Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium IVIember Libraries http://www.archive.org/details/theoryofmilitaryOOacem DEWEY HS3I Massachusetts Institute of Technology Department of Economics Working Paper Series A THEORY OF MILITARY DICTATORSHIP Daron Acemoglu Davide Ticchi Andreas Vindigni Working Paper 08-10 March 30, 2008 Room E52-251 50 Memorial Drive Cambridge, MA 021 42 This paper can be downloaded without charge from the Social Science Research Network Paper Collection http://ssrn.com/abstract=1 11 2865 HB31 .M415 at A Theory of Military Dictatorships* Daron Acemoglu and CIFAR Andrea Vindigni Davide Ticchi University of Urbino MIT Princeton March 2008 Abstract We investigate how nondemocratic regimes use tlie military and how this can lead to the emergence of military dictatorships. Nondemocratic regimes need the use of force in order to in power, but this creates a political moral hazard problem; a strong military may not remain simply work as an agent of the more own elite but may The turn against them in order to create a regime moral hazard problem increases the cost of using repression in nondemocratic regimes and in particular, necessitates high wages and in line with their objectives. policy concessions to the military. When political these concessions are not sufficient, the military can take action against a nondemocratic regime in order to create its own dictatorship. A more important consequence of the presence of a strong military is that once transition to democracy takes place, the military poses a coup threat against the nascent democratic regime until it is reformed. The anticipation that the military will be reformed in the future acts as an additional motivation for the military to undertake coups against democratic governments. We show that greater inequality makes the use of the military in nondemocratic regimes more likely and also makes it more difficult for democracies to prevent military coups. In addition, greater inequality also makes it more likely that nondemocratic regimes are unable to solve the political moral hazard problem and thus creates another channel for the emergence of We show that greater natural resource rents make military coups but have ambiguous effects on the political equilibrium in nondemocracies (because with abundant natural resources, repression becomes more valuable to the elite, but also more expensive to maintain because of the more severe political moral hazard problem that natural resources induce). Finally, we discuss how the national defense military dictatorships. against democracies more also likely, role of the military interacts Keywords: with its involvement in domestic politics. coups, democracy, military, nondemocracy, political economy, political tran- sitions. JEL Classification: H2, NIO, N40, P16. *We thank seminar participants at Brown, CIFAR, Rochester, and Toulouse. Acemoglu Much gratefully acknowl- work on this paper was done when Acemoglu and Vindigni were visiting Yale and Collegio Carlo Alberto. They thank the Economics Department and the Leitner Program in International Political Economy at Yale and Collegio Carlo Alberto for their edges financial support from the National Science Foundation. hospitality. of the "The class that bears the lance or holds the musket regularly forces that handles the spade or pushes the shuttle." Gaetano Mosca (1939 its rule upon the class p. 228). Introduction 1 Nondemocratic regimes almost always groups. This repression is rely on some degree of repression against competing often exercised by a specialized branch of the state, the military} Despite the prevalence of nondemocratic regimes throughout history and the important role played by the military in such regimes, the typical assumption is agent" of some social group, such as the little and how the military uses setting up a regime more many nondemocratic its elite. There has been that the military is a "perfect systematic analysis of why coercive powers to support a nondemocratic regime rather than in line with its own preferences. This question relevant since, while is regimes survive with the support of the military, there are also numerous examples of military dictatorships that have emerged either as a result of a coup against a nondemocratic regime or against the subsequent democratic government. In this paper, we take a first step in the analysis of the role of the military in nondemocratic regimes and develop a theory of military dictatorships. At the center of our approach agency relationship between the is elite in oligarchic that creating a powerful military a more powerful military is more is regimes and the military.^ a double-edge sword for the elite. effective in preventing transitions to On is The main the idea the one hand, democracy. On the other hand, a more powerful military necessitates either greater concessions to the military or raises the risk of a military takeover. will act as the agent of the elite in under which oligarchies will We investigate the conditions under which the military nondemocratic regimes (oHgarchies) and the conditions turn into military dictatorships. Our approach also sheds light on the role of the military in coups against democracy. If the elite create a powerful military to prevent democratization, then the military also plays an important role in democratic politics until it is reformed, and such reform is not instantaneous.^ In particular, we show that faced with a powerful military, a newly-emerging democratic regime will either need to make costly Throughout the paper, the military may be thought to include the secret police and other law-enforcement We also use the terms the military and the army interchangeably. "Since we draw a distinction between military dictatorships and nondemocratic regimes controlled by the elite, we use the term oligarchy to refer to the latter and nondemocracy as a general term encompassing both military dictatorships and oligarchies. 'This assumption is consistent with O'Donnell and Schmitter's (1986) emphasis that the power of the army plays an important role at the early stages of democratic transitions. For instance, a relatively weak army may be easier to reform than a stronger one. Our model will show that this is generally true, but also that weaker militaries may sometimes be more difficult to control. ' agencies. concessions or face a high probabihty of a coup. This coup tlireat disappears once the military is Interestingly, however, reformed. it as soon as the opportunity arises that early phases of a democratic regime the anticipation that the military wiU be reformed is makes —because this creates make impossible for democratic governments to difficult to control it the military dm-ing the a commitment problem, making credible promises to compensate it soldiers for not taking actions against democracy. More specifically, distingxiished by our model economy consists of two groups, the rich incomes (endowments). Democracy leads to redistributive their particular, to the provision of public goods, is which are beneficial policies, in and costly The only way they can prevent this state, the military, responsible for using force and unwilling to allow a transition to democracy. by creating a specialized unit of the repressing for the citizens citizens, Consequently, starting from an oligarchy in which they hold power, the for the rich elite. elite are and the elite demands for democratization.'' A powerful military, however, is not only effective in preventing a transition to democracy but also creates a political moral hazard problem because it can turn against the elite and take direct control of the to create greater redistribution towards its government (for own members). Consequently, example, in order the elite have three potential strategies in oligarcliy: (1) no repression, thus allowing a rapid [smooth) transition to democracy; (2) repression, while also paying soldiers an efficien.cy wage so as to prevent military takeovers; (3) repression without significant concessions to soldiers, thus opting for non-prevention or facing the risk of a military takeover. We characterize the equilibria in this environment and analyze the role of the military The presence in politics. politics. If it will also democracy of a strong military changes both democratic inherits a large military from the previous in democracy always wishes to prevent coups but not be possible. In particular, soldiers realize that reform the military reducing their the military when it rents. has the chance to do when so, it In principle, the elite may this may the opportunity arises democracy will Since democracy cannot commit to not reforming can only make current concessions to soldiers (since promises of future concessions are not credible) is, nondemocratic regime, then be confronted with a choice between making concessions to the military and facing a coup threat. The decisive voter that and nondemocratic and current concessions may not be prevent democratization by using some combination of "carrot" and of "sticl';," not only by using repression but also by making concessions and promising redistribution of income to The scope and Umitations of such promises of redistribution have been analyzed in previous Acemoglu and Robinson, 2000 and 2006) and here we ignore them for simplicity. We also ignore the possibility of cooptation of some subset of the citizens using various means, such as the distribution of public (some work of) the poor. (see jobs (see, for example, Acemoglu, Ticchi and Vindigni, 2006). compensate the sufficient to our first result shows that may have large militaries societies in why which nondemocratic regimes social conflict in is in the past not only intuitive but also provides us with a particular nondemocracy may create democracy may lead to worse economic performance during democracy because and citizens up a of the military and on the extent of inequality. is limited, the elite prefer to allow a large military When smooth when military repression is likely to be effective depends on the effectiveness the military not very effective or is transition to democracy, because such a regime will not be highly redistributive (while repression hand, of the conflict soldiers that this induces. In oligarchy, whether the elite prefer to set inequality More costs for (future) democracies. the desire of the rich to prevent democratization by bequeathing a laxge army to specifically, between have chosen democracy and may instead end up with difficulty consolidating military dictatorships.^ This result reason a mihtary dictatorship. Consequently, soldiers for the prospect of is On likely to fail). and inequality the other high, the elite prefer is to build a strong military for repression and deal with the political agency problem by paying the military an efficiency wage. This equilibrium configuration will therefore correspond to a situation in which the military is an agent of the (effectively) elite, which is the presumption in the existing literature. However, we also show that under certain circumstances the prefer to use repression but not their own regime persist due to the repression of the coup against the soldiers, thus allowing military elite elite and acts citizens, in its whenever there interesting implication of our is own interests. an opportunity As the These for interests involve attempting a doing so.'^ is When relatively low, level of inequality increases inequality is and most of low, the society it is in the Most military mocratic regimes, is likely is form of and the society remains democratic, the government (the amount of public good provision) increases. As inequality further, the society is model concerns the relationship between inequality and democratic, but the amount of spending size of the coups against though the military undertaking the repression the size and composition of government spending. public goods. may to take place along the equilibrium path. In this case, nondemocratic regimes not an agent of the An pay high wages to elite to be nondemocratic (because it fails rises to transition to democracy). dictatorships in Central and South America, which have resulted from coups against de- may be thought of as being of this type. These examples are further discussed in Section 6. ^Examples of military dictatorships that have resulted from coups against oligarchic/nondemocratic regimes include the majority of the mihtary regimes that have emerged in Peru between the 1820s and the 1930s, the army-backed regime created in Thailand after the 1932 coup that ended the rule of the Thai absolutist monarchy, the regime created in Egypt after Major Nasser's coup against the monarch,y, and the junta in Panama in 1968 that followed the coup by the National Guard under the leadership of Omar Torrijos. In this case, the level of taxes and the amount of government spending depend on the nature of the nondemocratic regime. If it is oligarcliic, taxes and spending are low. However, a military dictatorship, then taxes and government spending are high, but is on the military We all if it is of this spending —not on public goods. then enrich our baseline model by introducing natural resources. Natural resources increase the political stakes because soldiers realize that they will be able to capture the natural resource rents more if they take power. As a result, natural resource abundance makes democracies The likely to fall to military coups. On ambiguous, however. repression and military. On may be effect of natural resources on oligarchic regimes is the one hand, the oligarchic regime has a stronger preference for able to use the income from natural resources in order to the other, the military is more tempted buy off the to undertake coups against the oligarchic regime. Our baseline model abstracts from the national defense only use of military coercion is role of the military, so that the domestic pohtics. At the end of the paper, we also use our in model to investigate a potentially important interaction between international and domestic In particular, politics. zation. Somewhat we investigate the national defense role of the military on democrati- paradoxically, we defense, democratic consolidation find that when becomes more the likely. army is more important The reason of an international role for the military, the citizens are unable to strong military. credible, The presence of international threats that, in the absence commit makes the promises because democracy also needs the military, and The two is facilitates for national to maintaining a of the citizens more democratic consolidation. building blocks of our approach are that the military should be considered as a potentially self-interested body —or in fact, a collection of self-interested individuals — and that there should be a distinction between nondemocratic regimes controlled by the economic elite ("oligarchies") and military dictatorships. Both of these receive support from the qualita- self-interest of the torships, army emphasized by, are major themes among The self-interest of soldiers and the corporate of the political science literature on military dicta- tive evidence in the political science literature. others. Finer (1976), Nordilinger (1977) and Needier (1987).^ ^This contrasts with existing models of democratic transitions or coups, such as those in Acemoglu and Robinson (2001, 2006), where the miUtary is assumed to be a perfect agent of the ehte. This same perspective is often adopted in political science and sociology literatures with Marxian foundations (e.g., Moore, 1966, Luebbert, 1991). For instance, Barrington Moore's (1966) classic work suggests an explanation for divergent paths of pohtical development in terms of different types of class coalitions. The possibility that any state apparatus, including the army, may have some independent role as a coalition partner, is ignored. Subsequent contributions inspired by Marxian approaches, including work by Poulzantas (1978) and Skocpol (1979), pay greater attention to the "agency" of the state, but do not focus on the role of the military in politics nor note Nordilinger (1977 p. 78), for example, argues that "The great majority of coups are partly, primarily, or entirely motivated by the defense or enactment of the military's corporate interests." Similarly, Needier (1987 p. 59) observes: the military typically intervenes in politics from a combination of motives "... which defense of the institutional in interests of the military itself predominates, although those interests are frequently construed so as to be complementary to the economic interests of the economic The distinction elite." between oligarchic regimes and the military dictatorships in a large political science literature (see for references there). Huntington (1968) example the survey and Finer also well rooted in Brooker, 2000, and the (1976), for example, emphasize the prevalence of authoritarian elite-controlled regimes supported oligarchic regimes, is by the military, and contrast these with military dictatorships. which are similar to our Examples of the former include the dictatorship that Getulio Vargas established in Brazil in 1937, which was a mainly civilian authoritarian regime, relying on the support of the military for its political survival, and other Central and South American regimes formed discussion in Section 6 below). in the Philippines his More at roughly the same time (see also the recent examples include Marcos's long lasting regime and President Fujimori's regime's in Peru, which was established following de facto coup to extend his rule and powers beyond their constitutional limits. Both of these regimes were backed by the army, but the military establishment did not have important decision-making powers. Perhaps more common in practice of officers are in direct control. (2000), Welch and Smith are military dictatorships, Such military dictatorships are studied (1974), Perlmutter (1977, 1981), examples include the regimes established Muhammad after the where the mihtary or a subset in in detail by Brooker Nordhnger (1977).* Contemporary Pakistan by General Ayub Khan, by General Zia-ul-Haq and by General Pervez Musharraf, the regimes established in Turkey coups in 1960, 1971 and 1980, in Guatemala after the coup of 1954 under the leadership Carlos Catrillo Armas, in El Salvador in 1956 with Oscar Osorio's government, in Brazil after we emphasize here. 'a hybrid regime type, "bureaucratic the interactions authoritarianism," based on a coalition between the military and the economic elites, whose primary goal is the promotion of economic modernization, has also been identified by O'Donnell (1973). The practical applicability of this concept has appeared quite limited though, and we do not pursuit it in this paper. the overthrow of President Joao Goulart's government in 1964, and in Greece after the military coup of 1967. The miUtary has also been the dominant political force in 1932 coup and has repeatedly intervened in pohtics whenever power by nascent Our paper is it Thailand since the perceived a threat to its own civilian political institutions.