Abstract
This paper focuses on the process of coalition formation in presidential systems. It shows that party positions and institutions influence the formation of government coalitions. We argue that presidents will tend to include parties positioned close to their policy positions in their cabinets, and will be more inclined to do it when relative institutional authority is more advantageous to congress. We corroborate our arguments with data from 13 presidential countries in the Americas.
Introduction
For many years, conventional descriptions of presidentialism stressed the lack of incentives for coalition formation and called its occurrence a rare event (Mainwaring 1993; Linz and Stepan 1994). This view, however, is beginning to change. Since the 1980s, most presidential countries in the Americas have at some point had a multiparty coalition in government. In some countries like Brazil, Chile, and Uruguay, multiparty coalitions in government are now standard. In the absence of a single-party majority in the legislature, more often than not presidents build coalitions. 1 If we look only at the last decade, then the proportion is even higher. Although the increasing occurrence of coalition government in presidential systems has fostered a growing interest in the process of cabinet formation and the legislative performance of coalition governments, cross-national analyses remain scant. 2 Thus far, there has been relatively little theoretical discussion about the relevant aspects of government formation and coalition stability under presidential constitutions. This paucity of comparative studies on coalition governments in presidential systems stands in contrast to the wealth of sophisticated research on coalition governments in parliamentary systems.
About 54 percent of the time according to Cheibub et al. (2004).
Exceptions include Amorim Neto (2006) and Cheibub et al. (2004). There is, however, a growing number of works focusing on one country.
The constitutional structure of presidential systems creates a different set of incentives for coalition formation, since government survival is not dependent on the confidence of parliament and the formateur is always the president. However, there are usually strong policy motives for presidents to reach stable partisan agreements through portfolio coalitions, particularly when there is no single-party majority in the legislature. In this paper, we extend the analysis of coalition formation under presidentialism (where the most common scenario is the absence of a single party controlling a majority of seats in the legislature). Our focus is on political parties and the positional and institutional forces that affect participation in cabinet coalitions. We argue that parties positioned in the ideological proximity of the president are more likely to participate in government than those positioned further away, and when the balance of institutional power is less favorable to the president, members of non-presidential parties are more likely to participate in the government than when it is tilted in the president's favor. Furthermore, we posit that the effect of ideological distance is conditional on institutional power: non-presidential parties positioned relatively close to the president should be significantly more likely to be brought into the cabinet when the balance of institutional power is less favorable to the president. To investigate these propositions, we examine the formation of governments in presidential countries, more specifically, how ideological closeness to the president, the relative balance of institutional power, and the interaction of these variables impact the probability of particular parties entering the cabinet. 3
This focus on the probability of particular parties entering the cabinet is similar to Warwick's approach (1996, 1998). Altman (2000) also followed a similar approach to study coalition making in Uruguay.
This paper 4 is divided into four parts. The first part defines some fundamental traits of coalition formation under presidentialism, and expands on our hypotheses about the influence of ideological distance and institutional authority on cabinet formation. This section also highlights relevant similarities (e.g. policy motives of the formateur) and differences (e.g., on parties expectations of government participation) between government formation under presidentialism and under parliamentarism. The second part presents the empirical analysis. We use a dataset that includes information on 107 different cabinets in 13 presidential countries from the 1980s to the early 2000s, 5 in addition to information on every party with congressional representation in those countries and its ideological position. 6 To examine the probability of cabinet participation we utilize a random effects logit model. The results of the empirical analysis confirm the influence of positional and institutional forces and their interactive effect. The third part elaborates on the implications of some of our findings, and the fourth part provides a conclusion.
The authors would like to thank Ernesto Calvo, Javier Zelaznik, Octavio Amorim Neto, and the anonymous reviewers for their comments and suggestions.
Cabinet data comes from Amorim Neto (2006). The sample includes cabinet composition data for the following periods in 13 nations: Argentina (1983 to 2001), Bolivia (1982 to 2001), Brazil (1985 to 2001), Chile (1990 to 2006), Colombia (1978 to 2000), Costa Rica (1978 to 2002), Ecuador (1979 to 1999), Mexico (1988 to 2000), Panama (1990 to 2002), Peru (1980 to 1992), United States (1981 to 2001), Uruguay (1985 to 2003), and Venezuela (1979 to 1999).
Ideological data comes from Coppedge (1997).
