The Pennsylvania State University. The Graduate School. College of the Liberal Arts THE POLITICS OF INTERRUPTED PRESIDENCIES IN THIRD WAVE DEMOCRACIES

Transcription

1 The Pennsylvania State University The Graduate School College of the Liberal Arts THE POLITICS OF INTERRUPTED PRESIDENCIES IN THIRD WAVE DEMOCRACIES A Dissertation in Political Science by Young Hun Kim 2008 Young Hun Kim Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy August 2008

2 The dissertation of Young Hun Kim was reviewed and approved* by the following: Donna L. Bahry Professor of Political Science Dissertation Advisor Chair of Committee Head of the Department of Political Science Michael H. Bernhard Professor of Political Science Gretchen Casper Associate Professor of Political Science Quan Li Associate Professor of Political Science Mark Anner Assistant Professor of Labor Studies and Employment Relations *Signatures are on file in the Graduate School

3 iii ABSTRACT Comparative researches on presidential systems have largely focused on regime survival whether presidential regimes remain democratic and for how long compared to parliamentary systems in Third Wave democracies. And many scholars seem to have reached a conclusion that presidentialism overall is more vulnerable because of some classic assumptions of presidentialism: divided government, the potential for deadlock, and the chief executive s fixed term. However, the experience of Third Wave democracies suggests that some of these classic assumptions may need to be revised, especially those regarding the inflexibility of the president s term. While constitutions may prescribe a fixed term, in practice a substantial number of Third Wave presidents have left office prematurely. Out of 52 Third Wave democracies with popularly elected presidents, 19 (about 40 percent) have experienced the resignation or removal of a president between 1974 and 2003, some more than once. Since roughly 75 percent of the new democracies from 1974 on have presidential/semi-presidential systems, understanding the roots of presidential vulnerability is a critical issue. My project examines two related questions. Using original data on presidential interruptions in all new democracies in Asia, Africa, Europe and Latin America from 1974 through 2003, I first ask what makes presidents vulnerable to interruptions defined as when a duly elected president leaves office by impeachment or parliamentary vote, resignation, or force. I show that interruptions are more likely where presidents face divided government, a fragmented party system, and an imbalance of presidential powers

4 iv relative to other branches. A low share of the first-round presidential vote and discontent in the streets also make interruptions more likely. Poor economic performance adds another risk. I find, too, that presidential survival is separate from democratic survival in most regions of the world, with one significant exception. Democratically elected presidents in sub-saharan Africa tend to be removed by force, with the result that democracy itself is typically suspended. In contrast to presidential interruptions, the question of impeachment attempts remains almost unexplored, mainly due to high barriers to successfully removing a president by impeachment. But such attempts are not uncommon (about 45% of Third Wave presidential democracies have experienced some type of presidential impeachment attempts), and about 60% of interrupted presidencies faced some forms of impeachment attempts beforehand. In this regard, my second question asks when legislators attempt to initiate an impeachment and how and why their attempts develop. Using an ordered logit for five different levels of impeachment attempts, I find that legislators are more likely to challenge presidents, and the challenge often develops further, where the chief executive controls fewer seats in parliament and the level of party fragmentation is high. Presidents elected with a lower vote share in the first-round presidential election are also more vulnerable to impeachment attempts. Press freedom is critical, in the sense that more freedom decreases the odds of an impeachment attempt. Presidential scandal and protest add another risk, and legislators swiftly respond to the public discontents heard from the streets.

5 My findings in this research have several implications for the study of v presidentialism and democratic consolidation. First, as my analysis shows, political crisis in Third Wave presidential systems did not imperil the survival of democracy in many cases, contradicting expectations in earlier critiques of presidentialism. Governmental crisis may occur, but in most regions of the world outside sub-saharan Africa democratic regimes remain intact. The reality is that presidentialism is flexible. It is able to respond to political crisis by removing presidents who do not enjoy public and/or parliamentary support before the end of their term and most interruptions lead only to an interim presidency, and a subsequent election. Therefore, the critiques are mistaken in suggesting that political crisis in presidentialism is likely to lead to democratic breakdown. Similarly, the frequent attempts to impeach presidents should be understood in the context of overcoming political crises in presidentialism. I do not deny that presidentialism is exposed to a high probability of political crisis due to dual legitimacy, divided government, and high exclusivity. What I emphasize is that when political crisis emerges, interruptions and impeachment attempts can work to overcome the crisis. In this regard, the contrast between the inflexibility of presidentialism and the flexibility of parliamentarianism appears overemphasized.