^ a contribution to a number of distinct literatures. First, there is now a substantial literature on political transitions, but to the best of our knowledge, no paper in this literature models the relationship between the elite Robinson, 2006, for an overview of this literature). and the military (see Acemoglu and Consequently, this literature does not distinguish between oligarchic regimes and military dictatorships discussed above. Second, there is a large literature on political agency in democracies, where citizens try to control politicians using elections and other methods (e.g., Barro, 1973, Ferejohn, 1986, Besley and Case, 1995, Persson, Roland and Tabellini, 1997, 2000, Acemoglu, 2005, Acemoglu, Golosov and Tsyvinski, 2008, Alesina and Tabellini, 2007; see Persson and Tabellini, 2000 and Besley, 2006, for surveys). In contrast, the potential agency relationship between the and the military has not been modeled. In this regard, the literature nondemocratic regimes and other aspects of agency relationships elite on the organization of in dictatorships is closely related to our work. This area includes the early important work by Wintrobe (1998), as well as recent papers by Acemoglu, Robinson and Verdier (2004), Egorov and Sonin (2004), Dixit (2006) and Debs (2007a,b) focusing on nondemocratic agency relationships. These works neither address the issue of controlling the military nor provide a framework for thinking about the emergence of military dictatorships. which studies the The only exceptions effect of the organization of the military are Ticchi and Vindigni (2003), on the consolidation of democracy, and recent work by Besley and Robinson (2007), which investigates the relationship between the military and civilian governments, though focusing on a different set of questions than our paper. Third, the recent literature in comparative politics of public finance and Tabellini, 1997, 2000, Persson types of democratic institutions on (e.g., Persson, Roland and Tabelhni, 2003) investigates the influence of fiscal policy different and economic outcomes, but does not inves- (somewhat more speculative) interpretation to the possibilitj' of where the elite allow a powerful military to form but do not take steps to prevent coups can be viewed as an implicit support for military dictatorships by the ehte. An example of this is the experience of Peru in the early 1930s. The raise of the Alianza Popular Revolucionaria Americana (a violent revolutionary movement) led the Peruvian elites to support a coup d'fetat by Colonel Sanchez Cerro. Roquie (1987 p. 115) describes this as follows: "The ruling classes that had long been civilian in their orientation put aside their distrust of the military and supported the colonel's coup. The massacres in Trujilio in 1932 involving the army and the APRA were to establish a long-lasting defensive alliance between the military and the upper bourgeoise." ^It is also possible to give a slightly different non-prevention in oligarchy. According to this interpretation, the case tigate the impact of different types of nondemocratic regimes, and in particular, the contrast between ohgarchic regimes and miUtary dictatorships. Our paper adds to modehng the impact of different types of nondemocratic institutions on nomic outcomes. In fiscal policy and eco- particular, a distinctive implication of our theory in this respect nondemocracies should typically have more military spending than democracies. mentioned above, there Finally, as dictatorships, though is approach it is that ^"^ a substantial political science literature on military this literature does not provide formal the military and the ehte, nor does by this hterature it models of the relationship between an agency problem. Some of the important as contributions in this literature have been cited above. The rest of the 3 characterizes the results. paper organized as follows. Section 2 presents our basic model. Section is Markov Perfect Equilibrium in this environment and presents our main Section 4 shows that the equilibria characterized in Section 3 are also the subgame perfect equilibria. of Section Section 5 presents a number of extensions of the baseline environment including an analysis of the implications of natural resources and international 2, some threats on oligarchic and democratic equilibria. Section 6 illustrates of the main ideas of our analysis with a brief discussion of the emergence and consolidation of democracy and the role of the military in Central America. Section 7 concludes, while the Appendix contains the proofs omitted from the text. 2 Basic Model In this section, we We describe the economic and political environment. horizon economy in discrete time with a unique maximizes final good. consider an infinite Each agent j at time = t oo Eo^/3*(c,, + x,-,A), (1) t=0 where Eq terms of the is the expectation at time is final good, Gt > is the t = 0, Cj^t amount > denotes the consumption of the agent in of public good provided at time t, and Xj,t G {0, 1} an indicator function denoting the occupational choice of the agent. This variable determines whether or not the individual benefits from the public good. The total population of the society and can produce A^ > 0, . is normalized to 1. Of those 1 —n > 1/2 have low skills while the remaining n agents are highly skilled and can produce A^ > A^ We will often refer to the '"This result is (rich) high-skill agents as the "elite" since at the beginning consistent with the evidence presented in Mulhgan, Gil and Sala-i-Martin (2004), which finds that nondemocratic regimes spend more on the military than democracies. they will be in control of the political system, and we "citizens." There are two occupations: producer and we use the subscript j E {L, M, H} high-skill producers. Soldiers A^ or depending on their will refer to low-skill soldier. With a slight agents as the abuse of notation, to also denote low-skill producers, military (soldiers) do not produce any output, while producers generate income To skill level. simplify the analysis and reduce notation, throughout that only low-skill agents can become Furthermore, we also assume that Xj,t = soldiers. 1 for j and A^ we assume ^^ = L and Xj,t = otherwise. This implies that only production workers benefit from the public good, for example, because the public good corresponds to services that rich agents and the military receive through other means (such as health care or schooling) or because the public good is associated with roads and other infrastructures mostly used by low-skill production workers. This assumption to simplify the expressions The and has no size of the military at time necessary for national defense to politics. In particular, size of the military at time thy at £ for denoted by at at — I Xt- We results. normalize the size of the military thus the only reason for xt We is repression in domestic G {0,x}, where S > xt is denote the decision to have (build) an army corresponding to xt G {Oil}, except that, as > one of two values, xt takes necessary for repression. with on our qualitative adopted — x. The government be described below, will it (social may group) take a while a newly-emerging democracy to be able to reform (disband) an already-existing army by choosing We is we assume that {0, 1}, power chooses in 0, t effect is also at — (that is, there will be further constraints on the choice of at in democracy). assume throughout that X < 1/2 - n, (2) so that low-skill producers are always the absolute majority in the population. This imphes median voter that, given the policy instruments specified below, in majoritarian elections the will always be a low-skill producer. Aggregate (pre-tax) output at time Yt where is i/Jj G {1 — cp, t will be = ipt{{l-n- xt) A^ + 1} captures potential distortions 7iA^) from coups with no coup atteiupt (against oligarchy or democracy), we have a coup attempt, a fraction ''This is cp of production natural, since military wages are is lost (3) , (pt = is recruiting, more attractive to though we do not do G (0, 1). Instead, When when there there is due to the disruption created by the coup, low-skill than to high-skill agents. We could is greater than those of soldiers introduce an additional constraint to ensure that the income of rich agents whenever the military 1- (f) so to simplify notation. thus = (fit; — 1 < (j) 1. Let us also denote Y = {l-n)A'^+7iA^, which as the potential output of the economy, to will (4) apply when the size of the military is equal to raise revenue for public good and there are no disruptions from coups. Aggregate output can be taxed at the rate provision and to pay the salaries of soldiers. C (r^) Yt) We G [0, 1] model the distortion of the manner: when the tax rate in a simple reduced-form (thus a total of Tt will be lost is a fraction Tt, costs of taxation C (rt) of the output due to tax distortions. These may result from distortions resulting because taxes discourage labor supply or savings, or because of the administrative when costs of collecting taxes. This fiscal technology implies that government revenues per unit of production will the tax rate is equal to be rt-C{Tt). We assume that C : — M+ » [0, 1] is (5) a continuously differentiable and strictly convex function that satisfies the following Inada conditions (which will ensure interior solutions): C(0) (so that there are define f £ (0, 1) no distortions without taxation), C" as the level of taxation at of the Laffer curve) , which is which fiscal (0) = and C" (1) > 1. revenues are maximized = Let us also (i.e., the peak uniquely defined by T^{C')-\l), where (C")~ between is and 1 Tt, the inverse of the derivative of the C (6) function. This tax rate f is strictly because of the Inada conditions. Finally, without loss of any generality, we parameterize A^ = A^ and A^ as i^r and >1^ = -y, —n n (7) 1 for some 9 G We inequality. (n, 1). This parameterization implies that a higher 9 corresponds to greater do not make the dependence of A^ and A^ on 9 explicit unless this is necessary for emphasis. We will (citizens) represent the and the elite. economy as a dynamic game between soldiers, low-skill producers As explained further below, given the policy instruments, there conflict within the groups, so we can suppose without any loss of generality that is no a representative agent from each group (e.g., commander the of the army, a decisive voter in democracy, or a representative agent in oligarchy) makes the relevant policy choices. ^^ In principle, there are two state variables in this game. The second military from the previous period, xt-\ 6 {0, x}. which takes one of three values, democracy D, oligarchy St, size of the military The is first one the size of the is the political regime denoted by M. The E, military (elite control) from the previous period, X(_i, only matters because immediately following a transition to democracy (without a coup attempt) government to choose = at and reform the may it not be possible for the democratic military. Instead, we assume that they have will to wait for one period before being able to reform the military. This assumption captures the once a large army realistic featirre that is new and in place, a government may not be able to disband the army immediately. The period for to be able to do so it is weak democratic potentially fact that only for simplicity and in subsection it 5.3, takes only one we consider the case where the opportunity to reform the military arises stochastically. This structure simplifies not only the analysis but also the notation. In particular, instead of carrying xt-i as a state variable TD, which occurs and only if, we can if, X(_i democracy define an additional regime, "transitional" = = x, st-i E, and there has been no coup attempt. Loosely speaking, transitional democracy corresponds to a situation in which the majority (lowproducers) have de jure political power but this skill power of the = at is military. The only not possible. followed by St+i St gS = Key > 0, = M. = D, unless there Given policy decisions are wage In addition to i>t G G [0, 1] for soldiers, at, Tt, {0, 1}, '"Alternatively, all and st —D is that with a successful military coup, we = TD, st is in = TD is in which case it St represent the state of the dynamic made by the government with in power. on the income of the producers, the Wt game > 0, The policy decisions are a level of public and the decision regarding the good provision, army, size of the at 6 {0, 1}. Gt and Wt, the military (the military commander) decides whether or not ipf. = 1 whether or not to repress the racy is this description, to undertake a coup against the regime in by = TD {D,E,M, TD}. linear tax rate Tt Gt st Otherwise, in both regimes, a representative low-skill producer immediately followed by st+i by between constrained by the de facto political Since in our baseline model transitional democracy lasts for one period, power. is difference is power (if the military is corresponding to a coup attempt, and, citizens in oligarchy, we could consider a which citizen-candidate model (e.g., is active), if V't denoted by Pf 6 0, it also {0, 1}, elite vote. but the analysis requires additional notation. 10 is denoted decides with Pt — I Beslcy and Coate, 1997) in which in democ- agents vote, in military only the soldiers vote, and in oligarchy only the rich this case are identical to those presented below, — which The results in corresponding to repression. In addition, individuals should have a decision of whether to apply to an army position and also soldiers should have a decision to quit the military. Both of these decisions translate into a participation constraint that the value to a low-skill agent in the military should be higher than the value to a low-skill producer. Assumption that this is 1 below ensures always the case and to simplify notation, we do not introduce these additional decisions explicitly. ^^ =D particular, st+i smooth up a at — 0), 1 following transition, there is = St then there 0, is a smooth transition to democracy, in E. The important point here is if in oligarchy, the military pj = so that there 1, the elite choose an is and effective military it demands = TD. Therefore, [0,1]. of the military repression, with the case in which the repression of the democratic st+i is, Finally, if chooses to repress the citizens, then transition to democracy takes place with probability n e an inverse measure of the effectiveness set present but chooses not to repress is then the regime transitions to transitional democracy, that and when that no transitional democracy since the oligarchic regime has not In contrast, military. = (/Oj = oligarchy the elite choose at If in tt = n represents corresponding to by the military of the citizens is fully effective. We both also cases, assume that when the military attempts a coup against is denoted by i/'t = 1, it = M, then a military dictatorship, sj+i s = M is = D will st+i = D st we immediately have is i + St 1; = TD, if, To emerges. on the other hand, is = St E, then the of opportunity resulting only arises following failed repression. If automatically consolidated (and thus between the army and the suffers a military coup regardless of the regime at time coiiflict if t\ fails, then the regime at time between the military and the from the failure of the military. This description therefore implies that the transitional — a coup ever succeeds elite s there = is no repression, there D), and if there is In other words, loss it never a best response will and </?( = coup to reform the democracy regime, is s — TD, no army, so democracy a coup attempt, then the conflict enables the emergence of consolidated democracy (and thus, D). Finally, recall that when the military attempts a coup an income elite no effective repression and a democratic regime emerges immediately and can use the window again, s if if we assume that then the transitional period will be over and the army will be reformed at implies that there is the coup succeeds, simplify the analysis, be absorbing). However, also If (0, 1). absorbing, so the society will remain as a mihtary dictatorship (and in equilibrium, t succeeds with probability 7 6 either regime, which, in 1 — (/> < 1 in equation always be a best response 11 = 1), the society (3). for low-skill for soldiers to quit the military. {tpt producers to apply to the military and it is Throughout we adopt the convention that even there if is fiscal poUcies enacted at time are implemented t regime change and the new regime starts enacting policies from Moreover, the military wage Wt announced by any non-military regime repression and no coup attempt. These wages are withheld The government budget constraint at time wtXt aggregate income at time where Yt is from the linear tax Tt, t if there is is t + l onwards. conditional on both a coup attempt. can therefore be written as t + Gt<iTt-C{Tt))Yu given in and the left-hand side (3), (tj is (8) - C (ti)) Yt is total tax revenue resulting the total outlays of the government, consisting of public good expenditures, Gt, and spending on the military, wtXt- We now variable 1. st summarize the timing of At time events. t, the economy starts with the state 6 S, which determines the group in power. Unless St = TD, the group in power chooses at G {0,1} and announces a vector {rtjWtjGt) that satisfies the government budget constraint at—1 2. and the group In oligarchy, if there in power only chooses {Tt,wt,Gt). is no military consolidated democracy and St+i 3. When Xt = x, in democracy and or not to attempt a coup probability tpt = If & {0,1}. repress, then st+i 5. If 0), = repression 0, — = is then there = TD, then a transition to fully- is is commander attempted, established (and St+i it decides whether is successful with — M). If a coup fails, D, emerges next period. x), then fails a coup If 7 and a military dictatorship the elite have formed a military (xt citizens, pt — in oligarchy, the military e {0,1}. and does not attempt a coup Vt xt policy D. a consolidated democratic regime, St+i 4. = [at If St (8). fiscal then st+\ it = D. also decides (probability tt) or If the military remains active whether or not to repress the if the military chooses not to — TD. Taxes are collected and wages are paid according to the announced policy vector {Tt,wt, Gt) if there is no military coup attempt. If there is such an attempt, then wt = 0.