Coalition Formation in Presidential Systems
The idea that presidential democracies provide very few incentives for coalition formation became the conventional wisdom, partly as a result of the debate on the merits of parliamentarism versus presidentialism that took place among comparativists twenty years ago. A common view within this literature was that coalition making in presidential systems was “considerably more difficult than in parliamentary regimes” (Mainwaring 1990: 169) and as a result, multi-party government coalitions would “only exceptionally” form (Linz 1994: 19). 7 This argument was based on institutional reasons. Presidents do not need to form a legislative majority to take office, nor do they need a legislative majority to fend off potential challenges (i.e., confidence vote). Presidential systems, unlike parliamentary ones, do not institutionalize multiparty alliances through government investiture or dissolution. Presidents are the exogenously (popularly) chosen formateurs, who have the option of forming multiparty cabinets, but can hold on to the executive even with little legislative support. In addition, non-presidential parties that join the government are comparatively less committed to supporting the government's legislative program; after all, legislative defection does not put the current executive out of office. 8 In parliamentary systems, government parties are veto players (Tsebelis 2002): they need to agree in order for a significant policy change to take place; otherwise, either there will be no change or this change will be made by a different government (which will not include the parties that disagree). In presidential systems, the parties in government are not necessarily veto players; that is, they can vote against a bill and presumably still remain in government.
See Cheibub et al. (2004) for a review of this characterization and a powerful critique.
Moreover, according to Altman (2000: 261), given the influence that the president's reputation has on voters’ evaluations of government performance, “the president's coalition partners are less able to claim credible credit for good government performance.”
Coalition making, however, is not a rare event in presidential systems. More than a decade ago, Dehesa (1997) examined 123 separate administrations and found that multiparty coalitions governed in 56 percent of cases. These results agree with the more descriptive analysis of six presidential democracies presented by Lanzaro (2001). And the findings of Cheibub et al. (2004) provide evidence of both, the frequent occurrence of presidential coalitions and the rather successful legislative record of such governments. Earlier literature, it now seems clear, had underestimated the strength of those incentives that can lead to successful coalition formation in presidential systems. 9
A similar critique is advanced by Foweraker (1998) in his comparative analysis of presidential governments.
So far, analyses have focused mainly on whether coalitions emerge in presidential systems, and to our knowledge few empirical analyses exist that examine the positional variables and institutional prerogatives that affect coalition formation. A notable exception in this respect is the work of Amorim Neto (2006), which concentrates on the majority status of the cabinet and patterns of partisan/ non-partisan appointments. He finds some evidence that extremist presidents (dummy for leftist or rightist presidents) are less likely to form cabinets that command a majority in congress, and that the type of cabinets they form have a smaller share of partisan ministers (i.e., more technocrats). They are also less proportional in terms of the distribution of partisan support among coalition members (lower coalescence). 10
With regards to the legislative powers of the president, he finds that decree power (dummy for five countries) and high veto override tend to be associated with cabinets composed of a smaller share of partisan ministers.
These recent works have illuminated relevant aspects of coalition making in presidential systems, yet several important questions remain unanswered. If coalition making occurs frequently, which parties are more likely to join the cabinet? How does the distribution of institutional powers among president and congress affect parties’ likelihood of participation in government?
We advance four propositions in this paper (ceteris paribus). First, parties that are ideologically close to the president should be more likely to participate in a government coalition than parties with distant positions. This is because parties closer to the president give the government coalition more leeway in policymaking. Second, when the disparity in relative power between the president and congress is small, multiparty coalitions should be more likely to form than when the balance of power greatly favors the president. The third and most important argument posits that the effect of ideological distance from the president should be dependent on institutional influence: parties located close to the president should be more likely to participate in government when institutional authority is less disadvantageous to congress. Lastly, we expect the electoral cycle to impact the incentives to participate in the president's cabinet. As the last presidential election fades away and the next one looms on the horizon, non-presidential parties have fewer incentives to join or remain in the president's cabinet.
The Impact of Distance
Why is positional proximity so important? Presidents care about policy, whether intrinsically (i.e., because of their own positioning) or because of the electoral connection (e.g., fulfilling electoral promises or promoting the interests of their constituents), and as result they strive to reach legislative outcomes close to their own preferred policy. Presidents, like prime ministers in parliamentary regimes, must achieve policy victories. To be able to deliver such victories presidents need to engage in bargaining with congress. Centrally located minority presidents may be able to convince non-presidential parties located to their right and to their left to vote with the government often enough to achieve some favorable policy outcomes without giving out cabinet posts to a non-presidential party. In most instances this is similar to the logic behind the stability of minority governments in parliamentary systems, where centrist governments seek support sometimes on the left and sometimes on the right. But most often minority-presidents see necessary to trade cabinet posts for greater and more reliable partisan support. This is particularly true when a president's inherent weaknesses (partisan or institutional) threaten the achievement of the legislative agenda of the government. In the process of government formation, an ally that is closer rather than far away in policy terms improves the chances of achieving outcomes near the president's preferred position.
