WHY DO PARTIES COOPERATE IN PRESIDENTIALISM? ELECTORAL AND GOVERNMENT COALITION FORMATION IN LATIN AMERICA

I. WHY DO PARTIES COOPERATE IN PRESIDENTIALISM?[Up]

The purpose of this article is to explore coalition formation in presidential systems using evidence from Latin America. The conventional view has been that it is considerably more difficult to form coalitions in presidentialism than it is in parliamentarism, and as a result, parties have only exceptionally been expected to form coalitions in the former. However, since the 1990s, most of the democracies in the region have at some point elected a presidential candidate backed by an electoral coalition or have been ruled by a president supported by a government coalition. Only in the past decade did coalitions win in Argentina (2011, 2015), Bolivia (2014), Brazil (2010, 2014), Chile (2013, 2017), Ecuador (2013, 2017), Guatemala (2008), Panama (2009), Paraguay (2008, 2013), and Peru (2016). So, why do parties form coalitions despite institutional constraints?

This puzzle has not yet been resolved by literature. There is still a discrepancy between the conventional view and the actual behavior of coalitions. The gap is in part a consequence of a comparative research program set forward by Juan Linz in the early 1990s, in which he compared the “virtues of parliamentarism” to the “perils of presidentialism” ( ‍Linz, J. J. (1990a). The perils of presidentialism. Journal of Democracy, 1 (1), 51-69.Linz, 1990a;  ‍Linz, J. J. (1990b). The virtues of parliamentarism. Journal of Democracy, 1 (4), 84-91.Linz 1990b). In his landmark work, Linz looked at both types of regimes, comparing their essential characteristics, such as the division of powers and term lengths. Among other things, he found that parliamentary systems were more effective than presidential systems when it came to the survival of democracy. Linz’s findings would go on to spark two additional waves of literature comparing parliamentary democracies to presidential ones ( ‍Elgie, R. (2005). From Linz to Tsebelis: three waves of presidential/parliamentary studies? Democratization, 12 (1), 106-122. Available at: https://doi.org/10.1080/1351034042000317989Elgie, 2005).

The second wave was triggered by the critiques of David Horowitz ( ‍Horowitz, D. L. (1990). Comparing democratic systems. Journal of Democracy, 1 (4), 73-79. Available at: https://doi.org/10.1353/jod.1990.00561990), whose main argument was that Linz had mistakenly based his inferences on a “highly selective sample of comparative experiences” (p. 74). Matthew Shugart and John Carey ( ‍Shugart, M. S. and Carey, J. M. (1992). Presidents and assemblies: Constitutional design and electoral dynamics. Cambridge: Cambridge University Press. Available at: https://doi.org/10.1017/CBO97811391739881992) built on this and criticized that the notion that parliamentarism was superior had been because they were “considering only a dichotomous classification of regimes” (p. 283). This view was complemented by Scott Mainwaring and Timothy Scully ( ‍Mainwaring, S. and Scully, T. (eds.) (1995). Building democratic institutions: Party systems in Latin America. Stanford: Stanford University Press.1995), who suggested that the consequences of presidentialism inherently vary from one democracy to the next, and by Mainwaring and Shugart ( ‍Mainwaring, S. and Scully, T. and Shugart, M. S. (1997a). Juan Linz, presidentialism, and democracy: A critical appraisal. Comparative Politics, 29 (4), 449-471. Available at: https://doi.org/10.2307/4220141997a), who suggested that presidentialism functions relatively well when presidents have weak legislative power and parties are moderately disciplined.

In the third wave, Robert Elgie ( ‍Elgie, R. (2005). From Linz to Tsebelis: three waves of presidential/parliamentary studies? Democratization, 12 (1), 106-122. Available at: https://doi.org/10.1080/13510340420003179892005) argued that the comparison should be guided by more general theories of political science. In his view, the research objective should be to study and compare the balance of power of political actors within both parliamentarism and presidentialism. In this vein, George Tsebelis and Kaare Strøm applied veto player and principal-agent theories to explore power-sharing relations within each type of regime. While Tsebelis ( ‍Tsebelis, G. (1995). Decision making in political systems: Veto players in presidentialism, parliamentarism, multicameralism and multipartyism. British Journal of Political Science, 25 (3), 289-325. Available at: https://doi.org/10.1017/S00071234000072251995;  ‍Tsebelis, G. (2002). Veto players: How political institutions work. New York: Princeton University Press. Available at: https://doi.org/10.1515/97814008314562002) argued that the multi-party setting in parliamentarism and the executive-legislative balance in presidentialism were what shaped partisan strategies, Strøm ( ‍Strøm, K. (2000). Delegation and accountability in parliamentary democracies. European Journal of Political Research, 37 (3), 261-290. Available at: https://doi.org/10.1111/1475-6765.005132000) argued that the chain of delegation was the decisive factor behind the motivation of political parties.