6 vi TABLE OF CONTENTS LIST OF FIGURES... viii LIST OF TABLES... ix ACKNOWLEDGEMENTS... xi Chapter 1 Introduction Interrupted Presidencies in Third Wave Democracies Introduction Sources of Presidential Vulnerability Prior Studies of Presidential Interruptions How Does This Study Differ? Attempted Presidential Impeachments in Third Wave Democracies Introduction Attempted Presidential Impeachments and their Development Sources of Attempted Presidential Impeachments Prior Studies of Attempted Presidential Impeachments How Does This Study Differ? The Structure of This Study Chapter 2 Data and Methodology Rules on the Case Selection Why All Third Wave Countries? Identification of Democratic Transition Why Polity IV? Definition of Presidentialism Operationalization of Democratic Transition in Presidentialism Beginning and Ending of a President s Term Other Issues List of Countries in the Analysis Description, Operationalization, and Sources of Key Variables Interrupted Presidency Attempted Presidential Impeachments Stage One: A Call to Impeach the President Stage Two: Endeavoring to Put the Question on the Parliamentary Agenda Stage Three: Parliament Holds a Vote Stage Four: Parliament Passes an Impeachment Vote Independent Variables... 63

7 The First-round Popular Vote for President: Presidential Mandate Parliamentary Support: Divided Government Effective Number of Parliamentary Parties: Multipartism Executive Constraints: Presidential Powers Economic Performance Scandal Protest Protest Timing Press Freedom Time Spent in Office Mixed System Latin America Descriptive Statistics and Correlation Matrix of Variables Methodology Logit and Ordered Logit Analysis Survival Analysis Chapter 3 Interrupted Presidencies in Third Wave Democracies Introduction Assessing the Roots of Presidential Interruptions: Logit Analysis Survival Analysis Different Modes of Interruptions Competing Risks Model Conclusion Chapter 4 Attempted Presidential Impeachments in Third Wave Democracies Introduction Assessing the Sources and Development of Attempted Presidential Impeachments Conclusion Chapter 5 Conclusion Bibliography vii

8 viii LIST OF FIGURES Figure 3-1: Predicted Probabilities of Interrupted Presidencies Conditional on Seat Share of Presidential Party Figure 3-2: Predicted Probabilities of Interrupted Presidencies Conditional on First-round Popular Vote Figure 3-3: Predicted Probabilities of Interrupted Presidencies Conditional on Executive Constraints Figure 3-4: Predicted Probabilities of Interrupted Presidencies Conditional on Effective Number of Parliamentary Parties (ENPP) Figure 4-1: Predicted Probabilities of Attempted Presidential Impeachments (API) Conditional on the Presidential Party s Seat Share in Parliament Figure 4-2: Predicted Probabilities of Attempted Presidential Impeachments (API) Conditional on the First-round Popular Vote for president Figure 4-3: Predicted Probabilities of Attempted Presidential Impeachments (API) Conditional on Effective Number of Parliamentary Parties (ENPP)

9 ix LIST OF TABLES Table 1-1: The Distribution of Presidents Parliamentary Support... 7 Table 1-2: The Distribution of Presidential Scandals across Region Table 1-3: The Distribution of Presidents First-round Popular Vote Table 1-4: A Cross-tabulation of Attempted Presidential Impeachments and Interruptions Table 2-1: Third Wave Presidential/Semi-Presidential Democracies Included in the Analysis Table 2-2: Interrupted Presidencies in Third Wave Democracies, Table 2-3: List of Attempted Presidential Interruptions in Third Wave Democracies, Table 2-4: Brief Definition and Summary Statistics of Variables Table 2-5: Correlation Matrix of Variables Table 3-1: Distribution of Presidential Interruptions by Regions Table 3-2: Distribution of the Mode of Interruptions by Regions Table 3-3: Determinants of Interrupted Presidencies in Third Wave Democracies, (Logit Analysis with Country Random Effects) Table 3-4: Cross-tabulation of Economic Downturns and Interruptions Table 3-5: Cross-tabulation of Protest and Presidential Interruptions Table 3-6: Determinants of Interrupted Presidencies in Third Wave Democracies, (Survival Analysis Using Lowess Fits) Table 3-7: Determinants of Interrupted Presidencies in Third Wave Latin American and Sub-Saharan African Democracies, Table 3-8: Determinants of Interrupted Presidencies in Third Wave Democracies, (Competing Risks Model) Table 4-1: Distribution of Democratic Country-years by Attempted Presidential Impeachment and Region

10 Table 4-2: Determinants of Attempted Presidential Impeachments in Third Wave Democracies, Table 4-3: Percentage Change in the Odds of Higher Impeachment Attempts Outcomes for Standard Deviation Increase in the Key Variables Table 4-4: Cross-tabulation of Attempted Presidential Impeachments and Press Freedom (All Countries) Table 4-5: Cross-tabulation of Attempted Presidential Impeachments and Press Freedom by Region (except for Southern Europe) Table 4-6: Frequencies for Different Levels of Impeachment Threshold Table 4-7: A Cross-tabulation of Attempted Presidential Impeachments and Interruptions Table 4-8: Determinants of Attempted Presidential Impeachments in Third Wave Democracies, (Logit Analysis with Country Randon Effects) Table 4-9: Determinants of Attempted Presidential Impeachments in Third Wave Latin American and Sub-Saharan African Democracies, (Logit Analysis with Country Randon Effects) x

11 xi ACKNOWLEDGEMENTS I am deeply indebted to my advisor, Donna Bahry, for her mentorship and guidance during the last several years. She has been very generous with her time, and has always encouraged me with her advice and suggestions. Most importantly, she actively demonstrated the pleasure of doing research, its enthralling virtues. I consider myself very fortunate to be counted one of her students. I would also like to thank my committee members, Michael Bernhard, Gretchen Casper, Quan Li, and Mark Anner. Their invaluable comments helped me fine-tune the theoretical arguments and empirical tests in this dissertation. In addition to my committee, I would like to thank the faculty, my colleagues, and the staff at the Political Science Department at Penn State. They have contributed to my dissertation in various ways and made my life at Penn State enjoyable. And Sushmita Chatterjee kindly took pains to read this dissertation several times. Finally, I reserve my deepest gratitude for the most important people in my life, my parents and my little daughter, Ga-Hyeon Kelly. Their unwavering support is immeasurably important to me, and is the key reason I have been able to accomplish what I have. And I will never be able to sufficiently thank my wife, Hyeon Ju for the personal sacrifices she made for me. It is to her that I dedicate this work.

