Mass Revolutions vs. Elite Coups Ruth Kricheli and Yair Livne ∗ [email protected] [email protected] February 8, 2010 Abstract Most of the contemporary autocracies were not replaced by a democracy after their collapse. Instead, they were more likely to be replaced by a new autocracy headed by the group that initiated the autocrat’s overthrow. The paper provides a theory of authoritarian breakdown emphasizing the relationship between an autocracy’s economic performance and the probability that it will be replaced by a democracy and the probability that it will be replaced by another autocracy. Our theory is based on the interaction between three political classes: the dictator, who is interested in surviving in power; the elites, who are interested in overthrowing the dictator and in establishing a new autocratic regime wherein they enjoy more power; and the masses, who are interested in overthrowing the dictator and in establishing a democracy. The theory predicts that mass revolutions are likely in bad economic periods, authoritarian stability in good economic periods, and, surprisingly, coups in normal economic periods, not in bad ones. This prediction still holds when we extend our model to an incomplete information environment allowing the dictator to hold less accurate information about his own vulnerability than the elites do. We conclude by showing that this prediction is consistent with available data on authoritarian breakdowns from 1960 − 2000. Keywords: autocracies, political transitions, revolutions, coups JEL Classification Numbers: D72, D74 ∗ We thank Jim Fearon, Matt Jackson, Antoine Lallour, Beatriz Magaloni, and participants of Cowbell and the Comparative Politics Workshop at Stanford for helpful conversations and comments. 1 1 Introduction When a contemporary authoritarian regime collapses, it is not likely to be replaced by a democracy. A more probable scenario is that a new authoritarian regime, headed by the group that initiated the autocrat’s overthrow, will be established. This was indeed the case in almost 72% of authoritarian breakdowns from 1950-2006, while the alternative scenario, in which a democracy was established, took place in only 28% of the cases.1 Why are some authoritarian regimes replaced by democracy, while the majority of them are not? What is the relationship between economic performance and the probability that an autocracy will be replaced by a democracy or the probability that it will be replaced by another autocracy? This paper provides a theoretical framework for analyzing these questions and some preliminary evidence in support of our main theoretical findings. Our theory focuses on adverse regime changes wherein either the masses or the elites attempt to overthrow the autocrat. It emphasizes that under authoritarian rule, military and political elites have an interest in overthrowing the dictator and establishing a new autocratic regime wherein they enjoy more power. The masses, on the other hand, have an interest in overthrowing the dictator and establishing a politically open regime. Elites can attempt a coup, which, if successful, establishes another authoritarian regime, whereas citizens can attempt a revolution, which, if successful, completely destabilizes the current political order. The crucial difference between the elites’ ability to threaten the dictator and the masses’ is that in order to overthrow the regime, citizens have to coordinate mass uprisings while elites can threaten the regime almost single-handedly. This contrast is strikingly evident throughout history: four senior military officers backed by the country’s security forces ousted the Mauritanian president in 20082 ; the 2008 military coup in Guinea was set forth by a small number of soldiers and officers;3 and a small group of young officers overthrew the Egyptian King Farouk in the 1952 coup. The same was true in many other coups, including the 1985 coup in Nigeria, the 1981 coup in the Central African Republic, and the 1957 coup in Thailand. Even ancient dictators such as King Superbus, Julius Caesar, and Caligula were overthrown by only a few elite members. Conversely, about a million East German citizens participated in the 1989 fall of the Berlin Wall and another half million protested in Prague in the 1989 revolution in Czechoslovakia.4 1 Figures are taken form Table 1 in Magaloni and Kricheli (2010). See ”Army Officers Seize Power in Mauritania” (The New York Times, August 7, 2008) 3 See ”Guinea Coup Plotters Announce Curfew, New Leader” (Voice of Africa, 24 December 2008) and ”’It’s our turn for power’: Guinea coup leader tightens grip” (AFP, Dec 25, 2008) 4 See ”The Berlin Wall, 20 Years Later” (The American, November 9, 2009) and ”The fall of the Berlin Wall: 20 years later” (Wood, 2009) 2 2 This asymmetry makes elite coups and mass revolutions likely in different economic periods. Our theory suggests that the likelihood of coups and revolutions depends crucially on an autocracy’s economic performance: under autocratic rule, political stability is likely during economic prosperity; mass revolutions similar to the 1989 revolutions are likely during economic crises; and, surprisingly, elite coups similar to the 1952 coup in Egypt and the 2008 coup in Mauritania are likely in normal economic times, not during economic crises. The relationship between economic performance and adverse regime changes in autocracies holds even when we extend our theory to an incomplete-information environment. We follow the literature on autocrats’ difficulty in obtaining reliable information about their own vulnerability (Tullock’s 1987; Schatzberg 1988; Wintrobe 1998) and find that even under such circumstances, coups are only likely in normal economic times, and revolutions in bad economic times. This relationship between economic performance and adverse regime changes is consistent with the general patterns in available data on coups and revolutions in authoritarian regimes from 1960 − 2000. We examine this data and show that revolutions are most likely during periods of economic hardship, that coups are most likely in periods of normal economic performance, and that authoritarian stability is most likely in periods of economic prosperity. In addition to this relationship, our theory proposes two more implications: First, instead of inducing them to distribute rents and resources only to the elites, autocrats’ interest in survival in fact encourages them to provide goods and benefits to the masses in an attempt to build a basis of support among the citizens. This pattern is increasingly discernible in contemporary autocracies: most of the autocracies today use political parties to distribute rents and privileges to the people in return for their political support (Diaz-Cayeros, Magaloni, and Weingast 2001; Lust-Okar and Gandhi 2009; Magaloni and Kricheli 2010). The party controls work permits, land titles, jobs, housing, food, and many other privileges whose distribution becomes a key tool for building mass support (Geddes 2006,2008; Brownlee 2007; Magaloni 2008). Second, the dictator will distribute privileges to the citizens, but only to a pivotal subset of the citizens, whose identity is fixed over time. By constantly rewarding the same set of citizens, the dictator induces them to support the regime even in bad economic times, because they expect to receive benefits from the regime in the future. This expectation for future benefits enables him to distribute them with less resources in each economic period without risking a revolution. Indeed, contemporary dictatorships rely on the citizens, but not on all of them. Instead, dictators privilege only a subset of the citizens in exchange for their political support—party members in Communist regimes (e.g., Dickson and Ruble 3 2000; Dickson 2003) or members of an ethnic group in ethnic dictatorships (e.g., Bates 1981, 1983, 1989; van de Walle 2007) are examples of such pivotal groups. Our paper relates to several burgeoning strands of literature. First, the theoretical literature on regime transitions sparked by Acemoglu and Robinson’s (2000, 2001, 2006) work on transitions form democracy to autocracy and vice versa (Boix 2003; Boix and Stokes 2003; Acemoglu, Ticchi, and Vindigni forthcoming). Second, the formal literature on authoritarian regimes and the strategies autocrats follow to minimize threats to their stability (Wintrobe 1998; Kuran 1991a, 1991b; Lohmann 1994; Gandhi and Przeworski 2006; Debs 2007; Acemoglu, Egorov, and Sonin 2008; Boix and Svolik 2008; Gandhi 2008; Egorov, Guriev, and Sonin 2009). Third, the empirical literature on regime transitions and the relationship between political stability and economic performance (e.g., O’Donnell, Schmitter, and Whitehead 1986; Przeworski et al 2000; Boix and Stokes 2003; Epstein, Bates, Goldstone, Kristensen, and OHalloran 2006). Lastly, the early empirical literature about the relationship between economic growth and the likelihood of political revolutions (Davies 1962, 1969; Haggard and Kaufman 1995, 1997) and coups (Londregan and Poole 1990). To the best of our knowledge, this is the first paper that focuses on both elite coups and mass revolutions using a single theoretical framework. In many respects, the paper that is closest to ours is Bueno de Mesquita and Smith (2009), wherein the authors use Bueno de Mesquita et al.’s (2003) framework to examine occasions in which leaders risk deposition by challengers within the existing political rules and by revolutionary threats. Their convincing theory explains endogenous political change, as opposed to adversary regime changes wherein conflict is costly yet elite coups or mass revolutions in fact occur, which are the focus of this paper. The remainder of this paper is organized as follows: the next section discusses the asymmetry between the masses’ ability to threaten the dictator’s survival and the elites’. The third section presents our model of authoritarian breakdown, and our main results. The fourth section extends this model to an asymmetric information environment. The fifth section presents preliminary evidence in support of our theory based on data from 115 authoritarian regimes during 1960 − 2000. Concluding remarks are discussed in the sixth and final section. 