Consider, for example, a choice of allies along the same dimension. Figure 1 shows the president (
Choosing Coalition Partners I
The examples assume that no actor has a majority (cannot decide unanimously), and that the president can create a winning majority by allying with just one of the two parties.
Players have ideal policy positions (ideal points), and prefer policy less the further it moves from their own ideal point.
If
The situation becomes significantly more complicated in multiple dimensions, as Figure 2 indicates. Consider a situation where parties
Choosing Coalition Partners II
Consider, however, a more complicated situation, where the position of the status quo with respect to the president and the other parties is different in each dimension, and the political salience of policy issues depends on exogenous conditions (e.g., a nuclear disaster or an oil shock may bring into the agenda energy or environmental issues, a bank scandal may similarly raise to the forefront financial issues, etc.). In this context, there is high uncertainty regarding which issues will surface to the top of the political agenda and what will be the relative position of the relevant parties vis-à-vis the status quo. If the position of the status quo is not known in advance, then the associated strategic calculations cannot be performed. An efficient strategy for a president in such a context is to select allies that are positioned relatively close to him, expecting that the position of the status quo on the policy issues that become central to the government does not fall between his preference and those of his coalition partners (in which case, no change will be possible). The logic is similar in one dimension: positional closeness to the president should reduce potential policy disagreements while increasing the set of possible policy changes favorable to government parties (i.e., the winset of the status quo would tend to be larger – the deadlock region smaller – with an ally positioned nearby).
To sum up, centrally located minority executives may be able to achieve favorable policy outcomes without entering into a formal cabinet coalition, as the literature on parliamentarism has highlighted, but most often than not, minority-presidents see necessary to bring non-presidential parties into the cabinet for more reliable legislative support. Why the president would want any parties in government instead of making ad-hoc coalitions (what Sartori has called “retail politics”) is not the focus of our article – we do not advance a new general theory of coalition formation. Our argument is more modest: when searching for government allies, presidents will find that coalitions with ideologically proximate parties give them more leeway to achieve favorable policy changes than coalitions with parties positioned further away, and, as we discuss in the following section, president will more concerned about ideological congruence when congress is more powerful.
Institutional Influences
What aspect of the institutional balance of power does influence the president's propensity to form coalitions? The answers can be grouped into three camps. One perspective tends to point to the president's formal prerogatives. Shugart and Carey's (1992) seminal work differentiated presidential systems according to executive authority, and argued that high legislative powers gave presidents opportunities to sidestep congress, opening the door to regime instability. According to Cox and Morgenstern (2002), it is when presidents have little partisan support in congress that they make full use of executive powers and go to constitutional limits. Consequently, presidents with institutional means of achieving policy goals may or may not care about forming multiparty coalitions. They may prefer to set policies by using their institutional prerogatives alone, or they may rely on their political skills to form ad-hoc alliances around specific policy proposals. However, if presidents lack sufficient authority to control the legislative agenda, they have high incentives to build lasting coalitions. Thus, this first perspective sees executive powers as supplanting for partisan support in congress.
Executive authority entails agenda-setting tools, such as introduction, amendment, and veto prerogatives, that presidents can use to promote their policy goals. By manipulating the policy choices faced by members of congress, presidents with high formal powers can influence policy outcomes in their favor. The most extreme of these prerogatives is the executive decree; although its regular usage as a law-making device is circumscribed to very few Latin American countries (and applied to only some areas of policy) and in all instances congresses have the power to reject the decrees. The proposition that the power of the president vis-à-vis congress influences coalition incentives is consistent with prior institutional works (Cox and Morgenstern 2002; Foweraker 1998; Amorim Neto 2006). Shugart and Haggard (2001) argue that it is no accident that constitutionally stronger presidencies are found in countries where unified government is less likely. Constitutional designers, they argue, have given these greater powers to the executive branch in an effort to partially compensate for the absence of unified government.
While formal presidential authority has been frequently underscored, an exclusive focus on the executive misses the countervailing impact of congressional prerogatives. An alternative answer to the question of the balance of institutional power, thus, stresses the importance of congressional authority. Since measures based solely on formal presidential authority do not account for the institutional strength of congress, they can fail to properly gauge relative power. Congressional oversight mechanisms, for instance, influence the balance of power between branches, giving legislators tools to affect political outcomes that matter to the president (Alcántara et al. 2005). Some of these mechanisms include the ability to form investigative committees, veto presidential nominations, impeach presidential appointees, compel information from state agencies, screen budgetary maneuvers, or force the interpellation of government officials. Minority presidents tempted to push the envelope of constitutional power to achieve policy victories on a case-by-case basis instead of seeking out more reliable coalition partners, have to consider the broader costs of attempting to bypass a legislature with relevant retaliatory powers. Thus, from this perspective, congresses with high formal authority should make president more inclined to build reliable alliances with legislative parties.
