained by the ability of the dictator's allies to credibly threaten a rebellion that would replace the dictator should he violate the power-sharing agreement (p. 300)." Once the ruling coalition observes the dictator's transgressions or policy failures, it will launch a coup to remove the dictator (Aksoy et al., 2015;Sudduth, 2017). In this way, reshuffling coups serve as an accountability mechanism in dictatorships and help the ruling coalition attain their goals by punishing opportunistic or incompetent dictators and deterring future dictators from accumulating power.
Political institutions decrease the probability of reshuffling coups because they provide the ruling coalition with alternative mechanisms to hold dictators accountable to other elites by providing ex ante and ex post constraints (Weeks, 2012). That is, political institutions in dictatorships diminish incentives for coalition insiders to attempt reshuffling coups. First, they can deter the dictator's opportunism against coalition insiders by facilitating collective actions among ruling elites (Boix & Svolik, 2013;Gehlbach & Keefer, 2012). Gehlbach and Keefer (2012) argue that party institutionalization increases capacity for collective actions among regime elites by increasing information flows within the party and weakening the dictator's capability to prohibit coordination among regime elites. Authoritarian legislatures and parties also provide an institutional forum for elites to meet and coordinate defenses against the dictators' opportunistic behavior (Albertus & Menaldo, 2018;Gehlbach & Keefer, 2012). By enabling ruling elites to credibly threaten to stage a coup against the dictator, political institutions make the dictator more likely to comply with power-sharing deals (Magaloni, 2008).
Second, political institutions enable regime elites to monitor the behavior of dictators because they institutionalize regular interactions between dictators and regime elites in high-level, deliberative, and decision-making bodies within authoritarian parties and legislatures (Boix & Svolik, 2013;Svolik, 2012). Enhanced transparency, in turn, facilitates the dictator's compliance with power-sharing rules and promotes regime stability, as any opportunistic behavior to violate the power-sharing arrangement could be detected by members of the ruling coalition. To the contrary, imperfect information about the dictator's behavior can fuel ruling elites' misperceptions and leads to unnecessary coups. Political institutions thus resolve informational problems that plague the task of maintaining power-sharing pacts between a dictator and the ruling coalition.
Last, authoritarian parties and legislatures allow members of the ruling coalition to participate in policymaking and exercise some influence on policy and personnel decisions (Boix & Svolik, 2013;Geddes, 2003). This may reduce the likelihood that a dictator chooses policies that disfavor ruling elites. Furthermore, when leadership selection and personnel appointments are made according to institutionalized rules, the need to resort to violent coups for replacing leaders is diminished (Boix & Svolik, 2013). For instance, in Mexico under the Institutional Revolutionary Party (PRI), the top leadership of the ruling party controlled the nomination processes for leaders, which enabled ruling elites to govern their country without suffering violent internal struggles.
The above-mentioned roles of institutions in enabling coalition insiders to deter dictators' opportunism apply not only to civilian members of the ruling coalition, but also to military elites who are most critical in executing a coup. According to the data by Geddes et al. (2018), for example, 61% of country-year observations with support parties include military officers as members of the party executive committees, which is the most powerful decision-making body within a party institution such as politburo, standing committee and presidium. This number is even higher (66%) if we include the party executive committees that include military officers who had been members of insurgent or revolutionary armed forces before the regime's accession to power. 5 By allowing them to participate in the high-level, deliberative, and decision-making bodies, political institutions allow not only civilian elites but also military elites to coordinate actions, monitor dictators' behaviors and reflect their policy preferences in the decisionmaking processes.
As an illustration of our proposed mechanisms, consider how the institutional reforms promoted by Deng Xiaoping have alleviated commitment and monitoring problems within the Chinese political system. Departing from politics under Mao, under Deng's leadership, key Party bodies began meeting regularly, based on formal rules of consultation and consensual decision making in the 1980s. The Politburo Standing Committee began meeting weekly, and since 1987, the full Politburo has held regular monthly meetings. New institutional rules were set that required the General Secretary of the Party to report about the work of the Politburo Standing Committee to the full Politburo and similarly to report on the full Politburo's work to the Party Central Committee (Miller, 2008). These institutional reforms allowed more frequent interactions among the ruling elite members and helped them coordinate their actions. They also increase transparency helping the ruling coalition detect dictators' potential opportunism (Svolik, 2012, p. 92). Indeed, scholars on Chinese politics agree that such institutional reforms initiated by Deng mitigated policy disputes among ruling elites, prevented power accumulation by any single individual, and stabilized leadership transitions (e.g., Li, 2010;Miller, 2008).
