SciELO - Scientific Electronic Library Online

vol.27 número2Dádivas durante las elecciones mexicanas de 2018La nacionalización de los partidos mexicanos índice de autoresíndice de materiabúsqueda de artículos
Home Pagelista alfabética de revistas  

Servicios Personalizados




Links relacionados

  • No hay artículos similaresSimilares en SciELO


Política y gobierno

versión impresa ISSN 1665-2037

Polít. gob vol.27 no.2 Ciudad de México jul./dic. 2020  Epub 17-Feb-2021



Criminal Contagion. How Governor Detentions Weakened the PRI

Milena Ang1 

1Profesora asistente en el Departamento de Ciencia Política y Geografía de la UT-San Antonio. The University of Texas at San Antonio, Department of Political Science and Geography, Attn: Milena Ang. One UTSA Circle San Antonio, Texas, 78249. Tel. (210) 458 5600 Correo-e: milena.ang@utsa. edu.


Recent years have witnessed an unprecedented rise of Mexican governors prosecuted and incarcerated for corruption, most of which were from the ruling party PRI (Partido Revolucionario Institucional). Could these judicial cases partly explain the resounding electoral loss suffered by the PRI party in 2018? In this paper, I argue that the criminal behavior of PRI governors who were prosecuted affected the PRI’S most recent presidential electoral outcome. First, I argue that this criminal behavior affected the partisan reputation of the PRI. I illustrate this claim by showing that these criminal actions were often referenced by the media as evidence not only of individual malfeasance but of networks —embodied in the PRI party— that had enabled the criminal behavior. Second, I argue that these discourses affected electoral choice. I propose that if these discourses were indeed affecting the PRI vote, we should observe lower support for PRI among voters that are most exposed to these discourses such as voters of places with criminal governors and voters in districts with higher access to internet. Using a difference-indifferences setup and a cross-sectional analysis, I show that both of these groups indeed voted less frequently for the PRI. I conclude the paper by discussing the aftermath of the election and what the existence of this contagion means for partisan accountability.

Keywords: corruption prosecution; elections; political parties; governors


En los últimos años se ha producido un aumento sin precedentes de gobernadores mexicanos procesados y encarcelados por corrupción, la mayoría de los cuales pertenecían al partido gobernante, el Partido Revolucionario Institucional (PRI). ¿Podrían estos casos judiciales explicar en parte la estrepitosa pérdida electoral sufrida por el PRI en 2018? En este artículo, sostengo que el comportamiento criminal de los gobernadores priistas que fueron procesados afectó el más reciente resultado electoral presidencial del PRI. Primero, argumento que este comportamiento criminal afectó la reputación partidaria del PRI. Ilustro esta afirmación mostrando que estas acciones criminales fueron a menudo referidas por los medios de comunicación como evidencia no sólo de malversación individual, sino de redes -encarnadas en el PRI- que habían permitido el comportamiento criminal. En segundo lugar, sostengo que estos discursos afectaron el voto. Propongo que, si estos discursos estaban efectivamente afectando el voto del pri, deberíamos observar menor apoyo al PRI entre los votantes que están más expuestos a estos discursos, como los votantes de lugares con gobernadores corruptos y los votantes de distritos con mayor acceso a internet. Utilizando una estrategia de diferencias en diferencias y un análisis transversal, muestro que ambos grupos en efecto votaron con menos frecuencia por el PRI. Concluyo el artículo discutiendo las consecuencias de la elección y lo que significa la existencia de este contagio para la rendición de cuentas partidista.

Palabras clave: escándalos de corrupción; elecciones; partidos políticos; gobernadores


Although Mexican governors have long been accused of being corrupt, it was only until recently that they were frequently investigated and charged for it. Recent research reports that during the last sexenio before democratization —from 1994 to 2000— only four governors were investigated at the federal level (Ang, 2017), whereas during the administration of the PRI president Enrique Peña Nieto (2012-2018), over 15 local executives have been investigated and at least eight have been issued an arrest warrant. These detentions occur in a country where 48 per cent of respondents in a nationally representative survey mentioned corruption as one of the issues to be solved, just after insecurity and unemployment (INEGI, 2017).

