Challenging Corrupt Politicians? Audits, Electoral Selection, and Accountability in Municipal Elections * Current version: January 2015

Transcription

1 Challenging Corrupt Politicians? Audits, Electoral Selection, and Accountability in Municipal Elections * Luis Raúl Cámara-Fuertes 1 and Gustavo J. Bobonis 2 Current version: January 2015 Abstract: Access to information on corruption does not always promote electoral accountability. We argue that electoral selection, the strategic decisions by incumbents and challengers to enter a race, may be important intervening factors. To study the consequences of information on corruption throughout the complete electoral cycle, we use municipal audit reports issued by the Comptroller of Puerto Rico to examine whether incumbents and potential challengers factor available information on political corruption in their calculus to enter a race; if this information influences the citizen s vote; and if this impact comes directly from the information provided by audit reports or indirectly because the reports ushers the entrance of an experienced challenger. We find that negative audit reports are detrimental among experienced versus junior incumbents, but that the entrance of quality challengers is not relevant to electoral outcomes. Keywords: corruption; information; electoral selection; challenger entry; accountability * We are especially grateful to Hon. José M. Díaz Saldaña, Comptroller of Puerto Rico, for providing us access to the municipal government audit reports of the Office of the Comptroller of Puerto Rico, and to Fred Finan for invaluable help and advice in the construction of the audit reports data. Aileen Cardona, Laura Delgado, María del Mar Ortiz, Zorimar Rivera, Vilma López, Adriana Robertson, and Julie Wilson provided superb research assistance. Research support from the Canadian Institute for Advanced Research (CIFAR), the Social Sciences and Humanities Research Council of Canada (SSHRC), and the University of Puerto Rico Social Research Center and Office of the Dean for Graduate Studies and Research is gratefully acknowledged. We are responsible for any errors. 1 Corresponding author: Professor, Department of Political Science, University of Puerto Rico-Río Piedras. Address: P.O. Box 23445, University of Puerto Rico, Río Piedras Campus, San Juan, PR 00931, USA. Tel: , ext luisraulcamara@gmail.com. 2 Associate Professor, Department of Economics, University of Toronto, BREAD, and CIFAR. Address: 150 St. George St., Toronto, Ontario, M5S 3G7, Canada. Tel: Fax: gustavo.bobonis@utoronto.ca.

2 Introduction Citizens access to information to evaluate a politician s actions is considered crucial for electoral accountability (Manin, Przeworski, and Stokes, 1999). 3 In contrast, a growing empirical literature shows that voters access to information on political corruption does not always promote electoral accountability. An important question in the design and implementation of transparency initiatives is then: under what circumstances does information, and audit programs in particular, work? In this article we argue that electoral selection the strategic decisions by incumbents and challengers to enter a race may be crucial for understanding the effectiveness of transparency initiatives. Specifically, we study the consequences of disseminating information on political corruption throughout the complete electoral cycle, taking advantage of a unique setting that provides us with the opportunity to examine such relationships. The government of Puerto Rico has established an independent body that systematically conducts municipal government audits, the findings of which are made publicly available and disseminated to media sources. We first examine whether incumbents and potential challengers factor publicly available information on political corruption in their calculus to enter a race. Second, we reexamine whether this information on corruption impinges on citizen s voting decisions, and aim to disentangle whether this impact comes directly from the information provided by the audit reports or indirectly because the report signals a weakness in the incumbent, thus ushering the entrance of an experienced challenger. Our empirical analysis conveys two main findings. Perhaps surprisingly, we find strong evidence that the effects of these negative audit reports for electoral accountability are particularly detrimental among experienced as opposed to more 3 More specific details on the data, along with additional analysis mentioned in the text or footnotes can be obtained from the author/s at ( s deleted for review process). 1

3 junior incumbents. In addition, we find that the entry of a quality challenger is not a very relevant dimension of electoral competition, even when corruption is an issue. The study contributes to the literature examining the heterogeneous role of information and transparency for electoral accountability. First, we show that electoral selection related to seniority, suggestive of the incumbent s quality, strongly mediates the effectiveness of information for electoral accountability. This result complements a small literature arguing that the political environment is crucial for the punishment of incumbent misbehavior. Most notably, Nannicini et al (2012) show evidence from Italian parliamentary elections that incumbent misbehavior is punished more strongly in districts with higher levels of civic or social capital. Second, a growing literature argues that the quality of an opponent can respond to the vulnerability of incumbent politicians due to for instance the level of observable rent seeking that is, there is strategic challenger entry. 4 However, in spite of that fact that the literature on voting has clearly shown that challenger quality is important in determining electoral outcomes, this has been an important omission from existing empirical work on the strategic interactions resulting from voters access to information and electoral accountability. 5 The paper is organized as follows. First, we discuss the literature on corruption, voting and challengers. Then we provide a background on Puerto Rico politics, the electoral cycle, and the Office of The Comptroller of Puerto Rico (OCPR) auditing program. We follow with an exposition of testable hypotheses and a description of the data and research methodology. We present the main empirical results and robustness tests, and finally the conclusion. 4 The literature argues that challenger entrance is a strategic decision, especially on the part of quality challengers (Jacobson, 1980; 1989; Jacobson and Kernell 1983; Lublin, 1994; Cox and Katz 1996; Stone, Maisel, and Maestas, 2004; Gordon, Huber and Landa, 2007; Lazarus, 2008a; 2008b; 2008c; Ashworth and Bueno de Mesquita 2008; Maestas and Rugeley, 2008). 5 The literature on voting has uncovered several key factors in explaining the entrance of quality challengers such as: A weaker than usual vote, usually the most powerful explanation, the normal vote in the district and its partisan trend, national conditions, and the incumbents legislative vote track (Bond, Covington, and Fleisher, 1985; Krasno and Green, 1988; Jacobson, 1989; Lublin, 1994; Lazarus, 2007; Krasno and Green, 1988). 2