In sum, parties and legislatures help neutralize coup threats from coalition insiders by increasing commitment by the dictator to power-sharing deals. When well-institutionalized political parties and legislatures exist, they can discipline and constrain dictators and/or deter their opportunism against ruling allies, reducing the need for reshuffling coups to punish dictators for reneging on power-sharing agreements. However, when the institutionalized accountability mechanism is absent, the ruling coalition has little choice but to rely on coups to hold the leader accountable for his violations of power-sharing or policy failures. The theoretical logics mentioned above suggest that authoritarian institutions reduce the risk of reshuffling coups in dictatorships.

H1
Authoritarian institutions, such as political parties and legislatures, decrease the likelihood of reshuffling coup attempts.

Why Authoritarian Institutions Do Not Diminish the Probability of Regime-Change Coups
While plotters in reshuffling coups have an interest in preserving their power and position within the regime, plotters in regime-change coups aim for a total regime overthrow (Aksoy et al., 2015). 6 Regime-change coups would change the set of formal and informal rules that identify the group from which leaders can be chosen and that determine who influences policy decisions. Plotters in regime-change coups seek to oust the ruling coalition in power and install a new group of ruling elite. Therefore, regime-change coups are more likely to be conducted by coalition outsiders who have positions within the regime but are outside the ruling coalition. Most crucially, coalition outsiders have no chance of rising into leadership positions as they are not members of the ruling group from which leaders are chosen.
For example, regime-change coups that oust a monarchy and replace it with a military junta are conducted by militaries outside the group of ruling elites that control leadership selection and policy making (i.e., a despotic family). A regime-change coup also occurs when a group of lower-ranked officers overthrows the incumbent military regime ruled by high-ranked military officers, or when a military regime ruled by certain ethnic groups is ousted by military officers from rival ethnic groups. The coup that replaced the military dictatorship led by Colonel Saye Zerbo in Burkina Faso with another military dictatorship led by Captain Thomas Sankara is an example of a regime-changing coup since the two military factions represented different ethnic groups (Geddes et al., 2018, p. 7).
We claim that political institutions, such as political parties and legislatures, do not affect the likelihood of regime-change coups for several reasons. First, unlike in reshuffling coups, plotters' goals in regime-change coups are incompatible with the existing regime; thus, activities within political institutions of the current regime cannot help plotters achieve their goals. As we elaborated above, political institutions reduce the necessity of reshuffling coups because potential plotters' goals are attainable through their representation within existing institutions. However, when the primary goal of plotters is to change the group of ruling elites who select the leader and make key decisions, it is incompatible with the values and rules of existing institutions (Aksoy & Carter, 2014). Institutionalized accountability or commitment mechanisms in dictatorships are unlikely to help potential plotters replace the entire group of ruling elites as the existing ruling coalition controls the way these institutions function. These mechanisms are created to improve the regime's prospects for survival (e.g., Brownlee, 2007;Magaloni, 2006) and are not suitable for ousting the regime. Authoritarian parties and legislatures do not provide a peaceful alternative to regime change and thus do not diminish the necessity that plotters resort to violent methods of achieving their goals (i.e., regime-changing coups).
Second, even the role of political institutions in co-opting potential coup plotters does not help to reduce the risk of regime-changing coups. Co-optation is usually considered to appease potential elite challengers and discourage their rebellion. Co-optation through ruling parties and legislatures, however, does not fully address grievances of discontented coalition outsiders. Imagine, for example, military officers are unsatisfied with the current regime where royal family members enjoy privileged access to key decision-making bodies and resources. Though participating in legislature or parties might increase the size of political and economic benefits enjoyed by these unsatisfied officers (Gandhi, 2008), most political and economic powers, including the ability to make key decisions about how political institutions work, are still controlled by royal family members. 7 Most crucially, regardless of their participation in existing institutions, these officers have no chance of rising into state leadership positions because they are not members of the ruling group from which leaders are chosen. Unless they change the group of ruling elites from one of royal family to the one of military officers, the officers' desires for a central role in leadership selection and exclusive access to political power would not be satisfied.