12 Chapter 1 Introduction 1.1 Interrupted Presidencies in Third Wave Democracies Introduction Since the global trend of democratic transition dubbed the Third Wave (Huntington 1991), much comparative research on democratization has focused on the survival of different regime types whether and for how long presidential and parliamentary system remain democratic (Cheibub 2002a; Colomer 2001; Linz 1990, 1994; Mainwaring 1993; Mainwaring and Shugart 1997b; Power and Gasiorowski 1997; Przeworski, Alvarez, Cheibub and Limongi 2000; Shugart and Carey 1992; Stepan and Skach 1993; Valenzuela 1993, 1998). And until recently, many scholars seem to have reached a conclusion that presidentialism overall is more vulnerable to the breakdown of democracy because of some inherent characteristics of presidentialism (cf. Linz 1990, 1994). In principle, presidentialism relies on an executive with a fixed term of office, which is supposed to increase government stability and guard presidents against the uncertainty so characteristic of parliamentarianism. The constitutionally guaranteed term presumably allows presidents to see their programs through to the end of their term with little threat of unexpected removal. Besides, presidentialism is supposed to provide clear

13 accountability, as well. Because voters can easily identify that presidents are mainly 2 responsible for national performance, particularly in the economic area, presidents have strong motivations to try to perform well in office. Also, presidentialism in theory is supposed to encourage the creation of governing coalitions out of disparate blocs of parties and voters because presidents are both the head of state and the head of government. 1 But in practice, presidents often face a number of problems. They may confront situations of divided government where presidents fail to control a majority in parliament, making it difficult to find consistent legislative support. And they may not forge coalitions, or forge stable ones, because their survival is largely independent of parliamentary support. These problems are especially worrisome due to a president s fixed term. Chief executives remain in office no matter how little legislative and public support they command (Linz 1990, 1994; Shugart and Carey 1992; Valenzuela 1998, 2004). Given this rigidity in presidential terms, political elites with dual legitimacy are likely to be in conflict and search for extraconstitutional means to resolve the impasse. This is why some suggest that presidential democracy is relatively short-lived in new regimes. However, the experience of Third Wave democracies suggests that some of these classic assumptions about presidentialism need to be revised, especially the one regarding the inflexibility of the president s term. Even though presidents have a constitutionally 1 Some suggest that this is not always the case. For example, Amorim Neto (2006) argues that presidential coalition building is dependent on the president s preferences, institutional incentives, and economic conditions.

14 prescribed term, many leave office prematurely. Indeed, out of 52 Third Wave 3 democracies with popularly elected presidents, 19 (over a third) countries have experienced the resignation or removal of a president (which I refer to as an interrupted presidency in this research) between 1974 and For example, in Brazil Collor resigned in 1992 after the lower house of parliament voted to impeach him and before the senate could take further action (Pérez-Liñán 2007). He had been accused of criminal charges of racketeering, forging documents, influence-peddling and numerous other corruption-related activities. Ecuador s president Bucaram was removed by parliament in 1997 for mental incapacity (Pérez-Liñán 2007). Paraguayan President Cubas resigned in 1999 following allegations that he was connected with the assassination of the vice president who was backed by the opposition parties in the lower house (Abente-Brun 1999). Interrupted presidencies also extend beyond Latin America. Philippine President Estrada was forced to leave office early in 2001 after an impeachment vote in the lower house, an inconclusive trial in the upper chamber, and massive street unrest calling for his removal (Kasuya 2003). In the post-communist region, Georgian President Eduard Shevardnadze resigned in the aftermath of tainted elections and mass demonstrations in 2003 (Fairbanks 2004). In Africa, the pattern of presidential interruption was more violent. President Ialá in Guinea-Bissau was one of many African presidents removed from office after a military coup, leading to democratic breakdown. One exception was Malagasy President Zafy in Amid economic decline and allegations of corruption 2 For a list of Third Wave presidential countries included in this study, see Table 2-1 in Chapter 2. And Table 2-2 provides a list of interrupted presidencies.

