Annu. Rev. Polit. Sei. 2002. 5:151-79 DOI: 10.1146/annuiev.polisci.5.102301.084508 Copyright (c) 2002 by Annual Reviews. All rights reserved Democratic Institutions and Regime survival: Parliamentary and Presidential Democracies Reconsidered Jose Antonio Cheibub1 and Fernando Limongi2 'Department of Political Science, Yale University, New Haven, Connecticut 06511; e-mail: jose. cheibub @yale. edu 2Departmento de Ciencia Politica, Universidade de Sao Paulo, Sao Paulo, SP 05508-900, Brazil; e-mail: fdmplimo@pop.usp.br Key Words democratic stability, minority governments, coalitions, legislative institutions ■ Abstract We review arguments and empirical evidence in the comparative literature that bear on the differences in the survival rates of parliamentary and presidential democracies. Most of these arguments focus on the fact that presidential democracies are based on the separation of executive and legislative powers, whereas parliamentary democracies are based on the fusion of these powers. The implications of this basic distinction lead to radically different behavior and outcomes under each regime. We argue that this perspective is misguided and that one cannot deduce the functioning of the political system from the way governments are formed. Other provisions, constitutional and otherwise, also affect the way parliamentary and presidential democracies operate, and these provisions may counteract some of the tendencies that we would expect to observe if we derived the regime's performance from its basic constitutional principle. INTRODUCTION The idea that the form of government influences the survival of democracies was a major point of contention among students of comparative politics in the late 1980s and 1990s. The argument, first developed by Linz (1978), about the superiority of parliamentary over presidential institutions guided much of the discussion about the prospects of democracies born in the wake of the so-called third wave of democratization. The new conventional wisdom among comparative politics scholars was that, if these democracies were to succeed, they should adopt parliamentary institutions. Indeed, parliamentary democracies seem to outperform presidential democracies in many key aspects, notably in their capacity to survive under a wide set of conditions. Between 1946 and 1999, one in every 23 presidential regimes died (that is, became a dictatorship), whereas only one in every 58 parliamentary regimes 1094-2939/02/0615-0151 $ 14.00 151 152 CHEIBUB ■ LIMONGI died. At very low levels of economic development, say at the level observed in sub-Saharan Africa, neither parliamentary nor presidential democracies are likely to survive; under these circumstances, one in every eight democracies, of either type, dies. At higher levels of development, however, things are different. Not only are parliamentary democracies consistently more likely to survive than presidential democracies, their chances of survival under economic crisis are at least as good as presidential democracies' chances of survival under economic expansion.1 Moreover, although presidential democracies are more likely to emerge from military dictatorships than from civilian dictatorships, and thus are more likely to become dictatorships themselves, when origin is held constant we still find that presidential democracies die sooner than parliamentary democracies (Przeworski et al. 2000). This fact has commonly been interpreted as evidence that the instability of presidential democracies stems from the principle of separation between executive and legislative authorities, which distinguishes presidentialism from parliamentarism. Several implications are derived from this basic difference that would explain why survival rates differ between these democratic regimes. The fusion of powers characteristic of parliamentarism is supposed to generate governments capable of governing because they are supported by a majority in parliament, composed of highly disciplined parties prone to cooperate with one another, which, together, would produce a highly centralized decision-making process. Presidential regimes, however, frequently generate presidents who cannot count on a majority of seats in congress. Congress is composed of individual legislators who have little incentive to cooperate with one another, with their parties, or with the executive. As a consequence, decision making under presidentialism is highly decentralized. Presidential regimes, therefore, are characterized by weak political parties and frequent stalemates between the president and congress in a context of loose decision making. Because presidential regimes lack a mechanism to resolve conflicts between executives and legislatures (such as the votes of confidence or censure of parliamentary regimes), minority presidents, divided government, and deadlock would drive actors toward extraconstitutional means of resolving their differences, thus making presidential regimes prone to instability and eventual death. This view, now widespread, was originally formulated by Linz (1978, elaborated in 1994). Other proponents include Stepan & Skach (1993), Gonzalez & Gillespie (1994, p. 172), Hartlyn (1994, p. 221), Valenzuela (1994, p. 136), Jones (1995a, pp. 34, 38), Mainwaring & Scully (1995), Linz & Stepan (1996, p. 181), Nino (1996, pp. 168-69), Huang (1997, pp. 138-39), and Ackerman (2000, p. 645). This view, however, is problematic. Parliamentary and presidential regimes are indeed founded on different constitutional principles, and this is a central choice in any democratic constitution. However, the operation of the political system cannot be entirely derived from the mode of government formation. Other provisions, constitutional and otherwise, also affect the way parliamentary and presidential 'These figures are based on the Regime and Development Database, available at http:// pantheon.yale .edu/~j ac236/. DEMOCRATIC REGIME SURVIVAL 153 democracies operate, and these provisions may counteract some of the tendencies that we would expect to observe if we derived the regime's entire performance from its basic constitutional principles. Moreover, even if these principles were the main factor in shaping incentives under parliamentary and presidential systems, it would not be sufficient to simply assert that they are different and that hence outcomes should also be different. One would need to specify which institutional features are affected by which incentives and with what consequences. We review arguments and empirical evidence in the comparative literature that bear on the differences in the performance of parliamentary and presidential regimes that emerge from the alleged differences in incentives that these constitutional frameworks generate. We focus on three areas that, according to the traditional view, give an advantage to parliamentary regimes: legislative majorities, incentives for cooperation, and the centralization of the decision-making process. Although we believe that parliamentarism does outperform presidential-ism in survival, we have good reasons to doubt that we understand what causes this difference. This is why we advocate a return to a theme about which some believe political science already knows everything there is to know. THE "MAJORITARIAN IMPERATIVE" There is a majoritarian imperative in parliamentarism, or so the conventional view implies. This imperative seems to follow from the very definition of parliamentary democracies. Parliamentarism, according to this view, is a regime in which the government, in order to attain and retain power, must enjoy the confidence of the legislature. Because decisions are made by majority rule, no parliamentary government can exist without the support of a majority. Minority governments could occasionally emerge, but these would be relatively infrequent and necessarily ephemeral, since they would simply reflect the temporary inability of the current majority to crystalize. This inability is temporary because the system contains automatic correctives for these situations: Either a new government supported by a majority will be formed, or, if this is not possible, new elections will be held so that such a majority may emerge. Presidential regimes, in contrast, lack the majoritarian imperative. But majorities also matter under presidentialism. Voters have two agents who, by design, do not necessarily represent the same majority. These agents have fixed terms in office and do not depend on each other to exist. If elections result in a situation in which the presidential party does not control a majority of legislative seats, there is no mechanism to solve the conflicts between the two legitimate majorities. The most likely outcome, it is believed, is stalemate and impasse between the executive and the legislative branch, which can ultimately result in the collapse of the democratic regime. However, several theoretical and empirical arguments suggest that the majoritarian imperative that supposedly distinguishes parliamentarism and presidentialism is neither an imperative nor sufficient to distinguish them. 154 CHEIBUB ■ LIMONGI In the first place, as conclusively demonstrated by Strom (1990), parliamentary governments do not necessarily produce majority governments. We consider this issue in more detail below when we discuss coalition formation. Here it is sufficient to say that minority governments existed 22% of the time in parliamentary regimes from 1946 to 1999 (Cheibub et al. 2001). Other counts (see, e.g., Strom 1990, Cheibub 1998), based exclusively on industrialized democracies, find that about one third of governments formed under parliamentarism have a minority status. More important, Strom's (1990) analysis shows that minority governments are not necessarily a sign of political instability. Rather, the emergence of minority governments can be explained in terms of the calculus made by party leaders about the costs and benefits of participating in government, given that they are concerned not only with achieving office but also with the policies to be implemented by the government. This calculus, Strom argues, is affected by the degree of policy influence parties can exert when outside the government, as well as the competitiveness and decisiveness of the electoral process. Out-of-government policy influence, in turn, depends essentially on the organization of parliament (existence of standing committees, degree of specialization, scope of action, allocation rules). Electoral decisiveness and competitiveness depend on the clarity of the electoral alternatives presented to voters (identifiability), the degree to which the distribution of seats fluctuates from party to party between elections (competitiveness), the direct relationship between electoral success and government participation (responsiveness), and the proximity of elections to government formation. Thus, it is simply not true that majority governments are the expected outcome of government formation under parliamentarism. Whether they are depends on institutional traits that are not part of what defines a democracy as parliamentary. Furthermore, there is evidence that minority governments under presidential-ism, although frequent, are not as widespread as we would expect them to be. Cheibub (2002), from whose work Table 1 was adapted, shows that about 40% of the years of presidentialism between 