^'* timing of events makes it clear that when a particular vector of policies, (rt, u'l, Gt), announced, it is not known for sure whether this will satisfy the government budget constraint, (8), because there might be a coup attempt reducing income and tax revenues, and also removing the burden of military expenditures. In such cases, we assume that the amount of public good provision, Gt, is the "residual claimant," so that if {Tt,Wi,Gi) docs not satisfy (8) as equality, Gt adjusts up or down to ensure this. '""This description of the is 12 . Finally, with we assume that the = a:_i society starts with s = E, an oligarchic regime, and i.e., = so the elite are free to either form a military of size x or leave xq 0, also in the initial period. Characterization of Equilibria 3 We now characterize the section. Markov Section 4, Markov in the previous Perfect Equilibria are both simple and natural in the current context. we show that of om- results generalize to all Subgame In Perfect Equilibria. Definition of Equilibrium 3.1 We first focus on pure strategy the dynamic be the H^''' Markov game described above up set of state variables, here € fcf to time t MPE is on the entire MPE is a natural equilibrium because present in a very similar fashion formally, let i^*''^. and let Markovian on the payoff-relevant history, condition only it we when we /i**'^ G H^'''. will see that the focus on for :5x/C-^ In the context of the subgame xE^ [0, 1] game commitment problems directly introduces the same commitment problems are perfect equilibria. be a Markovian strategy mapping, that cr cr which assigns a value historj' in t, defined as a set of Markovian strategies that are best that are central to our analysis. However, More denote the history of and stage k of the stage game of time responses to each other given every possible history here, h*'^ 6 S, and on the prior actions within the same stage game, denoted st Consequently, a /C. (MPE). Let Perfect Equilibria such histories. Strategies assign actions for any strategies, instead of conditioning by game described Perfect Equilibria of the is, X {0,l}^ each of the actions, the tax rate tj e [0, 1], the military wage wt G R+, the level of the pubhc good Gt G M+, the decision of whether to create or reform the military at G for {0, 1}, each value of the state variable given by j-i^t,k and the coup and repression decisions of the g /C. //*.*;. An MPE is S and is To characterize the dynamics induced by the MPE st 1p^ G conditional on a best response to itself at of political institutions, its (st I st-i) -.SxS and pt G {0, 1}, every possible history we define the one-step past value and the limiting distribution of s strategy profile a* as p {0, 1} each combination of prior moves in the stage game a mapping a* that transition probability of military, -^[0,1] and g 13 (s) : 5 ^ [0, 1] These concepts will be useful in describing how regimes change and what the in equilibrium likelihood of different regimes are in the long run. We next proceed to characterizing the MPE by first determining (the net present dis- counted) values of different individuals (groups) under different regimes. Values in Democracy 3.2 Let us start with the values of the three groups in democracy, are made by = st D. Decisions in democracy majoritarian voting and the median voter will always be a low-skill producer. This implies that democracy will be an absorbing state as long as future periods, and clearly, aj = is democracy. In view of this, = a^ chosen in is all always a dominant strategy for a low-skilled producer in we obtain the Y following proposition (recall that is defined in (4)). Proposition 1 The unique democracy) involves tax rate t^ is [at = MPE 0,Tt — starting in — T^,wt any subgame with — G^) -jGt s — D at each date t, (i.e., a consolidated where the democratic given by r^^{Cr'{l-^), and the level of public good is (9) given by G^ = (r^-C(r°))F. Moreover, consolidated democracy Proof. Clearly, the an absorbing state so that is median voter is p{D D) = \ democracy would never choose in would reduce the tax base and potentially create a coup democracy (10) tlireat against 1 and q [D] at = 1, ~ 1. since this democracy. Therefore, an absorbing state and the optimal policy can be characterized by the solution to the following static maximization problem max u'^(D)= (1 - t) A^ + G (11) r6iO,l],GeR+ subject to G< Given the Inada conditions and the convexity first-order condition of this of - C (r)) Y. C (r), (11) has a unique interior solution. The problem then gives the unique equilibrium tax rate corresponding level of public good as The (r (10). values of low- and high-skill producers are then given by 14 as (9), and the . and Here and throughout, D denote the identity of "values," that The tlie Af or TD) in parentheses denotes the regime, while superscripts V u denotes per period returns and In addition, agent. denotes the net present discounted values. value of an ex soldier in this regime now work will is, (or £, as low-skill producers. a^ At = also is this point, (1 V^' {D) it is = V^ [D), since former soldiers also useful to define - r^) A^ + G^ (14) as the net per period return to a low-skill producer in democracy. This expression will feature frequently in the subsequent analysis. Finally, recall that according to our parameterization evident that is r^ does not depend on due to the well-known linear taxes (e.g., A^ = [1 — 9)Y/ [\ Y and is a strictly increasing function of 9. effect of inequality on redistribution in — n) . This then It is last result models of majority voting on Romer, 1975, Roberts, 1977, Meltzer and Richard, 1981). We note this as a corollary for future reference (proof omitted) Corollary 1 The democratic tax inequality parameterized by 3.3 t^ , given by (9), is strictly increasing in the extent of 0. Values under Military Rule A military regime, s xt G = M, {0, S}, the military disbanding this rate itself can only occur when xt = x and it is government has no option to change the (and obviously it an absorbing state. size of the military would not want to expand the military even if it Since without could, since would imply dissipation of the rents captured by the military among a greater number of soldiers). Consequently, the military government wall simply maximize the utility of soldiers Moreover, we assume that there subject to the government budget constraint. is a natural seniority system in the military, so that current soldiers are not fired in order to hire applicants. Then, provided that the participation constraint of soldiers current soldiers will remain in the in state M the military can never army lose forever is satisfied (see below), and receive the military wage. power, this problem boils down Finally, to that of because maximizing the static utility of a representative soldier subject to the government budget constraint. 15 new At this point, consider the maximization problem of the military government. This takes the form u^(M)= max w (15) TG[0,l],ujeM+,G6R+ towx + G subject where the objective function incorporates the good (i.e., = XM,t f defined in {Y - xA^) fact that soldiers r does not feature in (1)). Since (15) involves taxing at rate <{t-C (r)) (6) to wage of w^'^ given do not benefit from the public in the objective function, the solution to maximize tax revenues (thus maximizing the — constraint set) and also setting to zero the public good, Gt soldier , This generates a unique per 0. by X where Y again denotes the potential output defined in government as possible is in and power it undertaking a coup which is a^ (which will is ensure is tax rate f extracting as is the one where tu^ an attractive option this. much revenue from the producers proceeds to the soldiers. As noted above, the interesting will redistribute all the case for the current paper 1, will set the Evidently, as long as the military (4). is high that entering the military and sufficiently We for low-skill producers. next impose Assumption Recall that the per period utility of the low-skill agents in democracy simply a function of the underlying parameters, given by defined uniquely in (9) and (10)). The (14), following assumption ensures that with w^ is r^ and G^ sufficiently greater than a^ so that the participation constraint of soldiers (of being in the military both during military dictatorships, democratic regimes and oligarchies) Assumption is always satisfied. 1 7 > -3-737—3 ^7. In particular, this assumption immediately implies that the expected value of a coup for soldiers is always greater than the value of a low-skill producer in permanent democracy, that is, V^^ (coup) which tion is 1 is 7 > stating sufficient to =p [jV^^ (M) -f (1 - 7) V^ {D)] guarantee the participation constraint in > V^ {D) all , regimes. Note that Assump- not very restrictive. For example, as the size of the military, x, becomes small, any satisfies this assumption. We maintain Assumption it explicitly. 16 1 throughout the paper, without , The following proposition describes the equilibrium under a military rule (proof omitted). MPE in Proposition 2 The unique any subgame starting with — ship) involves the following policy vector at each date: r^^ Moreover, military dictatorship w^'' given in (16). and starting with s M — Given the unique continuation equilibrium values for individuals who are currently for high-skill elite agents are given (i.e., f as defined in military dictator(6), G^ = = \ and 0, M) = 1 1. and of the military, for low-skill non-military by ' ^ M in Proposition 2, the (net present discounted) members yM (M) = l^^lm = 1-^ ~ ^ = an absorbing state so thatp{M is any point, we have q (M) at s ir-C ^_ 1-/? (y - x^^) (f)) ^^'' ' X -^^ ^ l^' (M) = ^^^^ and Notice that the expressions for u^ (Af ) and becoming a of low-skill producers would like to become a V^ (M) do This soldier. not include a term for the option value because even though each low-skill producer is = not be hiring any more soldiers, since xt soldier, the military will x already and there are no quits from the army. Values in Transitional Democracy 3.4 We now regime turn to the analysis of transitional democracy, where will emerge when st^i Moreover, this regime fails at this point, is = — E, xt-i indeed "transient" then Si+i = D, and if ; x, Pi_i there if there is St = TD. Recall that this — 1 is no coup attempt or the coup attempt and repression fails (probability a successful coup attempt, then st+i depending on the subsequent regime, either the equilibrium of Proposition Clearly, of Proposition 1 will apply. We now investigate policy choices tt). — M. 2 or that and the reaction of the military during the transitional period. Suppose that the democratic government during the transitional period has announced the policy vector [t^^,w'^^,G'^^), and to start with, let us ignore the participation constraint of soldiers, Now if soldier which ensures that they prefer not to quit the military the military chooses Vt = (that (will return to this below). no coup attempt), then the value of a typical is, is V^'^ {TD no I coup) = u;™ + 17 (3V^ {D) (20) . which incorporates the do not receive utility fact that the soldiers will receive the military We from the public good. value to the soldiers, which, from Proposition = V^ consolidated democracy V^' {D) have also substituted and equation 1 = t/'^ the value of a soldier 1, V^^ {TD coup) = receive the {M) low-skill as in (17) is (cf. given as the value in a former soldiers all V^ (iW) + - (1 will (D)] at the become (21) , the military undertakes a coup, the soldiers do not Following a successful coup, 7. established and the continuation value of the soldiers Proposition 1, {D). V^ 7) wage w'^^ and the coup succeeds with probability a military dictatorship V^'^ when fact that the continuation is [7^^ /3 I which incorporates the for {D). In particular, in Una with Proposition low-skill producers, accounting for the continuation value of if is (12), next date the government will choose to reform the army and In contrast, wage w"^^ today and the coup If 2). fails (probability 1 — is given by 7), soldiers become producers and simply receive the continuation value of the low-skill producer in a consolidated democracy, V'^ {D) as in (12).^^ Comparing these two expressions, we obtain the no-coup constraint: V^ [TD Using (20) and (21), the no coup) > I w^'^ is \ coup) (22) no-coup constraint can be written as w^'^ where V^ {TD ^7 > {u^'' - a") (23) , defined in (16) and a^ in (14). Constraint (23) defines the minimum military wage that democracy must offer to soldiers in order to prevent a coup attempt. In what follows, will use w'^^ for the and t"^^ wage for the "transitional prevention" ( corresponding tax for the military to induce Clearly, this rate. them ) wage that makes This wage level can be thought of as an efficiency is, not to undertake a coup). of the coup 7, and the gap between the value that soldiers will receive in a military dictatorship, V'^ value in democracy, The question is V^ whether ^ There no (1 — will satisfy the depend on two ly^^/ (1 this — factors. minimum The government budget constraint, desirable for low-skill producers to differential taxation or further = and their military wage /?), /?). coup attempt. This pay such a wage (and it is is — a^/ (M) whether a democratic government would pay in order to prevent a feasible to {D) this constraint hold as equality to take the right action (that wage depends on the success probability we pay this wage. The (8)). whether it is The second is feasibility condition requires punishments that are possible on former 18 first is soldiers. minimum wage this can be raised, that times the number of soldiers to be than the maximum revenue that is, w'^^x The right-hand less side is <{t-C (f)) maximum the (y - xA^) (24) . revenue that can be raised, since it involves taxation at the revenue-maximizing rate f and the tax base consists of the entire population except the x soldiers. < w"^^ Using the expression lu^^. Alternatively, constraint, (24), w^^ using the expression for satisfied is for ly^^ in (16), the condition (24) if and only from can be seen to be equivalent to (23), we find that the feasibihty if w^^ -/3 1 7< . ^7- ..,A./_,L This condition states that preventing a coup attempt is (25) feasible only if 7 is than some less Otherwise, only very high wages will discourage soldiers from attempting critical threshold 7. a coup and such high wages cannot be paid without violating the government budget constraint. Note also that Assumption Next, suppose that low-skill producers to coup? To answer satisfied it 1 pay this this question, —and compare the pay the necessary 7.-^'' wage or is them better for it . beneficial for Is it to face the risk of a military suppose that a wage of w'^^ can indeed be paid low-skill producers' utilities no coup) = (l I — (25) When under the two scenarios. - r^^) A^ + G^^ + i.e., is they is I3V^ [D) (26) , the utility-maximizing tax rate for the low-skill producer subject to the no-coup is constraint, (23), and G^^ > the transitional phase. is the utility-maximizing level of public good spending during Alternatively, if there democracy, the value of a low-skill producer V^ {TD < efficiency wage, the value of low-skill producers V^ {TD where t^^ thus 7 democracy to pay the wage w^^ feasible for is < w^^ and ensures a^ coup) I - (1 - T™) since in this case there are no -(t>)A^ fact. no coup prevention during the transitional is + G™ + payments to the utility-maximizing policy choices expression incorporates this (1 is /3 military, when coup attempts Current output is (1 - 7) V^ (£>)] (27) , and now t"^^ and G'^^ denote the are not prevented. '' a fraction '"This also implies that the participation constraint of soldiers + [jV^ (M) 1 — in this case, The rest of the of potential output (4) V^' (TD) > V^ (D), is always satisfied. ''The full maximization program when the proof of Proposition low-slcill producer chooses not to prevent coups 3. 19 is given in the . because of the disruption caused by the coup attempt and there coup and the regime from tomorrow on successful is V^ (M) Now to the representative low-skill producer. we obtain will is a probabihty 7 that the be a military dictatorship, giving value combining the previous two expressions, that a low-skill producer will prefer to prevent coups during transitional democracy if V^ {TD no coup) > {TD V'^ I The next proposition shows that whenever (25) is \ coup) (28) satisfied, (28) is also satisfied, so that coups against democratic governments are always prevented when prevention Assumption 1 is assumed Proposition 3 Let 7 s — TD • If is to hold throughout, thus be defined by (25). J G [7,7], then the transitional military is MPE in 7 > 7. any subgame starting with At the next date, = D s' (i.e., p{D TD) = 1), \ the implemented. The long-run equilibrium in this case involves democracy 1 is with probability J 6 democracy chooses the policy vector (^t^^,w'^^ ,G'^^) reformed and the policy vector in consolidated democracy characterized in Proposition (7, 1], 1, i.e., q{D) = 1. then transitional democracy chooses the policy vector (r^^ ,w'^^ ,G'^^^ and the military attempts a coup, p Then, the unique to fiscally feasible. as follows. and prevents a military coup. • If we Hmit attention is [M TD) = 7, I i.e., ip = 1. and thus starting with s Consequently, we havep{D TD) = = TD, — 7 and q {D) = 1 \ q (M) = 1 — 7 and 7. Proof. See the Appendix. The essence of Proposition 3 ducers are better off when coups is that condition (28) are prevented. coup prevention may require excessive when condition (25) is future, both low-skill in Proposition 1, if to Markovian MPE after s Thus the =D of this inefficiency is made better off. is Nevertheless, as involves reform of the military in the shown and thus no effi- inability of the democratic regime to commit to future We will see in Section 4 that the restriction the source of coup attempts. strategies In particular, this will be the case The source . producers and soldiers could be ciency wages for the soldiers. is not be possible because the democratic regime could promise high wages to soldiers in the the unique rewards to soldiers may a clear inefficiency in equilibria involving coups is (because of the economic disruption that they cause) commitment problem; alwajj^s satisfied, so low-skill pro- However, this efficiency wages. not satisfied. There is not important here. 20 Instead, the commitment problem emerges from the underlying economics of the interaction between democracy and the mihtary fact that the military, A number whose main function here repression, is of other features related to this result are then transitional democracy may pay even not needed in democracy. is worth noting. higher wages (and thus we the military than an oligarchic regime would (which will —the First, if (25) make is satisfied, greater concessions) to study in the next subsection) . This again reflects the commitment problem; because democracy has no use for a large military, cannot commit to not reforming and thus it needs to make greater concessions today in order Second, the extent of income inequality influences whether there to prevent coup attempts. will it be coup attempts against democracy. In particular, the threshold 7 the inequality parameter 9 via its effects on a^ and w^'^ . It —a higher 9— increases w^^ and reduces a^, thus reducing is lower, soldiers, more unequal societies. Intuitively, this and thus the prospect who of more tempted are becoming a is 7. This makes coups more producer is is V^ {D) attractive for the current is less coup to secure a military dictatorship. The rents that soldiers can appropriate in the military regime are also greater in a society because net output depends on because in a more unequal society, low-skill to undertake a in (25) can be verified easily that greater inequality likely in it greater (a smaller fraction of the potential output more unequal Y given in (4) foregone as a result of the fact that some of the potential producers are joining the army). We state this result in the next corollary (proof in the text): Corollary 2 Higher more inequality (higher 6) reduces 7 and makes coups in transitional democracy likely. Finally, elite in it is compute, also useful to a transitional democracy. not depend on whether there is for future reference, the values to soldiers First, the value to soldiers in transitional coup prevention or not. This is because in and the democracy does both cases soldiers receive the value of a coup against democracy, either as expected return for undertaking a coup or as a result of the efficiency no-coup constraint. This value is wages paid by the democratic government to given by V^ {TD) = In contrast, the value to the elite 13 [)V^ (M) + (1 - 7) V^ depends on whether 7 is determines whether transitional democracy can prevent coups. coup attempt in transitional V" {TD coup) i = democracy and the (1 - satisfy their <^) (1 elite's - T™) A" + 13 21 value [71/^ is {D)] (29) . greater or less than 7, which When 7 > 7, there will be a given by (M) -)-(!