15 and abuse of power, he left office prematurely after the legislature authorized an 4 impeachment trial and the High Constitutional Court confirmed it (Allen 2003). Also, it is worth noting that some countries such as Dominican Republic, Ecuador, Bolivia, and Argentina experienced interrupted presidencies more than once (see Table 2-2 in Chapter 2). In this research, I examine two key questions. The first is why some presidents are more vulnerable to interruptions than others. What factors help to explain why some complete their terms while others fall? Since about two-thirds of the new democracies from 1974 on have presidential or semi-presidential systems, understanding the roots of presidential vulnerability is important for students of democratic transition and consolidation. Moreover, interrupted presidencies may have various effects on the new democracies. Interrupted presidencies might hinder democratic consolidation and economic development because removing presidents involves a sudden governmental change, if not a regime breakdown, and this could deepen the political crisis and instability. Alternatively, interrupted presidencies may have positive effects because many presidents who are removed are allegedly involved in political scandals and incapable of running the national economy properly. Therefore, by replacing presidents with a presumably cleaner and more capable political figure, a country may increase its chances of consolidating democracy and laying the foundation for economic development. Also, it is reasonable to expect that interrupted presidencies may have an effect on presidentialism per se by lessening term rigidity and changing executivelegislative relations (Marsteintredet and Berntzen 2006).

16 5 Existing research on presidentialism thus far has focused more on the duration of newly established democracies, and much of the evidence appears to suggest that presidentialism is more fragile (Cheibub 2002b; Cheibub, Przeworski and Saiegh 2004; Linz 1990, 1994; Przeworski et al. 2000; Samuels and Eaton 2002; Stepan and Skach 1993). 3 Attention to interrupted presidencies is much more recent (Cheibub 2002b; Edwards 2007; Hinojosa and Pérez-Liñán 2003; Hochstetler 2006; Negretto 2006; Pérez- Liñán 2003, 2007; Valenzuela 2004). 4 But, the existing literature (to be discussed in greater detail below) offers divergent conclusions on several key factors particularly on the impact of economic conditions and of divided government; and it neglects some important variables. In addition, much of the work focuses on Latin America. 5 While that provides us with some important evidence about presidential vulnerability, it leaves open the question of what drives interruptions in other new democracies elsewhere. My first aim in this study is to reassess the roots of presidential vulnerability across a broader range of new democracies in Europe, Asia, Africa 6 and Latin America, from 1974 through I reexamine several core arguments that have yielded 3 There also has been a growing body of research that emphasizes different levels of fragility of presidentialism across countries, types of presidential systems, and successive presidents in the same country (e.g., Amorim Neto 2002; Cheibub 2002b; Cox and McCubbins 2001; Jones 1995; Mainwaring 1993; Mainwaring and Shugart 1997; Negretto 2006; Samuels and Eaton 2002; Shugart and Carey 1992). 4 There have also been a number of case studies of interrupted presidencies in individual countries. See, for example, Coslovsky (2002) and the essays in Baumgartner and Kada (2003). 5 Cheibub s (2002b) study appears to be the only exception, covering all presidential countries. 6 Africa in this study is limited to sub-saharan countries and I use the terms Africa and sub-saharan Africa interchangeably.

17 6 inconsistent results in previous studies; and I add new data on the impact of presidential scandals and protest. I also explore the impact of a key presidential resource that has been omitted in most assessments of presidential vulnerability the president s share of the popular vote Sources of Presidential Vulnerability The existing literature on presidentialism discusses several factors that may make some presidents more vulnerable than others. First of all, the president s partisan resources in congress are held to be especially important. The inability to command stable majority party support in parliament (divided government) is taken as a major impediment to passing legislation and thus to carrying out the president s agenda (Hochstetler 2006; Linz 1990; Mainwaring 1993; Pérez-Liñán 2003). Without a stable majority, the incumbent may be less likely to govern by statute (Cox and Morgenstern 2002), and more likely to rely on extra-constitutional methods and unilateral powers to implement policies (Jones 1995). Indeed, the situation of divided government is fairly common: about 68 percent of chief executives in Third Wave presidential democracies have lacked a majority in the legislature (see Table 1-1 below).

18 7 Table 1-1: The Distribution of Presidents Parliamentary Support Parliamentary seats held by the president s party Less than 25% 20.95% (31) 25% to less than 50% 46.62% (69) Proportion of Seats 50% to less than 65% 23.65% (35) 65% or more Total 8.78% (13) % (148) Note: i) A president s last year in office was used to calculate the share of parliamentary seats held by the president s party; ii) Numbers in parentheses are frequencies. Sources: Adam Carr s Election Archive, African Elections Database, Election Guide, Election Results Archive, Political Database of Americas The reason why divided government raises the prospect of opposition parties ranged against the chief executive can be found in inherent characteristics of presidentialism: dual legitimacy and a zero-sum game (Linz 1990). Both the legislature and the president enjoy legitimacy derived from general elections. Theoretically, the separation of powers between the executive and the legislature is devised to protect individual liberty of each institution and promote checks and balances (Lijphart 1992), but it can also justify any attempt by the parliament to confront a president in the name of the people, especially when the opposition parties control a majority of the seats. For instance, the legislature may return bills suggested by a president by arguing that the bills do not reflect the will of people, and a president in turn might resort to unilateral and extra-constitutional powers to promote his agenda. Political deadlock would emerge and the legislature might be more disposed to solve this stalemate by removing a president or forcing a president to resign. In addition, presidentialism is often characterized as a system of high exclusivity : the almost zero-sum nature of presidential elections means that winners in presidential governments have less incentive to compromise with the losers because government survival is not necessarily dependent on majority support in the legislature (Valenzuela 1998). Losers preferences could be excluded until the next