These waves are highly relevant to understand the gap in literature, since one common finding was that parties that would normally form coalitions in parliamentarism would only exceptionally form coalitions in presidentialism ( ‍Linz, J. J. and Valenzuela, A. (eds.) (1994). The failure of presidential democracy. Baltimore: The Johns Hopkins University Press.Linz and Valenzuela, 1994: 19). Indeed, academics across all three waves concurred that the institutions found in parliamentary regimes generated stronger incentives for parties to cooperate with one another than those found in presidential regimes (see  ‍Linz, J. J. and Stepan, A. (1996). Problems of democratic transition and consolidation: Southern Europe, South America, and post-communist Europe. Baltimore: The Johns Hopkins University Press.Linz and Stepan, 1996: 181;  ‍Mainwaring, S. (1990). Presidentialism in Latin America. Latin American Research Review, 25 (1), 157-179.Mainwaring, 1990;  ‍Mainwaring, S. and Scully, T. (eds.) (1995). Building democratic institutions: Party systems in Latin America. Stanford: Stanford University Press.Mainwaring and Scully, 1995: 33;  ‍Stepan, A. and Skach, C. (1993). Constitutional frameworks and democratic consolidation: Parliamentarianism versus presidentialism. World Politics, 46 (1), 1-22. Available at: https://doi.org/10.2307/2950664Stepan and Skach, 1993: 20). As a backlash, coalition formation in presidential regimes was relegated to a secondary role in subsequent scholarship.

In a more recent contribution, José Antonio Cheibub ( ‍Cheibub, J. A. (2007). Presidentialism, parliamentarism, and democracy. New York: Cambridge University Press. Available at: https://doi.org/10.1017/CBO97805118133442007) revived this debate (also see  ‍Cheibub, J. A., Przeworski, A. and Saiegh, S. M. (2004). Government coalitions and legislative success under presidentialism and parliamentarism. British Journal of Political Science, 34 (4), 565-587. Available at: https://doi.org/10.1017/S0007123404000195Cheibub et al., 2004). In the process, he made a major breakthrough regarding the role of coalitions in presidential regimes. He showed that institutions found in presidential regimes did not, as initially suggested, hinder coalition formation. After controlling for the distribution of seats and policy preferences, parliamentary and presidential systems were equally likely to host government coalitions, and if these were less frequent in the latter it was only in cases in which legislatures were relatively weak (or presidents were extraordinarily strong), suggesting that both formateur power and the party system are relevant determinants.

Latin American presidential regimes are ideal for exploring these questions. In contrast to the 1980s, the period Linz and other authors of the three waves of parliamentary/presidential studies looked at, today half of the democracies in the region constantly undergo electoral and government coalition formation ( ‍Chaisty, P., Cheeseman, N. and Power, T. J. (2018). Coalitional presidentialism in comparative perspective: Minority presidents in multiparty systems. Oxford: Oxford University Press. Available at: https://doi.org/10.1093/oso/9780198817208.001.0001Chaisty et al., 2018). The evolution in the frequency of coalitions provides a perfect backdrop to explore when, how and why parties decide to form coalitions. Data stemming from almost four decades of democracy is bound to hold important hints that will help uncover the dynamics behind one of the most important forms of political organization in the region today.

The tension between the conventional view and the actual behavior of coalitions in the presidential regimes requires a revision. While the overall trend of coalition formation in the region has fostered a growing scholarly interest in regime survival ( ‍Cheibub, J. A. and Limongi, F. (2002). Democratic institutions and regime survival: parliamentary and presidential democracies reconsidered. Annual Review of Political Science, 5 (1), 151-179. Available at: https://doi.org/10.1146/annurev.polisci.5.102301.084508Cheibub and Limongi, 2002;  ‍Neto, O. A. and Samuels D. (2011). Democratic regimes and cabinet politics: A global perspective. Revista Ibero-Americana de Estudios Legislativos, 1 (1), 10-23. Available at: https://doi.org/10.12660/riel.v1.n1.2010.4123Neto and Samuels, 2011), the process of cabinet formation ( ‍Altman, D. (2000). The politics of coalition formation and survival in multiparty presidential democracies: The case of Uruguay, 1989-1999. Party Politics, 6 (3), 259-283. Available at: https://doi.org/10.1177/1354068800006003001Altman, 2000,  ‍Martinez-Gallardo, C. (2012). Out of the cabinet: What drives defections from the government in presidential systems? Comparative Political Studies, 45 (1), 62-90. Available at: https://doi.org/10.1177/0010414011421306Martínez Gallardo, 2012, 2014;  ‍Camerlo, M. and Martínez-Gallardo, C. (2017). Portfolio allocation in the Americas. In Government Formation and Minister Turnover in Presidential Cabinets (pp. 17-36). London: Routledge. Available at: https://doi.org/10.4324/9781315466491Camerlo and Martínez-Gallardo, 2017), legislative performance ( ‍Alemán, E. and Saiegh, S. M. (2007). Legislative preferences, political parties, and coalition unity in Chile. Comparative Politics, 39 (3), 253-272.Alemán and Saiegh, 2007), executive-legislative crisis ( ‍Pérez-Liñán, A. (2005). Democratization and Constitutional Crises in Presidential Regimes: Toward Congressional Supremacy? Comparative Political Studies, 38 (1), 51-74. Available at: https://doi.org/10.1177/0010414004270888Pérez-Liñán, 2005), bicameralism and federalism ( ‍Albala, A. (2017). Coalition presidentialism in bicameral congresses: How does the control of a bicameral majority affect coalition survival? Brazilian Political Science Review, 11 (2). Available at: https://doi.org/10.1590/1981-3821201700020002Albala, 2017) fundamental questions on the determinants of coalition formation remain unanswered. Why do parties cooperate with one another in presidentialism? And, more specifically, why do parties in presidential regimes form electoral and government coalitions?