Finally, while the institutionalized regime's ability to mobilize mass opposition to attempted coups can deter regime-change coups, this diminishing effect of political institutions on regime-change coups would be offset by institutions' effects in increasing the feelings of relative deprivation and grievances among coalition outsiders. As failed coups are costly for plotters, individual elites would participate in coups only when they expect that the coup is most likely to succeed. For coups to succeed, the public's perception of the regime would be crucial. Tactically successful coups can be overturned by widespread disapproval among the general public (Galetovic & Sanhueza, 2000). Mass support toward the regime would discourage regime-change coups, while widespread mass dissatisfaction with the government would increase the risk of regime-change coups (Aksoy et al., 2015). In this vein, Geddes (2005) argues that political parties can deter coups in dictatorships by building mass support for the incumbent regime and increasing the likelihood of widespread post-coup protests. Ruling parties can increase popular support by providing political concessions and supplying benefits. Party members, tied to the pre-existing networks and relationships with the incumbent regime, play a critical role in mobilizing and organizing post-coup protests since they enjoy targeted benefits in return for support and thus develop a vested stake in the regime's survival (Geddes et al., 2018;Magaloni & Kricheli, 2010). Further, some ruling parties, such as the United Socialist Party of Venezuela and the Islamic Republic Party in Iran, have paramilitary wings whose coercive capacity can counterbalance coup threats. Authoritarian legislatures also mobilize public support toward the regimes by allocating economic rents and local public goods to citizens (e.g., Geddes et al., 2018). Deputies and regime supporters who compete for deputies' offices play a central role in increasing mass support toward the regime. They are motivated to extend their distributive networks to citizens and gather and report information from citizens to regimes to secure deputies' offices (Geddes et al., 2018). Authoritarian regimes can also use legislative debates to increase mass support toward the regime by, for example, allowing public debate on salient issues in the legislatures (Schuler, 2020) and signaling the regime's attentiveness to citizen concerns by passing a law on salient issues (Truex, 2020). In short, both ruling parties and legislatures can exert a deterrent effect on regime-changing coups as they mobilize mass support toward the regime and increase the likelihood of mass protests against attempted coups. 8 At the same time, political institutions would have positive effects on the likelihood of regime-change coups as they increase the feeling of relative deprivation and grievances held by coalition outsiders. As we discussed above, political institutions increase the share of political and economic benefits for those who participate in the institutions. In many instances, however, coalition outsiders do not have opportunities to participate in political institutions and receive political and material benefits associated with the institutions. Potential plotters who oppose the regime in the military, for example, tend to be based outside the capital city because a dictator who has cause to fear their threats to his tenure will send them away (Aksoy et al., 2015, p. 431). Their location outside the capital city makes it almost impossible for them to participate in political institutions and receive the associated benefits. Relatedly, the majority of seats in the legislatures and key positions in the regime party are typically distributed to high-level civilian officials and top-ranking military officers. Therefore, middle-or low-ranked military officers, who are often the key plotters organizing regime-change coups in some cases, are unlikely to have opportunities to have representation within political institutions. 9 Seeing that political institutions increase the share of material and political benefits for the group of ruling coalition, coalition outsiders who do not receive such benefits would perceive relative deprivation and increase their grievances against the regime (Woo & Conrad, 2019). Political institutions increase the power and wealth disparities between those who would receive benefits associated with the institutions and those who do not receive such benefits. As coalition outsiders are typically unable to obtain such benefits, institutions would increase their frustration and incentives to challenge the regime.
In sum, authoritarian political institutions produce countervailing effects on regime-changing coups: they increase the disposition to attempt a regimechange coup by increasing individual grievances among coalition outsiders against the regime, while the institutions' ability to mobilize mass opposition to attempted coups would deter coalition outsiders from attempting a regimechange coup. We expect that these two opposite effects are likely to cancel out each other and, thus, the total effect of political institutions on regimechanging coups should be negligible.
The above theoretical discussion produces our second testable hypothesis: H2 Authoritarian institutions, such as political parties and legislatures, have no effect on the likelihood of regime-change coup attempts.