4 Political Corruption, Voters, and Challengers We start our discussion with a simple fact: corrupt politicians routinely win elections. The reason assuming fair elections is quite basic: people do vote for corrupt politicians and corrupt politicians routinely win elections. Even when corrupt politicians are punished by voters through a loss of votes, they have often amassed such large margins of victory that the punishment is not large enough to ouster them from office (Peters and Welch, 1980; Welch and Hibbing, 1997; Chang, Golden and Hill, 2010; Jacobson and Dimock, 1994). Why do people vote for corrupt politicians? One reason is that voters do not know that the candidate that they are voting for is corrupt. Recent research shows that in places where there is greater distribution of information and thus people know more about the corrupt acts of politicians, the fortunes of incumbents charged with corruption is worse (e.g., Ferraz and Finan, 2008). However, another possibility is that people filter corruption information depending on their partisan identification: individuals who favor the incumbent party perceive the political system to be less corrupt than those who favor opposition parties (Anderson and Tverdova, 2003). This likely occurs due to confirmatory bias individuals filter the news they receive through their previously held beliefs. When this information contradicts previously held information it is discarded (Zaller, 1992). Thus, partisan voters may simply refuse to believe information that their representatives are corrupt, especially if such accusation comes from an opposing party or candidate. However, even some voters who do know that politicians are corrupt will end up voting for them. One important reason is that some people vote for corrupt politicians because they perceive corruption simply as another variable in their calculus of voting. In this tradeoff view, 3

5 accusations of corruption compete with issue positions as well as with voters party identification, and it is thus incorrect to assume that voters weigh corruption more heavily than other factors (Rundquist, Strom, and Peters, 1977; Peters and Welch, 1980; McCann and Domínguez, 1998; Winters and Weitz-Shapiro, 2013). This means that voters will not simply switch their vote from their corrupt candidate to a clean candidate who is on the wrong side on all the important issues and/or belongs to the wrong party. This is especially relevant with relation to core issues and partisanship. Yet another reason voters vote for corrupt candidates is because they receive goods and favors from them, especially in systems of patronage politics (Fackler and Lin, 1995; Manzetti and Wilson, 2007). In countries where patron-client relationships are strong and government institutions are weak, citizens are more likely to support a corrupt representative from whom they expect to receive tangible benefits (Manzetti and Wilson, 2007). A possible intervening factor between corruption and politicians electoral outcomes is the mediating effect of challenger quality/experience (we use the terms interchangeably). While there is an extensive literature on the entrance of experienced challengers and challenger experience and electoral results, few studies of electoral accountability take into consideration the effect that quality challengers attracted by serious corruption allegations or accusations can have on the incumbent s political fortunes (for an exception to this see Jacobson and Dimock, 1994). This is an important omission because challenger experience is important in determining election outcomes. One of the most consistent arguments in the literature is that challengers are rational actors that enter a race strategically (Osborne and Slivinski, 1996; Besley and Coate, 1997). In general, the literature states that quality challengers enter the race when conditions are favorable, 4

6 whereas when conditions are unfavorable, low quality amateurs enter (Jacobson, 1989; Jacobson and Kernell, 1990; Lublin, 1994; Stone, Maisel, and Maestas, 2004; Gordon, Huber and Landa, 2007; Maestas and Rugele, 2008). The entrance of experienced challengers is important because they do significantly better against incumbents than low quality, inexperienced challengers (Jacobson, 1989; Jacobson and Kernell, 1990; Lublin, 1994; Lazarus, 2008). Experienced challengers are better known and thus have higher name recognition. This means that they do not have to spend many resources to make voters know who they are. They also have a short learning curve when the campaign starts; they have experience running successful campaigns and winning. Some also argue that they can raise more money and thus have more resources (Maestas and Rugele, 2008; Lazarus, 2008). The issue of strategic challenger entry is theoretically important in the study of electoral accountability for two main reasons. The first is from the perspective of the challenger himself: experienced challengers choose races that they can win (Lazarus, 2008). A second reason this time from the perspective of the voter is that experienced candidates influence a race by affecting the voter s uncertainty calculus (Gordon, Huber and Landa, 2007). Voters have incomplete information, mainly due to inattentiveness and to the complexity of politics. Given this uncertainty, a costly challenge can provide information to the voter. The mere presence of a quality challenger represents a signal to voters about the opposition s willingness to engage in electoral competition, and can generate doubts about an incumbent s political strength (Gordon, Huber and Landa, 2007). Moreover, their entrance can lead to a more costly and intense campaign which are associated with more media coverage (Sulkin, 2001). Thus, an audit report that may have passed with limited exposure can now be diffused more widely; and, as mentioned above, where the distribution of information is better, incumbents charged with corruption tend 5

