The effects of combating corruption on institutional trust and political engagement: evidence from Latin America

Abstract While a number of high-level figures around the world have been prosecuted and even jailed for corruption in recent years, we know little about how such anticorruption efforts shape public opinion and patterns of political engagement. To address this question, we examine evidence from Argentina and Costa Rica involving the unprecedented sentencing of two former Presidents on corruption charges. Exploiting the coincidence in timing between these cases and fieldwork on nationally representative surveys, we find that citizens interviewed in the aftermath of these events expressed lower trust in institutions and were less willing to vote or join in collective demonstrations. Overall, these findings suggest that high-profile efforts to punish corrupt actors may have similar effects as political scandals in shaping citizens’ relationship to the political system.

Design (UESD, Muñoz et al., 2020) to estimate the causal effect of prosecutions on popular attitudes. 2 Our analysis focuses on two sets of outcomes that are widely acknowledged as crucial for improving governmental accountability: institutional trust and political engagement. Regarding institutional trust, previous studies consistently show that corruption has a negative impact on how people view implicated politicians as well as associated institutions (e.g. Seligson, 2002;Anderson and Tverdova, 2003;Bowler and Karp, 2004;Morris and Klesner, 2010;Ares and Hernández, 2017;Solé-Ollé and Sorribas-Navarro, 2018). 3 However, we know little about whether efforts to combat corruption can restore institutional trust, particularly with respect to institutions such as prosecutors or the courts (Maier, 2011). For one, since oversight institutions are tasked with ensuring 'top down' accountability, they may be viewed as opposing corruption emanating from other parts of the political system. In addition, successful prosecutions could signal that enforcement agencies have both the institutional capacity and the determination to respond effectively to instances of abuse (Manion, 2004). As such, while we expect anti-corruption efforts to decrease trust in actors and institutions implicated in corruption, we examine whether such efforts might nonetheless build trust in oversight institutions and strengthen confidence in the state's anti-corruption efforts.
Our second set of outcomes capture political engagement. Here, some prior studies contend that corruption generates popular frustration which is then translated into political action, with public pressure being exerted either through electoral (De Vries and Solaz, 2017) or non-electoral channels (Bauhr and Grimes, 2014). In contrast, other work argues that political participation depends upon feelings of political efficacy (Levi and Stoker, 2000, 486). In particular, if citizens come to believe that 'nothing can be done' about corruption, they are likely to adopt the 'rational' response of withdrawing from the political arena (Persson et al., 2013;Bauhr and Grimes, 2014). Against this backdrop, we examine whether anti-corruption efforts-in weakening the perception that 'nothing can be done' about corruption-may increase citizens' willingness to take part in both electoral and non-electoral forms of political activity.

Evidence from Argentina
On June 13, 2013, an Argentine court sentenced former President Menem to seven years in prison for orchestrating illegal weapons sales to Ecuador and Croatia during his first term in office between 1991 and 1995. This sentence was widely hailed as the culmination of prosecutors' longstanding attempts to punish the ex-President for wrongdoing. 4 The court's announcement generated high levels of interest in the news media and was followed closely by the public. An analysis of Argentina's most popular newspapers, Clarín and La Nación, shows that the outcome of Menem's trial was the most important news item on June 13 and 14, 2013. 5 A similar picture emerges when looking at Google searches ( Fig. 1): while citizens sought information on Menem in the days surrounding the sentencing, interest spiked on the day of the announcement itself.
Menem's sentencing was particularly noteworthy because it represented an instance of long-sought-after accountability in Argentina. Clarín noted for example that Menem was the 2 The UESD assumes that individuals surveyed immediately prior to the judicial decisions are, on average, identical to individuals surveyed after, except that the latter group is (potentially) exposed to information about the criminal sentences. Given the rollout of the Latinobarómetro (see Appendix S1 for details), we contend that this ignorability assumption holds within localities/cities (CIUDAD), and thus include CIUDAD fixed effects in our estimates (see Appendix S5 for balance tests). 3 That said, we do not expect that the negative impact of corruption will spill over to institutions not connected to the case at all. See Appendix S10. 4 See Appendix S3 for a summary of the case history. 5 Other news stories reported in national newspapers included a train crash, clashes between soccer fans and the police, ongoing coverage of a murder and disappearance case, and political disagreements over the nomination of magistrates. first official to be sentenced to prison in over 700 cases involving high-level corruption (Clarín, June 14, 2013). In reflection of the unprecedented nature of the case, the tenor of media coverage was almost celebratory. For instance, La Nación characterized the event as the beginning of an unprecedented 'new era' of accountability (June 18, 2013). 6 Finally, it is worth noting that, while Menem remained (in)famous for his political exploits as President between 1989 and 1999, by 2013 he had become a less influential political figure.
Therefore, it appears highly unlikely that the timing of his sentence (or the sentence itself) was politically manipulated or anticipated. Indeed, we uncover no indication in the newspaper coverage that political considerations influenced the disposition of Menem's case. We therefore regard this case as a relatively 'clean' setting for examining the public opinion effects of anti-corruption efforts.