Yet a third perspective casts doubts on operationalizations based solely on aggregating various constitutional and other formal prerogatives and instead stresses the institutional capacities of congress. This view places greater emphasis on instruments capturing legislative competence and sees electoral rules, nomination procedures, and, more generally, political ambition as key in fostering a more or less competent congress. This insight underlies Santiso's (2005) so-called paradox of legislative budgeting in Latin America: while congresses possess a wide range of formal budgetary powers, they often fail to exercise them effectively or responsibly, thereby rendering executive dominance.
Consider the situations of Chile and Mexico. There is agreement among all types of classifications that the Chilean president enjoys much greater formal authority than the president of Mexico. However, the Chilean congress has a highly professionalized membership working within a strong system of committees, while the Mexican congress has a weaker committee system composed of members that cannot be re-elected and whose term is also shorter. It can therefore be argued that, all else equal, the Chilean president has to bargain with a mightier institution than does the Mexican president, and as a result parties in the Chilean congress are more likely to join a government coalition. Consequently, if capacity is a more appropriate gauge of institutional relevance, then presidents should be more likely to form government coalitions in congresses with more professional memberships. Both tenure and service time demands are measures traditionally employed in the legislative literature to capture institutional capacity and membership professionalization. 14
See, for instance, Polsby's classic work (1968), or Squire's (1992) work on legislative professionalization.
To sum up, we argue that a balance of institutional power more favorable to the president should decrease presidential incentives to form coalitions. While there is no consensus on the comparative presidential literature about the best proxy to capture the institutional balance of power, we reviewed three alternative operationalizations of such a concept.
This takes us to the third proposition of this paper: we argue not only that that the probability that a (non-presidential) party participates in a government coalition increases when the balance of power is less disadvantageous to congress, but also that in such institutional context the probability that a party close to the president joins the government should increase even more. In other words, when the balance of power is less favorable to presidents, they are more concerned about the ideological composition (how close partners are to each other) of the coalitions they form. When anticipating a greater emphasis on coalition bargaining rather than unilateral mechanisms, presidents should be particularly concerned about enhancing the cohesiveness of government coalitions. So, the effect of positional closeness to the president is conditional on the institutional context of the country.
Presidents as Formateurs
One could argue that we have presented an argument about presidential reasoning, or wishes; not about coalitional outcomes, because so far we have omitted to study the reaction of non-presidential parties to the president's offer. Some particular party may be close to a given president, and the president may want to include it in his coalition but, what if the party refuses? And here we come to one major difference between presidential and parliamentary systems in terms of coalition building. Most non-cooperative game theoretic models of government formation in parliamentary systems, starting with Baron (1991), analyze a two-step process: in the first step a formateur is selected, and then she makes a series of proposals to other parties. Some of these proposals may be accepted and others rejected. As a result, the formateur may or may not be able to form the government she initially wanted or any government at all. In the case that she is unable to form a government, another formateur will be selected and the procedure is repeated. As a result, the coalition formation game is solved backwards on the basis of the “continuation value” of this two-step procedure where the formateur has to calculate what will happen if she fails and another formateur is selected, and the other parties have to calculate on the basis of both, what will happen if the formateur survives making the offer to another party, and what will happen if she fails and another formateur is selected.
In presidential systems, however, the process follows a one-step game, where the president is always the formateur. As a result, the parties invited to participate in the cabinet have to compare the benefits from participating in government (i.e., receiving the portfolios, policy and agenda influence, and the spoils that may be related to the offer), to those they would obtain by rejecting the president's offer (i.e., receiving nothing during the presidential term while waiting for the next election hoping to be electorally rewarded). Among a clear and present benefit and a conditional long-term gain, we believe that the choice is heavily tilted towards the first.
There are particular circumstances, however, where parties may be more inclined to reject such offers. In these instances, the perceived costs of joining the president's cabinet must be high if parties are to forgo the temporary gains in policy influence and office benefits. Political contexts considered unfavorable for the government, such as an economic downturn or a period of low approval ratings for the president, are likely to make some parties more reluctant to join the cabinet. But these same circumstances are likely to make presidents more prone to reshuffle the cabinet, creating greater opportunities for parties to join the government.
One variable that can make parties more or less reluctant to accept a presidential offer is the electoral cycle. The benefits parties receive from joining the cabinet are presumably lower further into the president's term, when new programmatic initiatives are less likely to be brought forward and office perks can be enjoyed for a shorter period of time. This scenario is consistent with the literature on presidentialism that has often noted a tendency of presidents to gain opponents and lose support with time in office. After the president's election fades away in voters’ memories and the proximate contest gains attention, the incentives parties have to wait until the next round of government formation may prevail. It is in this context that parties will be less prone to accept the president's offer of a cabinet position, waiting for a better deal after the next presidential election. Thus, we expect parties to be less likely to enter the government when a cabinet is reshuffled late in the president's term.