Data
We test our hypotheses with time-series cross-sectional data at the country- To measure the existence of political institutions, we use the data on political parties and legislatures obtained from Geddes et al. (2018). As coups are organized by individuals who have positions within the state apparatus, we want to capture political institutions that primarily benefit regime members, not opposition groups excluded from power. Support Party is a dichotomous variable that identifies whether a support party exists in a country-year. It is coded as one if a party exists for the explicit purpose of supporting the current regime, defined as the current set of rules for choosing leaders and policy, or its leaders (Geddes et al., 2017, p. 10). Legislature is a dichotomous variable to capture whether a legislature exists. It is coded as one if a legislature, defined as "a body capable of enacting laws that is different from the executive," exists (Geddes et al., 2017, p. 17). It is coded as zero if a legislature does not exist in a country-year. Note that, unlike opposition groups, regime members, including ruling party members and front groups, always occupy a majority of seats in legislatures (Geddes et al., 2017, p. 17). Last, we also use a binary variable that is coded as one if both a party and a legislature exist, and zero otherwise. As we explain subsequently, we also try several different measures of political institutions in our robustness tests and show that the results are consistent. 10 We also include several control variables in our analyses. To incorporate the literature's finding that the economic well-being of countries affects the probability of coup attempts (e.g., Kim, 2016;Londregan & Poole, 1990), we include Ln(GDP/capita) and Economic Growth. Ln(GDP/capita) measures the natural log of a country's gross domestic product per capita, lagged by one year. Economic Growth captures the level of economic growth, lagged by 1 year. These economic indicators are obtained from the Penn World Tables dataset (Feenstra et al., 2015). We also include the 1-year lag of the natural log of a country's population, taken from the Penn World Tables dataset.
As coups are typically organized and executed by the military or other security apparatus, a dictator whose base of support is a group of high-ranking military officers might be more vulnerable to coup threats. To capture this intuition, we include Military, a dichotomous variable that identifies whether the regime leader was a member of the military before assuming power. This variable captures the above-mentioned vulnerability because it identifies the regime leader's career experience where he is most likely to have developed his most useful support network (Geddes et al., 2017, p. 25). The variable is obtained from Geddes et al. (2018).
Last, we include a binary post-Cold War variable capturing the post-Cold War period. The period witnessed a decline in the number of coups as well as an increase in the proportion of autocracies with democratic institutions. The literature also argues that countries that have experienced a coup in the recent past are more likely to experience a coup in the present (e.g., Belkin & Schofer, 2003;Londregan & Poole, 1990). To capture "the coup trap" phenomenon, we include Time since last reshuffling coup and Time since last regime-change coup to measure how many years have passed since the last coup of either type occurred in the same country. Following Carter and Signorino (2010), we include the cubic polynomial functions of years since the last coup, such as t , t 2 , and t 3 . Table 1 presents the results of nine slightly different logit models of coup attempts with standard errors clustered by country. Before testing our arguments about the divergent effects of political institutions on different types of   coup attempts, we first estimate models that pool all coup attempts without distinguishing between regime-change and reshuffling coup attempts (Columns 1-3 in Table 1). The results show that the coefficients on both Party and Legislature are negative, although only Party is statistically significant at the 5% level. As some authors focus on how institutions shape the likelihood of dictators being ousted by coups (e.g., Boix & Svolik, 2013), in the supporting appendix, we also explore the impacts of political institutions on the likelihood of successful coups. Supplemental Table A4 reports that both Party and Legislature have negative and significant effects on successful coups. Thus, the results confirm the existing literature's claim that powersharing political institutions, such as political parties and legislatures, diminish the threats of coup d'etats in general. Below, however, we reveal that this diminishing effect of political institutions on the likelihood of coup attempts is primarily driven by the effect of institutions on reshuffling coups, and that the effects of political institutions are dependent on the type of coup. We explore how political institutions impact different types of coup attempts and test our hypotheses. Columns 4 to 6 examine occurrences of reshuffling coup attempts, while Columns 7 to 9 analyze regime-change coup attempts as the dependent variables. Consistent with our hypothesis 1, the results show that the existence of political institutions reduces the likelihood that a dictator experiences reshuffling coup attempts. This can be seen by the negative and statistically significant coefficient estimates on Party and Legislature in Columns 4 to 6. Compared to models for aggregated coup attempts reported in Columns 1 to 3, the negative effects of both Party and Legislature on reshuffling coups are stronger in terms of both statistical significance and substantive importance. For example, the coefficient on Party for aggregated coups is −0.35 in Column 1, while the coefficient on Party for reshuffling coups is −0.81 in Column 4. Similarly, the coefficient on Legislature for aggregated coups is −0.14 in Column 2, while the coefficient on Legislature for reshuffling coups is −0.56 in Column 5.