Analysis
Menem's sentence coincided with fieldwork on the Latinobarómetro survey conducted in Argentina between June 1 and 30, 2013. We use the June 13 announcement of the court's decision to define 'treatment' and 'control' groups among respondents. In order to ensure a sufficiently large number of observations, we analyze respondents surveyed within ±7 days of this cutoff date. 7 In Appendix S7, we show that our results are substantially unchanged using a larger 14-day window. Unfortunately, extending our analyses beyond 14 days is not feasible due to the short fieldwork phase of the survey.
Figures 2-4 display our main results. 8 We plot the estimated average treatment effects in the main text, while reserving the tables for the Appendix (S6). We show estimated treatment effects While the newspaper coverage may be thought of as a 'collateral' event (in the UESD terminology), it is, in a sense, a 'constitutive element' of the event being analyzed. 7 Our analytic sample consists of respondents interviewed within ±7 days of the cutoff residing within localities that have observations on both sides of the threshold (428 observations; 181 observations before the cutoff and 247 observations after). 8 Full descriptive statistics of the outcome variables are provided in Appendix S4.

Political Science Research and Methods
with and without controls for observable pre-treatment respondent-level characteristics. 9 In general, the inclusion of controls does little to alter our estimates.

Effects on institutional trust
We first consider measures of TRUST IN  Our focus on TRUST IN CONGRESS and TRUST IN POLITICAL PARTIES is informed by earlier research showing that, when presented with evidence of individual wrongdoing, citizens are likely to generalize perceptions of malfeasance to associated institutions (Bowler and Karp, 2004;Maier, 2011). We therefore expect that the sentence underscoring Menem's guilt cast a negative shadow on not only Menem himself, but also the specific political institutions in which he is embedded.  . Effects are estimated with controls and CIUDAD fixed effects within ±7 days of Menem's sentencing. The main error bars present 95% confidence intervals for the means; the small whiskers represent additional 90% confidence intervals. Outcomes are scaled on a 4-point scale. 9 We control for AGE, GENDER, EDUCATIONAL ATTAINMENT, LABOR FORCE PARTICIPATION, HOUSEHOLD POVERTY, and PREVIOUS ELECTORAL PARTICIPATION.
In contrast, Menem's sentencing may instead increase TRUST IN THE JUDICIARY, as citizens should positively evaluate this institution in light of its efforts to hold Menem accountable.
We observe that, as expected, Menem's sentencing erodes trust in Congress and parties (Fig. 2). Moreover, the court's announcement also led to lower trust in the judiciary. This latter result stands in contrast to the positive media coverage of the event. Further, it is unlikely that this result is driven by Menem's supporters objecting to the sentencing as (i) Menem had become a minor political figure by 2013, and (ii) supporters of his Justicialist Party (Partido Justicialista; PJ) only accounted for 6.1% of all respondents. More formally, we show in the Appendix that the negative effects are not driven by supporters of Menem's party (see Table A11) or residents of his constituency of La Rioja (see Figs. A4-A5). Rather than a partisan reaction, we instead observe a broad negative effect on TRUST IN THE JUDICIARY across demographic subsets of the population. At the same time, as expected, we find no negative impact on institutions not connected to the case at all, such as the military or the church (see Table A21).
To explore this negative finding further, we consider three additional outcomes: perceptions of PREVALENCE OF CORRUPTION in government, judgments about the state's PROGRESS IN REDUCING CORRUPTION in state institutions, and estimates of the state's ABILITY TO SOLVE CORRUPTION. The items are coded on a 4-point scale, with higher values representing more positive evaluations. If Menem's sentencing generated greater optimism about the ability of the judiciary to hold corrupt officials accountable, we should expect to observe positive effects on all three variables.
In line with our results for trust in the judiciary, we observe that Menem's sentencing actually renders citizens more pessimistic about the state's PROGRESS IN REDUCING CORRUPTION (Fig. 3). The coefficients relating to PREVALENCE OF CORRUPTION and the state's ABILITY TO SOLVE CORRUPTION are also both negative, although the effects are just insignificant at the 5% level. Taken together, these results indicate that the announcement of Menem's sentence served to erode confidence in enforcement agencies and increase pessimism about corruption in Argentina.