Empirical Analysis
To empirically examine how positional and institutional factors influence coalition formation, we follow Warwick (1996, 1998) in analyzing the probability that a party participates in a government coalition (i.e., the cabinet). The data we utilize to test our propositions includes information on 106 governments (59 presidents) in 13 presidential countries from the 1980s to the early 2000s. 15 Data on the partisan composition of cabinets comes from Amorim Neto (2006); according to his coding, a new presidential cabinet is defined as (a) the inauguration of a new president, or (b) a change in the party composition of the cabinet. We also collected data on the share of seats held by all parties represented in the legislature (with at least 1 percent of seats), and their ideological positions. Positional data on these parties comes from Coppedge's (1997) left-right classification of political parties. Presidents are classified as their parties, 16 and from this basis we build measures of presidential non-centrality and party distance from the president.
See footnote 4 for a list of countries and time periods included in the analysis.
We understand that this is not an ideal situation, but no alternative dataset exists that includes data on such a large number of parties and time periods as well as separate classification for parties.
The variable distance equals the squared distance between the president and the party divided by the sum of all parties’ squared distances from the median legislative party. We expect distance from the president to be negatively associated with government participation. 17 Presidential non-centrality is calculated as the squared distance of the president from the median legislative party divided by the sum of all parties’ squared distances from the median legislative party. This control variable should capture whether parties are more likely to enter a coalition government when there is a non-centrist president. Both of these measures come from Warwick (1998), who took the non-centrality measure from Crombez (1996). 18 These measures of relative distance pick up variations in ideological closeness given the ideological polarization of the party system (denominator). For example, holding the squared distance from party X to the formateur (i.e., the president) constant, greater fragmentation and ideological dispersion lower the relative distance score, while fewer parties and little ideological dispersion augment it. Omission of the denominator would produce the same score for distance, despite the fact that in one context the president found a close ally from among parties holding distant ideological positions, while in the second the president found a rather distant ally from among a few parties positioned nearby. So, both variables capture ideological distance given the polarization of partisan positions in the country's legislature. 19
The values range from 0 to a maximum of 1.8. While most parties are positioned not far from the president, a few have values above 1 (e.g. the distance between Venezuelan president Luis Herrera Campins of the center-right COPEI and the leftist Movement of the Revolutionary Left was 1.8 in 1978).
By replicating these measures, our results can be compared with those from parliamentary systems.
Upon the suggestion of referees, we replicated our analysis using just the squared distance, without the denominator. Our results regarding the impact of ideological distance to the president, institutional authority, and the electoral cycle were robust to this change in measurement. However, the interaction variable designed to test our third hypothesis did not achieve statistical significance.
Following from our earlier discussion about institutional influences on coalition formation, we use three different proxies for the balance of power. The first one measures the formal powers of the executive. It is a ranking developed by Samuels and Shugart (2003), which classifies countries according to the president's power over legislation. The second one measures the formal powers of congresses. It is a ranking developed by Alcántara et al. (2005), which classifies countries according to nine parliamentary oversight mechanisms. The third one measures congressional effectiveness. This institutional indicator is a perception-based index of congressional effectiveness provided by the World Economic Forum's Global Competitiveness Report. This organization polls chief executive officers or top-level managers in each country from a sample of companies including domestic firms that trade in foreign markets, units of foreign firms operating in the domestic market, and enterprises with significant government ownership. This indicator is constructed from the survey question “How effective is your national parliament/ congress as a law-making and oversight institution?”, and scored from 1 = “very ineffective” to 7 = “very effective, the best in the world”. 20 While we report results only for the three institutional indicators previously mentioned, we also examined the effect of several alternative measures of relative authority. 21
We use the log of the country effectiveness score, and for each country the score is the average response. We use data from five surveys (2002-06), with data made available by the Inter-American Development Bank at http://www.iadb.org.
As alternatives for formal presidential power, we try a comprehensive ranking of constitutional powers of the president developed by the United Nations Development Program (2005: 92), and a ranking associated with presidential agenda-setting powers derived from veto rules (Tsebelis and Alemán 2005). As alternatives for formal congressional authority, we try a categorical variable indicating whether ministers can be censured and forced to resign, and a ranking developed by Fish and Kroenig (2008), which classifies legislatures according to a variety of formal powers. As alternative for congressional competence, we try an indicator of congressional experience based on the average number of years spent in office provided by the Inter-American Development Bank's Economic and Social Progress in Latin America (2006) report.