This article examines coalition formation in presidential regimes. It attempts to explain why parties cooperate with one another in the form of coalitions, despite the institutional characteristics of presidentialism that have been found to theoretically hinder coalition formation. It also attempts to explain why some parties form electoral coalitions and others form government coalitions. Because this particular body of literature is burgeoning, this article borrows from theories of coalition formation in parliamentary regimes. It uses established concepts from proven power sharing theories to fill in the gaps, and tests a theories of hypotheses that have been so far treated independently.

It also borrows from the work of Cheibub ( ‍Cheibub, J. A. (2007). Presidentialism, parliamentarism, and democracy. New York: Cambridge University Press. Available at: https://doi.org/10.1017/CBO97805118133442007). It uses his contribution as a point of departure to develop a more sophisticated theory of coalition formation in presidential regimes. It aims to advance his original research agenda by providing additional conceptual context and explanatory power related to the specific types of coalitions found in Latin American presidential regimes. Furthermore, in addition to looking at coalition formation at the government level, it looks at coalition formation at the electoral level. It builds on previous work by adding a significant number of cases to the sample as well as a larger set of independent variables—classified into three different theories—to provide a more comprehensive account. By contextualizing and expanding the scope it provides a stronger framework for future studies.

The remainder of this article is structured as follows. The second section reviews theories of coalition formation. It differentiates between the traits of coalition formation under parliamentarism and presidentialism, and shows that contrary to conventional wisdom, parties have several reasons to form coalitions in presidential regimes. The second section outlines three theories that aim to explain coalition formation in presidentialism, related to formateur power, electoral structures and party systems. The third section breaks the research questions into hypotheses. The fourth section presents the data and methods. The fifth section presents the findings, and the sixth section discusses them. The final section makes some concluding remarks.

II. THEORIES OF COALITION FORMATION[Up]

In most democracies, rational parties compete for votes, both to control office and for the opportunity to influence policy ( ‍Strøm, K. and Müller, W. C. (eds.) (1999). Policy, office, or votes? How political parties in Western Europe make hard decisions. Cambridge: Cambridge University Press. Available at: https://doi.org/10.1017/CBO9780511625695Strøm and Muller, 1999). In multi-party systems, however, a single party is often unable to garner a majority of support. Thus, parties that wish to win elections typically cooperate with other parties and form coalitions. This is normally the case in parliamentary democracies where parties cooperate to avoid minority governments, and sometimes the case in presidential democracies where parties cooperate to avoid minority presidents ( ‍Laver, M. and Schofield, N. (1998). Multiparty government: The politics of coalition in Europe. Ann Arbor: University of Michigan Press. Available at: https://doi.org/10.3998/mpub.8290Laver and Schofield, 1998;  ‍Strøm, K. (1990). Minority government and majority rule. Cambridge: Cambridge University Press.Strøm, 1990). William Riker ( ‍Riker, W. H. (1962). The theory of political coalitions. New Haven: Yale University Press.1962) argues that coalitions attempt to form with the minimum number of parties possible to maximize office spoils.

Not all coalitions are alike.  ‍Cheibub, J. A., Przeworski, A. and Saiegh, S. M. (2004). Government coalitions and legislative success under presidentialism and parliamentarism. British Journal of Political Science, 34 (4), 565-587. Available at: https://doi.org/10.1017/S0007123404000195Cheibub et al., 2004) distinguish between legislative and government coalitions and define the former as a set of legislators belonging to parties that vote in the same direction with the same intensity in the legislature, and the latter as a set of politicians belonging to parties that share cabinet posts in the executive. Similarly, Royce Carroll and Gary Cox ( ‍Carroll, R. and Cox, G. W. (2007). The logic of Gamson’s Law: Pre‐election coalitions and portfolio allocations. American Journal of Political Science, 51 (2), 300-313. Available at: https://doi.org/10.1111/j.1540-5907.2007.00252.x2007) distinguish between legislative and presidential coalitions, using the same definition for the former, but adding that their strength will be mediated by their capacity of extending their cooperation in parliament to a formal agreement to back the same presidential candidate.