Effects on political engagement
We next turn to measures of political engagement. Specifically, we examine whether respondents stated that they would refrain from casting a valid vote in a future election (INVALID VOTE). 10 This variable is coded dichotomously. Second, we analyze five items eliciting respondents' willingness to participate in demonstrations and protests about various political issues that capture intentions . Effects are estimated with controls and CIUDAD fixed effects within ±7 days of Menem's sentencing. The main error bars present 95% confidence intervals for the means; the small whiskers represent additional 90% confidence intervals. Outcomes are scaled on a 0-1 scale. 10 We create this variable by combining the stated intention to either cast a 'spoiled ballot' or abstain from voting all together. Given that voting in Argentina is theoretically mandatory, abstaining from voting altogether might be considered a costly form of disengagement. See Figure A3 in the Appendix for a separate analysis of these forms of disengagement.
to DEMONSTRATE for political causes more broadly. Respondents provided answers on a 1 to 10 scale. We average responses over all five items and rescale them on a [0 1] interval.
We observe a consistent demobilizing effect on vote intention and willingness to demonstrate (Fig. 4). The effects are most striking for voting: results indicate that Menem's conviction causes an additional 8.6% of respondents to intend to cast an INVALID VOTE. Overall, our results are consistent with our previous findings on trust in institutions: information about Menem's sentencing not only increased popular pessimism about accountability but also reduced citizens' willingness to take part in both electoral and non-electoral forms of political activity.
We briefly outline here some additional analyses we conducted to assess the robustness of our main results. First, we expanded our sample to include all respondents interviewed within ±14 days of Menem's sentencing (see Appendix S7). We also conduct sensitivity tests dropping individual provinces one-at-a-time (see Appendix S8). Our substantive conclusions remain unchanged under these additional analyses. We also conduct a series of analyses to explore potential heterogeneous treatment effects, but the main findings are largely similar when dividing respondents by characteristics such as gender, age, prior voting status, and region (see Tables  A13-16 in the Appendix). Importantly, while we are cautious about examining the role of partisan identities-as partisanship itself could arguably be a post-treatment outcome-we find that the results remain largely unchanged when removing supporters of the PJ from the sample.
Finally, we test for pre-existing time trends during the control period and conduct a placebo test within the control group treating June 6th as the cutoff date (Muñoz et al., 2020). 11 We find no effect at this alternative date for most outcomes (see Fig. A10). 12

Additional evidence: Costa Rica
Our analysis so far yields evidence of wide-ranging negative effects of Menem's sentencing on institutional trust and political engagement. However, these results also raise an important question: given that every anti-corruption case is in some respects sui generis, are these results driven by the particulars of Menem's case, or might they hold more broadly across other contexts?
To address this question, we examine the case against former Costa Rican President Calderón as an 'out of sample' test of our hypotheses. Calderón was sentenced on October 5, 2009 to five years in prison on embezzlement charges in relation to an illegal loan obtained during his term in office from 1990 to 1994. Although Calderón was first arrested in relation to these charges in 2004 (and even served a five-month jail sentence), the 2009 verdict was significant in that it represented the culmination of the first public trial of a former President in Costa Rican history (Tico Times, October 2, 2009).
Our consideration of Calderón's case is motivated by the extent to which his circumstances differed from Menem's. Following a 'most-different' design, if we can observe similar negative effects in reaction to Calderón's sentencing, we gain confidence that our findings in both cases are not due to the differences between them (or the unique circumstances surrounding either case.) Here several key particulars are noteworthy. First, while Menem enjoyed immunity from incarceration as a sitting senator, Calderón had no recourse to such protections. 13 Second, while Menem had become relatively insignificant politically in Argentina by 2013, Calderón remained a powerful public figure in Costa Rica following his term in office, and had even announced his candidacy for the 2010 Presidential Elections. Taken together, these features 11 Additionally, we conduct placebo tests using three randomly-selected 'cut-off dates' (June 23rd, 27th, 29th) (see Appendix S11). 12 We find a small positive effect for WILLINGNESS TO DEMONSTRATE and a negative one for PROGRESS ON CORRUPTION. Given the total number of placebo tests conducted (10 outcomes x 4 alternative dates), one would expect such a number of significant 'effects' by random chance. 13 However, Calderón also appealed his sentence to the Supreme Court, which reduced his sentence from five to three years in a 2011 ruling.
suggest that Calderón's sentencing had more 'serious' implications than Menem's, and thus arguably sent a stronger signal about the state's anti-corruption efforts.
Data for our analysis come from the Latinobarómetro survey in Costa Rica which was fielded between September 30 and October 22, 2009. As before, we use the announcement of the court's decision on October 5 as the critical cutoff date to define treatment and control groups. As a manipulation check, Figure 5 shows that internet searches for Calderón also peaked on that date. 14 Before turning to our substantive results, we again assess the ignorability assumption. Balance tests of observable respondent characteristics show that respondents have similar profiles across the treatment and control groups (see Table A6).