- 7) V" [D)] . (30) , In contrast, when 7 < 7, V" {TD where r^^ and t'^^ = no coup) - r^^) A" + I3V" (D) (l I refer to the tax rates defined in Proposition 3. Values in Oligcirchy 3.5 We now turn to the analysis of subgames starting with s that unhke democracy, an oHgarchic regime is (31) may = E. The key economic insight here benefit from having a military used repression (thus preventing democratization). Counteracting this problem mentioned is the political moral hazard whereby the military may turn against the in the Introduction, for and elite try to establish a military dictatorship. In oHgarchy, the 1. choose Xt — choose Xt = 3. choose Xt = strategy by The = X and allow coups. We denote this strategy by P if elite is lemma, which next compute the values to the cases, the elite always supply smooth accept the fact that denote this is, = p 1 this is costly for them if in 1. This will give them a value the value of the th high-skill (elite) (13). 22 In all terms of taxes they choose the transition, they will set the lowest possible tax rate {t^ \ at the whenever corresponding to these three strategies. no public good, since p{D E) — is be stated and proved at the end. will elite which 1, indeed viable. Throughout this subsection we V^{E,S) = A" + I3V^{D), by We direct (financial) costs Lemma and they do not obtain any benefit from public good. Consequently, is "non-prevention;" shows that the military always prefers repression, that of the result of this ^ {D) where V'^ (£>) denote the military chooses not to repress, would obtain no benefit from having the military and pay both strategy, that of We for "prevention." E, thus the third strategy for the We 0). X and pay high enough military wages to prevent coups. this section, make use = x_i N for indirect costs (in terms of the risk of a military dictatorship). end of s (recall that by S, "smooth transition;" third strategy would not be attractive for the elite since they and have three possible strategies: and allow immediate democratization this strategy 2. elite = 0) first and of (32) individuals in consolidated democracy given The second strategy for the ehte, non-prevention, mihtary coups. In and V" where (1 V^ (M), from their value - (f)) A^ is this strategy = [E, iV) (1 - <^) can be written as A^ + /3 ['yV'' (M) + (1 - 7) given by (19), fails V" {D)] (33) , the flow payoff to the elite (net of the disruption caused by the coup), is the value to the elite under military dictatorship, which occurs is and successful (probability 7), finally the value to the elite in consolidated democracy, which occurs military if an army, but not to prevent Consequently, zero taxes are again feasible and optimal for the coup attempt by the military is to create equiUbrium, soldiers attempt a coup against the ohgarchic regime and this therefore receive zero wages. elite, is (probability 1 — 7). Recall that here a coup attempt against the oligarchic regime we given by (13), the coup attempt by the making use are fails, this if V^ {D), the if of the assumption that immediately leads to a consolidated democracy. Finally, if the elite set up an army wage to prevent military coups, then V^ {E, P) = where is V^ {TD) is (1 - r^) and to repress the citizens their value A^ + p [(1 also pay the required can be written recursively as - n) V" [E, P) + uV" {TD)] (34) also incorporates that to prevent military coups the elite have to impose a tax rate of t^ on all incomes the elite repression the value fails V^ {TD). and the 1 political state switches to transitional Rearranging and (to finance the military efficiency wage), the state of oligarchy with prevention recurs next period with probability tt, (34) , given by (30) or (31) in the previous subsection depending on whether 7 greater than or less than 7. probability efficiency — tt, whereas with democracy, giving we obtain (34), where, for future reference, we have written this value as a function of the probability that repression is fails, tt, i.e. as V^ {E, P it). \ The important observation here continuous and strictly decreasing in n. (34), by noting that To easiest way to see that it is that V^ {E, P decreasing is \ tt) from V" {E, P\n) = V" {E, P) > V" {TD). characterize the equilibrium starting in a value of the tax rate elite The is t^ necessary subgame with to allow for prevention, 23 = E, we need to compute the and then compare the values to the from the three possible strategies outlined above. Let us military. s first look at the values to the The They military has three possible strategies. oligarchy and, if decide whether to attempt a coup against they do not undertake a coup, whether or not to repress the citizens. participation constraints of soldiers and oligarchy will chosen. We repression) , make sure that the last option refer to the military's strategies as "coup", "repression" (to and "non-repression" (to denote no coup and non-repression) when they attempt the value to the soldiers a coup (i/' = 1). The not is denote no coup and Let us . first consider Their value can then be written as: V^ {E = p [tF^ coup) I (A/) + (1 - 7) V^ (D)] (36) . This expression incorporates the fact that when they attempt a coup, soldiers will not receive a wage and that the coup will succeed with probability 7 giving them a value of F^^ (A/) and fail with probability 1 — democracy with a value The and p = of V^ value to the soldiers {D) repression) = + /3 [W^^ [TD) + u;^ reflects the fact that soldiers receive successfully carried out the necessary repression. fact that the same succeeds). instead repression If state will recur is (probability fails (1 - tt) by V^" {E w^ the wage tt), V^'^ repression) {E is repression)] — Q . today because they have 1 — tt (i.e., repression) and V^'^ = they will receive the continuation value of we obtain that vjP + pT^V^^ {TD) —-^—-^—^. " (37) I- P[1-7T) therefore repression) {E | for the as long as repression > V'^" {E coup) (38) \ I with V'^ [E [tjj (29) (recall that this expression applies regardless I no- coup constraint in oligarchy \ The continuation value then accounts greater than 7 or not). Solving the recursion above I/^^i? The a coup and choose repression tomorrow with probability transitional democracy, V^'^ [TD), given of whether 7 a fully consolidated ^^ when they do not attempt I This expression the soldiers. for will transition to the recursion 1), satisfies V^' [E which case the regime 7, in coup) given in (36) and (37).^^ The I expressions in (37) is the same as the value from transitional coup) = V^^ (TD), since a coup gives them the same value regardless of which regime it is attempted against. Furthermore, recall also that in the MPE of the subgame beginning in transitional democracy, soldiers have the same value regardless of whether they attempt a coup (since when a coup is '^Notice that the value to soldiers from a coup against the oligarchy democracy, i.e. V'^' {E \ prevented, the no-coup constraint ^'We can is binding). also verify at this point that the participation constraint of soldiers in oligarchy This participation is given by V'^ (E | repression) > V^ {E, S), since, in view of Lemma is always satisfied. whenever it exists, the be smooth transition to democracy. Using the fact that V^ {E,S) = A^ + PV^ (D) together with the no-coup constraint (38) and Assumption 1, ensures that the above inequality and thus the participation constraint is always satisfied. military prefers repression to no repression. Without the 1, 24 military, there will . and (36) immediately imply that the constraint (38) The minimum w^ > (Sjw^ +/3 equivalent to is wage consistent with coup prevention military . — (1 7) a^. is w^ = l3^w^ + I3{l-j)a^. To (39) finance this military wage, the elite will impose a tax rate of t^, which government budget constraint (8), u,Px Combining satisfy the thus ={tP and using (40) with (39) must -C (r^) we (16), (y - xA^) ) (40) . find that this tax rate is implicitly and uniquely defined by = ^7(f-C(f)) + /3(l-7)^7^^+C(r^), r^ and moreover t^ G (0,f). convex and C (0) = positive. strictly satisfies Thus at The uniqueness 0, of most one value of t^ can decreasing in inequality. satisfy (41) Therefore, when of inequality on the no-coup constraint. more is likely collected for any level of r effect of inequality greed effect and r^ Finally, satisfied. it on is V^'^ side are strictly unique solution is inequality is is t^ greater, a lower tax rate strictly is is sufficient the outcome of two counteracting effects become democracy becomes low-skill producers. A the military attempts a coup against the oligarchic reduces effect of inequality w^ and t^ . Counteracting when V^^ {M), increases xt — x are also increasing > is at f. fiscal [E repression) > V^ C(0) = {E, S) 0. in fact proportional to the always dominates the effect 7 and is always that, strictly increasing in 7, the = A^ + j3V^ {D) To estabhsh that t^ < by the definition of f, r — C Moreover, when 7 = 1, we have t'' — C (r''') is revenues given by follows from the fact that strictly increasing in since t'' and {M). Consequently, the outside option I maximized in 9 However, the can be verified that the participation constraint of soldiers in oligarchy This constraint ^"That T^ (recall (17)). decreasing in 9 (in the extent of inequality) V'^'' 7 strictly is the "greed" effect resulting from the fact that as inequality increases, the value to soldiers in military dictatorship, 1; this First, as inequality increases when regime and thus this "outside option" = and moreover, the soldiers because in democracy they will consolidated democracy is C {) fact that also establishes that the tax rate prevent coup attempts. This result less attractive for from the and t?^ between for the elite to follows and the remaining terms on the right-hand Differentiation of (41) with respect to this t^ (41) (t) = is note from (41) that t'' t for all r < f and strictly increasing in {t same conclusion holds 25 f, (42) — C (r)), for all which implies that 7 6 [0, 1]. , r'' < is is f when The no-coup sumption combined with 1, V^ (14). Since We end repression oligarchy constraint, (38), implies tliat = is always \ repression) coup) \ > V^ > V^'' {E {D), where coup), while As- \ V^ [D) is defined in satisfied. by showing that the military never chooses no coup and non- = The remainder 0) in oligarchy. of the analysis of equilibrium in presented in the next subsection in the context of the complete characterization of MPE. the Lemma and pi = 1 In any MPE starting = subgame with in a St = E, the military never chooses ipi — 0. Proof. The value tpi {E,S), (42) and p V^ [E (36), implies this subsection {ip is > V^ (D) V'^ {E and receive no p^ = 0, is V^'^ {E \ V^ {TD) obtain the value V'^ {E footnote IS), \ = no repression) wage and the regime receive the value when they choose not to the soldiers in the to attempt a coup and not repress, /3V^^ [TD), since, without repression, soldiers next period will be transitional democracy where they given in (29). When they choose to attempt a coup, coup) given in (36), and since this value we have that V'^ {E \ > F*^ {E no coup) \ is ip^ = they 1, equal to V^'^ {TD) (see This imphes that in repression). oligarchy there exists a strategy giving the soldiers a strictly higher equilibrium payoff than no coup and non-repression. Equilibrium Dynamics 3.6 We now combine the elite's decisions in oligarch)', Two and from the previous four subsections, complete the analysis of the results and present a full characterization of MPE. observations facilitate this analysis. First, a direct comparison of V^ {E, N) in (33) V^ {E, S) shows that is always greater, so the transition to down V" when such a solution, n, exists. ^'As noted above, V'^ e political elite prefer transition to repression (with prevention) can be determined in a simple manner. For this purpose, let us define the threshold if fr to allowing democracy versus building a strong military and dealing with the moral hazard problem by paying the required efficiency wage. Second, whether the smooth in (32) never choose elite will the non-prevention strategy. Thus the choice of the ehte in oligarchy boils smooth V^ {E, S) (0, 1), then V" {E,P {E, S) \ tx) = V" is n G {E,S) When [0, 1] = V" {E,P\7r = a solution exists, strictly decreasing in {E, P \ it) as the unique solution whenever tt 26 = tt, tt. it is whereas between and 1 (43) TT) uniquely defined."^ V^ to When {E,S) does not depend on such a n. Thus, solution to (43) does not exist, then either or V" {E, S) < V" [E, P latter case, = ^ tt) for all described above has a unique MPE, ifn G [0,^), then whenever T — T^ and and suppose be defined as in (43) w= u)-^ , n set an army for repression the elite build and prevent military while oligarchy persists with probability AdPE 7 £ (i.e., The military chooses coups. 2. if TT G (tt,!], p [E = 1—7 = and q (M) the elite do not build = TD, — racy obtains with probability q{D) E) \ = and 1, = — tp (35) on S and P such that we have when tt V" {E, P < it). tt) \ Moreover, V^ {E,P\tt ^1) < V" {E,S). second condition is equivalent to the subgame obtaining in state with T^-^ given in Proposition i.e., if T^ > Remark /3r-°, 1 When Proposition 4 interior, i.e., then is it it ^ V" r^ > s it \ (D) q elite The /? first = not interior relevant. — i.e., tt = The necessary and or equivalent to is is any \ tt if > = 1 V" 0, {E,S) defined in tt) {E, G (0,1), are specified in the proof of Proposition conditions are satisfied when A^ is P \ sufficiently small). 27 4. G there will be > V" r^ < is {E, S) /3t^. The no coup tt ^ 0) < V" which implies that —only is 0) tt in t^ > P [r^ — t^^) tt {E, S), = 0. one of the two parts of guarantee that sufficient conditions that the set of parameters that satisfies both of these conditions is, (when there equivalent to tt = \tt 1. does not depend (that tt V" {E,P if can also be verified that for tt, when 1 V^ between V^ {E,P ^ tt V^ (D)] /A^ this condition V" {E, P\tt) <V" {E, S) is only condition [V^ (TD) — = TD, 3), It (0, 1) if 1 never choose non- V^ {E, S) and tt, {E, S) whenever G — {TD E) — Therefore, with the conventions adopted in the definition of (43), there exists repression only and and p run consolidated democ- (i.e., Next, as noted above (recall also footnote 21), tt)). 1), set with allocations as described in Proposition 1 a continuous and strictly decreasing function of is n. [0, 1] \ — a Transition to consolidated 0). in the long prevention, so their choice depends on the comparison of V^ {E,P game (7, 1]; a (i.e., we have Proof. The argument preceding the proposition establishes that the and MPE. Proposition 3 characterizes the so that 7 when 7 G an army democracy occurs with probability p {D = l—n. E) \ of starting in any subgame s and q {D) [7, 7] [0, 1] in the Then, the political y^ n. (no coup and repression). Transitional democracy arises with probability p unique and == tt e tt characterization of full that for all tt) | where =E s we In the former case, [0, 1]. These observations enable us to provide a 1. Proposition 4 Let n 1. G tt I V" {E, S) > V^ {E, P we have It can be nonempty (for it is verified that example, both Note also that the requirement When proposition simpler. ir — tt 7^ tt only introduced to is make the statement of the the elite have two best responses, thus the equilibrium it, is not unique, but either part of Proposition 4 could apply in this case. Proposition 4 shows that the suiEciently effective, - probability 1 tt i.e., < tt choose to set up an army only elite will When tt. this happens, it aligns itself with the elite. the society becomes a transitional democracy. military. In contrast, find profitable to set In this case there is has to It either by the it When 71" is with repression (probability fails Proposition 3 then implies that, in this when make significant concessions to the military or risk a repression not very effective, is i.e., tt up an army and pay the high wages necessary > tt, the elite coup do not to co-opt the soldiers. a smooth transition to a consolidated democracy. Another important observation follows from the comparison that will persist, recurring is government pays the cost of the oligarchy having established a powerful case, the democratic military previously. repression each period. Moreover, as long as oligarchy persists, the military receives in high (efficiency) wages so that tt), oHgarchy when increasing in This implies that repression 9. is more of (32) likely in and (35), which shows a more unequal society (proof omitted): Corollary 3 Higher inequality (higher 9) increases tt and makes repression in oligarchy more likely. Now putting the results of Proposition 3 together with those of Corollaries rize the effects of inequality 2 on equilibrium political dynamics and economic 1-3, we summa- choices. Corollary shows that greater inequality makes coups against transitional democracy more likely, so that (conditional on a democratic regime emerging) greater inequality makes military dictatorships more likely. Moreover, Corollary 3 shows that greater inequality also makes democracy more costly for the elite and thus encourages repression. Consequently, greater inequality has three related effects on political dynamics. First, powerful military more likely. more likely makes repression and the formation Second, by encouraging repression, of oligarchic regimes (rather than a dictatorships it smooth both because early stages of a democratic regime and it it transition to democracy) of a large and leads to a longer duration . Third, it makes military leads to the presence of a powerful military in the also because it makes it more difficult for democracy to convince the military not to undertake a coup. In addition, however, inequality also affects fiscal policies. As noted in Corollary 1, once a fully consolidated democracy emerges, greater inequahty leads to higher taxes and greater public good expenditures. 