Importantly, however, not all the parties were equal contributors to the pool of prosecuted governors. Of the eight prosecuted local executives that governed during Peña Nieto, seven of them were his copartisans. This overrepresentation of priistas was noted, emphasized even, in the 2018 electoral campaigns that portrayed the PRI as an institution built to protect or enable corruption. Indeed, the presidential candidate who would ultimately win the election, Andrés Manuel López Obrador (AMLO), centered his campaign around fighting corruption and criminal behavior and getting rid of the mafia del poder, political mafia, that had encroached in the federal and local administrations. During the campaigns it was also not uncommon to find newspaper editorials and pundits echoing the idea that the ruling PRI party had enabled or protected local executives that had illegally diverted millions of pesos.

Of course, these prosecutions have had important repercussions for the political figure of the governors. Once dubbed the new virreyes (Meza, 2009; Zuckermann, 2003), a term that signaled their almost absolute power and impunity, governors nowadays seem to be more and more vulnerable. In an unprecedented move, former Veracruz governor Javier Duarte, stepped down from his administration when faced with accusations of corruption (only to flee from the country). Former Durango governor Jorge Herrera Caldera requested an amparo (legal protection) against his possible detention, a request that was denied because he was not being investigated in the first place (Lastra Guerrero, 2016). This political weakening of the gubernatorial figure is also tangible in electoral results: governors that have been prosecuted have cost the governorship to their parties.

Despite the abundance of anecdotal evidence connecting criminal prosecution and party reputation or electoral results, research still needs to answer whether —and how— criminal corruption of a single individual affects party-level electoral results. Thus, the questions that motivate this paper are: What were the consequences of the criminal prosecutions of priista governors for the PRI party? Specifically, did the detentions of PRI governors polluted the PRI’S presidential electoral result? And if so, why?

This paper answers these questions by arguing that the criminal prosecution of local executives had two related effects. First, the public discussion that arose from the detention was less about the individuals and the crimes they had been accused of, and more about the political structure that produced the criminal behavior in question. Specifically, I present qualitative evidence to show that discourses in the media saw these detentions as evidence of a larger network of complicities that had enabled the malfeasance revealed by the prosecutions. Furthermore, I show that these networks were often embodied in the institution of the PRI, a discourse that is somewhat counterintuitive given that these prosecutions were conducted by a priista president.

Secondly, I argue that this public discourse portraying the PRI as a party that en-abled criminal behavior weakened the partisan brand, decreased the appeal of PRI candidates, and directly impacted the presidential electoral results. I demonstrate this point by presenting a difference-indifferences analysis of district-level electoral outcomes for the PRI presidential candidates in 2012 (Enrique Peña Nieto) and in 2018 (José Antonio Meade), and a cross-sectional analysis of PRI vote across districts with different access to such discourses. In this analysis, I show that although the PRI suffered considerably nation-wide, its electoral losses were even higher in states that had a criminal governor. I also find that districts with higher internet access voted less for the PRI, even after controlling for socioeconomic characteristics that might be related to internet access and PRI vote choice. Both of these groups of voters, I argue, are more likely to be exposed to public media and discourses of systemic corruption. I conclude the paper by discussing key implications of the existence of this mechanism: as corrupt behavior is made public and the electoral consequences of prosecution become more apparent for political parties, party leadership seems to be interested, and able, to discipline their members, or even publicly repudiate, for engaging in criminal behavior.

Corruption and criminal behavior

The questions addressed in this paper are relevant considering that free and fair elections are a way in which citizens can hold their government officials account- able for wrongdoings. Elections enable citizens to hire (elect) or fire (not reelect) public officials (Manin et al.,1999b; Adsera et al., 2003) based on their performance and actions in government (Manin et al.,1999a; Grzymala-Busse, 2007). Of course, the relationship between governmental actions and accountability might not be so straightforward. First, elections are a diffuse accountability mechanism because voters must consider a multiplicity of topics in a single election, topics that can and often are prioritized differently (Cheibub and Przeworski, 1999). Second, elections are also a weak tool for accountability because they entail choosing a candidate from a fixed set of alternatives. Thus, voting is not only a matter of looking at an incumbent’s performance, but also considering what the rest of the politicians can offer and how credible that offer is (Maravall, 1999).