Results
To the extent possible, our analysis focuses on the same outcomes as for our Argentina analysis, although some variables were not available in the 2009 wave of Latinobarómetro. We have a smaller analytic sample in Costa Rica, and thus report treatment effects estimated using a ±14 day window only. 15 We also present results with and without controls for observable respondent-level pre-treatment covariates. 16 For descriptive statistics of the outcome variables, see Appendix S4. Figure 6 displays results for institutional trust. Consistent with our prior results, we observe that Calderón's sentencing reduced trust in all parts of government, including the judiciary. The magnitude of the effects is, if anything, even larger than in Argentina (although our estimates are much less precise as a result of smaller sample size). 17 Next, we turn to two variables to examine the effect of Calderón's sentencing on political engagement. The first, INVALID VOTE, is constructed in the same manner as before. The second variable is constructed using an item asking whether voting or participation in protest movements is the most effective way to contribute to change, or whether it is not possible to contribute to change. We recode responses such that 1 represents an INCLINATION TO VOTE OR PROTEST, while 0 indicates that the respondent believes it is not possible to change things, regardless of what one does. Figure 7 shows the results. While we find no significant effect of Calderón's sentencing on the stated intention to cast an INVALID VOTE, we do find a significant negative result with respect to INCLINATION which is consistent with a broader pattern of political disengagement. Specifically, citizens interviewed in the aftermath of the court's announcement were significantly more pessimistic about their ability to effectuate change.
In summary, our analysis of the Costa Rica case yields results broadly in line with our previous findings. 18 Although the estimated treatment effects are not significant across all variables, this is most likely due to our smaller Costa Rica sample. It is also noteworthy that we do not find any evidence indicating a positive reaction to the announcement of Calderón's sentence. Instead, the pattern across both cases is consistent with a loss of institutional trust, increased pessimism, and political disengagement.

Discussion
Leveraging data from two Latin American countries, this study examined the effects of unprecedented judicial actions against corrupt officials on institutional trust and political engagement. 14 Balance tests are reported in Table A6. 15 Even with a larger window, we have only 125 observations with CIUDAD FEs. 16 The set of controls always consists of the variables listed in Table A6. 17 Questions about the PREVALENCE OF CORRUPTION and the STATE'S ABILITY TO SOLVE THE PROBLEM of corruption were unavailable in the survey, but we could examine views about the state's PROGRESS IN REDUCING CORRUPTION. Here, we find a slight negative but not statistically significant effect. 18 We report parallel robustness checks in the Appendix.

Political Science Research and Methods
Somewhat depressingly, we find consistent negative effects on a range of political attitudes including lower trust in the judiciary and (in Argentina) greater pessimism about anti-corruption efforts.
One explanation is that citizens may have been disappointed by the leniency or lack of finality in the courts' decisions, although the positive tone in media coverage speaks against this argument. Rather, to our minds, the more likely explanation is that information about the sentencing in both cases served to refocus public attention on the facts of the initial wrongdoing, rather than the judiciary's efforts to control corruption. Specifically, while details about the allegations against Figure 6. Trust in Courts, Parties, and Congress (Costa Rica). Effects are estimated with controls and CIUDAD fixed effects within ±14 days of Calderón's sentencing. The main error bars present 95% confidence intervals for the means; the small whiskers represent additional 90% confidence intervals. Outcomes are scaled on a 4-point scale. both Menem and Calderón were already well-known, they may not have been very salient (Solé-Ollé and Sorribas-Navarro, 2018), and thus the sentencing itself may have paradoxically 're-exposed' citizens to scandalous information, leading to a broad-based pessimistic response.
This outcome may have been particularly likely given the high-corruption contexts in which the judicial decisions were announced. Argentine politics especially has been plagued by chronic issues of malfeasance. And although Costa Rica has traditionally enjoyed a reputation as being one of the 'cleaner' countries in Latin America, this image had become significantly tarnished by 2009 following a wave of corruption scandals (Corbacho et al., 2016). Given this history, it may be the case that popular cynicism was too deeply entrenched in citizens' minds for beliefs to have shifted in response to singular judicial decisions.
These considerations raise the possibility that prosecutions may have more positive effects in low-corruption contexts. This portends to be a promising avenue for continuing research, especially as high-level prosecutions become a regular feature on the global political scene.