1946 and 1996 were under minority governments, a number not far from the estimates for parliamentary regimes. These cases, as he shows, do not occur randomly. The frequency of minority presidential governments is associated with the number of political parties (although not in the expected way), with the type of electoral system, and with the electoral cycle (Mainwaring 1993, Jones 1995, Shugart 1995). However, the occurrence of minority governments has no impact whatsoever on the survival of presidential democracies (Cheibub 2002). As for deadlock, the specter that supposedly haunts presidentialism, it is neither pervasive nor is it associated with regime breakdown (Cheibub 2002). Deadlock occurs only when the preferences of a majority cannot prevail. These situations depend both on the share of seats controlled by the party of the president in congress and on specific institutional features regarding the presidential veto and legislative override of the presidential veto: whether the president has veto power; the type of congressional majority necessary to override the presidential veto; whether the system is unicameral or bicameral; and whether, in bicameral systems, DEMOCRATIC REGIME SURVIVAL 155 TABLE 1 Minority presidents, deadlock situations, and transition probabilities in presidential regimes by type of legislature, effective number of political parties, electoral system, and timing of elections (source: Cheibub 2002) % Minority Deadlock Transition presidents3 (N) situations5 (N) probabilities0 All 40.22 (726) 33.52 (710) 0.0395 Type of legislature: Unicameral 36.46 (277) 29.67 (273) 0.0464 Bicameral 42.54 (449) 35.96 (437) 0.0353 Electoral system: Majority-plurality 39.04 (146) 36.99 (146) 0.0482 Pure proportional 39.42 (553) 32.96 (540) 0.0378 Pure prop. + mixed 40.52 (580) 32.62 (564) 0.0372 Effective number of parties (ENP): ENP<2 35.33 (150) 27.33 (150) 0.0458 2 < ENP < 3 33.45 (281) 31.49 (280) 0.0209 3 < ENP < 4 59.69 (129) 49.22 (128) 0.0714 4 < ENP < 5 28.17 (71) 28.17 (71) 0.0417 ENP>5 50.60 (83) 32.10(81) 0.0111 Timing of legislative and presidential elections: Non-concurrent 45.16 (124) 40.32 (124) 0.0318 Alternate 66.92 (133) 47.11 (121) 0.0548 Non-conc. + alternate 56.42 (257) 43.67 (245) 0.0429 Concurrent 31.34 (469) 28.17 (465) 0.0374 Political conditions Minority presidents 0.0462 Majority presidents 0.0293 Minority governments'* 0.0392 Majority governments'* 0.0377 Deadlock situations 0.0378 No deadlock situations 0.0318 a"Minority presidents" include the cases in which the party of the president does not control >50% of the seats in the legislature in a unicameral system; or where it does not control >50% of the seats in at least one of the chambers in a bicameral system. b"Deadlock situations" are defined by the number of seats held by the government, by whether presidents can veto legislation, by the requirements for legislative override of the presidential veto, by the legislative structure, and by whether, in bicameral systems, veto override is by a vote in each house or by a joint session of both houses. ^Transition probabilities" indicate the probability that a presidential regime will become a dictatorship (the number of transitions away from presidentialism divided by the number of cases of presidentialism). d"Minority" and "majority" governments are defined by the share of legislative seats held by all the parties that hold cabinet positions. 156 CHEIBUB ■ LIMONGI veto override is by a vote in each chamber separately or in a joint session of both chambers. In combination with the share of seats held by the government, these factors allow one to distinguish three situations: presidential dominance, opposition dominance, and legislative deadlock. The latter occurs only when the president is likely to veto a bill approved by a majority in the legislature but that majority is not sufficient to override the presidential veto. If arguments about the perils of presidentialism are correct, presidential democracies should face higher risks of dying when conditions for deadlock between the president and congress are met. Yet, as we see at the bottom of Table 1, this is not the case. The difference in the transition probabilities for deadlock and no-deadlock situations, although in favor of the former, is rather small: One in every 26 presidential democracies dies when there is deadlock, one in every 31 when there is no deadlock. This difference does not seem to warrant the level of concern with deadlock that is often expressed in the comparative literature on presidentialism. The belief that the survival prospects of presidential democracies are compromised when presidential parties lack a majority of seats in congress, or when deadlock situations exist, has no empirical basis. The conventional view is that deadlock will never occur in parliamentary regimes. Indeed, parliamentary regimes are designed so that whenever there is a deadlock between the government and the legislature, either the government changes or the legislature changes. Thus, although divided government may also exist under parliamentarism (for example with minority governments), the fact that the government in these systems exists only as long as there is no alternative majority that can replace it distinguishes them from presidential regimes. Ultimately, divided government in parliamentarism cannot produce deadlock, at least not deadlock in the same sense as in presidentialism. Yet, the fact that parliamentarism includes a mechanism that can be invoked in case of policy conflict between the government and the legislative majority does not mean that this mechanism will always be invoked, or that, once invoked, it will necessarily put an end to the disagreement that led to its use. For this reason, deadlock under parliamentary regimes may occur over time, as when governments succeed governments and no stable majority is formed even after new elections are held. The discussion so far has been guided by the supposition that the existence of a majority, either in parliamentarism or in presidentialism, automatically means the ability of the government to govern. However, the scope of action available to the government can also be reduced as it seeks to obtain majority status. What matters is both the number of parties that must come together in order to establish a majority and the cohesion of the party (or parties) that belong to the majority. In spatial terms, government action is necessarily limited to the area that contains policy proposals preferred by its supporters over the status quo. That area, however, may be small, and it may become smaller as the government attempts to broaden its base of support; that is, as it attempts to secure the support of a majority in the legislature. As Tsebelis (1995) has demonstrated, policy change in democratic regimes is associated with the number of actors who can veto a proposal. The DEMOCRATIC REGIME SURVIVAL 157 number of veto players, in turn, is a function of both institutional and ideological variables: Policy stability increases with the number and the cohesion of, and the distance among, the parties that belong to the government. The reduction in the scope of governmental action as a by-product of majority building will certainly affect the policies pursued in a given system. Whether it will affect the survival of the regime and whether its impact will be larger under presidentialism than under parliamentarism cannot be specified a priori. The point is that governments may have to pay the price of policy immobility in order to form a majority, and this price may have a negative effect on the survival of the democratic regime. INCENTIVES FOR COOPERATION Parliamentary regimes are supposed to foster cooperation. Political parties have an incentive to cooperate with one another; parties in the government will support the executive, and parties out of the government will refrain from escalating conflicts because they may, at any time, become part of the government; individual members of parliament will align themselves with their parties. Consequently, the government is supported by a majority composed of highly disciplined parties, prone to cooperate with one another. Presidentialism, on the other hand, is characterized by the absence of such incentives and hence is likely to generate governments that, even if supported by a majority, are based on undisciplined parties that tend to compete fiercely with each other. There are, in fact, two distinct issues here: the discipline of political parties and their propensity to enter and stay in governing coalitions. Although related, these issues must be discussed separately. Party Discipline There are formal and nonformal arguments relating regime type to party discipline or cohesion in legislative vote.2 In the nonformal arguments, which originated with Linz and are reproduced by most critics of presidentialism, what matters is 2Conceptually, party discipline and cohesion are distinct (Ozbudun 1970, Tsebelis 1995). Empirically, however, we can only observe a group of legislators voting together, either as an expression of their true preferences or as the result of disciplinary measures. As Bowler et al. (1999b) note, this distinction matters only when cohesion is moderate. If it is very high, then disciplinary measures are not necessary. If cohesion is very low, it is unlikely that such a heterogeneous group of legislators will agree to any measure that will make them vote together. Only when cohesion is high enough that a group of like-minded representatives will accept a common set of constraining rules, but low enough that they will occasionally find it in their interests to vote against the party's position, do disciplinary mechanisms become relevant. In what follows, unless explicitly noted, we treat discipline and cohesion as synonymous. 