Existing literature has examined at length the link between corruption and corruption scandals and vote choice. Research has shown that learning about malfeasance of politicians negatively affects campaigning strategies and reelection rates (Rennó, 2008) and decreases vote intention for candidates accused of such malfeasance (Chang et al., 2010). These effects, however, are mediated by individual factors like risk propensity (Morgenstern and Zechmeister, 2001), the opinion on the state of the economy (Klašnja and Tucker, 2013), or whether the voters benefitted from corruption.1 A different literature also looks at how context shapes the effect of corruption on vote choice, highlighting the role of media and how the coverage of malfeasance and its intensity shape voters’ choice (Ferraz and Finan, 2005; Chang et al., 2010).

But not all findings suggest that voters will react to the abuse of public office. Comparative empirical evidence has found that corruption scandals can reach a point of saturation in which the voters become somewhat resilient to the wrongdo- ings of politicians, and stop punishing them electorally (Kumlin and Esaiasson, 2012). In addition, practices such as clientelism and patronage can prevent voters from freely exercising their right to choose, further weakening the link between information and electoral accountability. Studies of clientelism in Japan and Mexico have shown that clientelistic practices gives incumbents an excessive advantage (Scheiner, 2006; Magaloni, 2006), while also deteriorating the quality of opposition,2 and even rendering it incapable of positioning themselves as clear alternative in terms of differentiated policy proposals (Grzymala-Busse, 2007).

The existing research, however, has been mostly interested in corruption scandals, and has used corruption detentions as a proxy measure of the former, thus addressing a somewhat different set of questions that the ones set forth at the beginning of this section. The few studies that separate the effects of corruption scandals and corruption detentions on vote choice have found that indeed the latter have effects that are distinguishable from the former,3 a finding that suggests that these two are distinct empirical phenomena with different political effects. In addition, the literature reviewed above has mostly focused in understanding the electoral effects of corruption scandals and detentions as they affect politicians involved in corruption scandals, overlooking the effects that they might have on other political actors and institutions.

This last limitation has been partially addressed by the comparative literature that has looked into how corruption and its prosecution can affect public trust (Chandler, 2006). More specifically, evidence on China, Japan, and South Korea also reports that the perception that the government is efficiently fighting corruption can positively impact trust in institutions (Kim and Voorhees, 2011). These findings are promising for the use of the criminal justice system when it comes to political corruption, but they still cannot address whether prosecution of corruption affects political parties and their electoral performance.

Why should parties suffer the consequences of their members’ actions, particularly when these are already being prosecuted? Put differently: what mechanism might trigger consequences for copartisans of an indicted politician? To answer this question, I borrow from the established notion that parties can serve as informational shortcuts, fulfilling the crucial function of conveying information to voters to help them cast a vote when information about specific candidates is scarce or too costly to acquire. Specifically, I borrow from this literature to argue that in the same way that party labels might be associated with ideological leanings or specific policies that are taken as information about candidates, parties can also be associated with widespread corruption. If this is the case, parties can inform (or misinform) voters whether candidates are likely to engage in, or cover up, corrupt behavior. I further develop this argument in the remainder of this section.

The idea of parties as heuristics, developed mostly in the study of American parties, originates from the assumption that voters make choices in an environment of great uncertainty and where information acquisition is costly. In these environments, parties can be seen as ‘labels’ that convey information to the voters about policy views, qualities, and other attributes of their candidates. In his classical work, Downs convincingly argues that voters must acquire information in order to cast a vote, but that the cost of acquiring such information is, in many cases, larger than the expected utility from casting a well-informed vote (Downs, 1957). Thus, he concluded, voters in general will not make any additional efforts, instead relying on whatever information they can easily encounter. Empirical research in the US shows that in the absence of complete information, voters rely on ‘heuristics’ to make choices (Ferejohn, 1990; Iyengar, 1990), and that partisanship is one of the most accessible ones (Huckfeldt and Sprague, 1990; Schaffner and Streb, 2002). Studies conducted in the case of Mexico show that party cues are helpful for voters and that the usefulness of this heuristics increases as topics become more complex (Merolla et al., 2007; Guardado Rodríguez, 2009).