28 The combination of the effect of inequality on redistribution in consolidated democracy and dynamics implies a novel pattern sition of spending. When for the relationship inequality is its effects on regime between inequality, taxes and the compo- low, a consolidated democracy is likely to emerge, and thus a small increase in inequality starting from a low base leads to higher taxes and greater When public good provision. In such an environment, taxes are lower and likely. there is no public good provision). Therefore, the provision all tax revenue all is spent on the military (and effect of inequality on taxes and public good likely, it also increases the likelihood of a regime that will set high taxes and spend the proceeds on the military (again with no public good provision). might account for the lack of a across countries (e.g., Perotti, Subgame The analysis so far has focused monotonic relationship between inequality and redistribution Perfect Equilibria on Markov Perfect Equilibria, which make only on the payoff-relevant states. MPE is commitment problem that Nevertheless, useful to is in subgame know whether is perfect equilibria might prevent lead to significantly different results. In this section, some we between game up players in the to time t and stage k within the stage game. game can be represented by a mapping a H^'^ : the military Pi £ level of the pubhc good Gt 6 G A strategy profile a* {0, 1}, is a The next proposition shows that in {0, 1}. (i.e., if it is SPE if it is may /z*"'^ is MPE Subgame » [0,1], and Perfect the history of play A strategy profile for — [0, 1] x R^ x {0, 1} 1K+, the decision of and the coup and repression decisions at manner. social groups that briefly discuss the the range of the strategy profiles again refers to the tax rate Tt S Wt £ IR+, the captures the of the inefficiencies of Equilibria (SPE) of the above-described political game. Recall that of the it at the heart of our theory in a simple implicit promises depend strategies a natural equilibrium concept since political-economic it These implications 1996, Benabou, 2000). 4 be possible becomes more nonmonotonic. Furthermore, because greater inequality makes military dictator- is more ships inequality increases further, oligarchic repression all , the where the military wage whether to create or reform of the mihtary, 0^ G {0, 1} a best response to itself for all /i'''^ and G H*'^ sequentially rational). our political game the MPE and the SPE coincide, thus there was no loss of generality in focusing on Markovian equilibria. Proposition 5 Suppose identical to the MPE that n ^ ft. The political described in Proposition 4- 29 game described above has a unique SPE Proof. First consider some history St = M. Since military rule absorbing, is will involve the military maximizing a similar history where St = ^*'° (i.e., at D. their Now democracy t) where absorbing and the same argument implies is Next consider a similar history with 1. at thus Proposition 2 applies. Next consider utility, that the unique sequentially rational play after this history Proposition game clear that the unique sequentially rational play it is own the beginning of the stage st = TD. is identical to that described in Since there are no strategies that can be used to punish former soldiers in democracy and the continuation play after a successful down coup is after any history involving and this behavior, the Now and St uniquely, this implies that the unique sequentially rational play — in particular, soldiers will Given 3. already pinned TD TD will be the same as that characterized undertake a coup any history /i*'° is if the no-coup constraint in this case, (38), response for soldiers to undertake a coup. (12), since the regime will be S( = If it fails, D, and if it is is violated. provided by Proposition involving either of uniquely pinned down, the behavior after histories where In particular, st = E not satisfied, it is is st — M, D also unique. a unique best they receive exactly the same value as in succeeds, they receive the unique continuation value associated with subgames that start with is the no-coup constraint, (23) unique best respoiase of low-skill agents since sequentially rational play after is if in subsection 3.4, S( = A'l. Finally, the value to the elite from S which implies that the results of Proposition 4 hold as the unique subgame also unchanged, perfect equilibrium in this case. 5 Extensions In this section we provide a number of extensions that are useful to map our theory of military dictatorships to the data. First, recall that in Proposition 4 military dictatorships only emerge during transitions to democracy. In practice, there are many cases of military dictatorships that result from coups against oligarchic regimes, such as those discussed in footnote 6 in the Introduction. Our extension shows first how a slight modification of the baseline framework leads to coups both against democratic and oligarchic regimes. in the next subsection. number abundant about the in the is how and we would like to role of the military in politics economy. This is studied natural resource abundance affects regime djaiamics. empirically relevant because societies is Second, we introduce rents from natural resources and provide a of comparative statics about This extension This possibility many military dictatorships emerge in resource- understand whether the model has clear predictions when natural resources play a more important role investigated in subsection 5.2. 30 Third, we show how incorporating The key the national defense role of the military affects our results. when army has a national defense the role, becomes it easier for findings here democracy to commit reforming the military, making transition to democracy easier and more briefly illustrate that stronger oligarchic regimes make democracy more Throughout the on MPE and all T r = < exists C (r) = > f T and that taxation above f restriction choose to not Finally, less repression and > is for the tax distortion function C. In particular, same C [t) — as f defined in (6)) such that Q for This implies that there are no costs of taxation until t. prohibitively costly. Furthermore, we assume the following on the fundamental parameters of the model Y>{l + x)A^. We we simplify the notation and the analysis by focusing (not the t 1 for all we form also adopting a simpler f and may sometimes likely. that likely. rest of this section we assume that there is (44) maintain these assumptions throughout without stating them Faced with the new cost schedule for taxation, the elite still prefer zero taxes (except for paying the military wages when they have and military regimes would explicitly. moreover, by condition (44), both democratic to); f set taxes equal to simpler forms. For example, the value to the democracy, and in transitional Consequently, the value functions . elite in now take military dictatorship, in consohdated democracy are y^(M) = y^(D)= ^^^^y (45) , and + P)i^-r)A" rHr^r.^_ii^-l3)^t '-— [TD) = ^^ V" \_p ' The military wage and the instantaneous payoffs (46) to low-skill producers in consolidated democ- racy are also modified similarly and become ^K4 ^ "^{Y- ^-4'') ^^^d ^i X Also throughout these extensions. Assumption the participation constraint of soldiers next subsection, which 5.1 (1 _ ^) ^L + 1 is in effect and fY. is (47) sufficient to ensure that always satisfied (except with non-prevention in the discussed separately). Coups Against Oligarchy We now to a is is ^ modify the basic setup and continue to assume that the absence of a military leads smooth transition to democracy. When instead the oligarchic regime chooses to create a 31 } two additional states of the world, denoted by the variable military, there are state T], the elite are insulated from democracy. In contrast, in state and the probability 77^-'^, whose stochastic process, with Pr of and 77j = [77^ — T]^^] additional state variable elite particular, in chooses As = Tpt- MPE any and = Pf- 77^ is is, by V^'^ {E and 77^-^ and the military 77j = 77^-^ \ subgame with This constraint repression) repression) \ is repress. st is assume that the = E and binding only > V^' {E This = 77^ when repression) before they V"'^ it is is still know {E,P) coup), where V^'^ when the is 77''^-'^, (48) = 77^ 77^ = 77^'-^ \ coup) state variable because in state it is still is still 77^ and the mihtary repress 77^''^ given by (29) and V^"^ [E, P) 77^. {E the appropriate if This value = w^ + ^{{1-^) V^ V- jB (49), , (' defined in takes the value {E, P) is + p, is (1 - is ^) V'"' {E, P)] ^ g'^ d - m)) ^;- I defined recursively as [ttV^^ 32 + (48) the value of soldiers under prevention + {TD) {I - n) V^^ {E, P)] we obtain repression) given by with probability 7 as in the baseline ^ w^ + /S [nV^ (TD) + the realization of and In the military never therefore given by Combining can leave soldiers a consequence of the fact that coups are only possible this event, they succeed I V^ {TD) \ the value of a soldier and conditional on V^^ {E and also before the realization 77^. model. The value of a typical soldier when where We 77 j. t 0. satisfied. V^ [E (36) when and independently distributed over time the elite are insulated from coups and transitions to democracy and (38), that is evolves according to an exogenous applies without any change in this modified model. 1 starting in a — 77^ model, coups against oligarchy can be prevented only in the baseline no-coup constraint rj^ Lemma can be verified that does not, transition to democracy have to decide military wage at time the military only before the realization of It r]^^}. In depends on the repression decision of the military wage cannot be conditioned on this if it realizations are identically The IJ- , both types of transitions away from oligarchy can occur. of this event The as in the baseline model. {ri^ both the threat of a coup and from possible transitions to In particular, the military can attempt a coup and, possible e rj^ /»> d- ff d - ri) V" (TD) Prom (50), and the no-coup constraint (36) prevention has the same expression of tlie (38),^^ the mihtary wage consistent with coup wage efficiency w^ necessary to prevent coups against ohgarchy as in the basehne model, (39), with the only difference that, because of the change in the fiscal technology, the expressions for w^^ and a^ are now given by (47). With these changes, the tax rate for the oligarchy to be able to finance the military wages necessary to prevent coups becomes:^^ r- The = a'^x /37f + ^(l-7)^7—IT- (51) net present discounted value of the elite from prevention, starting with — st E, is recursively defined as V" {E, P) - (1 - r^) A" + p [(1 - V" {E, P) + iiT:V" (TD)] /i^) , and can then be rewritten as " where V^ {TD) The (^' '') = (^^) i-g(i-% given by (46). is net present discounted value of the elite from non-prevention, starting with st = E, is then given by V" {E, N) = a" + (3 The first term in this expression, V" {E, N) + fiV^ (coup)] [(1 - fi) {1 — fxcp) a^ = A^ , is the expected flow pa3'off to the ehte, which takes into account that with non-prevention there The thus income disruption. can take place only when ~ rj^'- In addition, . denotes the expected future value to the V^ (coup) = from (1 political — f) will be coups when possible and probability that a coup will take place 77^ A^ / (1 — /3). elite in V^ (coup) case there is = (1 a coup. In addition, with probability change today and the same . (1 political state recurs — — is 7) fi, V^ {D) + 7^^ (A/) Prom /i), because a coup (45), the we have that elite are tomorrow, i.e. insulated St+i = E. Therefore, The following conditions on the set of parameters are useful for the characterization of the equilibrium of the model. V" = = - 7) V^ {D) (see footnote 18 for details). V'^ {E, P) > V^ (E, S) and it is alwaj's satisfied. This follows by combining the no-coup constraint, (38), V'^^ (E repression) > V^'^ {E coup), which implies w'' >(1- P) V"' {E coup), and Assumption 1, which imphes that V'^' {TD) > V^ {D). "Recall that {TD) V"^ {E | coup) /S^V'^ (M) -f-/3(l "''The participation constraint of soldiers under prevention is \ I 33 | . Condition 2 < fi = JI /3f / Condition 2 ensures V" is, where ¥^^{£,3) (45). If this condition did not hold, the elite be indifferent between them when are independent of starting in s = Therefore, vr. /j. = Jl). defined in (32), with is would prefer S to V" [D) now given by N for any value of n This follows since both when Condition would be identical to that .B non-prevention to smooth transition, that tliat tlie elite strictly prefer V" {E,S), {E,N) > + /3f) {(f) V^ 2 fails to hold, the {E,N) and MPE in (or would V^ {E,S) any subgame Proposition 4 in the previous section and would in not feature coups against oligarchy.^"* The participation constraint of soldiers under non-prevention is V^''{E,N)>V^[E,S), where V variable N) {E, -q^. the value of soldiers under non-prevention before the realization of the is This value is defined recursively as V^ {E, N) = where V'^^ {E coup) \ is still il- (55) and V^ fj.) {E, {E, S) N) + /iV^ [E coup) I = A^ + j3V^ [D) V''{E\coup). in (54), satisfied is equal to it is ^_^''^^_^.^ under non-prevention straint of soldiers PV'^ defined in (36), and V''{E,N)^ Using (54) we obtain that the and only if (55) if participation con- the following condition is satisfied. Condition 3 H> If this is this condition is - (1 /3) A^ + means that this strategy always satisfied when P is is /3a-^ not feasible for the high enough, and parameters where Conditions 2 and 3 both hold set of Let us next define 7f = 1-/3 condition does not hold, the participation constraint of soldiers under non-prevention violated and this let H tt € [0, 1] in a similar fashion to is tt it elite. Moreover, notice that can be easily verified that the not empty. in the previous section. In particular, be the solution to the equation V" {E,P\Tr^7t) = V" {E,N), "Notice that ^ < is 1 and also that, except the simplification in the fiscal technology, the baseline a special case of this extended model with ^ = 1. 34 (56) framework when such a By solution exists, with the value of prevention for the elite defined as in (52). V^ the same argument as in the previous section, V^ and N) {E, independent of is V^ {E, P uniquely defined and any solution it G ^ V^ n) \ then we set [0, 1], {V" {E,P\tt=1)>V" when a n, so that = tt N) {E, for {E, any MPE in transitional democracy where y^ ]I, JL model presented the extended defined in Condition is 2. U Conditions 2 MPE as follows: above, then there exists a unique n G If set T then whenever [0,7f), = T^ and w = w^ = and p p(TD If IT is [E, N)) or = tt 1 given by Proposition 3 is still features = E, the elite build w^ and a^ defined = [7,7], arid q E (tt, 1], (D) — q (M) = /7-. where n tt, defined is = a (i.e., 1), = ip p {E E) \ TD, = — jm. l Proposition = so that q (D) when 1 1]. army for repression the elite build an The military chooses dictatorship arises with probability 1 7^ is Transitional democracy arises with probability MPE starting in any subgame s — = 1—7 and q{M) — 7 when 7 G (7, and do not prevent coups. probability tt that MPE The military chooses while oligarchy persists with probability fin, and suppose an army for repression and prevent military coups. (no coup and repression). 1 E) , s and 3 hold and 3 characterizes the unique 2. < V^ 0) it (56) does not have in this subsection I 7G When Condition 2' or 3 does not hold, the If identical to that in Proposition 4- 1. to (56) exists, n Using these observations, we obtain: Proposition 6 Consider fi (0, 1) . = now (again with the only difference that the threshold 7 in (25) (47)). tt | 7f strictly decreasing in {E, N)). Finally, notice also that the by n G ^ tt is vr) | solution V" {E, P (when P p{D E) — \ /i(l — p (Af E) | = ip = r = and (i.e., a — (coup) in state fi^' 1 1), set = 0, Military consolidated democracy arises with ji-y, while oligarchy persists with probability 7), . u; Consequently, the long-run likelihood of regimes are given by q [D) — p{E E) = \ (1 — 7) and tt 7^ tt 7. Proof. See the Appendix. Remark 2 The requirement that n in Proposition 4 above. so that the equilibrium Proposition 6. When n is it n not unique, but V^ emerges when tt an identical plaj's equal to is Also, the case where Propositions 6 since ^ {E, = ff then the tt, nature its A'' = | tt 0, 35 = 0) which is role to the elite will have two best responses, unchanged from that described = V^ is assumption that {E, P | tt = 0) is in not covered in ruled out by the restriction tt 7^ tt. Finally, the requirement that ^ ^ rules out the case /i where the elite obtain the same value from non-prevention and smooth transition. The important additional result in Proposition 6 arise along the equilibrium path. is that now coups against oligarchy also Previously, the threat of coups against oligarchy affected the equilibrium allocations, but such coups never took place in equilibrium (and military dictatorships always emerged from coups against democracy). is The introduction of such coups irnportant for two reasons. First, coups against oligarchy provide a clearer demonstration of the political moral hazard problem facing the elite in army building a strong can turn against them. Second, the model can now potentially account dictatorships we observe in practice for — that the army why the military have different origins; some resulting from coups against democracy, while others are preceded by non-military oligarchy regimes. Natural Resources 5.2 In this section, we extend the baseline which generates income equal to elite i? > (high-productivity agents) and Since there are now two model by assuming that there in all is a natural resource, each period. The natural resources are owned by the of the natural resource rents initially accrue to them. sources of income, we allow for two different fisca,l instruments; a tax rate on income at the rate r (with tax distortions specified as in the beginning of this section with the function C, so that there are no distortions until r natural resources C £ distortions. This is [O;!]- We = f) and a tax rate on income from assume that the taxation of natural resources generates no reasonable, in view of the fact that natural resources are typically supplied inelastically. The N characterization of the since it will is similar to before (in particular, the elite never choose gives a smaller equilibrium payoff than S), main observation that regime MPE choose ^ simplifies the analysis = 1, thus taxing all is and we provide fewer details. The that both a military regime and a democratic income from natural resources. Consequently, an analysis identical to that in Section 3 implies that military wages in a military dictatorship and instantaneous payoffs to low-skill agents in democracy are X and a^ = {l-f)A^ + G^, 36 (58) where G^ = fY + R. These expressions exploit the fact that r = f and C = 1 1 and thus they incorporate the additional revenues coming from natural resource rents either directly in the wage military or in the amount good provided of public Let us start with a subgame where s = TD. An in democracy. analysis identical to that leading to Proposition 3 immediately implies that low-skill producers are always better-off by preventing coups. In fact, now retaining political power has tional source of revenues military coups more of the addi- coming from natural resources. However, natural resources also make attractive for the soldiers. In particular, as in Section 3 a military coup can be prevented only if given the instruments and natural resource rents, democracy fiscal can raise enough revenues to pay soldiers a This become more valuable because feasibility constraint now sufficient amount to satisfy the no- coup constraint. takes the form uF^x <t{Y - xA^) + R. (59) In this expression, the revenues include the income raised by taxing production at rate f and from the taxes on natural resource rents prevent coups is is at the rate C = 1- The military wage necessary to no longer equal to w'^^ (defined as the wage satisfying (23) as equality), but instead ^^^ = with w'^ and a'" now ^4^7 {w'' defined by (57) and (58). making coup prevention more that the prevention of coups difficult. is now It is -a^), (60) straightforward to verify that w'^^ > w^^, In particular, the combination of (59) and (60) shows possible if 1-/3 u)^^ 7 satisfies the following condition, which is a simple generalization of (25), With this threshold replacing 7, Proposition 3 continues to apply.