Results
To evaluate the substantive effect of political institutions, we calculate the differences in the predicted probabilities of reshuffling coup attempts when we change the value of political institutions from 0 to 1. We adopt the observed value approach, setting all other covariates to the values observed for each observation and obtain average effects. The probability of reshuffling coups diminishes by 51% from .039 to .019 when we increase Party from 0 to 1. The effect of Legislature is similar to the effect of Party. Given the magnitude of the first differences, the impacts of both support parties and legislature on reshuffling coups are substantively meaningful. Table 1 also provides results that are consistent with our hypothesis 2 that authoritarian political institutions do not influence the probability of regime-change coups. The coefficient on neither Party nor Legislature is statistically significant (Columns 7-9). The coefficient estimate on Legislature is positive, while the coefficient estimates on Party and Party and Legislature are negative. They are also much smaller in magnitude than those in the models of reshuffling coups. Though a lack of statistical significance of the coefficients on political institutions is consistent with our hypothesis 2, we provide further support to the claim by testing whether political institutions have negligible effects. In doing so, we follow the methodological approach proposed by Rainey (2014). We calculate a substantive effect of political institutions on the probability of regime-change coups, construct a 90% confidence interval for the effect, and determine whether the confidence interval includes a threshold for a meaningful effect. For the threshold of a meaningful effect, we use a reduction in the probability of regime-change coups by 80% of the sample proportion (.042 .8 .034 × ≈ ). If political institutions decrease the likelihood of regime-change coups by at least 0.034, we determine that they have a meaningful effect on regime-change coups. On the other hand, if the lower bound of the confidence interval is smaller in magnitude than the threshold, it indicates that the effect of political institutions may be negligible. Figure 1 plots the 90% confidence intervals for the estimated effect of Party and Legislature and the meaningful impact thresholds defined above. In the left panel of Figure 1, we see that the confidence intervals lie entirely within the region of substantively negligible effects. It thus demonstrates that neither Party nor Legislature has a meaningful effect on the probability of a regime-change coup. This provides further evidence for our hypothesis 2. On the other hand, the right panel of Figure 1 shows that the confidence intervals for political institutions contain a meaningful impact threshold for reshuffling coups. Figure 1 provides us with strong evidence for our hypothesis 2. Political institutions, such as political parties and legislatures, do not have a meaningful impact on the probability of regime-change coups.
The negligible effects of political institutions on regime-change coups is an important finding because the literature typically posits that political institutions have diminishing effects on the likelihood that a dictator experiences rebellions from regime elites via coups (e.g., Boix & Svolik, 2013;Bove & Rivera, 2015;Geddes, 2005;Woo & Conrad, 2019). But our results demonstrate that political institutions, such as parties and legislatures, have reductive effects on reshuffling coups, but do not affect regime-change coups. Given that more than 62% of coup attempts are regime-change coup attempts, this finding has important implications for authoritarian regime survival and coups.
The results for the control variables also reveal the importance of distinguishing between reshuffling coups and regime-change coups. For example, GDP per capita is significantly associated only with a decreased probability of regime-change coups. This finding might imply that the level of economic development provides an opportunity for disgruntled regime opponents to overthrow the regime by affecting the regime's capacity to distribute patronage and co-opt regime opponents. However, it does not affect the likelihood of reshuffling coups, which usually serve as a mechanism for ruling elites to punish the leader. We also find that military dictators are more prone to reshuffling coups than civilian dictators, which is consistent with the previous finding (Geddes, 2003;Kim & Kroeger, 2018). On the other hand, there is little difference between military and civilian dictators when it comes to regime-change coups. Similarly, the post-Cold War period is statistically significant only in the models of reshuffling coups.  Figure 1. Effects of political institutions on regime-change versus reshuffling coups. We report the averaged marginal effects of political institutions with 90% confidence intervals around them, and the meaningful impact thresholds for regime-change coups (±.034) and reshuffling coups (±0.020). The effects of parties are estimated based on Columns 4 (reshuffling coups) and 7 (regime-change coup). Similarly, the effects of legislatures are estimated based on Columns 5 and 8, and the effects of the existence of both a support party and a legislature are estimated using Columns 6 and 9 in Table 1. As we mentioned in the text, any effect larger than the threshold is considered substantively meaningful.

Matching
One potential selection bias could arise if dictators establish parties and/or legislatures to neutralize threats to their political survival (Pepinsky, 2014). We should note, though, that if dictators are more likely to establish authoritarian institutions when they feel less secure in power, it will be more difficult to find supporting evidence for our first hypothesis. Additionally, we find that authoritarian institutions only influence the probability of reshuffling coups, not regime-change coups. Thus, the potential selection bias may not be serious here.