158 CHEIBUB ■ LIMONGI the legislators' desire to come to and remain in office. The postulate is that the threat of government dissolution and early elections is necessary and sufficient to induce party discipline. Here is how the argument works. Under parliamentarism, undisciplined parties may mean a failure to obtain majority support in parliament, the defeat of government bills, and consequently the fall of the government. In order to remain in government, political parties enforce discipline so that their members in parliament can be counted on to support the bills proposed by the government. Individual legislators, in turn, have an incentive to support the government in order to prevent the occurrence of early elections in which they would risk losing their positions. Under presidentialism, since the government and the legislature are independently constituted, office-seeking political parties have no reason to impose discipline on their members; their survival in office does not depend on the result of any particular vote in the legislature. Individual members of congress also lack any incentive to accept the discipline of political parties (if they were to try to impose it); voting against the party or the government would not make them any more likely to lose their mandates in early elections. Thus, given office-seeking politicians, the fusion of power that characterizes parliamentary regimes generates incentives for individual legislators and political parties to cooperate with the government, resulting in a high level of party discipline. The separation of powers that characterizes presidentialism implies very low levels of party discipline. Even if a president were lucky enough to belong to a party that controlled a majority of seats in congress, he or she could not necessarily count on the support of that majority in order to govern. On the contrary, the president should expect, at least under some circumstances, that no support would be forthcoming from that majority.3 There are several problems with this argument. At the highest level of generality, the assumption on which it is based—that politicians care only about office—is not tenable. If it were true, we should never observe minority governments, since the party in charge of forming a government would always be able to lure some party into the government in order to attain a majority. Yet minority governments, as we have seen, are not infrequent in either parliamentary or presidential regimes. At a lower level of generality, the standard argument that connects the threat of government dissolution with party discipline is inconsistent. If the argument 3One of the most-noted circumstances appears near the end of the presidential term. As presidential elections near, the argument goes, members of the president's party try to distance themselves from him or her in order to avoid paying the costs associated with the government's policies. This argument neglects the possibility of circumstances in which members of the president's party do want to identify with the president in order to share in the benefits of the government's policies. Implicitly, this argument assumes that control over the government brings no electoral benefit and that presidents are not able to transfer votes for the politicians who support them. This, however, does not seem to be the case. Data on the rates of presidential reelection in the absence of term limits show that incumbency is indeed a big advantage (see, for instance, Cheibub & Przeworski 1999). Note, in addition, that the effect of forthcoming elections on support for the government, if any, should also be expected under parliamentarism (Baron 1998). DEMOCRATIC REGIME SURVIVAL 159 assumes at one stage that individual legislators may gain electorally by providing specific benefits to their constituency, it denies the existence of these benefits at the next stage. Let us consider this point in detail. Suppose a key presidential initiative implies losses for a specific group and gains for the whole society. In the standard view, presidents have no means to induce legislators to support a proposal that removes some special privileges of a given constituency. Because legislators are office seekers, they have a clear preference for distributive policies, that is, policies that concentrate benefits on their constituencies and disperse the costs. Hence, legislators will do better if they vote against the presidential initiative and protect their constituency's narrow interests so that they will obtain its vote again in the future. They bear no costs of acting this way, and they will collect the benefits when elections are due. The same would not occur under a parliamentary regime. An identical situation would lead to a different result. Because dissolution and early election are possible, legislators will prefer to follow the party line and support the government so that they can guarantee their seats. There are costs to bear, since one may lose his or her seat. By calling (or threatening to call) an early election, the prime minister invites the electorate to judge the behavior of the legislator. But why would a legislator who voted against the party to protect the interests of his or her constituency be punished by that same constituency? If the legislator loses the seat in early elections because his vote helped bring the government down, it must be because the number of voters who benefited from that action is not large enough to prevail electorally. If this is the case, then the legislator should know that it does not pay electorally to go against the party line and in favor of those voters. In turn, if the number of voters who benefited from the government defeat is large enough to elect a representative, then the legislator can safely defy the party line to protect them, since the legislator can expect to survive an early election.4 Therefore, with purely office-seeking politicians, early elections are not a credible threat that would induce party discipline. Hence, one needs something else to argue that the threat of dissolution leads to party discipline. The standard argument includes an unstated presumption that voters in parliamentary regimes base their votes on party labels and not on individual politicians. In other words, it is presumed that under parliamentarism the electoral connection is necessarily different from the one that prevails in the US congress, as analyzed by Mayhew (1974). But whether voters vote on the basis of party labels or of legislators' personal attributes is not necessarily related to the form of government. As a matter of fact, electoral laws, and not the form of government, are usually seen as the main factor determining whether voters will vote according to personal or party attributes in a given system. In Carey & Shugart's (1994) attempt to rank electoral systems according to the kind of incentives they provide, the key factor is 4Government dissolution, in fact, is not always an undesired outcome. As Smith (1996) shows in a model of majority governments, early elections are more likely to be called when the times are good (see also Baron 1998). 160 CHEIBUB ■ LIMONGI the control parties exercise over a politician's chances to get elected or reelected. This control, in turn, depends on the access to the ballot, the rules for transferring votes within party lists, the choices offered to vote below the party level, and the district magnitude. Parties are said to be strong and capable of enforcing discipline if these factors allow them to affect the probability that a politician will get elected. Important as the electoral variables may be in affecting the degree of party discipline in a system, they cannot be the whole story. Electoral laws may provide the incentives for legislators to cultivate the personal vote (by seeking policies that have concentrated benefits and diffuse costs), but the decision-making process may deny them the means to do so (by centralizing decision making so that the preferences of the individual legislator are virtually irrelevant).5 Indeed, the personal vote in the US congress is closely related to the decentralized decision-making process that characterizes its committee system (Mayhew 1974). On the other hand, as Cox (1987) demonstrated in his analysis of nineteenth-century England, a centralized decision-making process may neutralize the electoral incentives for the cultivation of the personal vote. What matters is whether we have reasons to expect that parliamentary governments necessarily foster a higher degree of centralization in policy making. Is the process described by Cox inherent to parliamentarism? Are all presidential regimes like the one described by Mayhew? In other words, can we take the United States and England to be the paradigmatic cases of presidential and parliamentary regimes when it comes to policy making? We argue in the next section that they are not, and that we have no reason to expect a systematic variation in the centralization of policy making between the two regimes. Finally, some of the implications of the standard view concerning the relationship between mode of government formation and party discipline are not supported by the facts. The calculus of the individual legislator under parliamentarism cannot be solely connected with the risk of election because early election is not the necessary consequence, or even the most frequent consequence, of a government dissolution. Cheibub (1998) shows that 56% of all prime ministers in 21 industrialized democracies between 1946 and 1995 changed without elections. In the same data set, 38% of changes in the party of the prime minister, 46% of changes in the partisan composition of the government, and 24% of changes in the major party in the government occurred without elections. Elections are far from being the necessary outcome of government dissolution in parliamentary democracies; hence, the costs they represent are not necessarily high and uniformly distributed across these systems. This point is forcefully made by Mershon (1996 and 1999) in her studies of coalition formation in Italy and other countries. 5Here the case of post-198 8 Brazil, arguably the presidential system with the most permissive party legislation in the world, becomes relevant (Limongi & Figueiredo 1995). In any roll-call vote taken in the lower house of the Brazilian National Congress since 1988,9 out of 10 representatives voted according to the recommendation of their party leaders. As Limongi & Figueiredo argue, some of this unexpected level of legislative vote cohesion must be attributed to the organizational structure of congress. DEMOCRATIC REGIME SURVIVAL 161 On the other hand, as far as presidentialism is concerned, the standard argument assumes that voters use their two votes independently and that representatives are judged exclusively by what they do to defend the narrow and immediate interests of their constituency. Voters do not care about their representative's role in the success or failure of the executive. If this were true, the electoral performance of the presidential party in legislative elections would be entirely dissociated from the performance of the president. Yet, considerable evidence suggests that voters do tend to associate their vote in presidential and legislative elections. This is why concurrent presidential and legislative elections would work to reduce the number of political parties competing in a given political system (Shugart & Carey 1992, Jones 1995, Shugart 1995). Hence, if voters connect their votes in presidential-legislative elections, legislators will have incentives to support the executive in some key votes. Their seats may depend on the good performance of the president. Formal arguments linking parliamentary regimes with legislative vote cohesion have been recently developed by several authors. Huber's (1996b) spatial model of the interaction between the prime minister, the cabinet, and the prime minister's majority highlights the role of vote-of-confidence procedures in legislative outcomes. Baron (1998) and Diermeier & Feddersen (1998) use a model of legislative bargaining to show how confidence procedures that characterize parliamentary democracies affect legislative cohesion. These papers represent important advances in the understanding of the functioning of parliamentary democracies, but they do not necessarily provide a compelling argument to the effect that levels of legislative cohesion are higher in parliamentary than in presidential democracies. To begin with, the models proposed by Huber, on the one hand, and Baron and Diermeier & Feddersen, on the other, differ in at least one very important aspect. Whereas the latter models explicitly set up a situation entailing conflict of interests among political parties so that legislative cohesion is not a function of similarity of preferences, Huber's analysis does not. In his case, there is an area of the policy space in which the preferences of all the actors overlap, and what the model shows is that the agenda power of the prime minister will allow her to pick, in that area, the policy that she prefers. Legislators who go against the government do so in order to signal to their constituents that they are defending their interests. They do so, however, knowing that the prime minister will choose a policy that they prefer over the status quo. The contribution of Huber's model, in our view, is not to show that vote-of-confidence procedures induce high levels of party discipline in a context of conflicting preferences, but to show, as he himself notes (Huber 1996b, p. 279), that prime ministers are strategically well-positioned to obtain policies that are to their liking and that, hence, political parties are constrained in their ability to shape policies after the government is formed.6 6This suggests a curious, and unexpected, parallel with complaints about the limited role of political parties and the legislature in some new, presidential, democratic regimes (O'Donnell 1994). What is seen as a positive trait in parliamentary regimes takes on a negative tone when observed in presidential regimes. 