^^ Appendix shows that the threshold 7 7 in the itive; (i?) is strictly decreasing in R. greater natural resource rents raise the political stakes and attractive for soldiers (and the additional revenue available to compensate democracy is less likely to make of Proposition This is intu- military coups more democracy soldiers for the prospect of dividing natural resources sequently, transitional The proof among is not sufficient to themselves). Con- consolidate in natural resource abundant societies. can again be verified that the participation constraint of soldiers is satisfied in this case as well. In particdemocracy with natural resources is V'^' (TD) > V^ (D), where V'^' (TD) = V'^ [TD coup) = /JtV"^ (M) -h /3 (1 - 7) y^ (D), V^ [D) = a^ / (1 - /3) and V*' (A/) = w"' / (1 - 0), and holds under Assumption 1. ^'It ular, this constraint in transitional I 37 . We stakes, next turn to subgames starting with s though in this case their effects will = E. Natural resources will again raise political be somewhat more complex. Recalling that the value to the elite in the subgames starting with obtain the value to the elite s is M and s =D are given by (45) above, + ^/»)+g(l-t)X''^ (62) from natural identical to (32) in Section 3 above, except that the elite enjoy the rents resources for one period. we from smooth transition as V-'(i;.5)='^-'^".4'' which = on the other hand, they build a strong army and choose coup If, prevention, their value, as a function of n, will be given by the solution to the following program V"iE,P\.)= ^ ' ^ ' i^ max ' ^) A" + H - OR/n + ^nV- jTD) ' 1-/3(1-^) fe[o,f],ce[o,i] ^ ^ subject to the government budget constraint, which written as an equality takes the form - xA^) +CR = w^ f (y Expression (63) is {R) X. (64) obtained from the solution of a recursive equation analogous to (34), but now uses (46) and incorporates the fact that to prevent coups the elite have to pay the efficiency wage w^ (R). This efficiency wage has the same expression as in w'^ and a^ are now replaced by their counterparts w and a^ w^'^ in (57) and higher except that (39), (58), so that + (l-7)5^). (i?)=/3(7Ti'^ (65) Let us also denote the tax levels that solve the maximization problem in (63) by f ^ and Comparing the maximized value it in Proposition 4. The of this program to (62), we obtain a threshold tt . {R) replacing Consequently, rest of this proposition continues to apply.^^ (^ we have the following characterization of equilibrium in the presence of natural resource rents. Proposition 7 The extended model with natural resources has that in Proposition 4 (except that • a higher level of • there exists R R* > 7 (i?) replaces 7 and tt makes democratic consolidation and x > sion, starting in state s — 0, replaces (i?) K^ {E,S) = A^+ {E \ repression) PV'' {D). Assumption V'^ [E 1 identical to Moreover: = TD less likely; such that an increase in natural resources makes repres- E, more = it). starting in state s likely if x > x, for any "''The participation constraint of soldiers in oligarchy under prevention V'- {E, S), where V"' MPE, a unique \ coup), V" {E \ coup) is initial now is R* given by V''' = ^7^^^ (M) + again ensures that this constraint 38 R> always /3 (1 {E - satisfied. \ repression) 7) V^ > {D), and Proof. See the Appendix. The new results in this proposition are the comparative statics As already discussed above, of natural resource rents. more to gain size greater natural resource abundance and makes democratic consolidation more increases the political stakes military has with respect to the from taking control and democracy is difficult, because the not always able to compensate soldiers for forgoing these returns. The second part of the proposition shows that greater natural resource abundance also affects the likelihood of repression in oligarchy. On biguous, because of two opposing effects. Nevertheless, the effect is, in general, the one hand, greater natural resources am- make the elite more willing to use repression in order to prevent a transition to democracy again because of the greater political stakes. political On the other hand, natural resources also intensify the moral hazard problem because the military, once formed, to undertake a coup. Which With a natural resource rents. and thus when the effect size of the in natural resource rents dominates depends on the will size of the The size of larger army, per soldier rents in military dictatorship are lower army is larger than a threshold f makes repression more baseline model analyzed in Sections 2 and more importantly, the only likely. be reformed after one period. objective is and 3 > and R> R*, an increase ^^ was simplified by two assumptions; role of the military secondly, the transitional phase in democracy We now was repression politics. We will also is the army = 0, is is is is, xt = x, the invasion would succeed and in Section 2, if To capture the possible. Our substantive role of the military except that we are (e.g., such an invasion all still using the and more substantively, a threat of invasion from another country). is not possible. If, on the other hand, agents (including low-skill producers) receive a lower payoff than even in the military dictatorship. This implies that an invasion and always good for democratic consolidation. model an international threat present, that politics; use this extension to investigate the claim by O'Donnell identical to our baseline there firstly the military could i.e., and defense simplified fiscal technology introduced at the beginning of this section, we assume that domestic relax both of these assumptions. and Schmitter (1986) that a weaker military The model in lasted for one period, to investigate the impact of international threats on democratic Xt army and the National Defense ctnd Democratic Consolidation 5.3 If have stronger incentives all agents would like to prevent role of international threats in the simplest possible way, ^^Here the threshold R' provided in the Appendix, is such that when R on production income and natural resource rents to finance military wages. , 39 > R' , the ehte will use both taxes ' we assume that the and when power posing the threat of invasion may reform foreign happens, the international threat disappears. this We its own military do not model the behavior of the foreign power explicitly, and instead assume that such reform happens with probability A G (0,1) in each period. Until the foreign threat disappears (due to reform of the foreign military), a transitional democratic government cannot reform the military. After this threat disappears, the transitional democratic government will choose to disband the military (since there is no longer any foreign threat). In this light, the current model is also a generalization of the baseline setup in Section 2, since reforming the military takes potentially longer than one period. Finally, returning to the discussion in footnote 3 and to O'Donnell and Schmitter's (1986) conjecture, A can be interpreted as another (inverse) measure of the relative strength of the army, since when A is low, there purposes and the military The whether the foreign threat s = TD no foreign threat is is in this case active is still is very similar to that in Section is now an 2, except that additional state variable. Let us start in a and with the foreign threat active (the analysis of the case where there identical to that in Section 2 from attempting a coup, attempting a coup greater need for the military because of national defense stronger. ^^ MPE analysis of the subgame with is is [TD V^"^ \ coup), is still and is omitted). The value to the military given by (21), whereas the return from not is y^^ [TD ^vF^ no coup) I where we again use iiF^ +(i [(1 - A) to denote the military V^^ [TD wage coup prevention. Note, however, that the expression \ no coup) in transitional for this wage is + XV^ [D)] , democracy when there is no longer the same as in Section 3 (see below). This value function also takes into account that the same state will recur with probability 1 — A (when the foreign threat remains active and there is no opportunity to reform the military). Rearranging this expression, we obtain V^^ [TD no I where a^ is now defined in (47). The coup) ^ ^ '^ ^ (1-/3)(1-/3(1-A))' expression for w'^'^ in this extended environment can be obtained by solving the incentive compatibility equation, V^ [TD \ coup) = V^'^ (TD | no coup), as is whether xt = or x, and we have already investigated the implications on democratic consolidation. 'Naturally, an even starker measure of this mesisure of military strength 40 w^ where As in is the soldiers' wage in a military dictatorship given by (47). our analysis in Section are satisfied: first, 3, transitional democracies will prevent coups The necessary condition < w^'^ Let us start with the second Using the expressions . ^^ /3(1-^(1-A))(^M_,L) Condition (67) -7(A). high. Moreover, — as A > it can be verified that 7 (A) is — > this condition coincides 1, V^ {TD no coup) = {I in transitional - t) I V^ {TD, P) = V^ [TD fact that taxes will /? [(1 - G^^ = A) with when difficult (25)). this is feasible. If is V^ [TD, P) + XV^ f (F - xA^) - [D)] (68) , ui^^x incorporates the be equal to f (given the tax distortion technology adopted at the beginning and whatever of this section), democracy A^ + G^^ + no coup) and \ not too a strictly decreasing function of A and that 7 (A) next verify that low-skill producers prefer to prevent coups they prevent coups, their value where is be successful will This implies the interesting result that condition (67) becomes more 1. to satisfy as A increases (in the limit as A We (67) generalization of condition (25) and shows that transitional democracies is can prevent coups as long as the probability that coup attempts 7 two wage for these the condition for the prevention of coups in transitional democracy can be written as levels, > be democracy to pay high enough wages for the transitional to the military again takes the form ViF^ — two conditions low-skill producers should prefer to prevent coups; second, they should able to pay high enough wages to the military to achieve this. requirement. if is left over from paying soldiers the efficiency wage goes into public good expenditures."^ Alternatively, without prevention, the value to low-skill producers is V^ [TD where O"^^ w'^^ there = is 0). coup) - (1 I = t — - f) (1 - <^) (Y — xA^), A^ + G^^ + (3 [7^^ (M) since in this case zero + (1 - 7) V^ [D)] wages are paid to , (69) soldiers (i.e., This expression also takes into account that, as before, when a coup attempt fails, [l cj)) an immediate transition to The comparison of V^ {TD \ fully consolidated no coup) and democracy. V^ {TD \ coup) shows that the low-skill pro- ducers are better-off preventing military coups by offering a wage to the army equal to w'^^ if and only if A > A*, where A* G [0, 1) and is defined in the proof of Proposition S in the Appendix. The next proposition states the main result of ^^The hypothesis that coup prevention ""it is is possible, that again straightforward to verify that Assumption satisfied. 41 is, 1 w^'' < this subsection. w'^'' , ^° ensures that G^'^ > 0. ensures that the mihtary participation constraint is Proposition 8 In any subgame beginning with MPE s — TD, there exists A* G [0, 1) such that the transitional democratic government prevents coup attempts 7 < 7(A) andX > and a unique if and only if A*. Proof. See the Appendix. This proposition has two interesting imphcations, both following from the decreasing in A (which implies that transitional democracy when A consolidate is low). The first is here corresponds to lower A, democratic consolidation defense role, more into it less is credible emphasized in this paper: is more when the army likely. is when that is is prevent coups and tliere is a which in national defense, This (A) 7 intimately related army has an important national threatened by reform in transitional democracy and this translates commitments by undertaking a coup. likely to and more important implication stronger foreign threat and thus a more important role of the to the key economic force more fact that In contrast, transitional when A democracy to compensate close to 1 is army foreign threats and no important role of the (meaning that soldiers for not tl:iere are no serious in national defense), concessions transitional democratic regime are not credible for the military because they foresee reform and are more willing to attempt coups. by the imminent This result therefore highlights a new and potentially important interaction between international and domestic politics. The second implication of Proposition 8 is that, somewhat gests that over a certain range, democratic consolidation is model sug- paradoxically, the may be more likely when the military stronger (since a lower A corresponds to a stronger military). Therefore, the logic of com- mitment emphasized by our approacli shows that O'Donnell and Schmitter's (1986) conjecture about stronger militaries always making the survival of transitional democracies need not be comparing may true. xt = Note that a version of to xt = x. this conjecture was true in less likely when our baseline model, However, our extended model here shows why this conjecture not capture the complete set of interactions between the strength of the military and transitional democracies because commitments to a strong military are more credible. Persistence in Oligarchy 5.4 Finally, we can also incorporate persistence in oligarchy into the baseline model. stantive pm-pose of this exercise is to see whether better organized democratic The movements conversely less strong oligarchies) are always likely to lead to more rapid democratization. main result in this section Suppose that in the shows wliy this may not be (or Our tlie case. absence of repression, oligarchy persists with probability a £ 42 sub- (0, 1). instead, there If, — (1 a) TT < repression and is Thus, in this extension TT. dation of the power of the and the military an elite democratization happens only with probability fails, it a represents a direct measure of the degree of consoli- The elite in oligarchy. in oligarchy are similar to our analysis in Section 3. ^^°) ' (l-^)(l-/3a) and the value from prevention, as a function of tt, is similar analysis again shows that non-prevention fact, be Briefly, the value to agent from smooth transition can be computed using the expressions in (46) as ^ ^^'^^- A analysis of the interactions between the elite is dominated by smooth transition (in the persistence of oligarchy without repression makes smooth transition more desirable Consequently, we immediately obtain the analog of for the elite relative to non-prevention). Proposition the 4: otherwise, and elite choose prevention n if is below some threshold straightforward to verify that it is in the next proposition, concerns the tt e fr and smooth transition The main new (0, 1). result, described comparative statics of regime transitions with respect to the persistence parameter a. Proposition 9 Let subgame beginning the TT be defined by if in state s subgame beginning in is strictly increasing in — TD, state s a V^ then = TD, for any (p > = V^ {E, S) tt is strictly n then (j)* , {E, P (j)* If fr). no coups take place in the decreasing in a. If coups take place in is strictly where \ G decreasing in a for any (f) < (p* , and (0, 1]. Proof. See the Appendix. Proposition 9 implies that the more consolidated in transitional elite are less likely to (greater a) under two related conditions: both the value to the opposing tt forces. the elite are {(p) they are more first, when their power is when coups do not happen democracy; second, when coups take place in transitional democracy but they do not cause too much income disruption the threshold choose repression elite a increases, from smooth transition and from prevention increase. Whether (</> small). This result increases (and thus whether repression Whether coups take place likely to to a lower value of a). forces. likely) depends on two democracy and what When choose repression when their power When intuitive; as becomes more in transitional determine the balance of these two is their costs for coups do not take place, then is less consolidated (corresponding coups are possible after democratization, then the trade-off for the 43 elite depends on how disruptive these coups then a lower a when coups are. When they are not very disruptive {(j) < cj}*), consolidation of elite power in oligarchy) encourages repression. However, (less are highly disruptive likely after repression and the {(p > then a lower 0*), smooth elite prefer a makes transitional democracy more transition in order to avoid the potential costs of coups in the future. The substantive implication of Proposition 9 regimes are not necessarily more likely to organized politically and the oligarchy is is emerge that, somewhat paradoxically, democratic where the in societies citizens are better weaker. In particular, a lower value of a in this model corresponds to a weaker oligarchy (and thus to a stronger democratic movement). Proposition 9 shows that smooth democracy may be more transition to corresponds to societies where the democratic movement sufficiently consohdated, so that The Role 6 In this section we they feel less is weak and the power need to create an army high, which of the elite is for additional repression. of the MiHtary in Recent History number briefly discuss a empirical context for our theory. We first of salient cases of military dictatorships to provide contrast the role played by the military in various how economic Central American nations. This highlights both namics and the is when a likely role that the military plays in pohtics. inequality affects regime dy- At the end of this section, we provide a brief discussion of the impact of natural resource rents on the behavior of the military. Democracy Military and 6.1 The in Central America case of Central America, and of Costa Rica in particular, provides an interesting illustra- tion of the ideas emphasized in explaining regime by our theory. This case both highlights the dynamics and also shows how the role of the military initial distribution of income affects the subsequent evolution of political institutions. Among role played 1997). Latin American countries, Costa Rica stands virtually alone for the very limited by the army During the 19'^^ in politics, since the country gained independence in 1821 (e.g., century, Costa Rica did not experience predatory caudillos typically influential in the political and economic Between 1891 and 1948, there was a single coup in life in much Yashar, (who were of the rest of Latin America). Costa Rica, followed by a brief dictatorship. After this episode, in 1949, the Costa Rican military was demobilized and essentially disbanded. The relative absence of external and internal conflicts during the history of a major factor in the limited role that the military has played in this country's 44 Costa Rica politics. is Even before its demobilization, the Costa Rican army was small and weak, and did not possess the professional traditions and strength typical of many other South American militaries. key factor that has led to the presence of a weak military in Costa Rica social conflicts in this country have been relatively Hmited. This of factors, including the structure of the economy consisting of is is A that historically because of a combinations many small landholders; the absence of a large indigenous population at the time of colonization, which prevented the Spaniards from setting up major plantations in this area; the poor endowment of natural resources such as gold; and the high agricultural wages resulting from the scarcity of labor due to the absence of a large (indigenous) population. Moreover, the relative homogeneity of Costa Rica's society appears to be an important reason why Costa Rican ruling shown limited interest in the creation of a large and powerful military. "^^ The decimated during a short civil war.'