7 to do worse. Furthermore, experienced challengers air and debate issues more credibly (Jacobson and Kernell, 1990; Gordon, Huber, and Landa, 2007). Given the importance of challenger quality in determining electoral outcomes, we aim to enhance our understanding of their role as a mediating factor in the relationship between information regarding political corruption and electoral accountability. Corruption, Puerto Rican Politics, and the Electoral Cycle The theme of political corruption in Puerto Rico has a long history. Patronage directly tied to political parties is strong and pervasive, and the spoils system is highly developed (Santana Rabell, 1993). This is especially true at the municipal level: concerns over rent seeking and lack of constraints over the actions of the local executive are pervasive and have persisted through time. Corruption has for a long time been a front and center issue in political campaigns, and the public believes that corruption is an important problem in Puerto Rico. In every one of the polls conducted between 1995 and 2004 commissioned by the newspaper El Nuevo Día the percentage of people mentioning corruption as a problem ranged from 55 to 72 percent and corruption ranked between the fourth and the sixth place. Furthermore, when people were asked to mention the most important problem, corruption was among the top five issues mentioned in 1999, 2000, 2002 and Two political realities are important in Puerto Rico when studying the impact of corruption in elections: partisanship and the so-called status issue the political relationship between Puerto Rico and the United States. Parties are central to the Puerto Rican political process. Furthermore, Puerto Ricans have very strong levels of political identification. At least 6

8 in the 1990s Puerto Ricans had some of the strongest levels of partisan attachment among established democracies and this attachment had important electoral consequences in candidate choice and voting turnout (Cámara-Fuertes, 2005). The second reality is the status issue. There are three main status options: Puerto Rico becoming a U.S. state, an independent nation or continuing its current Commonwealth status. Status is the main cleavage of Puerto Rican Politics (Cámara-Fuertes, 2005). The three main traditional parties revolve around this issue: The New Progressive Party (PNP) favor statehood, the Puerto Rican Independence Party (PIP) favors independence and the Popular Democratic Party (PPD) favors the current Commonwealth status. The status debate is an ever-present subject that influences the campaign and electors both directly and indirectly. Partisanship and status are important in this context. Based on the tradeoff theory, corruption can be seen by voters at par with other issues. Partisanship is both an aggregator of issues and a strong psychological attachment. Thus, when a candidate from a voter s party is accused of corruption that means that in order for that voter to choose the honest challenger he must go against many of his issue preferences including the powerful status issue in addition to the more psychological partisanship attachment (Zaller, 1992; Dimock and Jacobson, 1995; Anderson and Tverdova, 2003). The same argument applies to the status issue. This is significant because Davis, Camp and Coleman (2004) argue that in ideologically polarized party systems there is a stronger perception of corruption of incumbents. Thus, the status issue and party polarization should make it easier to pin corruption charges on incumbents, but more difficult for his partisans to believe it and thus vote for the opponent. This background is important because many studies have been conducted in polities with relatively low levels of voters party identification and disciplined party voting (i.e., Brazil and 7

9 the United States.) However, since partisanship varies significantly across polities and since parties and partisanship remains strong in many countries, it is important to study the relationship between corruption and electoral accountability in highly partisan environments. Finally, we provide some background on the Puerto Rico electoral cycle, in order to explain the timing of candidate entry and electoral challenge. Elections are a cycle that in many instances begins long before the election itself. It starts with the decision by the incumbent on whether to retire or to run for reelection. This is important since often politicians choose not to run when facing the real possibility of defeat in the face of scandal (Kiewiet and Zeng, 1993; Jacobson and Dimock, 1994; Groseclose and Krehbiel, 1994; Hall and Houweling, 1995). This is complemented by potential challengers decisions on whether or not to run. The electoral rules in Puerto Rico at the time of the 1996, 2000 and 2004 elections established the month of July of the year preceding the General Election as the deadline for this decision. Thus, both challengers and incumbents make their decisions to file for candidacy sixteen months before the election. If an incumbent decides to run, she could face a primary. Although primaries are executed using the State Electoral Commission s infrastructure, they are run by the parties and are a strict partisan affair. If there is a primary, the incumbent can face a token opposition or a significant challenge. In the electoral cycles under study, primaries took place on the first Tuesday of the November of the year before the elections; that is, a full year before the General Elections. If the candidate survives a primary challenge, she can either face a quality (experienced) challenger or an inexperienced, unknown candidate in the General Election. The OCPR s Municipal Government Audit Program The Office of the Comptroller of Puerto Rico (Oficina del Contralor de Puerto Rico OCPR) is an autonomous government agency created by the 1952 Constitution of the 8