19 8 presidential election. In turn, opposition parties may come to believe that support for the president s policies in the legislature would redound to the benefit of the chief executive, not to that of his temporary congressional allies (Valenzuela 1998: 129). They may believe that their own political interests can best be guaranteed by the political failure of the incumbent (Valenzuela 2004). High fragmentation in the legislature is another liability to a president. 7 As Mainwaring (1993) and Bernhard et al. (2001) suggest, multipartism has negative effects on new presidential democracies because it can increase ideological polarization and the probability of deadlock between executive and legislature. 8 As a result, a chief executive is more likely to resort to extra-constitutional ways to implement policies as in divided government. Moreover, the fact that congress is highly fragmented makes it less likely that any party controls a majority, making bargaining between the executive and the legislature extremely complicated and the maintenance of a stable coalition more difficult (Jones 1995; Mainwaring 1993; Mainwaring and Shugart 1997b; Valenzuela 2004). Nonetheless, some may suggest that the argument is somewhat misleading, because presidents can attempt to form coalitions, especially when they experience a 7 As a matter of fact, fragmentation and divided government tend to be highly correlated (Jones 1995; Samuels and Eaton 2002), and for this reason Mainwaring (1993) uses party fragmentation as a proxy for divided government. But, it should be emphasized that they are conceptually distinct. Government might be divided even at low levels of ENPP, because an opposition majority controls congress. It is more likely to be divided, however, when ENPP is high. In my data, the correlation between the two measures is about However, Lehoucq (2005), after reviewing previous studies, suggests that the relationship between divided government/multipartism and democratic breakdown is inconclusive at best.

20 9 divided government and multipartism (Cheibub 2006; Cheibub, Przeworski, and Saiegh 2004). However, what they overlook is that the coalitions in presidentialism may not be effective. Amorim Neto (2002) suggests that presidential coalitions may not be cohesive, or translate into stable support in parliament, especially when opposition party members get cabinet posts but the party leaders do not agree to a coalition of parties. 9 In other words, simply appointing members of other parties to the cabinet does not bind the cabinet member s party to any kind of cooperation with the president. And appointments to cabinet posts often do not correspond very closely to the partisan distribution of legislative seats (Amorim Neto 2002; Samuels and Eaton 2002; Shugart and Mainwaring 1997). Presidents might also use appointments to divide and rule, rather than promote cohesive parties (Linz 1994). Thus far, I pointed out negative consequences that multipartism might have on presidents. Yet, a highly a fragmented party system may work to the advantage of presidents since the opposition may not be able to unite against them. Even where presidents are highly unpopular, the opposition parties in high fragmentation may find it difficult to overcome their ideological differences and build a coalition to unseat the chief 9 Amorim Neto (2002) calls this type of coalition non-coalescent. He suggests that presidents do this to co-opt members from other parties and undermine their party leaders. There is also another type of coalition called coalescent where the opposition party leaders agree to a coalition of parties, and their members get cabinet posts. In this situation, the opposition leaders promise to work together with the president to get their policies adopted. However, this type of coalition is rare in presidentialism due to the independence of government survival from parliamentary support.

21 10 executive. In this context, presidents under high fragmentation might be either more or less vulnerable to pressures to step down. While the theory assumes a linear (either positive or negative) relationship between presidential interruption and the levels of party fragmentation, some suggest that the effect of multipartism may not be linear. Cheibub (2002b), for instance, concludes that the odds of presidential crisis is highest in the lower-moderate range of fragmentation (effective number of parliamentary parties ranging from 2 to 5) because the distribution of strength among parties is almost equal, making compromise among them more difficult. I test for the non-linearity effect of multipartism in Chapter 3, as well. Along with divided government and a high degree of party fragmentation, strong formal presidential powers can also affect presidential vulnerability, in one of two ways. On the one hand, strong presidents may be able to stay in office through a combination of constitutionally endowed powers such as veto, decree, and agenda power regardless of unfavorable conditions. They, for example, may be able to rule by decree if they lack a majority support in parliament. 10 On the other hand, presidents facing few institutional constraints may be more vulnerable to pressures for resignation or removal because they are less likely to compromise with other branches (Baumgartner 2003; Shugart and Carey 1992). Given a winner-take-all logic and accordingly the high stakes of holding onto power in presidentialism, presidents may be less willing to compromise with opposition parties 10 This is closely related to the isolation or confrontation strategies presidents might adopt against the legislature when they lack parliamentary support (Pérez-Liñán 2007).