162 CHEIBUB ■ LIMONGI Baron (1998) and Diermeier & Feddersen (1998), in contrast, explicitly model a situation in which the preferences of the party or coalition members are in conflict. The mechanism that drives their model is the control over the legislative agenda enjoyed by the parties in the government. Because agenda power guarantees future gains, and because the vote-of-confidence procedure allows the government to link the vote on a policy with the survival of the government, and, hence, to control the legislative agenda, parties and legislators may find it in their interests to vote against their preferences. Underlying both models, as well as Huber's (1996b, p. 280), is the view of a presidential system such as the one in the United States, in which agenda-setting power lies with the legislature. However, if presidents can control the legislative agenda much in the same way as prime ministers can, then the mechanism that drives party cohesion in parliamentary regimes can also operate under presiden-tialism. We argue in the next section that if one considers the full range of existing presidential regimes, the United States is exceptional in granting little or no legislative and agenda power to the executive; hence, the United States is by no means representative of what presidents can do. Here it is sufficient to say that presidential regimes are compatible with executives that hold a high level of agenda and legislative powers (see Mainwaring & Shugart 1997b). The specific institutional procedure whereby this is achieved is obviously different from parliamentarism, but the end result may very well be the same. Finally, as Diermeier & Feddersen (1998) state, confidence procedures may be a sufficient institutional feature to induce legislative vote cohesion, but they are not a necessary feature. Other mechanisms may exist, some of which are institutional (e.g., centralized legislative organization and executive agenda and legislative powers) and some of which are not. In this context, Medina's7 analysis is particularly relevant because it shows how legislative voting cohesion can emerge from pure congruence of preferences. His analysis demonstrates that cohesion does not necessarily depend on disciplinary measures (such as the vote of confidence) and can be obtained under any institutional set-up. Thus, it is not at all clear that the existence of cohesive legislative blocs is endogenous to the regime type. We should not presume that presidential regimes invariably generate low levels of party discipline in the legislature. Coalition Government The basic argument about coalitions is that presidentialism, unlike parliamentarism, does not offer incentives for political parties to cooperate with the government. This fact is supposed to give rise to legislative paralysis or some other kind of "ungovernability," with all its attendant tragedies. Stepan & Skach (1993, pp. 17-18) summarize the argument well: 7MedinaLF. 2001. Legislatures vs. political parties: endogenous policy with strategic voters. Unpublished manuscript, Dep. Polit. Sci., Univ. Chicago. DEMOCRATIC REGIME SURVIVAL 163 The essence of pure presidentialism is mutual independence. From this defining (and confining) condition a series of incentives and decision rules for encouraging the emergence of minority governments, discouraging the formation of durable coalitions, maximizing legislative impasses, motivating executives to flout the constitution, and stimulating political society to call periodically for military coups predictably flows. Presidents and legislatures are directly elected and have their own fixed mandates. This mutual independence creates the possibility of a political impasse between the chief executive and the legislative body for which there is no constitutionally available impasse-breaking device. Coalition governments, thus, are considered rare and unstable in presidential regimes and frequent and stable in parliamentary regimes. Mainwaring (1993) added the complicating factor of party system fractionalization. Whereas parliamentary regimes are equipped to deal with such situations—cooperation, remember, is inherent to the regime—the problems of presidentialism are only compounded by a multitude of political parties in the legislature. This argument is problematic in many respects. Most fundamentally, it assumes that the institutional differences between the two regimes are sufficient to create divergent incentives for coalition formation. This, however, is not the case. Cheibub et al. (2001) show that the circumstances under which portfolio coalitions are likely to be formed are identical under the two systems. The crucial difference between parliamentarism and presidentialism, they argue, is the "reversion point," that is, the situation that emerges if no coalition is formed. In parliamentary regimes, the reversion point is an early election; in presidential regimes, since mandates are fixed, it is a situation in which the president keeps all the portfolios. The implication is that whereas in parliamentary regimes every portfolio government enjoys the support of a legislative majority, under presidentialism it is possible for a legislative majority to hold no portfolio. This difference, however, is not sufficient to generate different incentives for the formation of portfolio coalitions. Cheibub et al. (2001) show that if there is a large distance in policy space between the party of the formateur and the party closest to it, portfolio coalitions are formed in both parliamentary and presidential systems. In these cases, the formateur uses portfolios to bring the policy closer to its own preferences. Alternatively, when the distance in the policy space between the party of the formateur and the party closest to it is small, and together these two parties would form a majority, no coalition is formed, again in both parliamentary and presidential systems. Because policy preferences are close, the formateur can allow policy to be chosen by a party other than itself. Given that, by definition, mandates are fixed in presidential regimes, what needs to be explained is not so much why coalitions are not formed under presidentialism (which, as we show below, is not true anyway), but why presidents do not always form majority coalitions. According to Cheibub etal. (2001), the reason majority coalitions are notalways formed under presidential regimes has to do with the opposition's beliefs about 164 CHEIBUB ■ LIMONGI how much it would gain electorally from opposing the president. If the opposition believes that its vote share will increase in the next election, it may be willing to stay out of the government, in which case a majority of legislators unite against the president and the president remains in office for the duration of the term. This outcome is structurally unavailable under parliamentarism. What is fundamental, though, is not that this outcome may occur but that, if it occurs, it does not invariably imply impasse or deadlock between the executive and the legislature. Legislative paralysis is likely to occur only under very specific institutional configurations, namely, if the legislature cannot initiate legislation or if the president can veto legislation without being overridden. These are relatively infrequent scenarios: Of 20 democratic constitutions that existed in Latin America, 11 contained no provision regarding exclusive introduction of legislation by the executive, and the rest included partial restrictions—usually involving budget laws and/or the armed forces—on the legislature's ability to initiate legislation (Carey et al. 1997). As for veto, in only 4% of the years of presidentialism between 1946 and 1999 could the president veto legislation with no legislative override (Cheibub 2002). Thus, although possible under some specific circumstances, a generalized lack of cooperation between executive and the legislature or chronic legislative paralysis are not the outcomes that would naturally result from the structure of incentives in presidential regimes. Empirical patterns largely support these considerations. Table 2, based on the work of Cheibub et al. (2001), contains the frequency of majority and coalition governments for parliamentary and presidential regimes according to the share of seats held by the largest party in the legislature. Majority governments are those in which the share of seats held by all parties holding portfolios is >0.50; coalition governments are those in which there are at least two parties that hold portfolios. We can see that, except for the cases in which one party holds more than 50% of the seats in the legislature, the frequency of coalition governments is higher under TABLE 2 Majority and coalition governments in parliamentary and presidential regimes by the share of seats held by the largest party in the legislature Both Parliamentarism Presidentialism Majority government: All 0.7474 0.7871 0.6128 >50% 0.9598 1.0000 0.8736 <50% 0.5531 0.5908 0.2892 33.3%-50% 0.5293 0.5688 0.2778 <33.3% 0.6316 0.6857 0.3239 Coalition government: All 0.4326 0.4314 0.3034 >50% 0.1250 0.0920 0.1348 <50% 0.7141 0.7443 0.5122 33.3%-50% 0.6581 0.6949 0.4306 <33.3% 0.8982 0.9571 0.7606 DEMOCRATIC REGIME SURVIVAL 165 parliamentarism than under presidentialism. The difference, however, is not as large as we would expect if it were true that presidential regimes do not provide any incentives for coalition formation. When no party holds a majority of seats in the legislature, coalition governments emerge slightly over half the time. Furthermore, we find that the frequency of coalition and majority governments actually increases significantly when no party holds more than a third of the seats in the legislature. This pattern is identical to that in parliamentary regimes, suggesting, as discussed above, that the incentives for coalition formation are the same under the two regimes. Thus, contrary to the widespread fear expressed in the existing literature, the fragmentation of the party system does not make coalition formation more difficult in presidential systems. Even if we grant that incentives for coalition formation are the same under parliamentary and presidential regimes, one basic difference between them is that the number of possible coalitions under presidentialism is necessarily smaller than under parliamentarism. When the president is an "outsider" with very little support in congress, the power of his or her party is magnified by the fact that that party must be a member of the government. If this is a frequent occurrence in presidential regimes, then this limitation could become a serious problem. However, it is simply not the case that presidents tend to belong to small parties or be outsiders (Cheibub et al. 2001). The probability that the chief executive will belong to one of the two largest parties is 0.9492 in parliamentary regimes and 0.9279 in presidential regimes. As Figure 1 shows, the distribution of seats 0.2 a is a is 0.14 0.12 I 0.1 f t □I 1 J> J$> & & tA° & & 0 & & $ Figure 1 Parliamentary (light grey) and presidential (dark grey) governments by share of seats of party of chief executive. 