We use coarsened exact matching (Iacus et al., 2012) to create a comparison group that is as similar as possible to the set of country-years having authoritarian parties or legislatures. Using the matching analysis, we preprocess the data to minimize any potential differences between observations with and without authoritarian institutions before conducting the parametric analysis. This process helps to improve balance among observed features of treatment and control groups, although it does not address unobserved factors that may be associated with the assignment of authoritarian institutions. We match on military leaders, a log of GDP per capita, economic growth, a log of population, post-cold war, and years since the last coup attempt. Supplemental Tables A26 to A28 of the supporting appendix show that the coarsened exact matching significantly improves the balance across these covariates. We then re-estimate the models of Table 1 on the matched data set to control for remaining differences between the two groups.
Time Since Last Coup polynomials are included in all the analyses (for each type of coups), but coefficients are not reported in Table 2. Table 2 presents the estimates based on the matched data. The results are quite similar to the results reported in Table 1. Table 2 shows that our main results are robust to matching. Authoritarian institutions only significantly reduce the probability of reshuffling coup attempts, but not that of regime-change coup attempts.

Exploring Institutional Variations
We further consider variations in support parties and legislatures. The content of nominally democratic institutions rather than the existence of these institutions may be more important for explaining variations in coup risk (e.g., Meng, 2020;Miller, 2020). For instance, ruling parties can vary in the degree of independence from the dictator, the level of organizational power, the strength of ties to social organizations, and so on. Some legislatures only include ruling party members, while others allow opposition party    members. Therefore, we move beyond the presence of ruling parties and legislatures and explore whether the results are robust when we account for institutional variations. Table 3 investigates the effects of institutional variations using matched data based on coarsened exact matching. 11 We first examine variations in support parties. Columns 1 to 2 use a measure of autocratic ruling parties obtained from a new dataset developed by Miller (2020). Miller (2020) defines a ruling party "as a political party that is either the supreme ruling power or is used as a significant vehicle of power by the regime and is clearly preeminent among all parties" (7). This definition is thus more stringent than the definition of a regime support party (i.e., a party that seeks to maintain the current regime) used by Geddes et al. (2018). Columns 3 to 4 test the effects of Independent party in which "party executive committee has policy independence and autonomy from the regime leader" (Geddes et al., 2017, p. 24). 12 As independent ruling parties enable ruling coalition members to influence policy-making discussion and decisions rather than serving as a rubberstamp for the regime leader, we expect such parties to better facilitate power-sharing arrangements and exert greater reductive effects on reshuffling coups. However, we do not expect them to influence regime-change coups. Columns 5 to 6 use the Party with local organization variable obtained from the same Geddes et al. dataset. This variable captures the existence of regime support parties that have "local-level branch organizations that link party militants to citizens" (Geddes et al., 2017, p. 13). Such parties can increase mass support toward the regime by providing benefits to ordinary citizens and mobilize anti-coup protest utilizing local party officials and militants. Thus, they might possess a greater capacity to counterbalance the threat of regime-changing coups. Last, Columns 7 to 8 examine the impacts of Revolutionary party that is originally formed as a violent revolutionary organization using Miller (2020)'s data. As Levitsky and Way (2013) argue, armed liberation struggles and postrevolutionary state-building enhance revolutionary parties' abilities to mobilize mass support and facilitate elite cohesion, which might increase their abilities to deter regime-change coups.
Baseline controls are included in all the analyses, but coefficients are not reported in Table 3. The results in Table 3 show that the estimated coefficients on all party variables, except revolutionary party, are negative and significant for reshuffling coups. In contrast, none of these party variables have significant effects on regime-change coups. While parties with local organizations and revolutionary parties are better equipped to mobilize popular support, our results reveal that they do not deter regime-change coups. This finding seems to indicate that even for these parties with high counter-coup mobilization abilities, the deterrent effects of institutions on regime-change coups are not sufficiently large enough to outweigh the increasing impacts of institutions on grievances held among coalition outsiders. As we argued above, political institutions would increase the power and wealth disparities between coalition insiders who receive political and material benefits from the institutions and coalition outsiders who do not receive such benefits. This positive effect of institutions on coalition outsiders' grievances would offset the institutions' deterrent impact on regime-changing coups.