In addition to variables capturing distance and institutional strength, we include another variable indicating the time of the particular cabinet formation. This is measured as months since the president's election divided by the length of president's term in office. This electoral cycle variable should capture whether parties are less likely to join the cabinet during reshuffles that occur late in the president's term, as we hypothesized earlier in the paper.
The statistical model includes five additional control variables. The first measures the size of the president's party. 22 Presidents with parties that have sufficient seats to control congress should be less inclined to seek out other parties as coalition partners than minority presidents. This is consistent with evidence from parliamentary systems (size of formateur's party) and the arguments made about coalition making in presidential systems (e.g., Foweraker 1998; Amorim Neto 2006). In the sample examined here, the share of seats for the president's party goes from a minimum of zero to a maximum of sixty percent. We operationalize the president's partisan support in congress with a categorical variable indicating whether the president's party has at least 45 percent of seats in the lower chamber. This operationalization follows Jones’ (1995: 37, 193) proposed “cut-point for strong legislative support,” which considers presidents with near majorities (i.e., ≥45 percent) as able to act as majority parties due to the supposed easiness with which they can find support among small parties and independents. Using other measures of partisan strength, such as a dummy for majority support in congress or the share of partisan seats, does not change the main findings.
If it is a bicameral country, then we use information on the lower chamber.
The second measure controls for partisan fragmentation in the legislature (effective number of parties). It seems reasonable to expect that the probability of any one party participating in the cabinet would be less with more parties being represented in the legislature. 23 The third measure is a dummy variable that accounts for the president's party; we should note that although it is common to expect the president's party to be in the president's cabinet, this is not always the case. The fourth measure is a dummy variable indicating whether the party participated in the prior presidential cabinet. Prior membership in the cabinet has traditionally been conceived as a proxy for reliability as a government partner. But more importantly perhaps, this variable allows us to control for the re-selection of parties already in the cabinet during reshuffles.
This is the opposite of what we would expect had the dependent variable be whether a coalition is formed instead of whether the party participates in the government coalition.
Lastly, we control for small parties with a dummy variable for parties with a share of seats equal or below 5 percent. We expect these parties to be less likely to join the cabinet for two reasons: first, small parties cannot provide a substantial block of votes to the president, and second, small parties are more easily “bought” or “co-opted” by the president on a bill-by-bill basis than large parties, making portfolio payoffs unnecessary. While these partisan actors may play a relevant role in contributing to the success of governments on the floor of congress, their small membership, multiplicity, and coexistence with independents and opposition-party rebels, weakens their bargaining potential. This is the rationale, for example, of Jones’ (1995) 45 percent cut-point for strong-legislative support utilized in his comparative analysis of presidential systems. In regards to parliamentary countries, Warwick (1996) hypothesizes the opposite, which is that coalition partners would tend to be small. His argument is based on the notion that the formateur would tend to belong to a large party seeking smaller partners to reach a majority. In parliamentary countries, small parties with slim chances of themselves becoming formateurs if there is a next round of government formation, appear particularly receptive to such offers. But, in the absence of institutional votes of investiture or confidence very small parties are not likely to emerge as pivotal, and if they become sufficiently relevant for reaching agreements on the floor of congress, presidents usually have at their disposal other tools to reach ad-hoc alliances with them that do not include giving up cabinet portfolios.
To examine the probability of a party participating in the cabinet we utilize a random intercept model that accounts for intra-country correlation in the data. It assumes that groups (countries) are random elements of a larger population of groups (e.g., countries with presidential system). The model contains a dichotomous dependent variable γij indicating whether party i belonging to country j entered the cabinet. It is assumed to be distributed as a Bernoulli distribution (nij = 1), and the probability of success π = P(γij = 1) is modeled using a logit function. 24 The random intercept reflects the effect of omitted country-specific variables that cause some countries to be more prone to form cabinet coalitions. The hierarchical structure of the model allows us to evaluate the effect of the institutional variables that do not exhibit within country variation.
The random country effects are latent variables rather than statistical parameters, and are not estimated directly. However, we can calculate the posterior modal predictions and related standard errors after estimating the statistical model.
Results
Table 1 summarizes the statistical results for three alternative models. The first model includes two variables capturing formal powers, while the second model incorporates the congressional effectiveness variable instead of the one capturing formal congressional powers. The third model includes the interaction effect. Standard errors appear below each coefficient, with the log likelihood, Akaike information criterion (AIC), and number of cases shown in the last rows.
Political Parties and Cabinet Participation in Presidential Systems
1. Coefficient significance level:
p>.01 =
p>,05 =
p> .10 =
a. The total effect of distance (including its interaction effect) is significant.
Source: Authors’ own compilation.