The literature of party as heuristics, then, suggests that party labels help voters summarize relevant information to help them decide which candidate to vote for. I build from this logic to claim that if parties inform voters about their candidates, it is possible that the actions of candidates and incumbents also inform voters about the parties that these politicians belong to. My claim finds motivation in —and contributes to— scholarship studying why and how the behavior of party members can weaken party brands. Barrow (2007), for example, studies how candidates that switch parties weakens the relationship between voters and their parties, offering evidence from the Mexican state of Morelos to illustrate this claim. Lupu (2016) argues that conflict between members of the same party can dilute a party brand because voters will be less certain about what the parties stand for. Put differently, exiting literature has claimed already that some actions by party members can affect party brands, and can weaken the links between parties and their voters. My proposed mechanism posits that trials involving allegedly corrupt politicians can also affect the party label of those being prosecuted by associating such party with criminal behavior. Weakening the party brand should, in turn, affect the electoral results for candidates that share parties with the alleged criminal, but that have not been accused of any wrongdoing.

Importantly, however, I do not expect my proposed mechanism to work equally across all countries, nor even in all parties within a same country. Instead, I propose that some electoral characteristics can accentuate —or mitigate— the connection between the candidate’s (criminal) behavior and partisan-level consequences. Although originally developed to classify legislative elections, I draw from Carey and Shugart’s proposition that some electoral systems cultivate strong party reputations, instead of strong individual reputations, as a way to garner votes (Carey and Shugart, 1995), as well as from Samuel’s subsequent paper, who argued that parties’ strategies vary even within an electoral system (Samuels, 1999). The proposed mechanism —associating poor behavior of a candidate to his or her party— will be less likely when politicians are actively trying to differentiate themselves from other party members, and more likely when candidates want to be identified with their party precisely to exploit partisan reputation to their own advantage.

Based on the characteristics identified by these authors, I argue that Mexican presidential elections are likely to incentivize party reputations, fostering the connection between candidate actions and party behavior. For example, the PRI operates with a centralized ballot, meaning that parties must support the candidate and voters cannot ‘disturb’ this, thus strengthening candidate allegiance to party structures (Samuels, 1999: 492-493; Carey and Shugart, 1995: 420-421). In addition, Mexico publicly funds its parties, which means that the candidate’s access to campaign finance resources is mediated by the party, deepening the candidate’s dependence on keeping the party leadership happy. Similarly, political candidates often rely on the ample PRI structure to deliver pork, although such structure is mostly responsive to local leaders, not central ones, and has therefore been weakened during the last two decades (Cantú and Desposato, 2012). Finally, although there are some opportunities for electoral alliances, which might weaken the incentives to build a party reputation, the PRI tends to build alliances with the same small parties, minimizing damage to its brand. Therefore, the PRI’S structure and dynamics when it came to determining presidential candidates —a structure that had remained quite strong until this last electoral cycle— strengthened the connection between the party itself and its candidates, making it more likely to observe the posited mechanism.

Considering this, my argument is that criminalization of individual politicians, under certain conditions like the ones discussed here, can affect not only those directly involved in the criminal process, but also those who share the same party label. As discussed in previous paragraphs, this ‘contagion’ occurs partly because criminal behavior of politicians might affect party reputation. Since voters use parties as heuristics to convey information about candidates, belonging to the same party than someone accused of wrongdoing can impact one’s electoral performance. As I will show in the next section, I find evidence that the criminal cases involving PRI governors were often related to the PRI party structures in general, thus evidencing the contagion mechanism posited here. In this way, the Mexican 2018 elections suggest that putting a single politician in jail does not simply weaken him or her, but also the institutional networks that such politician is a part of, particularly parties. Specifically, I show that José Antonio Meade, the PRI presidential candidate, suffered electorally as a consequence of the criminal processes opened against his fellow copartisans.

So far, I have discussed criminal behavior broadly, but this paper focuses empirically on criminal behavior of Mexican governors understood as either political corruption or collusion with organized crime. Before introducing evidence to support my claim, I discuss why I focus on these types of criminal behavior among this particular set of actors. To be clear, I have no theoretical reasons to argue that specific criminal behavior is more or less likely to damage a party’s reputation, nor do I argue that prosecution of specific actors will always result in consequences for his or her party. Instead, I propose that contextual characteristics of each political system can illuminate the relevance of politicians —the more relevant the actors, the more likely are they to affect the party brand— and that the type of criminal behavior they are accused of affects their parties’ reputation differently. I discuss these two characteristics in the Mexican context to further clarify the relevance of the actors and behavior studied in this paper.