10 Commonwealth of Puerto Rico. The OCPR periodically audits state-level government agencies and public corporations, including the legislative and judicial branches, as well as municipal governments. The OCPR has been continuously carrying auditing municipal governments and generating and disseminating reports since Although municipal governments should be audited every other fiscal year, due to resource limitations, there may be some delay in the timing of the audit. The order of the audits follows a routine pattern as municipalities are audited following a pre-specified order. Once all municipalities have been audited, a new auditing round takes place following the same pre-specified order. The OCPR may publish multiple reports on a municipality for one auditing period depending on the size or complexity of the municipal government. Audits identify instances of corruption via fraud in procurement, the illegal hiring of employees, and over-invoicing for goods or services. A number of measures are taken to minimize potential biases in the conduct of the audits and in the dissemination of their findings. First, the Constitution provides the OCPR with a substantial degree of autonomy from the rest of the central government structures in order to isolate the agency from undue external interference. To help achieve this, the Comptroller is appointed by the Governor for a ten-year term with the advice and consent of both legislative chambers and can only be removed by an impeachment process. Second, the OCPR is technically under the purview of the legislative branch. Since most of the agency s activities are focused on the executive branch and municipalities, this gives it an additional layer of protection from undue influence. Third, the auditors, who are hired based on a competitive public examination, earn highly competitive salaries, and receive extensive training prior to visiting the municipalities. Finally, in order to reduce local-level conflicts of interest, individual auditors are precluded from participating in audits of their municipality of residence. 9

11 Once the audit reports are completed they are made public and disseminated to the public and to media outlets through press conferences. The reports are also published on the OCPR s website. Although the OCPR does not identify findings as corrupt acts, it can refer its findings of misuse of public funds to the Department of Justice and/or to the Office of Government Ethics. While we do not have specific measures of whether voters know or do not know if their mayor is corrupt, we believe that the percentage of uninformed voters is not substantial. The media tends to highlight the most headline grabbing reports. But perhaps more importantly, rival politicians often use information in the reports against their incumbent opponents in the campaign trail or through the press. [TABLE 1 ABOUT HERE] As Table 1 shows, the reports are routinely reported by the press. We searched El Nuevo Día arguably the most important national newspaper in Puerto Rico for any news coverage of the Municipal Audit Reports during the 2000 and 2004 election years (January 1 st through election-day on the first Tuesday of November). As Table 1 shows, in 2000, there were 127 news articles on OCPR municipal audit reports in 10 months of coverage relating to 37 municipalities. In contrast, in 2004 there were 35 news reports covering 22 municipalities. The difference may be due to the importance of the corruption issue for the 2000 elections, over a more normal election in Most of the coverage was from press conferences by candidates, mainly the mayoral challengers in 2004 and incumbents in This meant that the news information would further be diffused in the campaign trail by speeches, press releases and media interviews of the candidates. Hypotheses 10

12 Based on the literature and the previous discussion, we propose six hypotheses: H 1 : The higher the number of corruption findings, the more likely that an incumbent will retire. H 2 : The higher the number of corruption findings, the more likely a quality challenger will enter the mayoral race. H 3 : The higher the number of corruption findings, the more likely there will be a primary challenging the incumbent; and the challenge will be a serious one. H 4 : The higher the number of corruption findings, the more likely the incumbent individual or party will lose his/her reelection bid; reduce his/her vote proportion vis a vis other races in other municipalities; and/or reduce his/her margin of victory from the previous election. H 5 : In elections where quality challengers are present, the impact of negative audit reports will be more detrimental to incumbents than in elections with inexperienced challengers. Data and Research Methodology To test these hypotheses we look at the electoral process in four stages: incumbent s decision to run or retire, challenger entrance, primary challenges, and the general election. Each of these stages has a set of dependent and explanatory variables. We construct measures of political corruption using the Municipal Audit Reports issued by the OCPR. We code a report finding as a corruption violation if the mayor or vice-mayor obtained any personal financial or political benefit from a specific act. The OCPR does not label findings as corrupt. We label these based on our definition of corruption. For example, if a mayor received a kickback for a construction project or if municipal employees were used in political campaign activities we classify such actions as corrupt. On the other hand, instances 11