22 (Nel 2003). Moreover, less constrained presidents are typically more likely to rely on 11 executive prerogatives, and this may endanger relations with the legislature (Fish 2006; O Donnell 1994; Valenzuela 2004). There is some disagreement, though, about when this is most likely to occur. Cox and Morgenstern (2002) argue that strong presidents face more difficulty with the legislature when there is high fragmentation or the opposition controls a majority. Then presidents are more likely to press the limits of their authority. In contrast, Cheibub (2002b) offers a different formulation, concluding that presidents are at less risk when their formal powers and the balance of seats in congress create a deadlock. Economic adversity is another source of vulnerability, because the chief executive is assumed to be responsible for handling the national economy (Bernhard, Nordstrom and Reenock 2001; Lewis-Beck 2000; Remmer 1991; Wilkin, Haller, and Norpoth 1997). Thus, economic problems may mean eroding legitimacy of a president (Mainwaring 1999). High inflation and low or declining growth are particularly problematic, since they erode living standards, intensifying inequalities and conflicts over the distribution of resources (Haggard and Kaufman 1995; Mainwaring and Pérez-Liñán 2005). 11 Substantial budget deficits and high debt burdens can generate pressure presidents to adopt austerity measures, and such programs may intensify public and/or elite grievances (Barro 1996; Hochstetler 2006; Nel 2003; Pérez-Liñán 2003). In all, the odds of interruption should rise with economic adversity. 11 The absence of safety nets as in Latin America may magnify the negative consequences of any economic adversity (Amorim Neto 2006).

23 Another body of theory highlights the problems that arise when presidents are 12 embroiled in personal scandal and confront a mobilized and hostile public (Hochstetler 2006; Mainwaring and Pérez-Liñán 2005). As the examples of presidential interruptions presented earlier show, some presidents face accusations of violating the law and/or of personal corruption. Because presidents are considered as chief defenders of the constitution and the laws, personal scandal can be damaging to the incumbent. Charges of misuse of government funds and influence peddling appear to be particularly detrimental, since many presidents campaign on promises to root out government corruption. As Table 1-2 shows, presidential scandals were quite common in all regions except for Southern Europe (Portugal is the only new semi-presidential democracy in the region). But, there is a slight regional variation, as well: scandals were more widespread in Asia and postcommunist countries than in sub-saharan Africa. 12 Table 1-2: The Distribution of Presidential Scandals across Region Presidential Scandal Latin America Yes 8.85% (23) No 91.15% (237) Southern Europe 0.00% (0) % (28) Region Postcommunist region 12.00% (15) 88.00% (110) Africa Asia Total 5.37% (8) 94.63% (141) 13.64% (6) 86.36% (38) 8.58% (52) 91.42% (554) Total % (260) % (28) % (125) % (149) % (44) % (606) Note: The data indicate country-years when the incumbent president was publicly accused of personal malfeasance. Numbers in parentheses are frequencies. Source: LexisNexis Academic Universe 12 The fact that African democracies had fewer incidents of presidential scandals appears somewhat counterintuitive. But, a couple of factors such as the press freedom and political elite behavior might explain why this is the case (cf. Castagnola 2007).

24 13 Protest can be even more problematic for presidents. Protest calling directly for presidential resignation/impeachment indicates that the executive s mandate has eroded and constituents are not willing to wait until the next election. This means that civil society can also play a critical role in pushing presidents out, one that studies of elites alone often miss (Hochstetler 2006). Presidential electoral support may also be important in explaining presidential interruptions. 13 The vote can be an important signal for actors in other branches, to gauge the resources a president may bring to bear when conflicts arise. In this context, a low first-round mandate may persuade opposition leaders in parliament that there is little to gain in cooperating with the government. On the other hand, presidents elected with more votes in the first-round presidential elections can claim preeminence over the legislature when conflicts emerge. Thus, opposition parties may find it more difficult to challenge a president with stronger public support. 14 The vote may also be a factor in presidential calculations about whether and when to go public, for example, with a referendum. 13 Presidential electoral support can have indirect effects on party fragmentation and legislative seats a president will garner. I test this possibility in the analysis in Chapter The South Korean general election results in 2004 exemplify negative consequences of challenging a president with strong public support. The Grand National Party, then the largest opposition party, impeached popularly elected and popular President Roh for incompetence and economic mismanagement just one month before the general election scheduled in April 2004 (Faiola 2004). The subsequent election results were devastating for the opposition Grand National Party. They dropped from controlling nearly two-thirds to less than one third of the 299 seats in the legislature, while the ruling party s (Open Uri Party) seats rose from 47 to 152 (Adam Carr s Election Archives Retrieved from The Courts nullified the decision after the election.

25 14 Many presidents in new democracies come to office with less than a majority of the popular vote: about 64% of elected presidents failed to garner a majority in the firstround presidential elections (see Table 1-3 below). In the case of majoritarian (run-off) elections, a candidate may win the second round by a negative mandate, where many voters strongest preference is simply against the opposing candidate (Carey 2003; Pérez- Liñán 2003, 2006). But, a negative mandate can lead a president to overestimate his or her real degree of popular support when conflicts arise with other branches of government (Pérez-Liñán 2006). This is not to deny that support for the president can change once he or she takes office. 15 I simply want to note that a low mandate at the outset should make both the legislature and the public less willing to tolerate what they see as presidential failures. Table 1-3: The Distribution of Presidents First-round Popular Vote Less than 25% Proportion of first-round popular votes 25% to less than 50% 50% to less than 65% 65% or more Total President s firstround popular vote 8.72% (13) 55.03% (82) 28.19% (42) 8.05% (12) 100% (149) Note: i) The unit of analysis is the presidential term; ii) Numbers in parentheses are frequencies. Sources: Adam Carr s Election Archive, African Elections Database, Election Guide, Election Results Archive, Political Database of Americas Finally, presidents in semi-presidentialism might be less vulnerable than in pure presidentialism mainly because executive powers are shared between a president and a prime minister. In this system, the potential for conflicts between a president and congress could be lower. Even when serious conflicts are intense between them, a 15 Unfortunately, there is no data measuring presidential popularity over time across countries other than the initial support presidents received in the first-round of election.