Sona Golder ( ‍Golder, S. N. (2006). Pre-electoral coalition formation in parliamentary democracies. British Journal of Political Science, 36 (2), 193-212. Available at: https://doi.org/10.1017/S00071234060001232006) builds on this classification and distinguishes between coalitions that form before elections, and coalitions that form after elections. Recently, research has advanced on both electoral coalitions (see  ‍Chiru, M. (2015). Early marriages last longer: Pre-electoral coalitions and government survival in Europe. Government and Opposition, 50 (2), 165-188. Available at: https://doi.org/10.1017/gov.2014.8Chiru, 2015;  ‍Kellam, M. (2017). Why pre-electoral coalitions in presidential systems? British Journal of Political Science, 47 (2), 391-411. Available at: https://doi.org/10.1017/S0007123415000198Kellam, 2017) as well as government coalitions (see  ‍Chaisty, P., Cheeseman, N. and Power, T. J. (2018). Coalitional presidentialism in comparative perspective: Minority presidents in multiparty systems. Oxford: Oxford University Press. Available at: https://doi.org/10.1093/oso/9780198817208.001.0001Chaisty et al., 2018). This article intends to build on this line of scholarship. It focuses on these two categories: electoral and government coalition formation. In this line, the objective is to determine the reasons why parties decide to cooperate with each other in elections and during government. It follows the distinctions that stem from parliamentary studies and adds recent evidence that stems from presidential countries to generate a comprehensive account.

III. RESEARCH QUESTIONS AND HYPOTHESES[Up]

The research question that guides this article is: why do political parties form electoral and government coalitions in Latin America? While some literature has been advanced to answer these questions, none has taken a comprehensive approach within presidentialism. This subsection outlines the hypotheses that can be derived from these three theories. These hypotheses aim to explain why electoral and government coalition formation occurs. Some of the hypotheses correspond only to electoral coalition formation, and some correspond only to government coalition formation. Normally the same independent variables are used. However, this will depend on whether the dependent variable is electoral coalition formation or government coalition formation. The following hypotheses are put forward:

While the all hypotheses aim to explain government coalition formation, only the second two aim to explain electoral coalition formation. To capture variations, each of the hypotheses are further broken down into a series of independent variables associated with the theoretical expectations and empirical findings presented above.

The objective is to let the independent variables take different forms to see which elements of formateur power, electoral structure and party system size provoke the most influence. Naturally, these hypotheses are tested both independently and in combination with each other across both space and time. While it may be the case that not all independent variables within a specific hypothesis will line up in the expected direction with a strong statistical likelihood, understanding which elements are more salient will be considered to be a finding in itself.

IV. DATA AND METHODS[Up]

This article looks at eighteen democratic regimes in Latin America. Following Chaisty et al. ( ‍Chaisty, P., Cheeseman, N. and Power, T. J. (2018). Coalitional presidentialism in comparative perspective: Minority presidents in multiparty systems. Oxford: Oxford University Press. Available at: https://doi.org/10.1093/oso/9780198817208.001.00012018), it uses a ‘double threshold’ based on the two most commonly used regime classification datasets, Freedom House and Polity. It uses cases classified as at least as ‘partly free’ by Freedom House (less than five on a seven-point scale in which lower number indicate higher levels of freedom), but also as ‘open anacrocies’ or ‘democracies’ by Polity IV. This double threshold is useful since it makes case selection sensitive to incremental change. It allows for a more rigorous yet accurate selection of democratic years. For example, it leaves out Peru during the Fujimori years (1990-‍2000), which would have been included if only one of the indicators were used.

There are two dependent variables. The first is electoral coalition formation (when parties form coalitions to compete in elections), and the second is government coalition formation (when parties form coalitions to govern). Both dependent variables are dichotomous. In the former case, it is set to 1 when two or more parties form an alliance to compete in elections. In the latter case, it is set to 1 when the cabinet is composed by members of at least two different parties. Table 1 shows the list of independent variables that aim to explain when these events are more or less likely to occur. The independent variables are listed under their respective parent theories. The Table also shows the type of variable (continuous or dichotomous), the measurement units and the source when applicable.

In terms of formateur power hypothesis, three independent variables are tested to explore their impact on government coalition formation. The first is formal policy-making powers (such as veto power, decree power, and budgetary power), which is expected to increase the likelihood of coalition formation as it decreases (see  ‍Pereira, C., Power, T. J. and Rennó, L. (2005). Under what conditions do presidents resort to decree power? Theory and evidence from the Brazilian case. The Journal of Politics, 67 (1), 178-200. Available at: https://doi.org/10.1111/j.1468-2508.2005.00313.xPereira et al., 2005). The second is informal policy-making powers (such as appointment power, emergency power, and inter-branch conflict power), which is expected to increase the likelihood of coalition formation as it decreases (see  ‍Neto, O. A. (2006). The presidential calculus: Executive policy making and cabinet formation in the Americas. Comparative Political Studies, 39 (4), 415-440. Available at: https://doi.org/10.1177/0010414005282381Neto, 2006). The third independent variable is electoral support for the incumbent formateur, since the likelihood of coalition formation is expected to increase as the percentage of support for the winning president falls (see  ‍Cheibub, J. A. (2007). Presidentialism, parliamentarism, and democracy. New York: Cambridge University Press. Available at: https://doi.org/10.1017/CBO9780511813344Cheibub, 2007;  ‍Shugart, M. S. and Carey, J. M. (1992). Presidents and assemblies: Constitutional design and electoral dynamics. Cambridge: Cambridge University Press. Available at: https://doi.org/10.1017/CBO9781139173988Shugart and Carey, 1992).