^^ have tradition of low militarization of Costa Rica culminated with the formal demobilization of the in 1949, after these were elites armed forces This decision has not been reversed ever since. Consistent with the predictions of our model, the demobilization of the Costa Rican army appears to have contributed significantly to the stability of its nascent democratic institutions. Consequently, Costa Rica has become the most stable democratic country in Latin America. This is because of the absence of the military necessary in part "coup threat," perhaps in coalition with the economic for the elites. ^^ formation of an effective Roquie (1987 p. 192), for example, relates the success of Costa Rican democracy directly to the absence of a military and writes: "No doubt because it corresponded to deeper motivations hnked to the social equilibrium of the Costa Rican nation, a nonmilitary state has become today one of the bases of its democratic consensus." The - successful democratic consolidation of Costa Rica since the mid-twentieth century '''For instance, describing the economic and institutional reforms promoted during the "liberal" era (the period of Latin American history roughly going from 1870 to 1930, following the phase of caudillo politics), Mahoney (2001 p. 266) writes that, "In Costa Rica, where the reform period was launched at the time of independence, liberals were not faced with the kinds of political threats that led reformers elsewhere to build large standing armies. ..The pattern of reformist Uberalism that had developed by the early twentieth century saw neither the creation of a powerful military coercive branch that commanded a prominent position in the state nor an associated rural economy marked by polarized class structures and a high potential for lower-class agrarian revolts." ^^The economic elite did oppose the abolition of the armed forces, but they were overruled. '''As emphasized in Bowman (2004), the abolition of the army in Costa Rica has helped to stabilize democratic institutions not only directly, but also indirectly, by allowing the government to spend more in welfare and public education. Indeed, the broad programs of income redistribution and of provision of goods and services implemented by the National Liberation Party (the preeminent party in Costa Rican politics in the second half of the 20"* century) have further alleviated income inequality over time, and thereby reduced the scope of social and political conflicts. 45 stands in contrast to the different political development paths pursued by other Central Amer- For example, highly repressive military dictatorships were established in the ican nations. 1950s both in Guatemala and in El Salvador. Honduras and Nicaragua, These regimes persisted instead, traditional oligarchic regimes led until the 1980s. by the civilian elite, In but supported by a significant military element, emerged during this time period. Consistent with our model, in both Guatemala and El Salvador power militaries, initially created by the societies, later elite in order to ensure the repression of the lower strata in these highly polarized became strong enough These mihtary regimes lasted for to seize power and establish their own dictatorships.^'^ extended periods of time in both countries and were highly authoritarian and autonomous. Their policies also led to the reproduction, or even the amplification, of the high income inequality present in these Honduras and Nicaragua have out the 20''' century. However, also societies at the time of the coups. ^^ been ruled by nondemocratic governments through- in contrast to the military dictatorships in El Guatemala, these were personalistic or oligarchic regimes, in which the military, Salvador and though actively involved in repression, did not have autonomous power. In fact, the survival of nondemocratic regimes in Honduras and Nicaragua depended more on the distribution of political patronage through the expansion of state bureaucracy rather than the harsh repression of opponents and "state terror," which have been part of the experience the same period. The different in Guatemala and in El Salvador over '^^ path of political development that Honduras and Nicaragua followed since 1930s (relative to Guatemala and El Salvador) may be traced back to the economic and insti- tutional evolution of the two countries since the previous period of liberal reforms. '"'in Guatemala, the liberal era started in 1871 Whereas with the triumph of General Justo Rufino Barrios, and was new major wave "Except for a brief democratic interlude in the early 1920s, Liberal military dictators ruled with an iron hand on behalf of the coffee ohgarchy and foreign investors for over seventy years," (Vanden and Prevost, 2002 pp. 255-256). ^^An indication of the differences in the underlying socioeconomic conditions is provided by the average number of economically active individuals in the agricultural sector relative to the number of farms around 1950 (see Needier, 1987, Table 5, p. 98). This ratio was 10.9 in Costa Rica, 38.2 in El Salvador and 48.1 in Guatemala. This comparison therefore suggests a much more equal distribution of income in the agricultural sector in Costa Rica than in El Salvador and Guatemala. It ^^ Overall, there was some variation in the military policies implemented in Honduras and Nicaragua. is possible that this variation might have contributed to the different timing of democratization in these two countries. In Honduras, the army was not a bastion of support of the Can'as regime, and militarization in Honduras remained very limited until the late 1950s. In contrast, in Nicaragua the Somoza regime promoted the development of the National Guard as an important pillar of its control over the society. Consequently, though a typical oligarchic regime, Nicaragua was significantly more militarized than Honduras. Nonetheless, the army in Nicaragua did not develop into an autonomous institution, but rather into "a separate military casta, loyal only to their own leader, not to the nation as a whole," (Millet, 1977 p. 198). characterized by a further increment of the concentration of land ownership (also through a of expropriation of communal lands cultivated by the indigenous population). 46 in Guatemala and in El Salvador, oificers tional capacity to attempt coups had both the and to setup structural basis for military authoritarianism saw a further entrenchment of the was liberal era left largely controlled Guard of the was own rule, in institu- Honduras and Nicaragua a Indeed, while the liberal period lacking. in Guatemala and in El Salvador, in behind a relatively weak agrarian by the ruling and the coffee elites, as well as the creation of powerful militaries capable of independent political action and Nicaragua the their political incentives elite or dictator (for Somoza Garcia regime in Nicaragua). class both Honduras and a military that example, as in the case of the National "^^ Natural Resources £ind Military Dictatorships 6.2 Our model's predictions on the role of the military in natural resource abundant societies is also broadly consistent with the historical evidence. In her authoritative work on natural resource booms, Karl (1997) argues that the two bated in political instability many oil price hikes in the early 1970s and 1980s exacer- and regime transitions toward more repressive forms of government "petro-states," in particular, in Iran, Nigeria, Venezuela and Algeria. Nigeria, example, experienced growing economic and ethnic tensions after the witnessed the reemergence of military rule in 1983. oil price increases for and This was followed by a new transition to a fragile democracy between 1986 and 1991, but was again interrupted by a military coup in 1993. Algeria experienced a severe tion of the president, Mohammed the early 1990s, which lead to the assassina- crisis in Boudiaf, to the cancellation of elections, and to frequent switches between military rule and weak civilian rule. dant endowments of other natural resources offered by the empirical evidence presented may in Further evidence that oil and abun- adversely affect democratic consolidation is Ross (2001), and by Jensen and Wantchekon (2004). In particular, Ross (2001) presents evidence consistent with the notion that increases in natural resource rents lead to increased repressive efforts by nondemocratic governments Jensen and Wantchekon (2004) show that in in order to prevent or delay democratization. African countries, the abundance of natural resources tends to and the consolidation of, democracy less likely. In fact, many make both the of the more transition to, successful examples of democratic transitions in sub-Saharan Africa are by relatively natural resource poor na^'' While, during the hberal period, the Zelaya autocracy had implemented economic and institutional reforms setting Nicaragua on a development path similar to that of in 1909, following a US Guatemala and El Salvador, things changed abruptly intervention which lead to the creation of an informal protectorate. After the fall of which took over power relying on the support of US Marines for security, showed little interest in preserving the coercive apparatus created by the previous regime. The Nicaraguan army was eventually disbanded, putting an end the militarization effort of Zelaya. Zelaya's regime, the conservative elite, 47 tions such as Benin, Madagascar and Mali. Instead, natural resource abundant countries have experienced greater political turmoil and have not been successful in establishing democratic Examples include Gabon, Cameroon, Togo, Zambia, regimes. Algeria, Nigeria, Congo and Sierra Leone. Botswana, which Naturally, there are exceptions to this pattern. abundant country, has been the most successful democracy boom was also significant for or nondemocratic in is a natural resource movements Botswana had one oil Norway, but does not seem to have led to any military action This in this country. likely reflects the fact that there is major interaction between the strength of underlying institutions and the resources. The sub-Saharan Africa. of the most participatory indigenous a effect of natural institutions in Africa and a very successful transition to democracy before diamonds were discovered (see Acemoglu, Johnson and Robinson, 2003). Norway such societies, greater natural resource is clearly a very strong, consolidated democracy. In wealth increases the income of the citizens, relaxes the government budget constraint, and may contribute to more productive spending. '^^ also consistent with the prediction of our model regarding the in consolidated democracy, which increases the level of public effect of This is a positive shock to R good provision without affecting political institutions. Concluding Remarks 7 In this paper, we presented the analysis of the emergence of military dictatorships and first the conditions under which the military will act as an effective agent of the eUte (as opposed to acting in its own interests for research in political and against those economy for of the elite). These questions are relevant a number of reasons. First, most nondemocratic regimes survive with significant support from the military, so understanding the objectives of the military is important in the study of political transitions. Second, practice are military regimes, and and persist for different reasons we need many nondemocracies to understand whether military dictatorships in emerge than oligarchic regimes and what their economic consequences are. An investigation of these questions necessitates a a set of individuals who themselves with the elite if this is act in their own interests model in which the military consists of (though they can be convinced to align consistent with their interests). We introduced this feature '^See Ross (2001 pp. 343-344) for more evidence that major oil discoveries have not had discernible adverse on democracy in such countries as Norway, Britain and the US. See also Mehlum, Moene and Torvik effects (2006) for evidence of a significant interaction between institutions and natural resources in growth regressions. 48 by assuming that the means if of violence in the society are in the monopoly and the elite decide to form a strong military, then they have to live with the political moral hazard problem that this causes. In particular, a strong military agent of the with their is of the military, elite, own but may them instead turn against objectives. One immediate "efficiency not simply work as an in order to create a regime more in hne implication of the political moral hazard problem that the cost of using repression in nondemocratic regimes need to pay may is now higher, because the elite wages" to soldiers to prevent coups (or make other social or policy concessions to the military). An to important consequence of the presence of a strong military democracy takes regime until it is place, the military poses a is that once transition coup threat against the nascent democratic reformed. Precisely, the anticipation that the military will be reformed in the future acts as an additional motivation for the military to undertake coups against democratic governments. Consequently, societies where the elite form a strong military prevent in order to democratization are more likely to later lapse into military dictatorships because the military some retains of its power during transitional democracy and can attempt a successful coup against democracy. This leads to a specific (and to the best of our knowledge, novel) channel for the emergence of military dictatorships, which appears to be consistent with the evidence. effects It also on the economic and Our analysis also mocratic how highlights repression during a nondemocratic era can have important political success of a later democratic regime. showed how, under certain circumstances, military coups against nonde- elites are also possible, thus creating another channel for the emergence of military dictatorships. In light of these results, one types of nondemocratic regimes. The the militar}' acts as an agent of the might wish to distinguish between three first is oligarchies where the rich elite are in incentives of the elite. The second dictatorship that emerges as a result of a coup against a democratic regime. from Central America discussed The third The examples likely. power and is om a military The examples in Section 6 ensure the relevance of this type of military dicta- a military dictatorship that results from coups against oligarchic regimes. of this type of military dictatorship were also discussed in the Introduction. Our model more is different This type of regime emerges endogenously in elite. model depending on the technology and the torships. historical also provides a range of In particular, we show that greater inequality makes the use nondemocratic regimes more military coups. Both comparative statics about when such dictatorships are likely and of these effects also make makes it more difficult for of the military in democracies to prevent military regimes following brief democratic episodes 49 more In addition, greater inequality also makes likely. it more nondemocratic likely that regimes are unable to solve the political moral hazard problem and thus creates another channel for the make We also show that greater natural resource military coups against (unconsohdated) democracies effects elite, emergence of military dictatorships. on the political equilibrium in more likely problem resulting from the high natural resource making the military necessary is more in our model; when there is rents. likely for national defense. international politics and domestic politics. economic force and have ambiguous nondemocracies, which become more valuable for the but also more expensive to maintain because of the more severe implies that democratic consolidation rents The when This is political More importantly, there a is the model also a potential foreign threat, new and logic of the result moral hazard is interesting link between very related to the main an international threat, concessions from democratic regimes to the military become more credible, because democracy also needs the military. We view our paper as a first step in the study of military dictatorships and the political agency problems that are ubiquitous between branches of the state that control the means of violence and the economic area deserve further study. elite, First, especially in nondemocracies. Many topics in this broad a systematic empirical analysis of policy and economic performance differences between different types of nondemocratic regimes is necessary. Second, the current framework can be extended so that an alliance between the military and the can be formed during democratic periods as Finally, the current well. elite framework already highlighted the important interactions between international and domestic politics. However, we did not endogenize the political economy equilibrium in other countries. A fruitful area for future research appears to be the international relations aspects of the interactions between the military and democratic regimes. In particular, it would be interesting to investigate how military or democratic reforms in one country affect politics in other countries. 