Similarly, we explore variations among authoritarian legislatures by focusing on who occupies seats in the legislatures. Non-opposition legislature captures the existence of legislatures that do not include deputies from opposition parties and thus primarily benefit the front groups and ruling party members. Opposition legislature captures legislatures that include deputies from opposition parties elected in competitive elections. These variables are obtained from Geddes et al. (2018). We separately conduct a matching analysis for each variable. Since we expect reshuffling coups to serve as an accountability mechanism in dictatorships, Non-opposition legislature representing only regime elites will be more effective than Opposition legislature in reducing the likelihood of reshuffling coups. We also expect that neither type of legislature affects regime-change coups. Both types of legislatures could increase mass support toward the regime by distributing private benefits to citizens (Geddes et al., 2018) and signaling the regime's attentiveness to citizen grievance (e.g., Truex, 2020). At the same time, they would also increase grievances among coalition outsiders against the regime. By increasing the shares of benefits for coalition insiders (Non-opposition legislature) and opposition party members (Opposition legislature), these legislatures increase the feeling of relative deprivation among coalition outsiders who do not have legislative seats, offsetting the deterrent effects of legislatures on regime-changing coups. Columns 9 to 12 provide supporting evidence for our expectations. We find that Nonopposition legislature has negative and significant impacts on reshuffling coups, while the effect of Opposition legislature on reshuffling coups is not statistically significant. The magnitude of the coefficient on Non-opposition legislature is also much greater than that of the coefficient on Opposition legislature. We also find that the effects of neither type of legislatures on regime-change coups are statistically significant, which is consistent with our expectation.

Robustness Tests
Several additional robustness tests provide further evidence for our hypotheses. Due to space constraints, the results of these robustness checks are discussed here briefly and presented fully in the supporting appendix. First, our results hold when we use successful coups as our dependent variables. The results in Supplemental Tables A4 and A5 show that political institutions reduce the probability of successful reshuffling coups, but do not affect regime-change successful coups. Second, the results are robust to controlling for the following additional variables: antigovernment protests, intrastate war, and interstate war (Supplemental Tables A6 and A7), coup proofing measures including military spending, military size and counterbalancing (Supplemental Tables A8 and A9), IMF programs (Supplemental Tables A10 and A11), and election outcomes and post-election protests (Supplemental Tables A12 and  A13). This holds true regardless of whether we use the matched data or not.
Third, we attempt to control for unobserved factors that may influence both authoritarian institutions and coup attempts. Supplemental Table A14 adds decade fixed effects to account for any time-varying shocks that are common to all countries. Supplemental Tables A15 and A16 attempt to control for timeinvariant country-specific factors by including region fixed effects or country random effects. Further, Supplemental Tables A17 and A18 estimate the within-between model by including the country-specific means of the explanatory variables and the deviations from the country-specific means (Mundlak, 1978). The deviations capture the within-country effects, while the country means estimate the between-country effects. We find that both the withincountry and the between-country effects are consistent with our main results.
Fourth, we investigate whether political institutions' impacts on regimechange coups depend on countries' regime types. In particular, some might argue that our proposed theory should only apply to non-military regimes. We address this issue by including the interaction terms between the political institution and the military regime (whether the country-year is military regime or not) variable. The results show that the coefficients on the interaction terms are insignificant, indicating that our proposed argument about the relationships between institutions and regime-change coups does not depend on regime types (Supplemental Table A20). The results are also robust when we control for regime types (see Supplemental Table A19).
Finally, the results are consistent when we simultaneously analyze reshuffling coups and regime-change coups using multinomial logit models. The multinomial logit models, including ones based on the matched data, are reported in Supplemental Tables A21 to A24 and support our hypotheses.

Conclusion
In this paper, we show that whether political institutions have diminishing effects on the likelihood of coup attempts crucially depends on the types of potential plotters and their goals. By providing an institutional means to address the ruling coalition's concern over a dictator's opportunistic behaviors, political institutions diminish the necessity of reshuffling coups where the ruling coalition punishes dictators who have reneged on the power-sharing agreements. However, when potential plotters aim to replace the entire group of ruling elites with a new group, political institutions do not provide the means to achieve such a goal and thus do not reduce the risk of regimechange coups. Further, while political institutions can increase mass support toward the regime and mobilize anti-coup protests, this deterrent effect of institutions will be offset by the institutions' increasing effects on the feeling of relative deprivation and grievances held by coalition outsiders. We test our arguments using the data on two types of coups and political institutions in authoritarian countries between 1946 and 2010 and find strong evidence to support our hypotheses. Political institutions, such as ruling parties and legislatures, are effective in reducing the probability of reshuffling coups, but not regime-change coups.