166 CHEIBUB ■ LIMONGI held by the party of the president and the prime minster is similar under the two regimes. There is nothing in this distribution suggesting that presidential regimes are more likely than parliamentary regimes to produce governments headed by outsiders. Furthermore, presidential regimes are not considerably more constrained than parliamentary regimes in the process of coalition formation. In both regimes, the government tends to be headed by the largest party, and this political factor, rather than any formal rule, seems to be sufficient to constrain the process of government formation in ways that make the two systems look alike. What are the consequences of coalition governments under the two regimes? If the cycle described by Stepan & Skach (1993) were true, presidential regimes would be less likely to survive when the party of the president does not hold a majority of seats in congress and no coalition is formed. The same should be true of parliamentary regimes. The empirical patterns, however, are complex. As Table 3 indicates, parliamentary regimes do, in fact, have better chances to survive when no party obtains a majority of seats in the legislature and coalitions are formed. Presidential regimes experience the opposite situation: Although the difference is not large (life expectancies of 24.5 versus 22.6 years), the formation of coalition governments reduces survival chances when no party has >50% of the seats in congress. This, however, is not the whole story. This effect is stronger in presiden-tialism when the largest party holds between one third and one half of the seats. (The result is the same if instead of the seats of the largest party we use the seats of the party of the chief executive to condition the probabilities reported in the upper panel of Table 3.) In these cases, the expected lifetime of the regime when coalitions are formed is 13.2 years, compared with an expected lifetime of 24.5 years when no coalitions are formed. When the largest party holds less than one third of the seats, that is, when the legislature is highly fractionalized, coalition formation sharply improves the survival chances of the regime: Expected lifetime with coalition is 54.9 years versus 23.9 when no coalitions are formed. Thus, it seems to be the level of party fractionalization that determines the effect of coalition formation on the survival of presidential democracies. When fractionalization is moderate (for instance, when the effective number of parties is between 3.5 and 4.5, as indicated in the second panel of Table 3), coalition governments are highly fragile: Their expected lifetime is <9 years, compared with 26 years when no coalitions are formed. When fractionalization increases, we find, as expected, that coalition governments increase the survival chances of presidential democracies. We saw above that in situations of high party fractionalization, coalitions are more likely to emerge in both parliamentary and presidential regimes. Now we see that, under these conditions, not only are they more frequent but they also significantly improve the chances that democracy will survive. The reason why coalitions tend to be less frequent and detrimental to the survival of democracy when the party system is moderately fractionalized is puzzling and deserves further investigation. The third panel in Table 3 presents some of the conditions that transpire under moderate and high party fractionalization and may suggest some clues. When fractionalization is moderate, the party of the president DEMOCRATIC REGIME SURVIVAL 167 TABLE 3 Transition probabilities by the share of seats held by the largest party and the effective number of political parties (source: Cheibub et al. 2001) All Parliamentary Presidential Seats held by the largest party: .SB S o §3 — —] * I 2d o -a 'o '£ o D Ö Cd >50% 0.0251 0.0139 0.0475 No coalition 0.0266 0.0137 0.0516 Coalition 0.0143 0.0159 0.0000 <50% 0.0254 0.0191 0.0429 No coalition 0.0301 0.0253 0.0408 Coalition 0.0172 0.0114 0.0443 33.3%-50% 0.0222 0.0131 0.0555 No coalition 0.0307 0.0265 0.0407 Coalition 0.0178 0.0071 0.0753 <33.3% 0.0170 0.0197 0.0224 No coalition 0.0256 0.0000 0.0417 Coalition 0.0161 0.0206 0.0182 Effective number of parties (ENP): ENP < 2.5 0.0267 0.0167 0.0480 No coalition 0.0308 0.0192 0.0532 Coalition 0.0000 0.0000 0.0000 2.5 < ENP < 3.5 0.0170 0.0121 0.0325 No coalition 0.0230 0.0165 0.0291 Coalition 0.0117 0.0087 0.0392 3.5 < ENP < 4.5 0.0292 0.0094 0.0886 No coalition 0.0127 0.0000 0.0385 Coalition 0.0342 0.0121 0.1132 ENP > 4.5 0.0163 0.0140 0.0390 No coalition 0.0278 0.0000 0.0555 Coalition 0.0148 0.0148 0.0339 Conditions under moderate and high party fragmentation (effective parties): 3.5 < ENP < 4.5 ENP > 4.5 Majority government 0.5760 0.5719 Coalition government 0.7690 0.8824 Party of head of government is the largest party 0.6569 0.5085 Share of seats of party of the head of government 0.3361 0.2319 Share of seats of largest party 0.3809 0.2845 Share of seats of second largest party 0.2512 0.2049 Share of seats of third largest party 0.1600 0.1523 Sum of seat share of first, second, and third parties 0.7921 0.6417 Sum of seat share of second and third parties 0.4112 0.3573 168 CHEIBUB ■ LIMONGI is more often the largest party, holds on average about one third of the seats, and is confronted with other legislative parties that are, themselves, relatively large. The relative strengths of legislative parties may be what makes coalition so detrimental to presidential democracies under these conditions. For the moment, however, we simply emphasize that it is not high party fractionalization, and the difficulties of forming majority coalitions that allegedly follow from it, that kills presidential democracies. Thus, it is not true that incentives for coalition formation are different in presidential and parliamentary democracies. It is not true either that presidential regimes with highly fractionalized party systems make the task of coalition formation even more daunting; to the contrary, the frequency of coalition governments increases with the fractionalization of the party system in both presidential and parliamentary regimes. CENTRALIZATION OF THE DECISION-MAKING PROCESS From the constitutional point of view, all legislators, whether in presidential or parliamentary regimes, have the same rights and duties. Their mandates are the same, regardless of the number of votes they received in the electorate, their party affiliation, their degree of seniority in the legislature, and so on. Each legislator has the same right to propose legislation, to amend proposals made by others, and to participate in the process of deliberation. In addition, each legislator's vote has the same weight. In principle, therefore, legislatures are egalitarian institutions. The reality, of course, is different. In order to manage their workload, legislatures organize themselves in a variety of ways and adopt internal rules that regulate individual legislative rights and access to resources (Kriebhel 1992, p. 2). Legislative rights and resources are not uniformly distributed. Individual legislators' chances to influence the order of business and to have a say in decision making depend on the legislative rights granted to them by the internal rules of their assembly. Hence, legislative organization affects the structure of the decision-making process and the weight of legislators in policy decisions. Discussions of legislative organization usually make reference to the paradigmatic cases of England and the United States: a centralized and a decentralized legislature, respectively, and, as we know, a parliamentary and a presidential democracy. Most arguments that we find about decision making in democracies contrast these two systems and assume, often implicitly, that all legislatures, and, for that matter, the decision-making process, are centralized under parliamentarism and decentralized under presidentialism. The English parliament is indeed characterized by the complete control of the cabinet over the legislative agenda. Government bills are appreciated under a special calendar that gives them priority over bills introduced by individual members of parliament and, as a consequence, parliamentary minorities have no way to "close the gates" to governmental proposals. In addition, individual members of DEMOCRATIC REGIME SURVIVAL 169 parliament are often restricted in their capacity to amend government bills. For instance, since the beginning of the eighteenth century the government has had the sole prerogative to initiate measures that increase expenditures (Lowemberg & Patterson 1979, p. 249). Nowadays, it is rare for the budget presented by the cabinet to be modified by the parliament. In fact, given the high expectations that it will be approved as submitted, "a provisional resolution places it into effect on the day it is delivered, though months may pass before its final enactment" (Lowemberg & Patterson 1979, p. 250). Because of the government's control over the agenda, legislative output is marked by a high rate of success for the executive's initiatives. Proposals made by the cabinet had a 0.97 chance of being approved during the 1945-1978 period, whereas bills introduced by back-benchers, irrespective of their party affiliation, had a close-to-zero chance of being approved (Rose 1986, p. 11). This means that the cabinet introduces almost all laws that are approved in parliament. The government legislative success rests on disciplined party support. Cabinet defeats are rare events. The cabinet entirely monopolizes the law-making process and, for that matter, all the decisions about policy. The US congress, in contrast, is supposed to be a decentralized body, organized as it is around a strong committee system. In this view, the committee system allows legislators to have a say in decisions related to policy areas that are important to their electoral survival. The story goes like this. Each committee has a monopoly on initiating legislation in its own policy jurisdiction. The committee reports a bill to the floor, and for reasons that are not entirely clear, the floor accepts the bill as reported [see the debate between Kriebhel (1987) and Shepsle & Weingast (1987)]. Political parties do not control the assignment of legislators to specific committees; this is described as a process of self-selection in