2 Revolutions vs. Coups Authoritarian leaders cannot rule single handedly. Every authoritarian leader therefore forms an alliances with elite members who assist him with the task of ruling the country. 4 To fulfill their role in this alliance, elite members gain access to exclusive political and military resources necessary to operate the government apparatuses. Thus, at the heart of the stability problem of authoritarian leaders lies the following dilemma: On the one hand, every authoritarian leader has to rely on elite members and to bestow upon them the social, political, and military resources necessary for ruling the country. On the other hand, these exclusive resources enable the elites to threaten the regime’s stability. Elite members can use the resources bestowed upon them in order to overthrow the very same regime that supplies these exclusive resources as a means for the sake of its own survival (Wintrobe 1998; Haber 2006; Debs 2007a, 2007b; Acemoglu, Ticchi, and Vindigni 2008; Magaloni 2008; Egorov and Sonin 2009). Indeed, the most perilous threats to authoritarian regimes’ survival are carried out by former close allies of the regime (Zolberg 1966; Geddes 2008; Haber 2006). Moreover, for the following reasons, the initial support of only a handful of elite members is essential for a successful palace or military coup. First, politically and militarily powerful elites hold access to the necessary resources for setting out a coup. Second, these elites are often members of a single organization (a political party, the military, etc.) which mitigates the coordination and collective action problems associated with attempting a coup. Third, once a small group of elite members sets out the attempt, other elite members have an incentive to join them (Geddes 2003; Haber, Razo, and Maurer 2003; Nordlinger 1977). Fourth, in many cases, elites that currently play important roles in the governmental apparatus are the very same elites that originally put the dictator in power, meaning that they have already proved their ability to overthrow a regime (Haber 2006). Contrary to elite members, individual civilians cannot single handedly threaten the stability of the regime. Civilians have no access to arms, military training, economic funds, or many of the other resources needed for a rebellion. In order to threaten the regime’s stability, civilians therefore have to cooperate. To be successful, civilian rebellions need to be supported by a sufficiently large portion of the population, yet such mass uprisings require that the public overcome severe coordination and collective action problems. When the public indeed manages to overcome these coordination problems, and mass insurrections are set forth, the public becomes extremely powerful. The military is often reluctant to use violence against the masses (Geddes 2006, 2008). Moreover, even if the military does support the regime and tries to control the masses using force, this task becomes impossible when enough people are in the streets. When they are numerous enough, citizens can even defeat the military, as Erich Mielke, the chief of Eastern German police, said to Erich Hoenecker, the party leader, after a mass demonstration: ”Erich, one just cannot beat that many people” (Przeworski 1991). In addition, contrary to elite coups that usually only entail 5 a leadership change that does not influence the social and political fabric, mass revolutions completely destabilize the social and political order (Geddes 2003, p. 66). Revolutions can change the structure of political hierarchies and the rules of political contestation entirely. There is thus an inherent asymmetry between the political power elite members enjoy and the political power citizens enjoy under authoritarianism: to survive, authoritarian leaders have to appease both the elites and the people, yet the elites are less subject to coordination problems than citizens are, and when the citizens are in the streets in their masses, almost no force can defeat them. 3 3.1 A Model of Authoritarian Stability The Basic Model Basics We consider an infinite horizon society, with an authoritarian ruler d, a group of n elite members, and m citizens. Elites are privileged members of society who hold government-related offices and thus have access to political, economic, military or intelligence resources which are not accessible to other members of society. The identity of these elites may change in different types of regimes. For example, in one-party regimes they are often party cadres whereas in military regimes they are often military officers. The ruler of the country can either be a single individual, as in the case of personalistic autocratic regimes, or a group of highly influential political leaders who rule the country together, as in the case of several military regimes in which the army’s top ranking officers are members of a ruling council. Citizens are non-ruling members of society who have no access to political or military resources. For the sake of simplicity we will treat all elite members as identical and all citizens as identical. The Economy In each period t the economy generates a surplus which can be distributed as rents by the ruler and is given by Yt , a non-negative random variable with density f , taking values between y and ȳ. We assume that this surplus is distributed identically and independently across periods. This surplus should not be thought of as the entire production of the country’s economy, but rather only the parts of the state’s budget which the ruler can freely allocate as spoils. Timing of the Game In each period, the ruler first decides on the allocation of available rents between himself, elite members, and ordinary citizens. Specifically, he decides on a 6 fraction wti ∈ [0, 1] of Vt to be allocated to citizen i, 1 ≤ i ≤ m, a fraction wte ∈ [0, 1] to be allocated to elite member e, m + 1 ≤ e ≤ m + n, and a fraction wtd ∈ [0, 1] of Yt to himself, with the ruler’s budget constraint being: m X i=1 wti + m+n X wte + wtd = 1 e=m+1 Following the distribution decision, Yt is realized. The reasoning behind this ordering is that rents in society are allocated through institutional mechanisms such as government jobs, party positions, and the allocation of franchises and licenses. These sorts of mechanisms allow the dictator to credibly commit in advance to allocating the pie of available rents, as they cannot be changed overnight. Establishing powerful courts can further assist the dictator in credibly committing to an allocation in advance.5 After the realization of Yt , both the citizens and the elites observe the state of the economy and the ruler’s distribution decision. First, each citizen decides sequentially whether or not to support a political revolution, where the sequence of citizens is picked at random and citizens’ choices are observable to subsequent citizens.6 Finally, each elite member can opt to initiate a coup.7 To clarify, as depicted in Figure1 the order of the one-period game repeating in every period is as follows: the ruler first determines the shares of the rents each member of society will receive; the stochastic realization of the state of the economy Yt is realized; citizens and elites receive payoffs through institutional mechanisms; citizens sequentially decide whether to support a revolution, and finally elite member simultaneously decide whether to initiate a coup. Revolutions A revolution occurs if at least k + 1 citizens decide to support a revolution in a given period. The ruler is then overthrown and society goes into turmoil, causing a fraction β of the period’s rents to perish, where 0 < β < 1. The revolution is assumed to utterly destabilize the current political structure, and therefore, after the revolution, society can either become a democracy or a new authoritarian regime supported by a new elite. To reflect the political turmoil associated with a revolution, we assume that ex-ante, before the 5 See Myerson (2008). We assume that citizens make their decisions sequentially to get rid of degenerate equilibria where no citizen supports a revolution, or all citizens always support one. Alternatively, we can assume citizens move simultaneously and act ”as if pivotal”, a common assumption in the voting literature 7 The results would not change if the coup decisions happened simultaneously with citizens’ decision to support a revolution, or if elites acted first. 