In terms of the electoral structure hypothesis, nine different independent variables are tested to explore their impact on electoral coalition formation and six independent variables are tested to explore their impact on government coalition formation. Some of them overlap. Four of these variables are related to presidential elections (simple majority plurality, incumbency status, term length, and simultaneousness with other elections) and five of them relate to legislative elections (proportional representation, thresholds, the size of the legislature, district magnitude, and the number of tiers in congress). All of these electoral structure variables can be ordered on a restrictive-permissive continuum. Theoretically, the more restrictive they are, the more likely they will be to generate incentives for coalition formation.

Among those related to presidential elections, the first variable is Simple Majority Plurality (SMP). When it is available, coalition formation is more likely to occur. When a presidential candidate has only one shot at an election (as opposed to two shots, when a run-off is available), he will attempt to maximize his chances of winning by forming a broad enough coalition to surpass the required threshold. In most cases, parties that are asked to join a coalition will seek a reward in exchange for their endorsement, such as cabinet portfolios or congressional power positions. Thus, while SMP offers strong incentives to join a coalition before an election, the formateur-elect will often decide to uphold the agreement for the duration of his administration

A second independent variable is related to the formateurs incumbency status. When the formateur is bidding for reelection, or when he is successfully reelected, coalition formation is more likely to occur. Because incumbents that seek re-election are usually considered front-runners, smaller parties that do not stand a chance will seek to back them in exchange for cabinet portfolios. Similarly, formateurs in their second term are more likely to see government coalitions form, since they are likely to uphold partisan agreement that stem from their first term. This is also reinforced by trends in presidential approval in the region, since popularity levels have been noted to drop in the second of two consecutive terms. In these cases, formateurs need to make strategic alliances in order to pass legislation.

A third independent variable is related to the formateurs term length. When it is a longer term, coalition formation is more likely to occur (see  ‍Neto, O. A. (2006). The presidential calculus: Executive policy making and cabinet formation in the Americas. Comparative Political Studies, 39 (4), 415-440. Available at: https://doi.org/10.1177/0010414005282381Neto, 2006). When term limits are short (say, four years), smaller parties will be willing to take their chances in competing against larger parties in elections, since they will have another chance to compete (or at least reassess their strategy) in just a few years. In contrast, when term limits are long (say, six years), smaller parties will not be as willing to take the same chances, since the distance to the next election is significantly longer. In this case, when the smaller parties opt out of competing on their own and join a coalition with a larger party, they maximize their chances of obtaining cabinet portfolios.

The fourth independent variable is related to the simultaneousness of presidential and legislative elections. When these two elections are concurrent, coalition formation is more likely to occur (see  ‍Borges, A. and Turgeon, M. (2017). Presidential coattails in coalitional presidentialism. Party Politics, Online First. Available at: https://doi.org/10.1177/1354068817702283Borges and Turgeon, 2017). Concurrent elections force parties to have a national strategy, in which larger parties will tend to negotiate with smaller parties for legislative support, and smaller parties will tend to negotiate with larger parties in exchange for cabinet portfolios. When elections are concurrent, coalitions are usually a win-win strategy for both large and small parties. In contrast, when elections are held separately, parties tend to have divergent electoral strategies. Large and small parties with different interests may converge in concurrent elections.

Among those variables to legislative elections, the first is the electoral system. When votes are translated to seats by means of proportional representation, coalition formation is more likely. This is based on the premise that PR systems encourage multi-party systems, and thus boost the chance that at least some of the parties will find a common platform to campaign on. As the party system becomes more crowded, the chance of finding like-minded parties increases. In turn, as the number of like-minded parties’ increases there will be more incentives for them to negotiate and form coalitions, or even merge. As explained below, in small party systems there are fewer incentives to cooperate, since more parties have a chance of winning on their own.

A second variable in this line is related to barriers of entry. When legislative electoral systems have thresholds to gain legislative representation, coalition formation is more likely to occur. The logic being that the survival instinct of small parties will force them to find a way of remaining active in the political system. If a small party anticipates a low vote share in the upcoming election, the incentive to cooperate with a larger party, increases. This is tied in with the psychological incentives of parties when strategically deciding how to approach elections (see Benoit, 2005). If a small party anticipates that the electoral system will harm their chances of entering congress, they will choose to negotiate with a larger party in order to secure access to the legislature.