which legislators pick the committee that has jurisdiction over the policy area that will bring them the highest electoral payoff. At the same time, electoral considerations dictate that politicians prefer distributive, pork-barrel policies. The committee system in the US congress provides the organizational means to make these distributive policies possible. This textbook (Shepsle 1989) view of the US congress has been challenged by Kriebhel (1992) and Cox & McCubbins (1993), to cite only the most important works. Yet, it is true that even in the informational and the party-as-a-cartel views, the committee system is a key and distinctive feature of the US congress. Hence, with England as the prototype of executive-legislative relations in parliamentary regimes and the United States as the prototypical presidential system, it follows that "[i]n parliamentary systems the executive (government) controls the agenda, and the legislature (parliament) accepts or rejects proposals, while in presidential systems the legislature makes the proposal and the executive (the president) signs or vetoes them" (Tsebelis 1995, p. 325). The prototypical parliamentary regime is one in which the government has complete control over the legislative agenda; the rights of the individual members of parliament are "expropriated" and monopolized by the cabinet. Hence, all individual legislators can do 170 CHEIBUB ■ LIMONGI is to support the party line. Voters know that this is all they can do, and thus have no incentive to cast their ballot on the basis of the candidate's personal characteristics. Through the control of the legislative agenda, therefore, parliamentarism would lead to party vote in the electorate and to party discipline in the parliament.8 The prototype of a presidential regime, in turn, is one in which the organization of congress preserves the right of individual representatives to have a say on policy decisions. Separation of powers leads to independent legislators who act on the basis of their individual electoral needs, and in response to these needs, they build personal ties with their constituencies. One is thus led to expect that legislatures in presidential regimes will have strong committee systems and representatives will be elected on the basis of the personal ties they build with their constituencies. But, as we have known at least since Shugart & Carey published Presidents and Assemblies (1992), presidential systems vary considerably in the degree of legislative powers they grant the president. The US president, as a matter of fact, is one of the few presidents in existing systems who cannot initiate legislation. Moreover, the US president only has the package veto, which weakens his capacity to oppose distributive bargains produced in congress. Hence, the US presidency is unique in that the president has only "reactive legislative powers" (Mainwaring & Shugart 1997b). Not only are presidents often endowed with the capacity to initiate legislation; often they also have the exclusive right to initiate legislation in some areas (such as appropriation and budgetary matters), whereas legislators are restricted in their capacity to amend bills in these areas. Hence, contrary to Tsebelis (1995), presidents can do much more than simply sign or veto bills proposed by the legislature. In addition, some presidents also have decree power; they are constitutionally able to unilaterally alter the status quo. Although there is considerable variation in the specifics (Carey & Shugart 1998), often presidential decrees enter into effect first and the legislature acts second. The legislature acts a posteriori, rejecting, amending, or accepting the new status quo brought about by the executive decree.9 A president with decree power can dictate the legislative agenda by forcing the legislature to make a decision on some matter it could not have appreciated 8High rates of executive success and low participation of individual members in law making are, indeed, characteristic of most parliamentary regimes. The data assembled by the Inter Parliamentary Union (Herman & Mendel 1976), covering 14 countries with parliamentary regimes during the 1971-1976 period, register only 3 in which government legislative success is below 80%. There is no case in which individual initiatives represent more than 20% of the laws passed. For three countries (Australia, Ireland, and Malta), there is no case of a bill introduced by an individual legislator that became a law. For the period 1978-1982, Herman & Mendel (1986) register 3 out of 16 countries with government success below 80% and only 3 with individual members' initiatives above 20% (Austria, Italy, and Portugal). 'Often, rejection of a presidential decree does not mean a return to the status quo ante. Even if there is a majority in favor of the status quo, once the decree has been in effect, rejecting it may have become an unattractive or unavailable alternative. DEMOCRATIC REGIME SURVIVAL 171 otherwise. Thus, no group in the legislature, not even the majority, can "close the gates" against a presidential initiative made by decree. Note that the power to impose an agenda does not imply that presidents always prevail against the will of the majority. In fact, since a legislative majority can always reject a presidential decree, a model of executive-legislative conflict cannot explain why the executive would ever make use of decrees (Huber 1996a). There are, of course, strategic advantages that the agenda setter may explore. But as Kriebhel (1988, p. 270) has argued, these are not properly antimajoritarian devices.10 In addition, the government's legislative and agenda powers, including decree power, need not be interpreted solely as means for solving "vertical" conflicts, that is, conflicts between the government and the opposition. The government's legislative powers are also means for solving "horizontal" conflicts, that is, conflicts between the government and its supporters (Huber 1996a). These powers enable the government to protect the cohesion of its coalition against the opportunistic behavior of its members.11 It follows from this that, because of presidents' legislative powers, separation of powers in presidential regimes is not as complete as it is usually considered to be. Presidential legislative powers are commonly interpreted in the context of the US constitution, that is, as means to create checks and balances. But, as we have just seen, the legislative powers of the executive are not only a mechanism for checking the power of the majority or imposing the will of the president. They are also weapons of the majority. Therefore, the fusion of executive and legislative powers is not absent from presidential systems. This interpretation is at odds with Shugart & Carey's (1992) view, according to which presidential systems that endow presidents with considerable legislative powers—creating what Shugart & Carey call strong presidents—have a greater probability of breaking down. They argue that strong presidents have smaller incentives to negotiate with congress, making paralysis and crisis more likely. This is so because strong presidents have the institutional means to impose their will on congress, whereas weak presidents know that they have no alternative but to negotiate. This argument, however, is based on the "vertical conflict" model and disregards the possibility that the president will try to organize a majority in congress. Once the possibility that the president and the majority have overlapping 10It is true that, since a decree immediately alters the status quo, decree power increases the power of the agenda setter. When legislatures vote on ordinary propositions, legislators compare the status quo (SQ) with the situation to be created by the proposition. In the case of an executive decree, the legislator compares the situation created by the decree (D) with the new situation created by rejecting a decree that has been in effect for some time (SQD). If the preferences of the majority are SQ > D > SQD, then the majority will approve the decree. If the preferences of the majority are SQD > D (assuming that SQ is no longer a viable alternative), then the majority will reject the decree. "There seems to be no association between minority status and the use of decree power (Figueiredo & Limongi 1998). 172 CHEIBUB ■ LIMONGI preferences is considered, then legislative agenda powers need not imply paralysis, crisis, and eventual breakdown. The organization of congress and the degree of control the executive has over the legislative agenda does influence the behavior of individual legislators. They act in a constrained environment. If they want to influence policy, they have to do so according to the procedural rules of their legislative body and the terms set by the president. For example, the incentives to cultivate the personal vote that stem from the electoral arena may be entirely neutralized in the legislature through a distribution of legislative rights that favors the executive. For this reason, we cannot deduce, as is commonly done, the behavior of legislators from electoral and party legislation alone. In this context, the case of Brazil is of central theoretical interest, for it demonstrates the far-reaching effects of the centralization of the decision-making process. The system produced by the 1988 constitution is frequently cited as a prime example of bad institutional design (Ames 2001). All of the institutional choices that should not be made, it seems, were made in 1988: a strong presidential regime [ranked among the strongest in the world by Shugart & Carey (1992, p. 155)]; a proportional representation formula for legislative elections with high district magnitude; very permissive party and electoral legislation (e.g., open-list and low party control over access to the ballot). Under such conditions, the party system is bound to be fragmented and presidents can be virtually certain that their party will not control a majority of seats in both legislative houses. Even if it did, parties would be highly undisciplined, making the majority status of the president a mere formality (Sartori 1994, p. 113; Mainwaring 1991). Hence, to have their agenda approved, presidents would use their strong legislative powers, which would lead to conflict and paralysis. To paraphrase Sartori (1997), the system created in 1988 was nothing but hopeless. Yet, the performance of the post-1988 Brazilian regime is completely at odds with what we would expect. Brazilian presidents of this period have had great success enacting their legislative agenda. Presidents introduced 86% of the bills enacted since 1988, and the rate of approval of the bills introduced by the executive was 78%. Presidents have formed coalitions to govern and have reliably obtained the support of the parties that belong to the government coalition in approving its legislation; the average discipline of the presidential coalition, defined as the act of voting in accordance with the public recommendation of the government leader in the floor, was 85.6%. This support is sufficient to make a presidential defeat in a roll call a rare event. Thus, despite the "centrifugal" characteristics of Brazilian presidentialism, as indicated by the party and electoral legislation, presidents have governed relying on the support of a disciplined coalition (Figueiredo & Limongi 