The institutional variables reveal interesting results. The variable measuring formal congressional powers is not significant, while the variable measuring formal executive power is significant in Model 1, but not in the expected direction (i.e., suggesting parties are slightly more likely to enter the cabinet when the power of the executive is high). When we replace the variable measuring formal congressional power for the one measuring congressional effectiveness (Model 2), the coefficient for formal executive powers loses significance. The variable capturing congressional effectiveness is highly significant and in the expected direction, which we take as an indication that parties in more capable legislatures are more likely to participate in government coalitions than parties in weak, amateurish legislatures. 25
Replacing this variable for the one measuring membership experience (i.e., tenure) provides similar results. Other variables capturing formal presidential authority and electoral incentives, such as the measures of overall presidential power, agenda-setting power, ability to censure ministers, and election under majority-runoff, do not show significant results. Results are available upon request.
The results confirm our expectation that the greater the ideological distance between a party and the president, then the lower the likelihood will be that the party participates in the cabinet. The variable measuring distance from the president has a statistically significant and negative coefficient. The chance that a non-presidential party located at an average distance from the president will become part of the cabinet is around 21 percent when the president has only minority support in congress. 26 In the same context, the probability of participating in the cabinet is around 9 percent for a party standing far from the president (i.e., two standard deviations from the mean), while it is 27 percent when the position of the party coincides with that of the president.
This result is for a cabinet formation episode that takes place at an average time (1/5 into the president's term in office), and for parties that are not small and were not in the prior cabinet. The other variables are at their means. It is based on results from Model 3.
The results from Model 3 shed light on the interaction between institutional influence and relative ideological distance from the president. It includes the variables distance, congressional effectiveness, and the interaction term. As hypothesized, this interaction is statistically significant and in the direction expected, and the results are robust whether we use congressional effectiveness or membership experience as the indicator of congressional strength. The results lend support to the proposition that the impact of relative ideological distance varies with the institutional relevance of congress.
The effects of distance conditioned on congressional effectiveness can be seen in Figure 3. The solid lines indicate the predicted probability that a non-presidential party will join the cabinet when it is positioned at an average distance from the president and when it is positioned far from the president (i.e., two standard deviations from the mean). The dashed lines indicate the 90 percent confidence interval for each prediction.
Institutional Context and Cabinet Participation
For parties located at an average distance from the president, the probability of entering the cabinet increases markedly with the effectiveness of congress, while for parties located far from the president the probability of entering the cabinet is low and unchanged by congressional effectiveness. Moreover, the figure shows that the impact of distance is relevant at medium and high levels of congressional strength, but not at very low levels. When congressional effectiveness drops below 0.7, for example, the effect of distance begins to lose statistical significance. In Figure 3, this is the area where the confidence intervals for the predicted probabilities overlap.
The results also lend support to the view that parties are less likely to enter the cabinet late into the president's term of office. The variable measuring the impact of the electoral cycle on the likelihood of cabinet participation is significant and in the expected direction. The probability that a non-presidential party located at an average distance from the president will become part of the cabinet is close to 23 percent when the cabinet formation episode takes place 1/10 into the president's term. 27 The chance that the same party will join the cabinet 8/10 into the president's term is around 11 percent. This finding is consistent with the argument that presidents tend to lose support over time, which implies a lower tendency to accept presidential offers when a cabinet reshuffling occurs late into the president's term of office.
This result is for a cabinet formation episode that takes place when the president has minority support in the legislature, and for parties that are not small and were not in the prior cabinet. The other variables are at their means.
With respect to the control variables, three appear consistently significant and in the expected direction. Results show that (i) very small parties are significantly less likely to participate in the cabinet than larger ones, (ii) the president's party is significantly more likely to join in than other parties, and (iii) reshuffles tend to bring many of the parties already in the cabinet. These results lend support to the view that presidents in search of portfolio coalitions are likely to reach out to non-presidential parties with at least a moderate share of parliamentary seats.
The coefficient capturing whether the president's party has a working majority in congress is negative, as expected, but it is only significant in the final model (Model 3). Using a dummy variable for majority presidents or the share of seats for the president's party instead of a working majority does not improve results – both coefficients lack statistical significance. The coefficient for partisan fragmentation has the correct sign and is significant in Models 2 and Model 3, suggesting that all else equal, the greater the number of potential coalition partners, the greater the chance that one of them will join the cabinet. The coefficient for presidential non-centrality has the expected sign, but it is not significant. The tendency of non-centrist presidents to form coalition cabinets does not appear to be sufficiently different from the tendency of those positioned closer to the center.