6 7 Figure 1: Order of the One-Period Game new political structure emerges, all citizens and elite members have uniform expectations regarding how they will fare after the revolution. Thus, each one of them expects that in 1 each future period, his share of society’s rents will be m+n . This assumption is not crucial in any way to the analysis, and changing future expected payments after a revolution to any fixed amount would not substantively affect the results. For simplicity, we do not model an individual cost for citizens supporting a revolution, but adding one would not change the nature of the results as we discuss in section 3.2. Coups If one or more elite members attempts a coup, the attempt is successful with a probability p, independently of the realization of the economic variable Yt . This is to correspond with the asymmetry between the elites and the citizens, according to which elite members hold the power to single-handedly threaten the regime, either because elite members have an incentive to support a coup once it is initiated by other elite members (Geddes 2003), or because elite members have access to exclusive organizational, political, and military resources. Upon success, the autocratic ruler is replaced by a new autocratic leader, with one of the current elite members becoming the new ruler. We assume that once 8 a coup has been initiated each member of the elite is equally likely to become the new ruler, regardless of the identity of those who chose to initiate the coup. Following a coup attempt, a new autocratic regime is established in the country. Other than the new ruler, previous elite members retain their status as elite members. This is to correspond to the fact that coups usually result only in a change in leadership without any deep changes in the political structure and the rules of political contestation (Geddes 2003). Since the number of elite members represents the number of privileged office holders in the country, we assume a new player enters the game in the position of an elite member, and the number of elite members remains n. The previous ruler is completely excluded from society and its economic resources—he is either sent to exile, to prison, or to death row. In the following period, the new ruler—one of the elite members in the previous period—announces a new distribution of rents, and the game proceeds as in the previous periods. Coups do not cause wide scale damage to the economy8 . However, following an unsuccessful coup, the authoritarian leader prevails, and all members of the current elite suffer a cost reflected by a share c of their current period income. This cost reflects the leader’s retaliation and purges against the current elite due to the coup attempt. Revolutions Trump Coups If the elite attempts a coup in the same period in which at least k citizens support a revolution, then the coup attempt has no influence on the political outcome. The reason for this is that when a sufficient number of citizens supports a major political transition, we assume that neither the previous regime nor the newly established authoritarian regime can stop it, nor can the dictator impose any costs on the elites in the midst of a revolution. This stems from our understanding of a revolution in the model as an event which fundamentally changes the nature of the political order, while a coup against a dictatorship only replaces the identity of the ruler, but not the nature of the regime (Geddes 2003). Preferences All agents’ utilities are determined solely by the rents they consume, with all agents having the same utility function, represented by a linear indirect utility function over net income and a discount factor δ ∈ (0, 1). Since the ruler uses a credible commitment mechanism to transfer the announced distribution of goods to the other agents, this allocation is assumed to be ”sticky”—even if there is a coup or a revolution, the distribution of rents in the period of the revolution or 8 This assumption is not restrictive— if we assume that following a coup a share 0 < α < β < 1 of current surplus is destroyed, then the results do not change qualitatively, though the ranking between periods with coups and periods with stability may blur. 9 the coup remains proportional to the ruler’s initial distribution, taking into account the costs associated with coups or revolutions. In the following periods, however, the distribution of goods results from the new ruler’s distribution-decision. Thus, in a period of a successful revolution, citizen i’s expected utility from the present and future is, for example, δ δ E [Yt], and elite member e’s expected utility is βwteYt + (1−δ)(m+n) E [Yt]. βwti Yt + (1−δ)(m+n) The reason for this stickiness is that the mechanisms by which regimes transfers goods to citizens or to elites usually require an institutional infrastructure which cannot be abolished overnight. 3.2 Analysis and Main Result The solution concept we use is Markov Perfect Equilibrium (MPE). That is, we look for a profile of strategies for the players that constitute perfect equilibrium with respect to the payoff relevant history.9 Note that in our setting, the authoritarian ruler makes the distribution decision before the realization of Yt , which is an independent, identically distributed random variable across periods. Hence, a Markov strategy for him cannot be conditioned on the state of the world nor the history of play, and must be a fixed, pure or mixed strategy. Citizens and elite members, on the other hand, decide whether or not to attempt an overthrow after the state of the economy is realized. Hence, during a single dictator’s rule, their Markov strategies can depend only on Yt . To simplify the analysis and focus on the first order mechanisms behind the occurrences of intra-elite coups and mass revolutions, we regard the political positions – dictator, elite members and citizens – as the players in the game, in the context of solving for a MPE. Thus, we assume that all dictators throughout history use the same strategy in a MPE and elite members and citizens do not condition their strategies on the identity of the current dictator. This assumption is in line with much of the political transitions literature, which often considers political classes as the actors in the game (e.g. Acemoglu & Robinson, 2001). All proofs are provided in the appendix. We first prove that a MPE exists in the game we described. To do this we modify the proof of Theorem 1 in Jackson and Morelli (2009) to our model. Theorem 1. There exists a MPE in the above game. We can now proceed to our main result, the characterization of the structure of MPEs in which dictators play a pure strategy. 9 On the definition of MPE in non-stationary games see Fudenberg and Tirole (1991, section 13.2). 10 Theorem 2. Any MPE of the game, in which the dictator plays a pure strategy, is defined by two economic thresholds sR ≤ sC : when Yt < sR citizens start a revolution, when sR < Yt < sC the elites attempt a coup against the dictator, while citizens do not revolt, and when Yt > sC the regime is stable. In any such MPE, the dictator allocates a share w1 ≥ 0 of the rents to some privileged set of m − k citizens, while the remaining k citizens receive no rents, and allocates a fixed share w2 ≥ 0 to all elite members. While we present the full proof in the appendix, it is instructive to understand the main mechanisms behind it informally. In a given period, with the state variable (which is the realization of available rents) being the realized Yt , we first look at the elites’ decision. If a revolution has started earlier in the period, the elites can no longer effect the political outcome. If a revolution has not started, elites must decide whether to initiate a coup or not, conditioning on Yt . A successful coup carries with it benefits which are independent of the state variable – namely the possibility of becoming the new dictator and earning high expected rents in future periods, while a failed coup leads to costs to the elite which are proportional to the state variable, in the form of lost rents. Thus, elites will prefer a coup for low levels of the state variable, and prefer stability for high levels. Each elite member has the power to start a coup, and thus the elite member least satisfied with the current regime, in terms of the current share of rents he is receiving, will be pivotal. We will denote the level of the state variable where he is indifferent by sC . Ordinary citizens in turn choose, according to a random sequence, whether to go out into the streets and support a revolution. Like the dilemma facing