The size of the legislature is also relevant to explain why parties cooperate. When more legislators are elected to congress, coalition formation is more likely. When there are many seats available, larger parties will be looking to co-opt smaller parties and avoid potential veto-players. Since large parties are central in forming government or opposition, it is important for their purpose to have the greatest number of parties aligned on their side. If they do not negotiate with smaller parties, they risk internal dealignment. Likewise, smaller parties may seek to cooperate with larger parties, since their vote share will probably not give them a significant quota of power in congress anyway. From their perspective, it may be important to form a coalition with a larger party and piggyback on their share of power. At any rate, the size of the legislature has been found to be an important predictor of the number of parties ( ‍Shugart, M. S. and Taagepera, R. (2017). Votes from Seats: Logical Models of Electoral Systems. New York: Cambridge University Press. Available at: https://doi.org/10.1017/9781108261128Shugart and Taagepera, 2017).

A fourth variable is related to the size of legislative districts. When many legislators are elected in each district, coalition formation is more likely to occur. The logic behind this hypothesis is that because small parties face elimination if they decide to compete against large parties in districts where few seats are available, they will tend to maximize their chances of survival and join a coalition. Larger districts tend to favor larger party systems ( ‍Cox, G. W. (1997). Making votes count: strategic coordination in the world’s electoral systems. New York: Cambridge University Press. Available at: https://doi.org/10.1017/CBO9781139174954Cox, 1997;  ‍Duverger, M. (1954). Political parties: Their organization and activity in the modern State. Methuen.Duverger, 1954;  ‍Riker, W. H. (1982). The two-party system and Duverger’s law: An essay on the history of political science. American Political Science Review, 76 (4), 753-766. Available at: https://www.doi.org/10.2307/1962968Riker, 1982), which in turn tend to favor coalition formation. However, it is important to note that there is a district magnitude “sweet spot”, which tends to be large but not too large (see  ‍Carey, J. M. and Hix, S. (2011). The electoral sweet spot: Low‐magnitude proportional electoral systems. American Journal of Political Science, 55 (2), 383-397. Available at: https://doi.org/10.1111/j.1540-5907.2010.00495.xCarey and Hix, 2011). Parties have just as few incentives to join coalitions when district magnitudes are too small as they have when district magnitudes are too large (see  ‍Shugart, M. S. and Taagepera, R. (2017). Votes from Seats: Logical Models of Electoral Systems. New York: Cambridge University Press. Available at: https://doi.org/10.1017/9781108261128Shugart and Taagepera, 2017). Finally, the number of chambers in Congress are also an important element. When congress is bicameral, coalition formation tends to be more likely. In contrast to unicameralism, two-chambers are more complex. There are more veto players in bicameralism, therefore negotiations tend to be more prolific (see Tsebelis, 2000). As Albala ( ‍Albala, A. (2017). Coalition presidentialism in bicameral congresses: How does the control of a bicameral majority affect coalition survival? Brazilian Political Science Review, 11 (2). Available at: https://doi.org/10.1590/1981-38212017000200022017) suggests, bicameral majority constitutes a sufficient, but not necessary, condition for enduring coalition agreements. In anticipation of difficult legislative situations, presidential candidates will normally seek to form coalitions to boost the chances of passing their program if they are eventually elected and sitting presidents will do so naturally in order to manage stalemate or control presidential approval rating.

In terms of the party system hypothesis, two independent variables are tested to explore their impact on electoral and government coalition formation. The first is the size of the party system. Since timing is relevant, this variable is divided into two indicators: the number of electoral parties (ENEP), and the number of legislative parties (ENLP). The impact of the former will be tested on electoral coalitions, and the impact of the latter will be tested on government coalitions. The central idea is the same for both. When there are more parties in the system, coalition formation will be more likely (see  ‍Kellam, M. (2017). Why pre-electoral coalitions in presidential systems? British Journal of Political Science, 47 (2), 391-411. Available at: https://doi.org/10.1017/S0007123415000198Kellam, 2017). However, it is important to note that the ENEP will always be larger than the ENLP. Thus, it could be possible that because of a reductive mechanism, a lower ENLP will be a better predictor of government coalitions. A second independent variable is the formateurs legislative majority. When there are many seats in congress occupied by parties sympathetic toward the president, coalition formation is more likely (see  ‍Kellam, M. (2015). Parties for hire: How particularistic parties influence presidents’ governing strategies. Party Politics, 21 (4), 515-526. Available at: https://doi.org/ 10.1177/1354068813487109Kellam, 2015). At the very core of this hypothesis sits the idea that parties will first merge in congress and then work towards a government or electoral agreements. If likeminded parties cooperate and vote for policy in the same direction, it is only natural for them to evolve into a larger compromise. However, it can also be due to reverse causation. It may be the case that the president has a large working majority in Congress because he already has formal agreements at the electoral or government levels.

A series of additional variables are added to the models to control for external factors. One set of these controls are related to political variables, such as presidential crisis, left-wing government, outsider candidates, and the age of democracy. When they occur, coalition formation becomes more likely. Thus, when a democracy is going through crisis, the president belongs to a left-wing party, an outsider competed in the previous election, and the democracy is young, coalition formation is expected to be more likely to occur.