13 where the mayor did not obtain all the necessary permits to construct a sewer or if he kept insufficient records would not be coded as corrupt. Furthermore, we focus only on those findings that we coded as corrupt and that were referred to either the Department of Justice or the Office of Government Ethics. These tend to be the most serious or egregious cases and the ones that most likely received some type of media coverage. These variables are coded as the number of violations (i.e., 0, 1, 2, 3, 4 or more). All seventy-eight municipalities were audited multiple times during our period of study ( ). The use of OCPR Municipal Audit Reports has several advantages. First, by using routine audit reports done by professional auditors we avoid any impact that the newness of a procedure might have on voters. Voters have been exposed to these reports since 1953 and in a way have come to expect them. Municipal Audit Reports are, thus, a recurring stream of information. Second, the reports that we use here reflect the actions of the mayor, vice-mayor or department heads, not spectacular instances of group corruption like those reported by Chang, Golden, and Hill (2010) on the Italian Parliament. Third, and perhaps most important, we are not dealing with allegations of corruption but findings of fact by professional auditors. While auditors can certainly be wrong, we believe that they probably have more weight than an allegation thrown by a rival candidate or an enterprising reporter. More likely than not voters give these findings more weight than accusations from rival candidates. In addition, we believe that this is a more realistic measure of corruption than allegations by other candidates. We also employ data available from the Puerto Rico State Electoral Commission (CEE) containing the results of municipal and statewide ballots for the 1984, 1988, 1992, 1996, 2000 and 2004 primaries and general elections. Our measure of challenger quality is based on political (electoral) experience. This is a standard measure in the literature to measure candidate quality 12

14 (e.g., Jacobson, 1989; Jacobson and Kernell, 1990; Bond, Covington, and Fleisher, 1985; Krasno and Green, 1988; Lublin, 1994; Lazarus, 2008). The measure consists of three categories: low quality candidates are those with no political experience, who have never run for an elective office or held political positions; middle quality candidates are those with some political experience or visibility, who have run in the past but lost or are members of the Municipal Assembly; and, high quality candidates are those with significant political experience by holding elective positions as mayors or state legislators or who have occupied highly visible appointed political positions recently in the constitutional cabinet. We include those who have won elective office in the Municipal Assembly in the middle category because it is a low visibility position that is usually won through the campaigning effort of the mayor. In all three cases we use information on primaries and general elections in the three electoral cycles preceding each of the 1996, 2000, and 2004 elections to construct the measure. Using these data we construct two sets of models of electoral choice, one for events before the general elections incumbent retirement, challenger entrance, and primaries outcomes and another for the general election outcomes. The first empirical model seeks to explain whether an incumbent runs or retires, whether an experienced or inexperienced challenger enters the race, whether or not there is a party primary to challenge the incumbent, and whether these primaries are serious challenges or sure wins. The model is specified as follows: Y = β 1 (Corruption) + β 2 (Seniority) + β 3 (Seniority Corruption) + β 4 (Political Trend) + β 5 (Previous vote) + β 6 (2000 Election) + β 7 (2004 Election) + β 8 (Δ Unemployment) + β 9 (% College graduates) + β 10 (Municipal population) + ε, 13

15 where Corruption is the number of corrupt findings referred to the Department of Justice and/or Office of Government Ethics; Seniority is the number of terms that the incumbent has served (consecutive or non-consecutive); Seniority Corruption is an interaction term; Political Trend is whether the wave that favors a party at the island-wide level is aligned with the party of the incumbent mayor; Previous Vote is the percentage of the vote that the incumbent/party received in the previous election; 2000 Election and 2004 Elections are dummy variables comparing those elections against the 1996 election; Δ Unemployment is the change in the unemployment rate at the municipal level from the previous election to the current one; % College graduates is the percentage of college graduates at the municipal level, and Municipal population is the number of people living in each municipality, both based on the closest decennial population census. The measure of corrupt violations encompasses information published in the period between the first day of the incumbent s current four-year term and the event under study. Thus, in the challenger entrance equation and the incumbent s decision to run or retire, the measure includes referrals from reports released between the first day of the four-year term and the month of July of the year preceding the General Election (the deadline for filing candidacy). Since primaries were held on the first Tuesday of November preceding the year of the general election, for the primaries outcome models we include referrals until October of the year preceding the general elections. We allow for heterogeneous effects of the audit-based violations and the degree of seniority of the incumbent mayor with an interaction term. This incorporates the idea that less experienced mayors may be more strongly affected by negative corruption findings than more experienced ones, as they may have less of a political infrastructure such as political and/or clientelistic bonds with constituent groups to protect themselves against this information. In 14

16 addition, it is not unreasonable to think that mayors that have been elected many times have been reasonably good mayors, otherwise they would have lost in previous elections. On the other hand, due to a lifecycle effect, this information may induce experienced mayors to retire or to lose elections due to a wearing out effect or due to their divestment in political capital. The second empirical model seeks to explain the outcomes of the General Elections. We divide our dependent variables into two groups: one where we look only at incumbents who ran for reelection and a second group that includes everyone, incumbents and those who occupied the spaces left vacant by retiring mayors. Whereas the first group focuses on the damage (or benefit) done to an individual, second studies the effect of corruption on the party. For each group we focused on the change in the margin of victory/defeat from the previous election, whether they won or lost the election, and the margin of victory within that particular election. The model is specified as follows: Y = β 1 (Corruption) + β 2 (Seniority) + β 3 (Seniority Corruption) + β 4 (Challenger) + β 5 (Challenger Corruption) + β 6 (Political Trend) + β 7 (2000 Election) + β 8 (2004 Election) + β 9 (Δ Unemployment) + β 10 (% College graduates) + β 11 (Municipal population) + ε, where Challenger denotes the experience level of the challenger; Challenger Corruption is an interaction term; and all the remaining variables are the same as in the first model. In the General Election model, the measure of corruption referrals encompasses information published in the period between the first day of the incumbent s current four-year term and the month of October preceding the November general election. We estimate these models for races where incumbents survive to run in the general election. 15