elites, the benefits for citizens from a successful revolution are independent of the state variable, yet costs are increasing in it, since citizens suffer from the upheaval following a revolution. Thus, a citizen will support a revolution if the state variable is low enough. Since k + 1 citizens are needed for a revolution, the citizen k + 1-th least satisfied with the current regime will be pivotal, and will denote the level of the state variable where he is indifferent by sR . We thus immediately observe that in an MPE, the dictator will allocate 0 resources to the k citizens most unsatisfied with the regime. In equilibrium, it must be that sC ≥ sR . If it is otherwise, then sC < sR and whenever elite members prefer a coup to stability, a revolution is already taking place. Thus, coups never takes place , while the dictator does allocate some resources to the elites (otherwise we would have sC = ȳ ≥ sR ). The dictator can therefore decrease this share for the elites while still keeping sC < sR , and increase his own share of the pie, thereby increasing his rents while not changing his probability of survival. The dictator is better off after this change, and therefore the initial ranking between sR and sC could not hold in MPE. 11 Note that if citizens faced a fixed cost for supporting a revolution the ranking of political outcomes in a MPE would not change, as long as this cost was not too large. When making their decision, citizens would now take into account the cost of a revolution, and would thus need to have less resources available to them in order to prefer a revolution (recall that citizens make their decisions sequentially in a random order). They would thus choose a lower threshold sR for coordinating on a revolution. Other than this, the reasoning behind Theorem 2 stands as before, and thus the ranking result. 3.3 Ranking of Political Threats by Economic Outcomes Theorem 2’s most important and novel implication is the linkage it establishes between the state of the economy in authoritarian regimes on the one hand and the political transitions that are likely to follow on the other.10 As sC ≥ sR , coups are more likely than revolutions in periods with higher levels of distributable surplus, and vice-versa. In regimes where both threats to authoritarian stability are present, that is when sR > y and sC < ȳ, revolutions are likely when the economy is performing badly, coups are likely when the economy is performing moderately well, and stability is likely when the economy is performing well, as described in figure 2. Figure 2: The Ranking of Political Transition by Economic Outcomes The intuition behind this result is the following: because of the high costs associated 10 Note that Theorem 1 does not ensure the existence of a MPE where the dictator uses a pure strategy, which Theorem 2 requires. However, even if the dictator uses a mixed strategy in MPE, the ranking Theorem 2 establishes between political outcomes based on the underlying economic conditions holds in expectation. Thus, the main empirical prediction of Theorem 2 holds in any case. 12 with mass uprisings, citizens only opt to revolt when the state of the economy is sufficiently low to make the risk of a revolution worthwhile. In such periods, the elites have no incentive to initiate a coup since they expect the masses to go out to the streets. Moreover, because in MPE dictators will not allocate more resources to the elites than necessary to completely eliminate the coup threat, in intermediate economic times the elites find it worthwhile to risk a coup, whereas the masses do not find it worthwhile to risk a revolution. Lastly, in prosperous economic times, both the elites and the masses have too much to lose out of political conflicts. Indeed, many of the mass revolutions the twentieth century has witnessed occurred after a period of economic depression. The 1989 revolutions in East Europe followed a period of economic stagnation in the Eastern Block economies starting in the early 1980s. In Poland, for example, more than 60% of population lived in poverty, and inflation reached 1,500% in the eve of the revolution (Rachwald 1990). Similarly, the 1979 revolution in Iran followed a period of economic decay and of double-digit inflation rates wherein the regime’s attempts to dampen inflation proved unsuccessful. Authoritarian stability, on the other hand, is likely in periods of economic prosperity. As is the case in democracies, higher levels of per-capita income and stronger economic growth are also associated with stability of dictatorial institutions (Cox 2008; Geddes 2008; Magaloni and Wallace 2008). Lastly, surprisingly, coups seem to be associated with periods in which the economy is performing moderately well, not when it is performing badly. Sanders (2009) suggests that coups are likely to happen in periods wherein economic growth is below its ”normal” levels but is not dramatically low. Correspondingly, many of the coups occurring during the twentieth century took place in years wherein the economy was performing moderately well: the 1991 coup in Thailand, the 1999 coup in Pakistan, the 2003 coup in Nepal, the 2005 coup in Mauritania, and the 2006 coup in Fiji are all examples of coups occurring under moderately well performing economies. 3.4 Additional Implications In addition to ranking political transitions by economic outcomes, Theorem 2 also sheds light on the dictator’s behavior in equilibrium. In particular, the theorem suggests that instead of inducing them to distribute rents and resources only to the elites while fully neglecting the masses, autocrats’ interest in survival in fact encourages them to provide goods and benefits to the masses in an effort to create a basis of support among the citizens. The threat of a revolution makes it suboptimal for dictators to focus their efforts only on the elites while 13 refraining from distributing rents to the masses. Dictators would thus find it beneficial to create patronage networks whereby privileges are distributed to citizens in return for their acquiescence. Indeed, in many one-party dictatorships, dictators use the party machine to distribute rents to the people in return for their political support. The party controls work permits, land titles, jobs, housing, food, and many other privileges whose distribution becomes a key tool for building mass support. These privileges are distributed to them with the intention of creating an interest in the dictator’s survival among the masses. The Communist party in the former USSR (Havel 1978), the PRI in Mexico (Magaloni 2006; Greene 2007), and the NDP in Egypt (Blaydes 2008) are all examples of such clientelist parties. In these systems, the party distributes privileges to the masses in order to secure their political compliance, thereby making each citizen who takes part in this political-exchange an active contributer to the regime’s longevity (Diaz-Cayeros, Magaloni, and Weingast 2001). When they are well institutionalized, autocratic parties function as giant patronage systems that create a vested interest in the perpetuation of the regime among its citizens (Lust-Okar 2005, 2006; Magaloni 2006; Geddes 2006, 2008; Pepinsky 2007; Magaloni and Kricheli 2010). Autocrats also use many other social and political institutions to distribute rents to the people. Local and low-level political offices, kinship-based patronage networks, and legislature seats are frequently used by dictators as the infrastructure through which goods are provided to the public. In some cases, autocrats even use intermarriage between the property class and the ruling class as a way to secure long-lasting flows of rents to the citizens (Haber et al 2003). All these formal and informal institutions autocrats use to distribute rents to the people support our result that dictators never opt to focus all their efforts on minimizing threats from the elites while fully neglecting the people. Instead, autocrats seem to provide citizens with privileges in return for their political acquiescence. The common depiction of dictators as rulers who rely only on the elites and are completely divorced from the citizens is thus undermined. Theorem 2, however, also suggests that not all the citizens will receive these privileges. Instead, the dictator will privilege a pivotal subset of the citizens whose identity is fixed over time. By constantly rewarding the same set of citizens, the dictator creates an expectation for future rents among this pivotal group, thereby creating a vested interest in his stability among them. This vested interest allows him to distribute less resources than he would have otherwise need to, because the pivotal citizens are willing to accept less resources today in exchange for the assurance that they would keep receiving benefits from the ruler in the 14 future. Indeed, even in one-party autocracies resources are not distributed to all citizens. Instead, they are distributed only to a subset of the citizens, typically to those who are members of the party. Membership in the Communist Party in the USSR, for example, became a privilege which only a subset of the citizens enjoyed ( e.g., Dickson and Ruble 2000; Dickson 2003. The perquisites party-members enjoyed included, among others, access to foreign goods, visas for trips abroad, student enrollment in prestigious universities, and prestigious jobs. In other types of dictatorships the autocrat is often supported by a privileged ethnic or religious group. The Alawis in Syria, the Hutus in Rwanda during the military rule, and the Sunnis in Saddam Hussein’s Iraq are examples of such privileged groups. The civilian support of these regimes was concentrated in a fixed group of citizens who enjoyed a privileged status. Bates (1981, 1989) suggests that in many African dictatorships, dictators base their mass support on their own ethnic affiliation by discriminately transferring rents so as to privilege their own ethnic group. The clientelistic distribution resulting in these dictatorships was not redistributive and generally benefited only a relatively small proportion of the citizenry which was determined ethnically (van de Walle 2007). Bates (1983), Horowitz (1985), and Padró i Miquel (2004) further suggest that the bias in favor of the ruler’s ethnic group is prevalent and conspicuous. 