Contextually, Mexican governors are powerful actors that have unfettered access to large sums of money. Since the decentralization reforms that started in the 1990s, Mexican governors receive large federal transfers and can acquire public debt with little oversight which has led to unlawful mismanagement and appropriation of funds.4 Furthermore, governors also face almost no constraints when in power: the presidential “winner-take-all” arrangement means that executives are often popular politicians that had to win a direct popular vote and that, institutionally, lack a counterpart to restrain their use (or misuse) of power (Linz, 1985; Mainwaring and Shugart, 1997). Governor’s control of local bureaucracies and nominations of candidates for national legislators also means that parties often bow to them (Langston, 2010; Cantú and Desposato, 2012). In this way, governors in Mexico are likely actors to affect party reputations because of their importance within the hierarchy, an importance exacerbated by the amount of resources they control when in office.

These characteristics of the gubernatorial office, as well as the context of organized crime, increases the opportunities of governors engaging in criminal behavior of two types: political corruption and criminal corruption,5 both of which I argue are likely to trigger the “contagion” effect. By political corruption I mean that governors can divert funds from the public resources they command: the proverbial cookie jar in the form of almost unsupervised federal transfers. By criminal corruption I mean that governors can collude with organized crime by, for example, receiving bribes and kickbacks for letting criminal groups operate in certain parts of their jurisdiction. Crucially, both of these types of criminal actions are abuses that could only be committed by those with access to public power: any abuse of such responsibility would breach the mandate entrusted by voters when they elected these politicians. On the contrary, criminal behavior like shoplifting or stealing a vehicle breaks the law but will not necessarily damage public trust, and as such it is less likely to trigger the mechanism set forth here. Hence, criminal behavior that breaks the trust placed in an official elected to public office is more likely to affect the party reputation than breaking the law without abusing public office.

From this discussion, I hypothesize that the PRI governors that were criminally prosecuted affected the overall party standing in the 2018 elections in two distinct but related ways. First, I propose that the PRI’S reputation suffered considerable damages as a consequence of the prosecution of its governors. This phenomenon is observable in that the public discourses surrounding the criminal processes often transcend the specific cases that motivated the discourse by referring to networks that enabled the criminal behavior, and these networks are often embodied in the PRI. Second, I propose that this reputational loss directly impacted vote choice even in races where no candidates had been criminally prosecuted or investigated. This, I argue, means that voters considered the criminal corruption cases and used this to inform their vote choice. In this way, the reputational and electoral effects of criminal prosecution are not only limited to those involved in a legal case (Ang, 2019), but they can also damage political parties and their candidates in other races.

The remainder of this paper provides evidence to support this argument, and it does so in two analytically distinct, but complementary, steps. Next section shows the PRI’S reputational loss as a consequence of the arrests and criminalization of PRI governors. Specifically, I show how these arrests helped build and strengthen the narrative of the PRI as an institutional network that enabled corruption and provided impunity for candidates. Section “How arrests (and the PRI) affected its presidential candidate” develops two testable implications of the claim that these narratives had an electoral impact that affected the PRI presidential candidate, and presents quantitative evidence comparing the 2012 and 2018 elections to corroborate them. Section “Discussion and implications” concludes by discussing the immediate aftermath of the elections and the theoretical implications of the phenomena described here.

How arrests affected the PRI

As stated in the previous section, the first part of my argument is that criminalization of specific individuals can be taken as evidence that entire networks of institutions are complacent, if not complicit, with corrupt practices. I develop this argument by showing that the arrests shaped public discourse around the PRI. Specifically, I present qualitative evidence to illustrate that specific arrests were not seen only as cases of individuals abusing their public office, but as cases that evidenced how a party enabled the criminal behavior that detentions had revealed.