A second set of these controls are economic variables, such as inflation, growth and inequality. When the Consumer Price and Gini indexes are low while the GDP index is high (in the previous year), coalition formation is expected to be more likely to occur.

V. METHODS[Up]

This article uses data stemming from eighteen countries, 100 elections and 407 aggregate years of democratic government.^‍‍[1]As mentioned above, the dependent variable depends on whether coalition formation occurs at the electoral or at the government level. In the former case, it is a dichotomous indicator that tests whether both the winning and the runner-up formateur candidates are backed by coalitions of two or more parties. In the latter case, it is a dichotomous indicator that tests if the government portfolio is composed by members of at least two different parties. The independent variables are those mentioned above. Control variables are held constant across all models. The unexplained variation is manifested in the error component.

Data are organized and tested to account for individual heterogeneity. In these models, observations are treated as panel data to control for variables that could not be directly observed or measured. Thus, they are treated as cross-sectional time series with a time series logit function. This statistical technique is regularly used in studies that use data with multiple observations in multiple countries. Data is tested with both random effects and fixed effects models ( ‍McCullagh, P. and Nelder, J. A. (1989) Generalized Linear Models. London: Chapman and Hall.McCullagh and Nelder, 1989). The random effects models assume that the error term is not correlated with the respective independent variables, which allows for time-invariant variables to play a role as explanatory variables. The fixed effects models do not make the same assumption.^‍‍[2] The specification for all of the models is as follows:

ϓ_it = β₁χ_1it + β₂χ_2it + β₃χ_3it + β₄χ_4it + ζ_it + η_i + ξ_it

where ϓ_it is the value of the dependent variable for the _ith case (country) in the sample at the _tth time period (year); β₁ is the coefficient corresponding to each of the variables in the first vector (formateur power) and χ_1it it is the value corresponding to each of its time varying covariates. β₂ is the coefficient corresponding to each of the variables in the second vector (electoral structure) and χ_2it is the value corresponding to each of its time-varying covariates. β₃ is the coefficient corresponding to each of the variables in the third vector (party system) and χ_3it is the value corresponding to each of its time-varying covariates. β₄ is the coefficient corresponding to each of the variables in the fourth vector (control variables) and χ_4it is the value corresponding to each of its time-varying covariates. ζ_it is the unknown intercept for each entity; η_i is the between-entity error; and ξ_it is the within-entity error.

VI. EXPLANATIONS FOR ELECTORAL AND GOVERNMENT COALITION FORMATION[Up]

Table 2 shows the determinants of electoral coalition formation. It shows three models with different combinations of the independent variables. Model 1 shows the impact of the electoral structure, and Model 2 shows the impact of the party system, and Model 3 shows a combination of the two theories.

Bicameralism, proportional representation and size of the legislature are the most important elements of the electoral structure when it comes to predicting electoral coalition formation. When they are compared to each other, in Model 1, bicameralism and PR are significantly more important, suggesting that when congress is divided into two tiers, and legislators are elected with proportional representation formulas, parties have stronger incentives to cooperate. In essence, the probability of electoral alliances is higher in bicameral democracies that use proportional representation systems (such as Argentina or Brazil) in comparison to those that use majoritarian components (such as Mexico or Panama).

The number of electoral parties is also a crucial predictor for electoral coalitions. As the number of parties that compete in elections increases, the probability of electoral coalition formation also increases. It need not be this way; parties could very well work as a replacement for coalitions. The number of parties could grow at the expense of coalitions. Yet, evidence shows the contrary. The ENEP holds both its magnitude and direction of impact, as well as its significance, across all models. Irrelevant of the variables tested, the number of parties that compete in elections is the most powerful predictor of coalition formation.

Table 3 shows the determinants of government coalition formation. Model 4 shows the impact of formateur power, Model 5 shows the impact of the electoral structure, and Model 6 shows the impact of the party system. Model 7 shows a combination of the three theories. In comparison to Table 2, there are some differences in the sets of independent variables used. Most importantly, the effective number of legislative parties instead of the effective number of electoral parties.

Formateur power is a strong and consistent predictor of government coalition formation. As the power of the president falls, the probability of government coalition formation increases. This makes sense. Since all presidents seek to pass legislation, the real question is how they attempt to do so. If the president has ample constitutional power and high levels of popular support, she will likely be able to pass legislation unilaterally, and choose to govern alone with her party. Instead, if she has a combination of low power and low vote share, she will likely need to seek support from other parties to get things done. While formateur power is important across all contexts, informal powers seem to be a slightly better predictor than formal powers.

The electoral structure is also a defining element when it comes to government coalition formation. The manner in which the president is elected is defining. The probability of a president forming a multi-party cabinet significantly increases when the president is elected in a single round (in contrast to a runoff).

Term length is equally important. As the number of years in the term increases, the probability of coalition formation grows. Presidents that govern for six years are more likely to form coalitions than presidents that govern for four years. As discussed above, the logic is more bottom-up than top-down. It is not that large parties seek the support of small parties, but that small parties seek the support of large parties. Since the president has already secured a longer period in power, she is not as concerned as smaller parties in the immediate fate. In contrast, small parties are pressured to pact early on in the term so that they can secure political relevance. The logic is that a small party is worth more in government than in opposition, provided there are sufficient ideological affinities.