These findings are important precisely because they challenge the widely-held view among comparative authoritarian scholars that political institutions effectively help autocratic leaders and regimes stay in power by addressing threats from within the regime. Though dictators face multiple threats toward their political survival, an overwhelming majority of dictators lose power because of coup d'etats rather than popular uprisings or civil wars. Given this, the literature has highlighted the importance of roles that political institutions play in addressing coup threats-the most crucial threats toward leader and regime survival. Our findings, however, demonstrate that the literature's claim that political institutions reduce the probability of coups is true for reshuffling coups but not for regime-change coups. Given that more than 62% of coup attempts that have occurred in dictatorships from 1946 to 2010 are regime-change coups, our findings demonstrate that the creation of political institutions is not as effective in helping leaders and authoritarian regimes stay in power as the literature previously has argued.
Our paper also highlights the importance of carefully disaggregating political actors with distinct interests when evaluating political institutions' role in authoritarian regimes. Existing theories that explore authoritarian institutions' regime-preserving effects tend to exclusively focus on the power-sharing dynamics between dictators and their ruling coalition. Yet, more than 62% of coup attempts are led by individuals who have positions within the state apparatus but are outside the ruling coalition. Their goal is to replace the existing group of ruling coalition with a new group. Lumping together different actors with distinct interests within the government prevents us from fully understanding how institutions affect authoritarian regime survival. Future research should recognize this point and pay careful attention to significant variations among actors within the regime and variations among actors within the society in examining the roles of political institutions in authoritarian politics. 4. Singh (2014), Albrecht andEibl (2018), andDe Bruin (2019) show that coups executed by high-ranking officers have different causes and outcomes than from those executed by lower-ranking officers. Similarly, Aksoy et al. (2015) and Sudduth (2021) suggest the difference between coups conducted by members of the autocratic ruling coalition and those conducted by non-members. 5. We use the excomcivn variable to calculate these values. As Geddes et al. (2018) do not include military officers who transformed from former members of insurgent armed forces to regular military officers for the first 5 years since the regime's accessions to power, we exclude the first 5 years of regimes from our calculation. 6. We assume that coup plotters have different interests in terms of what they want to achieve via coups. Coalition outsiders who are outside the locus of power are more likely to want to transform the ruling coalition, compared to coalition insiders. See also Aksoy et al. (2015), Singh (2014), Albrecht and Eibl (2018), and Sudduth (2021) for the assumption that different coup agents have different incentives for coups. 7. Note that, as we argue below, the majority of coalition outsiders do not even have opportunities to participate in political institutions. 8. Some scholars, however, doubt the coup-deterrence effect of political institutions, arguing that plotters give little consideration to the public's response and coup attempts rarely face popular protests. See Singh (2014) and Brancati (2014). 9. As we explain above, regime-change coups are executed by high-ranked military officers as well as middle-or lower-ranked officers, depending on the type of the existing ruling elite. When the military is outside the ruling coalition (e.g., monarchy), regime-change coups are conducted by the top-ranked officers who represent the entire military organization. Meanwhile, when the ruling coalition consists of high-ranked officers, the group of middle-and lower-ranked officers could execute regime-change coups to replace the group of high-ranked officers. 10. To test our hypotheses, we need to compare the likelihood of reshuffling coups (or regime-change coups) (a) when political institutions such as support parties exist and (b) when there is no such institution. Using Geddes's regime type classification that compares one regime type (e.g., single-party regimes) with others (e.g., military regimes) is not appropriate in evaluating the roles of particular institutions (Svolik, 2012, 31). Because variations in the existence of certain institutions exist within the same regime type as well as across different regime types, the effects of institutions cannot be evaluated by comparing different regime types (Svolik, 2012, 21). As we report in the online appendix, our results hold when we control regime types. 11. The results based on matched data in Table 3 only report the coefficient estimates on "strong" institutions that are expected to have greater impact on coups, such as Independent party and Party with local organization since we treat them as a binary treatment. Accordingly, to present both coefficients on "strong" (e.g., Independent party) and "weak" institutions (e.g., Rubberstamp party), we report estimation results based on non-matched data in the supporting appendix. See Supplemental Table A2. The estimation results remain consistent. 12. In measuring this variable, Geddes et al. (2017) identify whether there is discussion of policy alternatives and disagreements and consider the absence of policy disagreements as an indication of the dictator's concentration of policy-making power. The excluded parties are those where the party executive committee serves as a rubberstamp for the regime leader.