2000). This outcome results from both the organization of the Brazilian congress and the president's control of the legislative agenda. The Brazilian congress is highly centralized. Legislative rights heavily favor party leaders, who are taken to be perfect agents of their caucuses (bancadas) regarding most procedural decisions, such as the request for roll-call votes, the closing of debates, and, most important, DEMOCRATIC REGIME SURVIVAL 173 the designation of a bill as urgent for purposes of appreciation. The urgency request is a kind of discharge petition: it removes the bill from the committee and forces its immediate (within 24 hours) deliberation by the floor. Bills that are appreciated as urgent cannot be freely amended; only amendments signed by 20% of the lower house are accepted, which implies that only amendments supported by party leaders will be considered. As Figueiredo & Limongi (2000, p. 157) have shown, the approval of the urgency petition is highly associated with the success of a bill. Centralization, thus, deprives members of congress of the legislative rights they need in order to influence legislation. The Brazilian presidents, thanks to their constitutional legislative powers, have a direct influence on the definition of the legislative agenda. Using its decree power, the executive places on the agenda the issues it deems most relevant and pressing. The president can also influence the pace of ordinary legislation by requesting urgency for the appreciation of specific bills (which will give each house 45 days to deliberate on them). The president also has the exclusive right to initiate legislation related to the definition of the budget, taxation, and public administration. Therefore, the executive monopolizes the legislative initiative on the most crucial areas of policy making. Hence, it is through participation in the government that individual legislators obtain access to resources they need for political survival: policy influence and patronage. Leaders bargain with the executive, exchanging political support (votes) for access to policy influence and patronage. The executive thus provides party leaders with the means to punish back-benchers who do not follow the party line: Their share of patronage may be denied. The executive, in turn, given the resources it controls, is in a very advantageous position. Party leaders become, in fact, the main brokers in the bargaining between the executive and the legislators. Contrary to what is currently assumed about Brazil, presidents do not need to bargain on a case-by-case basis. They are in a position to demand support for their entire legislative agenda. Once the government is formed and benefits are distributed among the members of the coalition, the president, with the help of party leaders, may threaten representatives and actually punish those who do not follow the party line. To say it once more, the actual pattern of legislative-executive relations in Brazil's presidential regime is rather different from what one would expect if one deduced it from electoral and partisan legislation. It should be clear by now that separation of powers does not necessarily imply decentralized decision making. Institutional analyses that stress the negative effects of separation of powers, and that point to specific, often restrictive, electoral laws as a corrective for these effects, miss the point. Presidentialism does not necessarily imply, or require, decentralized decision making and conflict between the executive and the legislature. Once one grants the possibility that coalition governments exist in presidential regimes, the degree of overlap between the executive and legislative majorities has to be adjusted. Presidential control over the agenda becomes a weapon to be used by the majority rather than against the majority. Thus, presidents are not necessarily as distinct from prime ministers as is normally assumed. As we showed above, 174 CHEIBUB ■ LIMONGI outcomes that are usually associated exclusively with parliamentarism, such as executive success and dominance over the legislative output obtained through disciplined parties, can be found even in "hopeless" presidential regimes such as Brazil's. Now, just as presidential regimes are not all alike, neither are parliamentary systems. Government control over the legislative agenda does not follow from the definition of parliamentarism. Neither is it necessary that the legislative rights of private members be curtailed in parliamentary regimes. Committees may have considerable powers in parliamentary assemblies and may erect barriers to the executive agenda.12 The weakness of individual members of parliament that characterizes England is not inherent to parliamentary governments, as illustrated by the cases of Italy after 1945 and France in the Third and Fourth Republics. In both cases, the government had no control over the definition of the legislative agenda, committees had considerable power, and the rights of individual legislators were not "expropriated." In France, until 1911, it was the Chamber presidents who defined the legislative agenda. As Andrews (1978) reports, after this date, a Conference of Presidents assumed control over the definition of the agenda. The government was represented in the Conference, but it was only in 1955 that internal rules were revised so that voting in the Conference of Presidents was weighted by the proportion of seats held by each party. The proposed agendas had to be approved by the Chamber, and this "often became an occasion for a vote of non-confidence through a device called "interpellation"' (Andrews 1978, p. 471). Hence, the government did not have firm control over the definition of the legislative agenda. On several occasions, interpellation led to judgment on the government agenda that caused the fall of the government. Besides, committees could act as "veto players," since a report from the committee was necessary for consideration of a bill by the floor. The government could expedite the committee report but could not avoid it. Therefore, committees could respond to government pressure with an unsatisfactory report. According to Andrews (1978), the Third and Fourth Republics placed few restrictions on the ability of private members to propose initiatives that would increase expenditures and reduce revenues. In his words, given the absence of serious restrictions, the government's financial projects were often "butchered in parliament" (Andrews 1978, p. 485). In Italy one finds the same pattern: the parliament's independence to set the legislative agenda, strong committees, and legislative rights that favor individual members' influence over decisions. In the Italian parliament, the presidents of each house, and not the government, define the legislative agenda. Bills introduced by the government have no special calendar or precedence over private members 'bills. Article 72 of the Italian constitution grants standing committees the authority to 12One of the variables used by Strom (1990) to explain the formation of minority governments in parliamentary regimes was the committee structure of parliaments. Obviously there must exist some variation in this structure. DEMOCRATIC REGIME SURVIVAL 175 pass laws. This capacity has been widely used. According to Di Palma (1976), the presidents of both houses decide unilaterally whether a bill will have to be considered by the floor. Di Palma labels these alternatives the centralized and the decentralized procedures. Bills scheduled for the decentralized procedure have a much greater chance of becoming laws. Hence, all a president must do in order to "kill" a governmental proposition is to schedule it for the centralized procedure. Besides being endowed with the power to pass legislation, Italian committees cannot be discharged when a bill follows the decentralized procedure. Although possible, discharges under the centralized procedure are rarely enforced. The committee chairmen are autonomous in defining their agendas and even in convening their committees. Therefore, committees not only are important decision-making bodies but also can act as veto players. As for individual members of parliament, until the 1988 reform, roll calls were secret and could be easily requested at any stage of the law-making process (Cotta 1990, p. 77). Hence, the government fell prey to the action of the franco attiratori. In other words, members of the majority could not be sanctioned, either by the government or their parties. These are obviously not examples of parliamentary regimes performing at their best. In fact, both systems are often cited as examples of pathological development of parliamentarism, and both have been considerably reformed. But this only emphasizes our point: The instability of these systems resulted not from the form of government but from the way decision making was organized. Although policy performance is important for the survival of a democratic regime, we cannot deduce it from the basic constitutional principle that defines the regime. Policy making under parliamentarism is not necessarily centralized, and consequently the government is not always successful in having its policy proposals approved. Similarly, policy making under presidentialism is not necessarily decentralized, and the government is not invariably immobilized in terms of the policies it can implement. CONCLUSION The difference between parliamentary and presidential democracies does not seem to attract the same attention today as it did 10 or 15 years ago. Then, as several new democracies were choosing their constitutional frameworks, there was a clear sense of urgency in discussions about which system was more likely to survive. Today, the scholarly imagination has been captured by more immediate concerns (it seems to have rested on federalism and corruption, at least momentarily), and an implicit consensus has emerged to the effect that political scientists know everything that needs to be known about "broad" constitutional choices. We disagree with this position. We hope this paper has shown that the difference in the survival of parliamentary and presidential regimes cannot be explained by the structure of incentives that supposedly follows from the regime's basic principles. Parliamentary systems do not operate under a "majoritarian imperative"; deadlock is not as frequent as supposed under presidentialism and is not absent from parliamentarism; coalition governments are not foreign to presidential systems and emerge for the same reasons as they do in parliamentary systems; decision making 176 CHEIBUB ■ LIMONGI is not always centralized under parliamentarism and is not always decentralized under presidentialism. The reality of both parliamentary and presidential regimes is more complex than it would be if we derived these systems' entire behavior from their first principles. So what explains the difference? We suspect that the main difference between the two regimes lies in the way the decision-making process is organized. Lack of coordination on policy making and the probability of deadlock increase with the decentralization of the decision-making process. More often than not, we find parliamentary regimes that have a centralized decision-making process, one in which the executive has a monopoly on the policy agenda. However, as the Italian and French cases