50 Appendix Proof of Proposition 3 7.1 By Assumption 1, 7 > and there will when 7 e [7,7], (28) When 7. If 7, then the feasibihty constraint, (24), be a coup attempt in equilibrium G^^ = 1). Thus we only have democracy chooses prevention, then which can be determined , st+i — D to show that and thus the Thus the only relevant w'^'^ that satisfies (23) as equality. choice of t"^^ and (V' necessarily violated is satisfied. is transitional must be paid 7 > soldiers decisions concern the as the solution to the following maximization program u-^ (rD = no coup) 1^ + G (1-t)A^ max I (72) re[o,i],GeR+ subject to G< Next, using the expressions (16) and (23), ' {t P w TP =3-:^7 - C (r)) (r - xA^) - w^^x. (r - C by the first-order condition of the A^ and moreover the utility of low-skill (1 If, government budget constraint ^j.p If, - utility of low-skill taxes are set at the level t'^^ defined program, which implies C (r)) {Y - xA^) (74) , producers during the transitional period (r^^)) {Y - xA^) - G^^ — (72) involves _ W =.. X 0, will be w^^'x. then taxes are determined by the ^ (JTP\ ^+C{r^^), Y -xA^ ^ producers in the transitional period will be instead, coups are not prevented, the tax rate chosen to maximize the 0, (73) (8) as T'- and the (1 G^^ > - r^^) A^ + (r^^ - C on the other hand, the solution to {Y - xA^) {t)) a^ Provided that the solution to (72) involves implicitly can be written as lu'^^ utility of and the (75) (l — level of public r-^^) A^. good provision are a representative low-skill producer in the transitional period, with no additional constraint, but taking into account the output disruption caused by the 51 coup, which will be forthcoming in this case. In particular, the tax good provision G'^^ in question, are the solution to the u^ {TD coup) = max I (1 ^ The low-skill where - /3 [jV^ {M) we can no coup) \ - -f (1 7) V^ ' - > /3) [(1 = u^ {TD {TD A^ + G no coup) \ - T^^) A^ + (r^^ coup) - (3-)' C when G^^ > and that depend on t-^^ does not {TD V'^ - coup), | u^ {TD \ (a^ - - (1 r) A^) (1 /? when G^^ > 0, - /3) w^^x (77) . — {TD /3) w'^^ is linear satisfied at /3 — /3, f3' , such that the left- no P' ^ C {f))Y < [0, 1] condition P e [0, 1]. is is sides are equal. Second, (77) at is since in this case this condition can be written as 1, = in this case w"^^ /3, imply that both the right-hand-side and and the right-hand {t-C (f)) Y-{f-C (f)) xA^ where (f — and coup) does not depend on \ the left-hand-side of (77) are strictly decreasing hnear functions of p. Therefore, there most one value of coup)-|- 0, as - xA^)] - (r^^)) {Y This observation, and the facts that u^ /3. > coup) | the result follows from three observations. First, from (73), (1 increasing in (76) J \ . V^ {TD no coup)+/3y^ {D) and \ I Now ^' {D)] as given by (26) and (27) above. Rearranging these expressions, {TD (1-/3) u^ 0) ^ toG<{l-(f>){T-C (r)) (F - xA^) write the condition for prevention to be preferred, (1 - r) (1 producers benefit from prevention when V'^ V^ {TD the level of public maximization program Telo,i],GeK+ subject r^^ and a^ — (1 — a^x < a^ - (1 - f) f) A^. Third, condition (77) also holds and thus prevention is A^, when /3 = 0, because These observations imply that there for free. exists such that the right-hand-side and the left-hand-side of (77) are equal, thus this always satisfied and we have This establishes that, if V^ {TD G^^ > 0, no coup) > \ coup prevention V^ {TD \ coup) for any value of always better for transitional is democracy. We applies G> next consider the case where when /? G in [/3*, 1], problem (72) is where G> (77) in this case is It is 0. defined as the binding (this constraint a strictly increasing function of the constraint /3* is G^^ = /3 slack for straightforward to show that this case minimum value of implicit in is and the tax rate defined by P < P* and binds for p > G 6 ^5 such that the constraint 1K+). Observe that (74) does not P* . The depend on w^^ /3, is hence equivalent of condition can be rewritten as T A^ < A^ - u^ {TD coup) I 52 + Y3-37 (a^ - (1 - 7^) ^ ) • (78) We now show that right-hand side depend on from P -^ if any Therefore, the right-hand side is Maximum Theorem (3* , where the a convex function extended at in the range b real line, h* G (6*, 6°°). = /3* in (14) a strictly convex function of is - /3/ (1 /3), whereas Clearly, condition (78) holds as h. (3 over by [0, 1] program (72) changes. Finally, observe than a linear function at two end points of an interval — / (1 = in 6 coup) does not \ Stokey, Lucas and Prescott, 1989 Chapter 3), condition (77) (e.g. less and A^ - u^ [TD term on the last a continuous function of set of active constraints in is Note that the [/?*, 1]. hnear (increasing) is Moreover, since the payoff to the citizens I. also holds at that /? by the definition of a^ positive and (73) t'^^ (75) Berge's /?. is any (78) holds for /3*) = 6°° and oo, then This establishes that (78) also holds it is for also less any j3 G in the than the same function [/3*, 1] and completes the proof. Proof of Proposition 6 7.2 V^ {E, S) and V^ [E, N), given in (32) and in (53) respectively, are independent of n, either V^ {E, S) > V" [E, N) or V" {E, S) < V" {E, N) for any value of (the case where V^ {E,S) = V^ {E,N) is ruled out by the assumption that y^ Jl). If Condition 2 does not hold, then V"^ {E, S) > V^ {E, N) and non-prevention is never chosen First, note that, because both IT fj, by the elite. In this case, the equilibrium is the same as in Proposition 4. When Condition 3 does not hold, the participation constraint of soldiers under non-prevention cannot be thus this strategy is not feasible and again the equilibrium from Proposition 4 apphes. Let us then focus on the case where both Conditions 2 and 3 hold. In this case, V^ [E, N) and non-prevention and preferred by the feasible is the argument in the text and the definition of any n ^ 7.3 Proof of Proposition 7 We it, satisfied, which establishes the tt, we have that elite to smooth V^ {E, P V transition. ^ V^ < {E, S) By N) for strictly decreasing in R. \ n) {E, result. begin by showing that the threshold 7 (R) defined in (61) is Straightforward differentiation of 7 (R) gives 7' (R) = '-^ ""''-"' P where w'^^ and a^ are defined in (57) and expressions of w'^ and a^ defined in (47), xw^ -a^ = xw'^'^ z(tt)™ - a^) in (58). Next, observe that taking into we obtain -a^ = -fxA^ 53 {I - t) A^ < 0. account the < Therefore, 7' (R) with equahty 0, Next consider the decision if and only of the ehte. if A^ = 0. R* First, define as the level of natural resources such that w^ {R*) X = f {Y - xA^) (79) . R = In other words, R* is wages necessary for coup prevention can be financed by taxing production income only maximum By possible rate f. substituting for w^'^ and fm w- (R) _„„P ^ w"^ + expression with (79), we have ,P Combining tliis when the level of natural resources such that fl7 13- , a'" in (65), + (l-7)S " ^ _ R*, total military at the we obtain R. (80) wM — wP ^* , R^^ir, T. r-T/3(7+(l-7)S) Claim R* 1 Suppose that = vention, the elite set f is R > given by (81) and f and choose Q > R* ^ Then . to balance the which implies C^^^(7+(l-7)^)-^ '" M _ in any MPE ' with coup pre- government budget constraint, P ^^ (82) . Proof. The expression of the government budget constraint provided by ,p_ w^{R)x-CR ~ [Y-xA^) ^ (81 (64) implies that '^ Using this expression, the per period utility of the elite in oligarchy This expression be everywhere decreasing in ( provided that is < the case, since x will ' set at the — [1 n) by assumption, maximum budget constraint, that Using the fact that in possible level f, and Q is, and since C, will can be written as nA^ + xA^ < Y Y = nA^ + (1 — n) which , A^ . is Therefore, We as given in (82). equihbrium f^ = f, and that (^ is given by (82), ^^'"^'"^^^ we have that • (1-/?)(1-/3(1-.)) also have ,->//, f^ be determined to satisfy the government can be written as "^ always p., {l-P){A" + R/n)+/3{l-r)A" 54 (63) Moreover, using n (R) (62), the threshold 1 n{R)^ -^ (1 - C^) R/n - (A - A= - (1 {A" + R/n)+p ^) - (1 f) \ n) is given by A") A-{l-(3)V"{TD) P where V^ {E, S) = V" [E, P which at - (1 A". Now f) since d - [{\ /dR = Q^) R) 1-/3 (7 + - (1 we have _ 1 -^ [1 /5 (7 + - (1 - 7) S) - (1 A- ^)] (1 - /3) - /3) V" {TD) %'{R) P A-{l-(3)V"{TD) !-/?[(!- C^) P The numerator of this R/n - (A - A- n expression is - (1 - (1 f) .4^)] (1 y^ {TD) /3) V^ (TD) decreasing in V^ and {TD) < (1 — f) ^^/ (1 — P). Therefore, - [1 is sufficient for vr' /3 + (7 (i?) > (1 - (A - 7) x)] Using the 0. (1 - f) which in turn is the - [1 same This establishes that when more likely. (7 + (1 - 7) x)] (1 (1 - - /?) f) (1 - A^ = C^) i?/n (83) (1-/3) {r^" + R/n) , equivalent to is tA" > X {w"'^ - w^) as the following condition: X repression /3 A - fact that substituting for (^ and rearranging terms, (83) n A") > > f (1 - - /37) (1 n) r(l-^7) + /3(l-7)(l-nf) R> R* and x > x, n' > {R) and thus higher resource rents make This completes the proof of the proposition. Proof of Proposition 8 7.4 V^ Using the expressions of that G*-^^ coups = f V^ {TD {I I — (f)) {D) = {Y — xA^), the value to low-skill producers coup) = i^-^i^-^0))^'-h^rY I Similarly, using the facts that V'^ {D) G^^ = in (18), of a^ in (47), and the fact when they do not prevent coup), given by (69), can be rewritten as V^ {TD that V^ {M) a^/ (1-/3), of f (y - xA^) — w'^^x > 0, = a^/ (1 that w'^^ 55 - is p), . ^^ )xA^ V^ {TD, P) = V^ {TD \ no coup), given by (66) and that w^' and a^ are 7) x), , ' when they prevent coups V^ {TD no given by (47), the value to low-skill producers V^ [TD coups if 1-/3 (TD no V'^ --^' ^ PifxA^ _ 1-/3(1 -A) 1-^ ^^l^ML = no coup) I Assuming that coup prevention coup) > V'^ condition A* {TD > 1. 0, - ma\ ,) 1-/3 ' X^xa^ (i_/5)(i_/3(i_A))' _ \ Combining the previous two expressions, coup). this equivalent to is fxA^ {^A (/37 +pxa^ - [(1 /3 + - (1 (1 /3) + /37) A - left-hand-side of this inequality = {l a feasible strategy, low-skill producers prefer to prevent is I at A coup), can be rewritten as (68), The \ is - (1 4>)) - /3) - (1 (1 - 7)] + /3) [/3 (1 - - (1 + 7) /3) (1 (1 - /3 - A for all < [0, 1) + /3A) 4>a^ > it is such that (84) holds F^ {TD A*, | no coup) (84) /3) </>]} a strictly increasing function of A and Therefore, there exists a unique A* e then (84) does not hold - strictly positive for all < V^ {TD 0. | > A A*. If coup) and the low-skill producers prefer not to prevent coups. Proof of Proposition 9 7.5 Using (70) and "" (defined as fr (1 there - is V" {E, S) = V" {E, P + ap {f - T^)) A" - Pf) A^] + a/3 [(1 - f ^^ - \ n)) can be written as {t^ - pf l-P ~ /3 If (71), the threshold a) [(1 - no coup P) V" {TD) - in transitional (1 ) democracy, (46) implies (1 V^ (TD) = - p) (1 — V" {TD)] f) A^ / (1 — /3) and ._ (l-Q^)T^ (l-a)/3f' which If is strictly decreasing in a. coups take place along the equilibrium path in transitional democracy, then (46) yields V^ {TD) = {i-f){l-^)A'' + Using this expression, the threshold Since T^ does not depend on (1 - f) A^. n becomes r-P /3 ^ - ^f + - T^) - f - (1 - a/3 (f - (1 - a, we obtain that the a) [aPcl) (1 f) (/) f)] derivative of this expression dn/da is proportional to B {<P) = [a (1 - P) r^ - (1 - a) (a/3 (f - r^) 56 (/3f - r^))] (1 - f) p^4>-{f + (1 - r) <P) (1 - P) Pr"" where (f) — any that all Q. 4> 71" p e B {(f)) is the numerator of dn/da. This expression Tlierefore, < is (j)* , it B {4>) < has at most one root (so that tt is strictly increasing in q). [0, 1] and we set p* = 1. 4> — (p* is over the interval strictly decreasing in a) Moreover, if B (c6) This establishes 57 Knear in all and (p, [0, 1]. for any has no root in and is negative This imphes (j) [0, 1], > cp* , then tliat for B > B {(p) < the claims in the proposition. {(j)) when (so for References Acemoglu, Daron (2005) Monetary Economics, "Politics and Economics in Weak and Strong States," Journal of 52, 1199-1226. Acemoglu, Daron, Michael Golosov and Aleh Tsyvinski (2008) ''Political Economy of Mech- anisms," forthcoming Econometrica. Acemoglu, Daron and James A. Robinson (2000) chise? "Why Did the West Extend the Fran- Democracy, Inequality, and Growth in Historical Perspective." Quarterly Journal of Economics, 115(4), 1167-1199. Acemoglu, Daron and James A. Robinson (2001) "A Theory of American Economic Review, Political Transitions," 91, 938-963. Acemoglu, Daron and James A. Robinson (2006) Economic Origins of Dictatorship and New Democracy, York; Cambridge University Press. Acemoglu, Daron and James A. Robinson (2006) "Persistence of Power, tutions," forthcoming Elites and Insti- American Economic Review. Acemoglu, Daron, James A. Robinson and Thierry Verdier (2004) "Kleptocracy and Divideand-Rule: A Model and Proceedings, 2, of Personal Rule," Journal of the European Economic Association Papers 162-192. Acemoglu, Daron, Davide Ticchi and Andrea Vindigni (2006) "Emergence and Persistence of Inefficient States," Alesina, Alberto NBER Working Paper No. 12748. and Guido Tabellini (2007) "Bureaucrats or Policy Task," American Economic Review, Politicians? Part I: A Single 97(1), 169-179. Barro, Robert (1973) "The Control of Politicians: An Economic Model," Public Choice, 14(1), 19-42. Benabou, Roland (2000) "Unequal American Economic Review, Besley, Societies: Income Distribution and the Social Contract," 90(1), 96-129. Timothy (2006) Principled Agents? The Political Economy of Good Government. Oxford University Press. Besley, Timothy and Anne Case (1995) "Does Policy Choices? Evidence from Gubernatorial Electoral Accountability Affect Term Economic Limits," Quarterly Journal of Economics, 110(3), 769-798. Besley, Timothy and Stephen Coate (1997) "An Economic Model racy," Quarterly Journal of Economics, 108, 85-114. 58 of Representative Democ- Timothy and James A. Robinson (2007) "A Theory Besley, of Mihtary Politics," work in progress. Bowman, Kirk S. (2004) Militarization, Democracy, and Development: The Perils of Prae- torianism in Latin America, University Park PA; Pennsylvania State University Press. Brooker, Paul (2000) Non- Democratic Regimes. Theory, Government and Politics, New Debs, Alexandre (2007a) "The Wheel of Fortune: Agency Problems in Dictatorships," MIT York; Palgrave Macmillan. mimeo. Debs, Alexandre (2007b) "Political Strength and Economic EfRciency in a Multi-Agent State," MIT Dixit, mimeo. Avinash K. (2006) "Predatory States and Failing States: An Agency Perspective," Economy Working Paper. Princeton Research in Political Egorov, Georgy and Konstantin Sonin (2004) "Dictators and Their Viziers: Agency Prob- lems in Dictatorships," CEPR Discussion Paper No. 4777. Ferejohn, John (1986) "Incumbent Performance and Electoral Control," Public Choice, 50, 5-25. Finer, Samuel A. (1976) The Man The Role of the Military in on Horseback: Politics, Piscataway NJ; Transaction Publishers. Huntington, Samuel P. (1968) Political Order in Changing Societies, New Haven CT; Yale University Press. Jensen, in Africa," Nathan and Leonard Wantchekon (2004) "Resource Wealth and Political Regimes Comparative Political Studies, 37(7), 816-841. Karl, Terry L. (1997) The Paradox of Plenty: Oil Booms and Petro-States, Berkeley CA; University of California Press. Luebbert, Gregory (1991) Liberalism, Fascism, or Social Democracy: Social Classes and the Political Origins of Regimes in Interwar Europe, New York; Oxford University Press Mahoney, James (2001) The Legacies of Liberalism: Path Dependence and in Central America, Baltimore, MD; The Political Regimes Johns Hopkins University Press. Mehlum, Halvor, Karl Moene and Ragnar Torvik (2006) "Institutions and the Resource Curse," Economic Journal, 116, 1-20. Meltzer, Allan H. and Scott F, Richard (1981) "A Rational Theory of the Size of Govern- ment." Journal of Political Economy, 89(5), 914-927. Millet, Richard (1977) Cuardians of the Dynasty: 59 A History of the U.S. Created Guardia Nacional de Nicaragua and the Somoza Family, Maryknoll NY; Orbis Books. Moore, Barrington (1966) Social Origins of Dictatorship and Democracy: Lord and Peasant in the Making of the Modern World, Boston, MA; Beacon Mosca, Gaetano (1939) The Ruling Class, New Press. York; McGraw-Hill. Mulligan, Casey, Ricard Gil and Xavier Sala-i-Martin (2004) "Do Democracies Have Different Public Policies than Nondemocracies?," Journal of Economic Perspectives, 18(1), 51-74. Needier, Martin C. (1987) The Problem of Democracy in Latin America, Lexington MA; Lexington Books. Nordlinger, Eric A. (1977) Soldiers in Politics: Military Coups and Governments, Upper Saddle River NJ; Prentice-Hall. O'Donnell, Guillermo A. (1973) Modernization and Bureaucratic-Authoritarianism: Studies in South American Politics, Berkeley CA; University of California Press. O'Donnell, Guillermo A. and Philipp C. Schmitter (1986) Transitions from Authoritarian Rule: Tentative Conclusions about Uncertain Democracies, Baltimore, MD; The Johns Hopkins University Press. Amos Perlmutter, (1977) The Military in Politics in Praetorians, and Revolutionary Soldiers, Perlmutter, Amos New Haven CT; Perotti, New Haven CT; (1981) Adodern Authoritarianism. A Modem Times: On Professionals, Yale University Press. Comparative Institutional Analysis, Yale University Press. Roberto (1996) "Growth, Income Distribution, and Democracy: What the Data Say," Journal of Economic Growth, 1(2), 149-87. Persson, Torsten and Guido Tabellini (2000) Political Economics: Explaining Economic Policy, Cambridge MA; The MIT Press. Persson, Torsten and Guido Tabellini (2003) The Economic Effects of Constitutions, bridge MA; The MIT Cam- Press. Persson, Torsten, Gerard Roland and Guido Tabellini (1997) "Separation Political Accountability," Quarterly Journal of of, Powers and Economics, 112(4), 1163-1202. Persson, Torsten, Gerard Roland and Guido Tabellini (2000) "Comparative Politics and Public Finance," Journal of Political Economy, 108(6), 1121-1161. Poulantzas, Nicos (1978) Political Power and Social Classes, London, Verso. Roberts, Kevin (1977) "Voting over Income Tax Schedules," Journal of Public Economics, 8, 329-340. Romer, Thomas (1975) "Individual Welfare, Majority Voting, and the Properties of a Linear 60 Income Tax," Journal of Public Economics, 7, 163-185. Roquie, Alain (1987) The Military and the State in Latin America, Berkeley CA; University of California Press. Ross, Michael (2001) "Does Oil Hinder Democracy?," World Politics, 53(3), 325-61. Skocpol, Theda Russia, arid China, Stokey, Economic Nancy in (1979) States and Social Revolutions: New L., A Comparative Analysis of France, York; Cambridge University Press. Robert E. Lucas and Edward C. Prescott {1989) Recursive Methods in Economic Dynamics, Cambridge MA; Harvard University Ticchi, Davide Press. and Andrea Vindigni (2003) "Democracy and Armies," Faculty of Eco- nomics, University of Urbino, Working Paper 83. Vanden, Harry E. and Garry Prevost (2002) Politics of Latin America: The Power Game, New York; Oxford University Press. Yashar, Deborah and Guatemala, 1870's Welch, Claude E. (1997) J. - Jr. Civil-Military .Relations, Demanding Democracy: Reform and Reaction 1950's, Stanford CA; Stanford University in Costa Rica Press. and Arthur K. Smith (1974) Military Role and Rule: Perspectives on North Scituate MA; Duxbury Press. Wintrobe, Ronald (1998) The Political Economy of Dictatorship, University Press. 61 New York; Cambridge 7