4 Asymmetric Information We now extend our model to include the possibility that the dictator is less informed than the elites are with respect to their ability to successfully overthrow him. This extension is of special interest since in many cases the dictator is at an informational disadvantage compared to elite members who are in control of the military, intelligence gathering institutions, or other security apparatuses. The information he is exposed to is filtered by these elites, whose incentives are often different than his (Schatzberg, 1988; Wintrobe, 1998; Tullock 1987). Even when these elites have an interest in the dictator’s survival in power, when passing information to the dictator, they often face an incentive to over-represent the extent of power he really enjoys (Schatzberg, 1988). This leads to what Friedrich and Brzezinski (1965) call ”the vacuum effect” surrounding the dictator—a state of asymmetry of information wherein the dictator holds less information than his bureaucrats about his own vulnerability. Specifically, we will maintain the same setup of the basic model, but now assume the dictator’s vulnerability to coups is determined stochastically—before each period, nature decides, with probability p, that a coup attempt will be successful during that period or, 15 with probability 1 − p, that coup attempts will fail during it. The ruler is aware of this prior distribution but receives no further information regarding the success probability of a coup. So far, this assumptions serves just as a micro-foundation for the informational structure in the original setting. Extending the model, we will assume that elite members may receive in each period an informative signal as to whether a coup attempt will be successful or not. If the elites receive such a signal, their posterior probability that the regime is weak and a coup will succeed rises to p < q2 < 1. If they do not receive a signal, they infer that the posterior probability of a coup’s success is q1 < p. We thus assume that all elite members have the same information in each period. The dictator is aware that the elite is better informed, and takes this knowledge into account when allocating free resources in the economy. Notice that the masses (as well as the other players) do not face incomplete information regarding the prospects of a revolution, but we do not make any assumption with regards to their knowledge of the dictator’s vulnerability to coups. Theorem 3. With asymmetric information, MPEs in which the dictator uses a pure strategy are defined by three economic thresholds: a revolution threshold sR and two coup thresholds s1C < s2C , with sR ≤ s2C . Two alternatives are possible: 1. If s1C < sR then: if Yt < sR , the masses will revolt; if sR < Yt < s2C and they receive a signal that the dictator is weak, the elites will attempt a coup; and if Yt > s2C neither a coup attempt nor a revolution will occur. All coups attempts will be successful with probability q2 . 2. If s1C > sR then: if Yt < sR , the masses will revolt; if sR < Yt < s1C , or if s1C < Yt < s2C and they receive a signal that the dictator is weak, the elites will attempt a coup; and if Yt > s2C neither a coup attempt nor a revolution will occur. On average, coups will succeed with some probability q1 < q < q2 . The introduction of asymmetric information exacerbates the challenges the dictator faces in his attempts to stay in power. Since elite members have better information regarding their chances of staging a successful coup, the dictator may make two types of errors—handing out too many resources to the elites when he is strong, or handing out too little resources when he is weak. Although the equilibria of this game can have two possible forms, we can deduce the following conclusion, which has important consequences regarding the empirical predictions of the model. 16 Corollary 1. In the presence of asymmetric information the relationship between the underlying state of the economy and the likelihood of regime transitions becomes noisier— revolutions are still likely when the economy performs badly and stability when it performs well, yet both stability and coup attempts may be possible at the same moderate realizations of the state of the economy. Note that the prediction presented by Corollary 1 changes the empirical prediction of the basic model. In the basic, symmetric information model, economic variables could be directly mapped to political outcomes—whether a revolution, coup attempts (which could be successful or not), or stability. The introduction of asymmetric information excludes this determinism, as in this setting coups and stability can happen in moderate realizations of the state of the economy. This implies that empirically, stability should be likely when the economy is performing well, revolutions when it is performing badly, and both coups and stability when it is performing moderately well. 5 Preliminary Evidence In this section we present preliminary evidence in support of the conclusions of Theorems 2 and 3. The aim of this section is not to fully test our theory, but rather to suggest that the theory’s main conclusions are compatible with available data from contemporary authoritarian regimes. The evidence will focus on our main result with regards to the relation between the state of the economy and the likelihood of coups and revolutions: H1 : revolutions are likely when the economy is performing badly, coups are likely when the economy is performing moderately well, and stability is likely when the economy is performing well Notice also that if the incomplete information version our model holds, the relationship between the state of the economy and the threats to the dictator survival is noisy— revolutions are still likely when the economy performs badly and stability when it performs well, but although coups are likely in periods wherein the economy is performing moderately well, they do no take place in every such period. 5.1 Data To test this hypothesis we use data from 115 authoritarian regimes during 1960 − 2000. Our data-set consists of 3, 388 country year observations, all coded as authoritarian regimes by Przeworski et al. (2000). To measure the state of the economy, we use the World Bank’s 17 (2009) measure of annual GDP growth, measured in percentages. The reason we use GDP growth rates is that in our model, the state of the economy Yt represents the available resources the dictator holds which can be distributed to the elites and the masses. Yt should thus be understood as a surplus of rents available to the dictator and can be proxied by annual GDP growth rates. To measure whether a palace or military coup was attempted by the elites, we code a dummy variable equaling one when the military or any other governmental governmental actor overthrows the ruler or attempts to do so based on the Archigos database (Goemans et al. 2004) and on the Coups d’Etat dataset (Marshall and Marshall 2007); and equaling zero when both datasets indicate that in a given country-year observation, neither a coup nor an attempted coup took place. To measure whether a citizen-driven revolution occurred, we code a dummy variable equaling one when the the ruler of the country is overthrown by popular protest or by civilian rebels based on the Archigos database (Goemans et al. 2004); and equaling zero otherwise. Notice that our variables indicate whether an attempted coup took place but not whether an attempted revolution took place. The reason for this is that our model predicts that an attempted coup—successful or unsuccessful—is likely when the economy is performing moderately well; but that only successful revolutions are likely when the economy is performing badly. 