17 In order to capture the mediating effect of challenger quality for the electoral accountability effects of the audit program, we allow for heterogeneous effects via an interaction between a quality challenger and corruption findings, in addition to an interaction term between seniority and corruption findings. The corruption-quality challenger interaction term allows us to test hypothesis five: in the face of a strong, quality challenger, negative findings will be more detrimental to an incumbent mayor or party. Some comments on these controls are warranted. The election-round indicators control for the peculiarities of each election; the 1996 election is the reference category. In addition each election was very different in its result: the 1996 was a clear sweep for the PNP, the 2000 election a clear sweep for the PPD and the 2004 resulted in a divided government. Coattails are a significant factor in Puerto Rican elections. The winning party tends to sweep all levels of government: the executive, the state legislature, and municipalities. Since the first gubernatorial election in 1948, only three times in sixteen elections has there been a divided government of some sort (1968, 1980 and 2004). Thus, a significant number of mayors may win or lose an election simply because of the general electoral tide. We created a dummy variable that identifies municipalities held by one party in an election where the trend favors the other. The unemployment trend control incorporates the idea that economic conditions have a significant effect on election results. For the general election model this variable is constructed by subtracting the municipal unemployment rate of the election year from the previous year. Because the decision by a challenger to enter the race or of an incumbent to run or retire occurs the year before the general elections and the primaries are held very early on that election year, in these instances the variable subtracts the unemployment rates the year before the general election to the year preceding it. 16

18 We estimate dichotomous dependent variable models using logit equations, while in the case of dependent variables with multiple categories we estimate OLS regression models. All dependent and independent variables are normalized to have support in the [0,1] interval. This normalization allows us to compare the magnitude of coefficients within and between equations. In the case of the logit models, we also convert the logit equations coefficients into probabilities and report these jointly with the estimated coefficients. Results Table 2 displays the results for the models examining incumbent mayor retirement and experienced challenger entry decisions, as well as primary election outcomes. Overall, negative OCPR reports do not affect these decisions/outcomes in a significant manner. This is true for both junior and senior incumbents, as implied by the violations term (β 1 ) and the interaction term with incumbent seniority (the β 3 term). [TABLE 2 ABOUT HERE] As highlighted in the literature, seniority (terms in office) is an important mediating factor in the early part of the electoral cycle. Not surprisingly, seniority is associated with a higher probability of retirement, all else equal. The implied effect from least to most seniority is an increase in the retirement probability of 37 percentage points. The empirical analysis implies that hypotheses H 1, H 2, and H 3 are generally not supported by the data in this context. That is, findings of corruption do not have an effect on the earlier parts of the electoral cycle: these are not related to the incumbent s retirement decision to retire or run, the quality of the challenger, or primary elections. Although these empirical results represent a puzzle for the literature, the analysis is quite informative for our study. Specifically, it 17

19 allows us to more credibly examine the role of quality challengers as a mediating factor in general election outcomes with less concern of endogeneity bias. As for the control variables the unemployment situation in the municipality is related to the decision to retire. All else equal, an increase in unemployment is associated with a higher probability of incumbent retirement. We also observe that, when there is a primary, the strength of the challenge is affected by the size of the municipality, with smaller municipalities being associated with stronger challenges. Strong challenges were less likely in the 2004 election, ceteris paribus. Turning to the General Elections, Table 3 displays the results for the models examining the impact of corruption on incumbent mayors. Table 4 shows the results for any candidate of the incumbent party (whether an incumbent candidate or not), thus focusing more on effects at the party level. In each table, the first column displays the coefficient estimates on the difference in the margin of victory/defeat from the previous election, the second reports the estimates for winning or losing the election and the third the margin of the victory or the defeat. [TABLE 3 ABOUT HERE] The public findings of corruption have heterogeneous effects based on the level of seniority of the incumbent mayor. In all models, corruption has a strong negative effect as mayors have more seniority. The estimated impacts are substantive in all cases. We find a 1.4 percentage point decrease in the change in the electoral win margin given a one-step change in the seniority and corruption scales (0.224 x 0.25 x 0.25 = 0.014). The analogous estimate in the second model implies a 2.5 percentage point decrease in the likelihood of winning reelection (0.392 x 0.25 x 0.25 = ). The estimated impacts are of similar magnitude in the win margin model (1.6 percentage point). In contrast, findings of corruption have small and 18