demonstrate, this is not a necessary feature of parliamentarism. And as the case of Brazil demonstrates, executive-legislative powers, in the presence of political parties, allow for a fusion of powers not predicted by the usual conception of presidential regimes. Presidents with active legislative powers need not impose their will on the congress; they can bargain from a very advantageous position with legislative majorities that can encompass more than one political party. Agenda powers that centralize the decision-making process may be the basis for the "efficient secret" (Cox 1987) of presidential regimes. Thus, if parliamentary regimes have a better record of survival than presidential regimes, it is not because they are parliamentary. ACKNOWLEDGMENTS We thank Adam Przeworski, Tasos Kalandrakis, and especially Argelina Cheibub Figueiredo, who has participated in many of the conversations that led to this paper. We also thank the Leitner Program in International Political Economy at Yale University for support for this research and the Fundacao de Pesquisa e Amparo a Pesquisa do Estado de Sao Paulo (FAPESP) for providing the conditions for us to work on this paper together. The Annual Review of Political Science is online at http://polisci.annualreviews.org LITERATURE CITED Ackerman B. 2000. The New Separation of Powers. Harvard Law Rev. 113(3):642-727 Ames B. 2001. The Deadlock of Democracy in Brazil. Ann Arbor: Mich. Univ. Press Andrews WG. 1978. Parliamentary procedures in France. Legis. Stud. Q. 3:465-506 Baron DP. 1998. Comparative dynamics of parliamentary government. Am. Polit. Sci. Rev. 92:593-610 Bowler S, Farrell DM, Katz RS, eds. 1999a. Party Discipline and Parliamentary Government. Columbus: Ohio State Univ. Bowler S, Farrell DM, Katz RS. 1999b. Party cohesion, party discipline, and parliaments. See Bowler et al. 1999, pp. 3-22 Carey JM, Amorin Neto O, Shugart MS. 1997. Appendix: outlines of constitutional powers in Latin America. SeeMainwaring & Shugart 1997b, pp. 440-60 Carey JM, Shugart MS. 1994. Incentives to cultivate a personal vote: a rank ordering of DEMOCRATIC REGIME SURVIVAL 177 electoral formulas. Elec. Stud. 14:417-39 Carey JM, Shugart MS, eds. 1998. Executive Decree Authority. Cambridge, UK: Cambridge Univ. Press Cheibub JA. 1998. Elections and alternation in power in democratic regimes. Presented at Annu. Meet. Am. Polit. Sei. Assoc., Boston, Sept. 3-6 Cheibub JA. 2002. Minority governments, deadlock situations, and the survival of presidential democracies. Comp. Polit. Stud. In press Cheibub JA, Przeworski A. 1999. Democracy, elections, and accountability for economic outcomes. In Democracy and Accountability, ed. A Przeworski, SC Stokes, B Manin, pp. 222-50. New York: Cambridge Univ. Press Cheibub JA, Przeworski A, Saiegh S. 2001. Government coalitions under presidentialism and parliamentarism. Presented at Conf. Brazilian Political Institutions in Comparative Perspective: The Role of Congress in Presidential Systems, Cent. Brazilian Stud., Oxford Univ., May 28-29 Cotta M. 1990. The 'centrality' of parliament in a protracted democratic consolidation. In Parliament and Democratic Consolidation in Southern Europe, ed. U Liebert, M Cotta, pp. 55-91. London: Pinter Cox GW. 1987. The Efficient Secret. Cambridge, UK: Cambridge Univ. Press Cox GW, McCubbins MD. 1993. Legislative Leviathan: Party Government in the House. Berkeley: Univ. Calif. Press Di Palma G. 1976. Institutional rules and legislative outcomes in the Italian parliament. Legis. Stud. Q. 1:147-79 Diermeier D, Feddersen TJ. 1998. Cohesion in legislatures and the vote of confidence procedure. Am. Polit. Sei. Rev. 92:611-22 Figueiredo AC, Limongi F. 1998. Institutional legacies and accountability: executive decrees in Brazil and Italy. Presented at Conf. Confronting Non-Democratic Legacies During Democratic Deepening: Latin America and Southern Europe. Buenos Aires, Arg., Aug. 17-19 Figueiredo AC, Limongi F. 2000. Presidential power, legislative organization and party behavior in the legislature. Comp. Polit. 32: 151-70 Gonzalez LE, Gillespie CG. 1994. Presidentialism and democratic stability in Uruguay. In The Failure of Presidential Democracy: The Case of Latin America, ed. JJ Linz, A Valen-zuela, 2:151-78. Baltimore, MD: Johns Hopkins Univ. Press Hartlyn J. 1994. Presidentialism and Colombian politics. In The Failure of Presidential Democracy: The Case of Latin America, ed. JJ Linz, A Valenzuela, 2:220-53. Baltimore, MD: Johns Hopkins Univ. Press Herman V, Mendel F. 1976. Parliaments of the World. A Reference Compendium. London: Inter-Parliamentary Union/Macmillan Herman V, Mendel F. 1986. Parliaments of the World. A Reference Compendium. Berlin/ New York: Inter-Parliamentary Union/De Gruyter Huang T. 1997. Party systems in Taiwan and South Korea. In Consolidating the Third Wave Democracies: Themes and Perspectives, ed. L Diamond, MF Plattner, Y Chu, H Tien, pp. 135-59. Baltimore, MD: Johns Hopkins Univ. Press Huber JD. 1996a. Rationalizating Parliament. Cambridge, UK: Cambridge Univ. Press Huber JD. 1996b. The vote of confidence in parliamentary democracies. Am. Polit. Sei. Rev. 90:269-82 Jones MP. 1995. Electoral Laws and the Survival of Presidential Democracies. Notre Dame, IN: Notre Dame Univ. Press Kriebhel K. 1987. Why are congressional committees powerful? Am. Polit. Sei. Rev. 81: 929-35 Kriebhel K. 1988. Spatial models of legislative choice. Legis. Stud. Q. 8:259-319 Kriebhel K. 1992. Constituency characteristics and legislative preferences. Public Choice 76:21-37 Limongi F, Figueiredo AC. 1995. Partidos 178 CHEIBUB ■ LIMONGI polfticos na cämara dos deputados: 1989— 1994. Dados 38:497-543 Linz JJ. 1978. The Breakdown of Democratic Regimes: Crisis, Breakdown, and Reequi-libration. Baltimore, MD: Johns Hopkins Univ. Press Linz JJ. 1994. Presidential or parliamentary democracy: Does it make a difference? See Linz & Valenzuela 1994, pp. 3-87 Linz JJ, Stepan A. 1996. Problems of Democratic Transition and Consolidation: Southern Europe, South America, and Post-Communist Europe. Baltimore, MD: Johns Hopkins Univ. Press Lowemberg G, Patterson SC. 1979. Comparing Legislatures. Boston: Little, Brown Mainwaring S. 1991. Politicians, parties, and electoral systems: Brazil in comparative perspective. Comp. Polit. 24:21^3 Mainwaring S. 1993. Presidentialism, multipar-tism, and democracy: the difficult combination. Comp. Polit. Stud. 26:198-228 Mainwaring S, Scully TR. 1995. Introduction: party systems in Latin America. In Building Democratic Institutions: Party Systems in Latin America, ed. S Mainwaring, TR Scully, pp. 1-32. Stanford, CA: Stanford Univ. Press Mainwaring S, Shugart MS. 1997a. Conclusion: presidentialism and the party system. See Mainwaring & Shugart 1997b, pp. 394-439 Mainwaring S, Shugart MS, eds. 1997b. Presidentialism and Democracy in Latin America. Cambridge, UK: Cambridge Univ. Press Mayhew D. 1974. Congress: The Electoral Connection. New Haven, CT: Yale Univ. Press Mershon C. 1996. The costs of coalition: coalition theories and Italian governments. Am. Polit. Sei. Rev. 90(3):534-54 Mershon C. 1999. The costs of coalition: a five-nation comparison. See Bowler et al. 1999, pp.227-68 Nino CS. 1996. Hyperpresidentialism and constitutional reform in Argentina. In Institutional Design in New Democracies: Eastern Europe and Latin America, ed. A Lijphart, CH Waisman, pp. 161-74. Boulder, CO: Westview O'Donnell G. 1994. Delegative democracy. /. Democr. 5:55-69 Ozbudun E. 1970. Party Cohesion in Western Democracies: A Causal Analysis. Beverly Hills, CA: Sage Przeworski A, Alvarez M, Cheibub JA, Limongi F. 2000. Democracy and Development: Political Institutions and Economic Performance, 1950-1990. Cambridge, UK: Cambridge Univ. Press Przeworski A, Alvarez M, Cheibub JA, Limongi F. 2001. Regime and development database, http://pantheon.yale.edu/~jac236/ Rose R. 1986. British MPs: more bark than bite? In Parliaments and Parliamentarians in Democratic Politics, ed. EN Suleiman. New York: Holmes & Meier Sartori G. 1994. Neither presidentialism nor parliamentarism. See Linz & Valenzuela 1994, pp. 106-18 Sartori G. 1997. Comparative Constitutional Engineering. An Inquiry Into Structures, Incentives and Outcomes. New York: New York Univ. Press Shepsle K. 1989. The changing textbook congress. In Can the Government Govern?, ed. JE Chubb, PE Patterson, pp. 238-66. Washington, DC: Brookings Shepsle K, Weingast B. 1987. Rejoinder—Why are congressional committees powerful? Am. Polit. Sei. Rev. 81:935^4-5 Shugart MS. 1995. The electoral cycle and institutional sources of divided presidential government. Am. Polit. Sei. Rev. 89:327-43 Shugart MS, Carey JM. 1992. Presidents and Assemblies: Constitutional Design and Electoral Dynamics. Cambridge, UK: Cambridge Univ. Press Smith A. 1996. Endogenous election timing in majoritarian parliamentary systems. Econ. Polit. 8:85-110 Stepan A, Skach C. 1993. Constitutional frameworks and democratic consolidation: parliamentarism versus presidentialism. World Polit. 46:1-22 DEMOCRATIC REGIME SURVIVAL 179 Strom K. 1990. Minority Government and Majority Rule. Cambridge, UK: Cambridge Univ. Press Tsebelis G. 1995. Decision making in political systems: veto players in presidentialism, parliamentarism, multicameralism and mul-tipartyism. Br. J. Polit. Sci. 25:289-325 Valenzuela A. 1994. Party politics and the crisis of presidentialism in Chile: a proposal for a parliamentary form of government. In The Failure of Presidential Democracy: The Case of Latin America, ed. JJ Linz, A Valenzuela, 2:91-150. Baltimore, MD: Johns Hopkins Univ. Press LiXs A nnual Review of Political Science Volume 5, 2002 Contents Bargaining Theory and International Conflict, Robert Powell l Experimental Methods in Political Science, Rose McDermott 31 Politics, Political Science, and Urban Governance: a Literature and a Legacy, Russell D. Murphy 63 democracy and taxation, Andrew C. Gould and Peter J. Baker 87 Forecasting for Policy Making in the Post-Cold War Period, Stanley A. Feder 111 The Origins, Development, and Possible Decline of the modern state, Hendrik Spruyt 127 Democratic Institutions and Regime Survival: Parliamentary and Presidential Democracies Reconsidered, Jose Antonio Cheibub and Fernando Limongi 151 Political Cleavages and Post-Communist Politics, Stephen Whitefield 181 Of Waves and Ripples: Democracy and Political Change in africa in the 1990S, Clark C. Gibson 201 How Conceptual Problems Migrate: Rational Choice, Interpretation, and the Hazards of Pluralism, James Johnson 223 The News Media as Political Institutions, Michael Schudson 249 The First Decade of Post-Communist Elections and Voting: What Have We Studied, and How Have We Studied It?, Joshua A. Tucker 271 Corporatism: The Past, Present, and Future of a Concept, Oscar Molina and Martin Rhodes 305 Landmarks in the Study of Congress since 1945, Nelson w. Poisby and Eric Schickler 333 Electoral and Partisan Cycles in Economic Policies and outcomes, Robert J. Franzese, Jr. 369 Toward a New Political Methodology: Microfoundations and art, Christopher H. Achen 423 viii contents Indexes Subject Index 451 Cumulative Index of Contributing Authors, Volumes 1-5 469 Cumulative Index of Chapter Titles, Volumes 1-5 471 Errata An online log of corrections to The Annual Review of Political Science chapters (if any have yet been occasioned, 1997 to the present) may be found at http://polisci.annualreviews.org/