26 15 president may divert their attack by simply replacing a prime minister with a new figure as occurred in several postcommunist countries (for example, Russian President Boris Yeltsin replaced three prime ministers in 1998 following the financial crisis and political turmoil). In addition to the factors that might increase presidential vulnerability, I also include two control variables. One is the time elapsed since the president s inauguration, to determine whether presidents enjoy a honeymoon period, or whether they are in fact more likely to face interruption earlier in their terms (Jones 1995). For reasons to be explained in Chapter 2, the time spent in office variable is also important methodologically. And the other is a regional control, since interrupted presidencies are unequally distributed across regions and Latin American countries have the highest rate of presidential interruption (see Table 2-2 in Chapter 2) Prior Studies of Presidential Interruptions Several authors have offered cross-national comparisons of interrupted presidencies in the past few years. Pérez-Liñán (2003) examines presidential vulnerability in five Latin American democracies (Brazil, Venezuela, Colombia, Ecuador, and Paraguay) from 1990 to 1999, 16 while Hochstetler (2006) and Edwards (2007) analyze interrupted presidencies across all South American democracies from 1978 to Cheibub (2002b), in the most comprehensive study on presidential 16 Pérez-Liñán s (2007) recent book builds on the previous work and includes all Latin American countries from 1990 to 2004.

27 interruptions, assesses the impact of economic performance and several political and 16 institutional variables on the duration of presidents terms and their parties duration in office in all presidential systems from 1946 to Edwards (2007), Hochstetler (2006) and Pérez-Liñán (2003, 2007) find that divided government raises the odds of an interrupted presidency, while Cheibub (2002b) concludes that having a minority of seats in the legislature does not make a president more vulnerable. If anything, he finds that divided government and deadlock, where neither president nor parliament can impose their preferences on the other, increase the odds of presidential survival. 17 Cheibub also finds that low-to-moderate, rather than higher levels of party fragmentation make presidents more vulnerable (in this case, to regime breakdown). Jones (1995), Mainwaring (1993), and Samuels and Eaton (2002), however, conclude that presidents facing higher levels of party fragmentation are more susceptible to pressure to leave office early. With respect to economic conditions, Edwards (2007), Hochstetler (2006) and Pérez-Liñán (2003, 2007) find that economic adversity makes presidents more vulnerable, 18 while Cheibub (2002b) finds that economic conditions have no effect. And 17 Cheibub s measure of deadlock draws on constitutional provisions for presidential vetoes and legislative overrides, the number of chambers, and the share of seats controlled by the president s party in each chamber. 18 In his 2007 book, Pérez-Liñán suggests that poor economic performance is associated with greater presidential vulnerability because it together with corruption leads to popular uprisings. However, he does not find statistical support for this argument. It turns out that only scandal, particularly when a president is involved, explains the occurrence of public mobilization.

28 17 Edwards (2007), Hochstetler (2006) and Mainwaring and Pérez-Liñán (2005) find that scandal and protest increase the odds of presidential crisis. The divergent findings seem to depend in part on the use of different time periods, samples, and model specifications. Cheibub s analysis is the most comprehensive, including presidential systems in new and established democracies and some authoritarian presidential countries over an extended time period (1946 through 1996). But the determinants of presidential vulnerability may be different in established democracies such as Colombia, Costa Rica, Finland, France, Iceland, Switzerland, and the United State versus newer and more fragile regimes (cf. Linz 1990; Valenzuela 2004; Mainwaring and Pérez-Liñán 2005). 19 And focusing on a relatively long time span might obscure changes in the impact of poor economic performance from earlier to later decades (Gasiorowski 1995; Mainwaring and Pérez-Liñán 2005). 20 Including both 19 Besides, interrupted presidencies are much rarer in established democracies during the period examined here, Out of seven established democracies listed in Cheibub (2002b), only the US experienced an interruption (Nixon in 1974). Finnish President Kekkonen resigned in October 1981, but that was due to ill health. Also, Cheibub (2002b: 137-8) seems to include authoritarian countries (Cameroon , Colombia , Djibouti , Gabon , the Philippines , Rwanda , and South Korea ) in his analysis (Alvarez et al. 2000), which could be problematic. In authoritarian regimes, presidents, for example, controlling less than half of seats in parliament would not find it difficult to pass laws as they desire. Authoritarian presidents can rely on other means such as threats or buy-outs to achieve their goals. Thus, presidents may face divided government in an autocracy, but in practice that would not necessarily affect presidential vulnerability. 20 Gasiorowski (1995: 892), for example, finds that the impact of high inflation on democratic transitions reversed from the 1950s and 1960s to the 1980s (reducing the odds in the earlier years, but increasing the