5.2 Revolutions, Coups, and Economic Performance Figures 3 and 4 display the level of GDP growth and the years wherein a citizen-based revolution or an elite-based coup took place for twelve countries over time. Years wherein a coup was attempted by the elites are indicated in blue, and years wherein a citizen-driven revolution occurred are indicated in red.11 [Figures 3 and 4 about here] Figures 3 and 4 support our hypothesis regarding the relationship between the state of the economy and the likelihood of a coup or a revolution. Revolutions seem to be likely when the economy is performing badly, whereas coups seem to be likely when the economy is performing moderately well. Additionally, when the economy is performing well, neither a coup nor a revolution is likely. 11 Some country-year observations are missing from the plots because the World Bank does not have an estimate of the annual GDP growth for these observations. 18 The Figures also support the incomplete-information extension of our model in which the elites hold more accurate information about the vulnerability of the dictator than the dictator himself. Coups seem to be likely when the economy is performing moderately well, yet a coup is not attempted in every year wherein the economy performs moderately well, as predicted by the incomplete information version of our model. One might worry that part of the reason we observe that revolutions occur in bad economic periods is that revolutions burden society with severe costs, which our measure of growth might be picking. The worry is that because our economic performance measures are based on country-year observations and because a revolution might occur early in the year, we might be picking the results of a revolution instead of its causes. One way to address this concern is to examine the relationship between the likelihood of a revolution and last year’s growth levels which cannot be affected by the revolution. When we examine this relationship in Figure 3, we find the same support for our hypothesis: expect of one case in Azerbaijan, all of the revolutions in our plots occurred after a period of economic decline. Figure 5 displays three box-plots of the annual levels of GDP growth, depicting the distribution of GDP growth in years wherein a revolution took place (right-hand box), years wherein a coup was attempted (center box), and years wherein neither a coup nor a revolution took place (left-hand box). The horizontal line in each box represents the median of the relevant distribution, and the upper and the lower limits of the box represent the 75% quantile and the 25% quantile respectively. The Figure also presents the mean value of GDP growth in each subset of the data (in blue). [Figure 5 about here] Figure 5 is also consistent with our hypothesis: On average, revolutions occur in periods with low growth levels, and coups occur in periods with moderate growth levels. Additionally, stability, i.e., a period wherein neither a coup nor a revolution take place, occurs in periods with high levels of growth. Notice also that the high level of variance in growth levels during stability and the fact that there is a significant degree of overlap between the lowest levels of growth during stability and the levels of growth during coups support the incomplete information version of our model, as they suggest that coups are likely when the economy is performing moderately well but that the elites do not attempt a coup in every year with moderate levels of growth.12 Interestingly, our findings support the empirical literature about the the conditions that foster revolutions, but not the literature about the conditions that foster coups. Early 12 We repeated the same analysis using lagged growth levels instead of growth levels in order to address the possible endogeneity problem described above. The results were similar to those depicted in Figure 5. 19 studies of revolutions highlighted that they are more likely to occur in periods of economic decline, when citizens’ expectations exceed actual economic performance (Davies 1962, 1969; Haggard and Kaufman 1995, 1997), which is consistent with our theoretical and empirical findings. However, the most systematic account on the influence of economic performance on the likelihood of coups suggests that coups are likely to occur in poor countries (Londregan and Poole 1990). At first sight, this finding seems to be at odds with our theoretical and empirical finding that coups are likely in moderately well economic periods, not in economic crises. However, Londregan and Poole’s analysis focuses on across-country variation while ours focuses on within-country variation. They ask what countries are more likely to suffer coups, while we ask when is a specific regime more likely to suffer a coup. 6 Concluding Remarks To survive, authoritarian leaders have to appease both the elites and the people. Yet citizensdriven threats to the regime’s stability are different than elites-driven threats. Citizens face coordination problems, but when they are in the streets in their masses almost no force can stop them from destabilizing the the entire political and social structure of society. Elite members, on the other hand, face less coordination problems, but they can overthrow the regime over night, without changing the basic social and governmental apparatuses in the country. Our theory is founded on the asymmetry between these two types of threats to authoritarian stability. The theory has three main implications. First, coups are likely under different conditions than revolutions are: specifically, revolutions are likely when the economy is performing badly, stability is likely when the economy is performing well, and, surprisingly, coups are likely when the economy is performing moderately well, not when it is performing badly. This relationship still holds when the dictator faces less accurate information than the elites about his own vulnerability to coups, but it becomes more noisy: stability is likely when the economy is performing well, revolutions when it is performing badly, and both coups and stability are likely when it is performing moderately well. Second, authoritarian leaders do not try to buy-off elite support without trying to do the same with regards to the masses. Elite support is not sufficient for authoritarian survival and the threat of a mass-driven revolution induces autocrats to distribute privileges and spoils to the citizens with the aim of building a basis of support among the citizens. Third, although dictators rely on the citizens’ support, they do not buy off all of their citizens. Instead, they transfer resources to a privileged group of pivotal citizens whose 20 identity is fixed over time. By constantly rewarding the same citizens, dictators manage to induce sufficient mass support at the lowest possible cost. From a theoretical perspective, our theory contributes to the literature by providing a single framework whereby both elite coups and mass revolutions can be analyzed. To the best of our knowledge, this is the first theoretical paper that examines when coups or revolutions are likely to occur using such a comprehensive theoretical framework. Looking at the relationship between economic performance and coups or revolutions separately is problematic, because, in reality, the political reaction to economic performance can come either from the elites or from the masses. From an empirical perspective, our paper suggests that examining the conditions that foster coups and the conditions that foster revolutions in the same analysis is a worthwhile avenue for future research. When we begin to do so, we find that, surprisingly, coups are not likely in times of economic crisis, as was previously believed, but rather, in moderately good economic times. 21 7 Appendix Proof of Theorem 1 The proof is a straightforward adaptation of Jackson and Morelli (2009) to our setting. The full proof will be added later. Proof of Theorem 2 First, since all player in the same positions play the same Markov strategies in all periods and the state variable Yt is i.i.d. across periods, being in a political position (citizen, elite or dictator) in a future autocracy is associated with a continuation value which is fixed across all periods. We will denote these values by Vi for citizen i, Ve for elite member e, and Vd for the ruler. Since we assume that revolutions trump coups, we can first look at the citizens’ decisions of supporting a revolution, despite the fact that elites move last. If a revolution occurs, it eliminates a share of the income a citizen receives in the current period, while giving a citizen more resources in all future periods. On the other hand, if the regime remains stable or is overthrown by a coup, citizens receive their original allocated rents, but receive the low continuation value Vi . Thus, citizen i strictly prefers a revolution for any realization of Yt below some threshold value siR , which can depend on i. Explicitly, a citizen strictly prefers a revolution to stability or coup if: (1 − β)wti Yt + δ E [Yt ] > wti Yt + δVi (1 − δ)(m + n) which happens if and only if13 : E [Yt ] δ − Vi Yt < βwti (1 − δ)(m + n) We denote the revolution threshold value δ E [Yt ] i sR := min − Vi , ȳ βwti (1 − δ)(m + n) (0) and order these bounds without loss of generality s1R ≥ s2R ≥ ... ≥ sm R . We claim that in a k+1 k+1 MPE a revolution will occur if Yt < sR , and would not if Yt > sR . Assume that Yt < sk+1 R , i and look at the last of citizens the citizens with Yt < sR to act. If by the time this citizen called to act at least k + 1 citizens have supported a revolution, then a revolution would occur in any case. If that number is exactly k, then in an MPE this citizen must support 13 If citizens were to suffer some cost k from participating in a revolution, then the last h i m − k + 1 citizens E [Yt ] δ k to act in period t would support a revolution in MPE if Yt < βwi (1−δ)(m+n) − Vi − δ each i among them. t 22 a revolution, and if less than k citizens supported the revolution before him, a revolution would occur in any case. Continuing by backward induction, if we now look at the k + 1-th to last citizen to act with Yt < siR (and by assumption there are at least k + 1 such citizens). If before him any citizen supported the revolution, then a revolution would occur regardless of his action, by the induction argument. If no citizen before him supported the revolution, then in a MPE he must support the revolution, since the backward induction argument shows that all subsequent citizens with Yt < siR will also support it, and thus it will occur. An analogous argument proves the case Yt > sk+1 R . Thus, in a MPE, the ruler need only pay-off a blocking coalition of m − k citizens. By equation 7, the cheapest m − k citizens to pay-off are the m − k citizens with the highest continuation values. Also note, that setting the weights wti such that the thresholds m sk+1 R , ..., sR are not equal would be a waste of resources for the dictator, as only the pivotal sk+1 effects the outcome. Thus, in a MPE the ruler will allocate wti = 0 to the k citizens R with the lowest continuation values from autocracy, and will allocate some wti ≥ 0 to the other citizens, such that the threshold siR is equal across this group. We will denote this threshold by sR . Since the dictator is using a pure strategy in this MPE, for the privileged group of m − k citizens who receive a non-negative amount of resources are fixed across periods. For these citizens, the above argument shows that those with a strictly higher continuation values from autocracy get strictly less resources from the dictator. This can only hold in equilibrium if all the continuation values of these m − k citizens are the same, and so are their weights of society’s product wti , which we will denote by w1 . Turning to the elites, first note that if a revolution has already started, and more than k citizens are in the streets, then by assumption the elite is powerless to stop them. Otherwise, elite members are free to contemplate the possibility of a coup. Elite member e prefers a coup attempt over the continued rule of the current dictator, if: wte Yt + pδ 1 n−1 Vd + Ve n n + (1 − p) (−cwte Yt + δVe ) > wte Yt + δVe which happens if and only if: Yt < p δ [Vd − Ve ] 1 − p ncwte 23 We denote the elite’s coup threshold by: seC := min p δ [Vd − Ve ] , ȳ 1 − p ncwte Since any elite member has the resources to start a coup, this shows that to prevent coups the ruler must worry about the elite member with the highest coup threshold SRe . Therefore, in equilibrium, all elite members will have the same threshold seC . Since continuation values are fixed across time, this implies that elite members with lower continuation values receive a higher share of the economy’s resources. As before, this is possible in equilibrium only if all elite members receive the same share of resources and thus have the same continuation values. We will denote the fraction elite members receive in equilibrium by w2 and the common threshold for them to prefer a coup by sC . To complete the proof, it remains to show that in equilibrium sR ≤ sC . To see this, assume the opposite. First note that this implies that sR < ȳ – otherwise the leader is always replaced by a revolution and is thus better off by setting wd = 1, making sC = ȳ = sR , contrary to the assumption. Now consider the following one-stage deviation by the ruler: to decrease we by > 0 for one of the elite members, and to increase wd by at the same time, making those changes only for period t. We know that this deviation is possible since sC < sR ≤ ȳ implies that we > 0 for one of the elite members. Since it increases wd , this deviation increases the ruler’s expected one-stage payoff by a strictly positive amount, yet if is small enough, it does not increase the ruler’s probability of being dethroned. This is because for sC < sR there is no real threat of coup, and a small change in we for one period does not change that. This completes the proof. Proof of Theorem 3 The proof of this theorem corresponds to the proof of Theorem 2 up to the analysis of the coup threshold. In the setting with asymmetric information the elites’ decision of whether to initiate a coup or not depends in this setting on the signal that they get. Ignoring the possibility of a revolution, if elite members receive a signal that the dictator is weak in period t, an elite member e will prefer initiating a coup if: wte Yt + q2 δ n−1 1 Vd + Ve n n + (1 − q2 ) (−cwte Yt + δVe ) > wte Yt + δVe which happens if and only if: Yt < q2 δ [Vd − Ve ] 1 − q2 ncwte 24 We denote the elite’s threshold upon receiving the signal by: se,1 C := min q2 δ [Vd − Ve ] , ȳ 1 − q2 ncwte and similarly the elite’s threshold in a period when do not receive a signal by: se,2 C := min q1 δ [Vd − Ve ] , ȳ 1 − q1 ncwte Note that for every elite member e these two thresholds are a multiple of one another by Therefore, as in the proof of Theorem 2, the dictator will just be concerned with the elite member with the minimal thresholds, as this member is the pivotal decision maker for coups no matter what the signal is in that period. Using the same logic, we can conclude that in equilibrium the dictator will set the resource weights we such that these thresholds are the same for all elite members, and we can denote them by s2C > s1C . Continuing the proof in the same manner as in the proof of Theorem 2, only replacing the ranking of sR and sC , with ranking of sR and SC2 , yields the required result. q1(1−q2 ) . q2(1−q1) 25 References [1] Acemoglu, Daron, Georgy Egorov and Konstantin Sonin. 2008. Coalition Formation in Nondemocracies, Review of Economic Studies, volume 75, pp. 987-1009. [2] Acemoglu, Daron and James Robinson. 2001. ”A Theory of Political Transitions.” American Economic Review, 91(4): 938-963. [3] Acemoglu, Daron and James Robinson. 2000. ”Why Did the West Extend the Franchise? Democracy, Inequality, and Growth in Historical Perspective.” Quarterly Journal of Economics, 115(4): 1167-1199. [4] Acemoglu D, and Robinson J. 2006. Economic Origins of Dictatorship and Democracy. New York, NY: Cambridge University Press. [5] Acemoglu D, and Robinson J. 2008. ”Persistence of Power, Elites, and Institutions”. American Economic Review 98(1): 267-293 [6] Acemoglu Daron, Ticchi Davide, and Vindigni Andrea (2008) ”A Theory of Military Dictatorships”. Forthcoming, American Economic Journal: Macroeconomics. [7] AFP. Dec 25, 2008 ”’It’s our turn for power’: Guinea coup leader tightens grip.” http://www.google.com/hostednews/afp/article/ALeqM5j8C7wO1oEjpQbAXSuldjCxbJMEjg [8] Arriola L R. 2008. Between Coordination and Cooptation: The Opposition’s Dilemma in Africa. Doctoral Dissertation, Department of Political Science, Stanford University, Stanford, CA. [9] Bates, Robert. 1981. Markets and States in Tropical Africa. Berkeley: University of California Press. [10] Bates, Robert. 1983. ”Modernization, Ethnic Competition, and the Rationality of Politics in Contemporary Africa.” In State versus Ethnic Claims: African Policy Dilemmas, ed. Olurunsola, V. and Rothchild, D. Boulder, CO: Westview Press. [11] Bates, Robert. 1989. Beyond the Miracle of the Market. New York: Cambridge University Press. [12] Blaydes L. 2008. Competition without democracy: elections and distributive politics in Mubaraks Egypt. PhD Thesis. Univ. Calif., Los Angeles. 26 [13] Blaydes L. 2009. Competition without Democracy: Elections and Distributive Politics in Mubarak’s Egypt. Book Manuscript. [14] Boix, Carles. 2003. Democracy and Redistribution. Cambridge, UK: Cambridge University Press. [15] Boix C, Svoliky M. 2008. ”The Foundations of Limited Authoritarian Government: Institutions and Power-sharing in Dictatorships.” Presented at Conf. on Dictatorships: Their Governance and Social Consequences, Princeton Univ., Princeton, NJ [16] Boix C, and Stokes S. 2003. ”Endogenous Democratization.” World Politics 55:51749. [17] Bolotnyy V, and Magaloni B. 2009. Authoritarian Regimes Classification. Dataset. [18] Brownlee J. 2007. Authoritarianism in an Age of Democratization. Cambridge: Cambridge University Press. [19] Bueno de Mesquita, Bruce and Alastair Smith. 2009. ”Political Survival and Endogenous Institutional Change.” Comparative Political Studies 42(2): 167-197. [20] Bueno de Mesquita, Bruce, Alastair Smith, Randolph M. Siverson, and James D. Morrow. 