Discussion
The empirical analysis shed light on the forces affecting coalition government in presidential systems. We showed that institutional influence exerts a significant effect on coalition making, with parties’ likelihood of cabinet participation rising with congressional strength. 28 More importantly, the analysis lends support to our argument that institutional influence affects the relevance of partisan ideology for cabinet formation. The results showed not only that parties standing ideologically far away from the president are significantly less likely to enter the cabinet than those positioned in the president's proximity, but also that the impact of ideological distance on cabinet formation is more pronounced in more relevant and professionalized congresses.
In the sample of presidential democracies studied here, the only cases where presidents facing strong congresses (ranked above the median in terms of legislative effectiveness) exclude members of non-presidential parties from their cabinet are those where the president's party had a substantial share of seats. The three Latin American cases in this group were Colombian and Mexican presidents with more than 49 percent of seats. The others were US presidents. Excluding the US from the sample does not alter the main findings of the empirical analysis.
While the statistical model provides a very good fit, 29 some outliers deserve further comment. Two unusual observations refer to parties that “should have entered” the cabinet but did not. Both outliers are parties associated with the president in office: one is the Peruvian party Cambio 90 (C90), which was excluded from a cabinet formed in 1990 by President Alberto Fujimori, and the other is the Ecuadorian party Partido de Unidad Republicana (PUR), which was excluded from a cabinet formed in 1993 by President Sixto Durán-Ballén. The strongest predictors suggest these parties should have entered the cabinet. Why were they excluded from government?
Model 3 correctly predicts 83.6 percent of cases.
Part of the explanation has to do with the limited supply of politicians with sufficient expertise and political weight to fill cabinet positions. Both parties were newly created organizations formed around the charismatic figure of presidential contenders. The Peruvian C90, created by Fujimori before his election as president in 1990, was a personalist party composed of a small and heterogeneous group of supporters without much organizational strength or valued technocrats in its ranks. Its exclusion from government, however, was limited to the first six months of the administration, when Fujimori brought some of his other electoral supporters into the cabinet. The Ecuadorian PUR was created by Durán-Ballén as his personal electoral vehicle. He had been a member of the Partido Social Cristiano (PSC) for many years, and created the PUR in 1991 only after failing to win the nomination of the PSC (Barczak 2001). Soon after creating the party, Durán-Ballén fought with a relevant faction of his supporters, which led to a series of party expulsions and the draining of militants even before taking office. While Durán-Ballén included members of the PUR in the first and third cabinets formed during his four years in office, no members appear in the 1993-94 cabinet (a period when the PUR performed badly in midterm elections).
The empirical findings of this paper are also relevant to long-standing normative concerns in the institutional literature focused on Latin America. Since at least the beginning of the twentieth century, academics and constitutional engineers have debated how to construct the proper combination of authority between the branches of government that would enable the president to act decisively but not govern around congress. This concern has continued with the new institutionalist approaches; Shugart and Carey (1992: 165), for instance, suggest that in legislative matters relatively strong assemblies provide more stable and effective government relative to strong presidents because assemblies serve as arenas of perpetual fine-tuning of conflicts. They remark that most of the problems commonly associated with presidentialism apply with the greatest force to regimes with great presidential legislative powers. Similarly, Cox and Morgenstern (2002: 453) argue that institutionally weak presidents, unable to pursue a unilateral strategy, are more likely to govern through, rather than around, congress.
In most cases, the institutionalist prescription for cooperative inter-branch relations has focused on decreasing the formal powers of the executive. The findings of this paper, however, underline the additional importance of congressional characteristics. Presidents (even those possessing a battery of formal powers) appear more likely to reach out to other parties and form coalition cabinets when they face strong congresses – composed of professional legislators capable of meaningful oversight – than when they face amateurish institutions. It is perhaps through measures that foster the professionalization of congress, and not solely through sweeping constitutional reforms, that a more workable inter-branch institutional structure may be achieved.
Conclusion
To conclude, this paper has shown that coalition building in presidential systems tends to follow a similar logic to parliamentary systems, despite the fact that in the first lawmaking is independent of cabinet survival, while in the second they are intrinsically linked. Actors in both systems are concerned with policymaking, and consequently use the properties of the political system (institutions as well as policy positions of the other actors) in an optimal way to promote their goals. Presidents dealing with weak congresses are not compelled to make coalitions, while presidents facing strong ones hold greater concerns about how their proposals will fare in congress, and pay more attention to coalition building. Once the coalition building phase begins, parties closer to the president are more likely to receive an offer to participate and accept such an offer, while distant parties seldom become members of a presidential coalition. Another relevant implication from comparing regimes is that in presidential systems the key institutional influence stems mainly from the importance of congress, while in parliamentary systems it centers in the agenda setting powers of the cabinet. This may be due to the lack of substantial variation of professionalism in parliamentary systems, while in presidential systems congresses have significant rates of variation along this dimension. As argued in this paper, this variation is highly consequential for coalition formation.