Another important predictor of government coalition formation is electoral coalition formation. If the president-elect was supported by an electoral coalition in the immediately previous election, she will almost certainly govern with a coalition. This echoes “Gamson’s Law,” in that parties that endorsed the president in the election will be rewarded with cabinet posts. Only rarely will a president that competed alone in the election decide to negotiate with other parties for support (one exception is the Violeta Chamorro administration in Nicaragua).

Perhaps the most interesting finding, however, is that as the number of legislative parties’ declines, the probability of government coalition formation increases. While this seems counterintuitive, especially considering the relationship between ENEP and electoral coalitions, is makes sense since there is a mechanical reductive factor occurring between elections and government formation. Since there will never be as many ENLP as there are ENEP, an implicit indicator of coalition formation is precisely a decline in the number of parties in this period. Furthermore, if there are fewer parties in congress, the president will have incentives to seek support elsewhere, to prevent facing an opposition cartel (see  ‍Slater, D. and Simmons, E. (2013). Coping by colluding: Political uncertainty and promiscuous powersharing in Indonesia and Bolivia. Comparative Political Studies, 46 (11), 1366-1393. Available at: https://doi.org/10.1177/0010414012453447Slater and Simmons, 2013).

One variable that is consistently significant across all scenarios is the age of the democracy. Government coalition formation is more likely in younger democracies. This makes sense, especially in the Latin American context. As democracies consolidate and party systems institutionalize, coalition formation stabilizes. While there is a larger share of coalitions in government today than thirty years ago, the countries in which they form are different. Countries that were young democracies three decades ago were more likely to see coalition formation then than they are now.

These findings suggest that the three hypotheses postulated above can be accepted, with some caveats. The probability of electoral coalition formation increases as the electoral structure becomes more restrictive and the party system becomes more fragmented. Some variables are substantially more important than others, such as the size of the legislature and the ENEP. Similarly, the probability of government coalition formation increases as the formateur falls in power, and the electoral structure becomes more restrictive. While there is mixed evidence for the hypothesis that a larger party system will lead to government coalition formation, it is likely that both variables are linked through a causal mechanism in which the number of parties are mechanically reduced in the process of translating votes into seats (see  ‍Benoit, K. (2002). The endogeneity problem in electoral studies: a critical re-examination of Duverger’s mechanical effect. Electoral Studies, 21 (1), 35-46. Available at: https://doi.org/10.1016/S0261-3794(00)00033-0Benoit, 2002).

VII. CONCLUDING REMARKS[Up]

This article contributes to the burgeoning body of literature preoccupied with building a theory of coalition formation particular to presidentialism. Through a wide theoretical review, it identifies some of the major variables previously syndicated as important determinants of coalition formation. It takes a confirmatory approach and tests them together in the same model. It finds that the data is generally consistent with the theories are hypotheses. As the president’s power decreases, electoral structures become more restrictive and party systems become more fragmented, the probability of coalition formation increases. This article complements previous contributions by taking a holistic to explain both electoral and government coalition formation.

This article also contributes by finding that within each of the theories put forward, some variables are more important than others, such as the president’s informal authority in the formateur power theory, bicameralism, proportional representation and the size of the legislature in the electoral structure theory, and the number of parties in the party system theory. While it establishes that the theories are generally consistent at the macro-levels, it also singles out the specific variables that are most influential in coalition formation. In this line, it also eliminates variables that have been mistakenly considered important determinants in the past, such as the concurrency of presidential and legislative elections and the incumbency status of the president.

The findings in this article are important insofar as yielding critical insight into partisan strategies in both the run-up to elections and the maintenance of governments. It is consistent with recent research that has found that coalition formation is more likely under some scenarios, such as bicameralism ( ‍Albala, A. (2017). Coalition presidentialism in bicameral congresses: How does the control of a bicameral majority affect coalition survival? Brazilian Political Science Review, 11 (2). Available at: https://doi.org/10.1590/1981-3821201700020002Albala, 2017) and formateur majority ( ‍Kellam, M. (2015). Parties for hire: How particularistic parties influence presidents’ governing strategies. Party Politics, 21 (4), 515-526. Available at: https://doi.org/ 10.1177/1354068813487109Kellam, 2015). Other relationships need to be further explored, such as the link between the size of the legislature and electoral coalitions as well as the link between the age of democracy and government coalitions.

These finding contribute to a general theory of coalition formation, which has thus far either been approached by findings made in parliamentary studies or in presidential regime survival. Further research should continue to explore the reasons of coalition formation in presidentialism, either by taking a deeper look at some of the variables that show likely causal links in this article or by comparing the factors that make coalition formation more likely across different regimes by crosschecking the variables found to be significant above. Future research could also look at the influence of additional variables, not tested here, that may impact coalition formation, such as the size and age of political parties and if they have clientelistic or programmatic tendencies.