2003. The Logic of Political Survival. Cambridge, MA: MIT Press. [21] Cox, Gary. 2008. ”Authoritarian Elections and Leadership Succession, 1975-2000.” Working Paper. [22] Davies, James C. 1962. ”Toward a Theory of Revolution.” American Sociological Review 27:5-19. [23] Davies, James C. 1969. ”The J-Curve of Rising and Declining Satisfactions as a Cause of Revolution and Rebellion.” Pp. 415436 in Violence in America: Historical and Comparative Perspectives, edited by Hugh Graham and Ted Gurr. Beverly Hills, CA: Praeger. [24] Debs A. 2007. ”On Dictatorships.” PhD thesis. Dep. Economics, MIT. [25] Diaz-Cayeros, Alberto, Beatriz Magaloni, and Barry R. Weingast. 2001. ”Tragic Brilliance: Equilibrium Party Hegemony in Mexico.” Hoover institution working paper. [26] Dickson Bruce. 2003. Red Capitalists in China. The P arty, Private Entrepreneurs, and Prospects for Political Change (Cambridge: Cambridge University Press). 27 [27] Dickson Bruce J. and Maria Rost Rublee, ”Membership Has Its Privileges: The Socioeconomic Characteristics of Communist Party Members in Urban China”, Comparative Political Studies 33(1) [28] Egorov G, S Guriev, and K Sonin. 2009. ”Why Resource-poor Dictators Allow Freer Media: A Theory and Evidence from Panel Data”. American Political Science Review 103 : 645-668. [29] Egorov G and K Sonin ”Dictators and Their Viziers: Endogenizing the LoyaltyCompetence Trade-Off.” Journal of European Economic Association, Forthcoming. Available at SSRN: http://ssrn.com/abstract=630503 [30] Epstein D L., Bates R, Goldstone J, Kristensen I, and OHalloran S. 2006. ”Democratic Transitions.” American Journal of Political Science 50( 3): 551569. [31] Fanon, Frantz. 1968. Wretched of the Earth. N.Y.: Grove Press. [32] Fudenberg, Drew and Jean Tirole. 1990. Game theory. Cambridge, MA: MIT University Press. [33] Gandhi, Jennifer. 2008. Political Institutions under Dictatorship. Cambridge, UK: Cambridge University Press. [34] Gandhi, Jennifer, and Adam Przeworski. 2006. ”Cooperation, Cooptation, and Rebellion under Dictatorships.” Economics and Politics 18(1): 1-26. [35] Geddes, Barbara. 2003. Paradigms and Sand Castles: Theory Building and Research Design in Comparative Politics. Ann Arbor, MI: University of Michigan Press. [36] Geddes, Barbara. 2008. ”Party Creation as an Autocratic Survival Strategy.” Working Paper. [37] Geddes, Barbara. 2006. ”Why Parties in Elections in Authoritarian Regimes?” presented in the Annual Meeting of the American Political Science Association, DC. [38] Goemans, Hein, Kristian Skrede Gleditsch, Giacomo Chiozza, and Jinhee L. Choung. 2004. Archigos: A Database on Political Leaders. Typescript, University of Rochester and University of California San Diego. [39] Goodwin, Jeff. 2001. No Other Way Out. States and Revolutionary Movements 19451991. Cambridge, UK: Cambridge University Press. 28 [40] Greene K F. 2007. Why Dominant Parties Lose: Mexicos Democratization in Comparative Perspective. New York, NY: Cambridge University Press. [41] Haber, Stephen. 2006. ”Authoritarian Government” in Handbook of Political Economy, ed. Weingast and Wittman, New York, NY: Oxford University Press. [42] Haber, Stephen, Armando Razo, and Noel Maurer. 2003. The Politics of Property Rights: Political Instability, Credible Commitments, and Economic Growth in Mexico, 1876-1929. Cambridge, UK: Cambridge University Press. [43] Haggard, Stephan and Kaufman, Robert R. The political economy of democratic transitions. Princeton, NJ: Princeton University Press, 1995. [44] Haggard Stephan and Robert R. Kaufman. ”The Political Economy of Democratic Transitions.” Comparative Politics, Vol. 29, No. 3, pp. 263-283 [45] Havel, Vaclav. 1978. ”The Power of the Powerless”. in Open letters: Selected writings 1965-1990 by Vaclav Havel. London: Faber and Faber. pp. 125-214. [46] Horowitz, D. 1985. Ethnic Groups in Conflict. Berkeley: University of California Press. [47] Jackson, Matthew O and Massimo Morelli. ”Strategic Militarization, Deterrence, and Wars”. Quarterly Journal of Political Science 4(4): 279-313. [48] Kricheli R. 2008. ”Information Aggregation and Electoral Autocracies: An Informationbased Theory of Authoritarian Leaders’ Use of Elections.” Manuscript. [49] Kricheli R. 2009. ”Rewarding Illiberal Elections? Electoral Autocracies and the International Community.” Manuscript. [50] Kuran, Timur. 1991a. ”Now Out of Never: The Element of Surprise in the East European Revolution of 1989.” World Politics, 44(1): 7-48. [51] Kuran, Timur. 1991b. ”The East European Revolution of 1989: Is It Surprising that We Were Surprised?” American Economic Review, 81(2): 121-125. [52] Lohmann, Susanne. 1994. ”Dynamics of Informational Cascades: The Monday Demonstrations in Leipzig, East Germany, 1989-91”, World Politics, 47(1): 42-101. [53] Londregan John B and Keith T. Poole. 1990. ”Poverty, the Coup Trap, and the Seizure of Executive Power.” World Politics 42(2): 151-183. 29 [54] Lust-Okar E. 2006. ”Elections under authoritarianism: Preliminary lessons from Jordan.” Democratization 13 (3):456- 471. [55] Lust-Okar E. 2005. Structuring Conflict in the Arab World. Incumbents, Opponents, and Institutions. Cambridge, UK: Cambridge University Press. [56] Lust-Okar E and J. Gandhi. 2009. ”Elections under Authoritarianism.” Annual Review of Political Science [57] Nordlinger, Eric. 1977. Soldiers in Politics: Military Coups and Governments. Englewood Cliffs, NJ: Prentice-Hall. [58] Magaloni B. 2008. ”Credible Power-Sharing and the Longevity of Authoritarian Rule.” Comparative Political Studies 41 (4-5): 715-741. [59] Magaloni, Beatriz. 2006. Voting for Autocracy: Hegemonic Party Survival and its Demise in Mexico. Cambridge, UK: Cambridge University Press. [60] Magaloni, Beatriz and Ruth Kricheli. 2010. ”Political Order Under One-Party Rule.” forthcoming Annual Review of Political Science. [61] Magaloni B, and Wallace J. 2008. ”Citizen Loyalty, Mass Protest and Authoritarian Survival.” Presented at the Conference on Dictatorships: Their Governance and Social Consequences, Princeton University. [62] Magaloni B, and R Kricheli. 2010. ”Political Order and One-Party Rule.” Annual Review of Political Science, forthcoming. [63] Marshall, Monty G., and Donna R. Marshall. 2007. Coup Détat Events, 1960-2006 database. Center for Systemic Peace. http://www.systemicpeace.org/inscr/inscr.htm. (January 2, 2009). [64] Padró i Miquel, Edward. 2004. ”Political Accountability in Divided Societies: The Politics of Fear.” Working Paper. [65] Myerson, Roger. 2008. ”The Autocrat’s Credibility Problem and Foundations of the Constitutional State.” American Political Science Review, 102(1): 125-139. [66] ODonnell G, Schmitter P C and Whitehead L (eds.) 1986. Transitions from Authoritarian Rule (4 vols.). Baltimore: Johns Hopkins University Press. 30 [67] Paige, Jeffery M. 1997. Coffee and Power: Revolution and the Rise of Democracy in Central America. Cambridge, MA: Harvard University Press. [68] Pepinsky T. 2007. ”Autocracy, elections, and fiscal policy: evidence from Malaysia.” Studies in Cooperative International Development. 42(12):13663 [69] Przeworski A. 1991. Democracy and the Market. New York, NY: Cambridge University Pres. [70] Przeworski, Adam, Michael E. Alvarez, José Antonio Cheibub, and Fernando Limongi. 2000. Democracy and Development: Political Institutions and Well-Being in the World, 1950-1990. Cambridge, UK: Cambridge University Press. [71] Rachwald, Arthur R. 1990. In search of Poland the superpowers’ response to Solidarity, 1980-1989. Stanford, CA: Hoover Institution Press. [72] Sanders, Kevin B. 2009. Coups D’états: An Economic Analysis of Factors that Contribute and Result from Coups. Honors Thesis, University of Oregon. [73] Schatzberg, Michael G. 1988. The Dialectics of Oppression in Zaire. Bloomington, IN: Indiana University Press. [74] The American. November 9, 2009. ”The Berlin Wall, 20 Years Later.” http://www.american.com/archive/2009/november/the-berlin-wall-20-years-later [75] The New York Times. August 6 2008. ”Army Officers Seize Power in Mauritania.” http://www.nytimes.com/2008/08/07/world/africa/07mauritania.html [76] Tullock, Gordon. 1987. Autocracy. Boston, MA: Kluwer Academic Publishers. In Patrons, Clients, and Policies ed. Kitschelt H and Wilkinson S. NY: Cambridge University Press. pp. 112-149. [77] van de Walle, N. 2007. ”Meet the new boss, same as the old boss? The Evaluation of Political Clientelism in Africa” [78] Voice of Africa. 24 December 2008. ”Guinea Coup Plotters Announce Curfew, New Leader” http://www1.voanews.com/english/news/a-13-2008-12-24-voa13-66738732.html [79] Wintrobe R. 1998. The Political Economy of Dictatorship. Cambridge: Cambridge University Press. 31 [80] Woods Alan. 2009. ”The fall of the Berlin Wall: http://www.marxist.com/fall-berlin-wall-20-years-later.htm 20 years later” [81] Zolberg, Aristide. 1966. Creating Political Order: The Party-States of West Africa. Chicago, IL: Rand McNally and Company. 32 Figure 3: Revolutions, Coups, and Growth 33 Figure 4: Coups and Growth 34 Figure 5: Growth and Political Stability 35
© Copyright 2026 Paperzz