20 insignificant effects among less experienced incumbents. These findings are stronger when we focus on incumbent mayors as opposed to incumbent party candidates more broadly defined. The estimated impacts imply that more information can mean the difference between victory and defeat in a moderately close election. We find that corruption does have an impact on whether the incumbent mayor or the incumbent party wins or loses and not just on a reduced margin of victory. Moreover, seniority among clean incumbents is significantly associated with positive electoral outcomes, such as a 1.8 percentage point increase in the win margin (0.072 x 0.25 = 0.018) and a 9.5 percentage point increase in the likelihood of winning reelection (0.38 x 0.25 = percent). Thus, findings of corruption can even reverse the influence of seniority. In other words, everything else being equal, clean mayors who have been in power a long time have better electoral outcomes than newer ones, while mayors who have been in power a long time and have OCPR audit reports that reflect corruption have worse electoral fortunes than junior ones. This evidence is consistent with lifecycle effects: information may induce experienced mayors to lose elections due to a weakening / wearing out effect or due to a depleted level of political capital. In contrast, we do not find significant evidence that facing an experienced challenger significantly reduces the likelihood of an incumbent winning or losing an election. The entrance of a quality challenger is associated with small and insignificant reductions in the change in the win margin and in the probability of reelection (columns 1-2). Only in the equations of margin of victory/defeat in an election (column 3) is this relationship significant, and positive. A possible explanation for this partial results is that voters perceive information regarding corruption as negative campaigning by a challenger, and this negative advertising may turn voters against who they perceive as a sponsor of it. This may lead to decreased support for the challenger. 19

21 This empirical analysis suggests that hypothesis H 4, the impact of corruption findings on electoral results, is supported by the data but almost exclusively through its interaction with seniority. H 5, the impact of a quality challenger receives only modest support. [TABLE 4 ABOUT HERE] Again, as an ancillary check, we evaluate the relationships of other control variables. A change in the municipal unemployment rate, and a negative municipal partisan trend in relation to the national political trend tend to favor the opposition party significantly predict the incumbent mayor or party electoral outcomes. The models explain between sixteen and twenty seven percent of the variation in electoral outcomes. The models that focus only on incumbent mayors display higher R 2. Turnout, Protest, and Third Party Votes While we have jointly studied significant aspects of the electoral cycle, we have ignored some other aspects of the electoral process such as voter turnout decisions and preference for third party candidates, which may impinge on the final outcomes of interest. Information on incumbents corrupt acts can affect voters turnout through a variety of mechanisms: voters may believe that voting will not benefit them because they lose trust in government or in its ability to respond to their needs, or because they become disenchanted with the entire political system upon learning about incumbent corruption (Aldrich, 1993). It could lead to abstention if individuals are unsupportive of the challenger s policy platform or unconvinced that the challenger is likely to be less corrupt. However, the existing evidence about the effect of exposing corruption on turnout is inconclusive. 6 In addition, since the act of voting 6 For studies based on observational data such as ours, see Bauhr and Grimes (2011), Caillier (2010), Davis et al. (2004), McCann and Domínguez (1998), Kostadinova (2009), Peters and Welch (1980). Recent field experiments have shown that 20

22 is so engrained in the political culture of the island, we focus on the percentage of spoiled or ballots left blank by voters (Cámara-Fuertes, 2005). This is perceived to represent a protest vote. We also consider the consequences for the electoral performance of third party candidates. This may be an important margin of accountability, as these parties in Puerto Rico often compete by focusing on a platform of honesty, transparency, and the imposition of accountability on the main parties. It is thus very plausible that information on a mayor s corrupt acts might drive some voters to choose these candidates because of their anti-corruption message, or because they view the opposition candidates as likely to be corrupt as the incumbent. We take these channels into account in the following manner. Since finding reliable population estimates in each municipality for each election to calculate age-based turnout is nearly impossible, we instead measured the vote drop-off from the gubernatorial vote to the mayoral vote. In other words, we measured the difference between the numbers of votes received by the gubernatorial candidate the most attention grabbing race and the mayoral candidates in each municipality. We calculated the protest vote by adding the number of spoiled and blank ballots as a percentage of the total votes. As for third party candidates, we estimate models that use their vote share as dependent variable (since these parties never win elections at the municipal level). Interestingly, in this context we find no evidence of a significant effect on support for third party candidates, protest vote or drop-off (turnout) either directly or indirectly. The results are not reported in the tables for the sake of conciseness (these are available from the authors upon request). Conclusion offering people information on incumbents performance has either no effect (Malesky, Schuler, Tran, 2012, Humphreys and Weinstein, 2012, Banerjee et al., 2010), or has a negative effect on electoral turnout (de Figueiredo, Hidalgo and Kasahara, 2011; Chong et al, 2013). 21