29 acting/interim and elected presidents could also be problematic, since acting chief 18 executives lack a popular mandate and may have a different relationship with other branches of government. Finally, Cheibub excludes presidents who are constitutionally barred from re-election because these cases are exogenous to the process that he wants to explore (Cheibub 2002b: 129). Doing so, however, excludes valid cases since even presidents with term limits may leave office early How Does This Study Differ? All told, prior research disagrees on the impact of several important factors, particularly of divided government and economic adversity. Moreover, most of it does not systematically examine some of the key factors such as the president s first-round vote, presidential powers, protest, or presidential scandals. Finally, much of the literature focuses heavily on South America, leaving open the question of what drives interruptions in other incipient democracies. In this study, I reassess the roots of presidential vulnerability across a broader range of new democracies in Europe, Asia, Africa and Latin America, from 1974 through Using original data on presidential interruptions in all Third Wave democracies, I examine why some presidents are more vulnerable than others. In addition to reexamining several core arguments that have proved controversial in the previous studies, I add new data on the impact of presidential scandals and civil society/public odds in the later period). Mainwaring and Pérez-Liñán (2005) suggest that the impact of economic conditions changed in Latin America after the late 1970s.

30 mobilization. I also explore the impact of a key presidential resource that has been 19 neglected in most analyses of presidential vulnerability the president s share of the popular vote. I examine this question using a combination of logit, survival analysis, and competing risks models. I find that presidents are more vulnerable than others when they face divided government, higher party fragmentation in parliament, and a low first-round popular vote. And an imbalance of executive powers with other branches of government works against the incumbent. Economic decline adds to the risk, as well. The probability of leaving office early also increases when the public are not willing to wait until the next presidential election. However, such political crises did not imperil the survival of democracy in many countries, as some earlier critiques suggested. Presidents may step down, but in most regions of the world outside sub-saharan Africa, electoral democracy continues (cf. Pérez-Liñán 2007). In contrast, interruptions are typically by force in Africa, and many lead to the breakdown of democracy. I define presidential interruptions as situations where a duly elected president leaves office prematurely by impeachment or parliamentary vote, resignation, or by force. This is how other studies of presidential interruptions define it (Cheibub 2002b; Hochstetler 2006; Marsteintredet and Berntzen 2006; Negretto 2006; Pérez-Liñán 2005). I define a presidential system as one where the president is both the head of government and the head of state and is elected for a constitutionally prescribed term, by popular vote (either directly or through an electoral college system), as in much of the literature on presidentialism (Lijphart 1992; Sartori 1994). I also include semi-presidential systems in the analysis. Semi-presidentialism is defined as one where a directly elected president

31 and a prime minister share executive authority. But, I use the term presidential systems in the rest of this research for the sake of simplifying terminology Attempted Presidential Impeachments in Third Wave Democracies Introduction The first question I address empirically is what makes some presidents more vulnerable than others. The second is where and when legislators attempt to impeach a president. While successful impeachments are rare in Third Wave presidential democracies (only one president (Malagasy President Zafy in 1996) left office as a result of successful impeachment proceedings from 1974 to 2003), impeachment attempts are fairly common. In spite of the high barrier to removing most presidents, about 45% of Third Wave presidential democracies have experienced presidential impeachment attempts (to be explained in detail below), some countries more than once (see Table 2-3 in Chapter 2). For example, a group of Nicaraguan legislators threatened to impeach President Chamorro in 1995 for blocking constitutional reform (Aleman 1995). And in Paraguay, some deputies in the governing Colorado Party publicly expressed their intention to impeach President Wasmosy for a series of political crimes of extreme gravity in 1997 (Gunson 1997). Attempted presidential impeachments are not limited to Latin America: In Ukraine, a parliamentary committee for Legal Policy and Court and Legal Reform 21 Chapter 2 provides more detailed definitions, coding rules of key terms, and country classifications.

32 21 attempted to impeach President Kuchma for his veto of a new law regarding the election of local administrations in 1997 (ITAR-TASS News Agency 1997). A group of deputies then collected signatures for a resolution to impeach him. In 2003, Zambia s opposition parties began impeachment proceedings against President Levy Mwanawasa over his unconstitutional appointment of a new vice president Nevers Mumba (Agence France Presse 2003). Although these impeachment attempts did not end in the constitutional removal of a president, they differ in their specific details of the failed outcomes. The first two cases of impeachment attempts in Nicaragua and Paraguay ended up with only a verbal threat. Neither of them developed further. However, the Ukrainian case is different. The impeachment attempt was more substantial than the previous two cases, since it involved collecting signatures of legislators necessary to submit an impeachment resolution to Supreme Council. They eventually failed to gather the required number of signatures. The Zambian case developed further than previous cases, with Zambia s parliament taking a vote on impeaching President Mwanawasa (which failed) (Agence France Presse 2003). Why would deputies even think about impeachment if it almost never leads to the constitutional removal of a president? One reason may lie in the declining attractiveness of coups to break through political crisis in new democracies. The number of successful coups has decreased significantly in Africa and Latin America, particularly since 1990 (Abente-Brun 1999; Clark 2007; Mainwaring 1999; McGowan 2003; Pérez-Liñán 2003, 2007; Posner and Young 2007; Valenzuela 2004; van de Walle 2003). Given that nondemocratic actors are not welcome to be involved in politics, the role of legislature