23 In this article we have comprehensively explored the impact of corruption on municipal elections. We studied the complete electoral cycle, from beginning to end; treated corruption not as an allegation by an opposing candidate or as a scandal but as report issued by a neutral, highly regarded autonomous government agency; and explored the impact that an experienced challenger could have in interacting with corruption reports. We found that reports of corruption do have an impact on the electoral fortunes of mayoral candidates. Specifically, corruption reports did not have an impact on the initial part of the electoral process: on the decision by an incumbent to retire or run; on the decision by a quality challenger to enter the race; and on the primary process. In contrast, we did find an impact on the general elections results, although mediated by seniority. The magnitude of the impact, however, was substantive, and strong enough to force a defeat for the incumbent. Finally, we found no effect of corruption reports on support for third parties, protest vote, or turnout. Accountability is considered to be one of the pillars of fair government because it helps to prevent corruption (Anderson and Tverdova, 2003). Through accountability the public s desires are supposed to be addressesed by their representatives in the ballot box. Assuming that the people do not want corrupt governments, this translates into a corruption deterrent since those who wish to govern will select policies in accordance to voter preferences. The extent to which transparency initiatives are quite imperfect at affecting accountability shows evidence of the limits of these for the betterment of democracy. 22

24 References Ades, Alberto, and Rafael Di Tella The New Economics of Corruption: a Survey and some New Results. Political Studies (45): Aldrich, John H Rational Choice and Turnout. American Journal of Political Science (37): Anderson, Christopher J., and Yuliya V. Tverdova Corruption, Political Allegiances, and Attitudes Toward Government in Contemporary Democracies. American Journal of Political Science (47): Ashworth, Scott, and Ethan Bueno de Mesquita Electoral Selection, Strategic Challenger Entry, and the Incumbency Advantage. Journal of Politics (70): Ashworth, Scott, and Ethan Bueno de Mesquita Is Voter Competence Good for Voters? Information, Rationality, and Democratic Performance. American Political Science Review (forthcoming). Banerjee, Abhijit V., Selvan Kumar, Rohini Pande, and Felix Su Do Informed Voters Make Better Choices? Experimental Evidence from Urban India. Typescript. Bauhr, Monika, and Marcia Grimes Seeing the State: The Implications of Transparency for Societal Accountability. Paper presented at the Midwest Political Science Association Conference. Besley, Timothy, and Stephen Coate An Economic Model of Representative Democracy. Quarterly Journal of Economics (112): Bond, Jon R, Cary Covington, and Richard Fleisher Explaining Challenger Quality in Congressional Elections. The Journal of Politics (47):

25 Caillier, J Citizen Trust, Political Corruption, and Voting Behavior: Connecting the Dots. Politics and Policy (38): Cámara-Fuertes, Luis Raúl The Phenomenon of Puerto Rican Voting. Gainsville: University Press of Florida. Chang, Eric C.C., Miriam A. Golden, and Seth Hill Legislative Malfeasance and Political Accountability. World Politics (62): Chong, Alberto, Ana de la O, Dean Karlan, and Leonard Wantchekon Looking Beyond the Incumbent: The Effects of Exposing Corruption on Electoral Outcomes NBER Working paper Cox, Gary W., and Jonathan N. Katz Why Did the Incumbency Advantage in U.S. House Elections Grow? American Journal of Political Science (40): Davis, Charles L., Roderic A. Camp, Kenneth M. Coleman The Influence of Party Systems on Citizens Perceptions of Corruption and Electoral Response in Latin America. Comparative Political Studies (37), de Figuereido, Miguel F.P., F. Daniel Hidalgo, and Yuri Kasahara When do voters punish corrupt politicians? Experimental evidence from Brazil. Typescript. Diermeier, Daniel, Michael Keane, and Antonio Merlo A Political Economy Model of Congressional Careers. American Economic Review (95): Dimock, Michael, and Gary Jacobson Checks and Choices: The House Banking Scandal s Impact on Voters in 1992, Journal of Politics (57): Fackler, Timothy, and Tse-min Lin Political Corruption and Presidential Elections, Journal of Politics. (57):

26 Ferraz, Claudio, and Frederico Finan Exposing Corrupt Politicians: The Effects of Brazil s Publicly Released Audits on Electoral Outcomes. Quarterly Journal of Economics, (123): Ferraz, Claudio, and Frederico Finan Electoral Accountability and Corruption: Evidence from the Audit Reports of Local Governments. American Economic Review (101): Gordon, Sanford C., Gregory A. Huber, and Dimitri Landa Challenger Entry and Voter Learning. American Political Science Review (101): Groseclose, Timothy, and Keith Krehbiel Golden Parachutes, Rubber Checks, and Strategic Retirements from the 102d House. American Journal of Political Science (38): Hall, Richard L., and Robert P. Van Houweling Avarice and ambition in Congress: Representatives decisions to run or retire from the U.S. House. American Political Science Review (89): Humphreys, Macartan, and Jeremy Weinstein Policing Politicians: Citizen Empowerment and Political Accountability in Uganda. Preliminary Analysis. Typescript. Jacobson, Gary C Money in Congressional Elections. New Haven, CT: Yale University Press. Jacobson, Gary C Strategic Politicians and the Dynamics of U.S. House Elections, American Political Science Review (83): Jacobson, Gary C., and Michael A. Dimock Checking Out: The Effects of Bank Overdrafts on the 1992 House Elections. American Journal of Political Science (38)