Nowhere was this clearer than during the detention of “the worst governor in history”, (Agren, 2018) Veracruz’ Javier Duarte (2010-2016). From the very beginning of his term, Duarte was publicly accused of irregularities6 and the intensifica-tion of violence under his government, including murdered journalists. But in May of 2016, a month before the elections for governor, an investigation denounced that Duarte’s administration had been embezzling millions of dollars using a network of shell companies (Animal Político, 2016). Faced with the accusations, Duarte stepped down just a few weeks before the end of his term, claiming that he would “face his accusations”. Two days after this resignation, he vanished, only to be arrested in Guatemala in April of 2017.

Undoubtedly, Duarte’s case impacted the figure of the PRI the most. Duarte’s figure has become a reference when talking about criminal corruption, partly because he embodied depredation to an almost deranged degree.7 Partly, however, the sheer size of his embezzlement operation suggests that he operated with the knowledge, if not the active participation, of dozens of public servants or politicians who turned a blind eye to his practices. Thus, it is not uncommon to find opeds and political analysis suggesting that Duarte’s detention not only evidences his personal criminal actions, but the existence of networks and institutions that en-abled it. Statements like “Duarte is proof that corruption is not the work of a single person, but of a network of links and complicities that involved both public and PRIvate people”(Peschard, 2017) or “[the] courts should call the former public servants, the front men and the accomplices that enabled [Duarte’s] big theft”8 describe how the case of Duarte evokes institutions and actors beyond himself.

But Javier Duarte was not the only politician to be described in this way. Similar statements are made regardin other PRI governors also prosecuted for corruption. An investigation surrounding Quintana Roo’s former governor Roberto Borge (2011-2016) was dubbed “The Pirates of Borge,” a reference to the networks involving more than 50 of his collaborators accused of participating in defrauding the public (Ibarra and Meza, 2016). César Duarte —former governor of Chihuahua who was arrested in Miami in July, 2020, after years of being fugitive— was also accused of operating within a network of politicians and operators that enabled his illicit enrichment (Torres Cofiño, 2017).

In all of these cases, the PRI is often identified as the network that enabled corruption and impunity. This association between the party and its candidates was not haphazard: right before the 2012 election, PRI candidate Peña Nieto appeared in the TV show Tercer Grado (2012), where he supported the “young actors of the new political generation” of the PRI, and listed Quintana Roo’s governor Beto Borge, Veracruz’ César Duarte, and Chihuahua’s César Duarte, all of whom were prosecuted years later. In doing so, the presidential candidate not only branded his party as renovated —a label that lost its persuasiveness a few years after— but also distinctly created a group of “new” or “renovated” politicians affiliated with the PRI. Therefore, claims like “the priisimo made of corruption a lifestyle and a way of exercising power” (Torres Cofiño, 2017) or “[PRI] governors that in their heyday symbolized the rebirth of the PRI party [… ] five months before the [2018] presidential elections, illustrate its decay”(Ferri, 2018), show that belonging to the PRI helped voters make sense of the accusations and arrests. Importantly, these association of the PRI party to its corrupt governors also extended to its presidential 2018 candidate José Antonio Meade, even when he was never accused of corruption: “[t]he Achilles heel of the PRI and its [presidential] candidate is corruption” (Zuckermann, 2018). Thus, at least in public media, criminal prosecutions of governors were not taken to be cases of individual malfeasance being uncovered, and instead they were depicted as cases that showed institutional-level practices, particularly of partisan (PRI) nature.

In the terms of the argument presented here, two clarifications are in order.First, the veracity or logical soundness of the claim that corruption was indeed en-abled by partisan networks is not important for the mechanism I propose. Instead, it is crucial for the public discourse surrounding the detentions of governors reproduced the idea that such detentions were instances of practices generalized in the PRI party. Second, my proposed argument goes beyond a “retrospective voting” claim. That is, my argument is not that governors ruled poorly and that triggers the punishment of the next gubernatorial candidate. Instead, I argue that because of improper and illegal behavior of governors while in office, voters might punish candidates of the same party that are running for different offices. Both of these clarifications are relevant because, as the next section will show, the association of criminal governors to the PRI in general directly impacted the electoral results of the PRI presidential candidate.

Before developing the second part of the claim set forth in this paper, one clarification is in order. This paper has focused PRImarily in the PRI governors that have been prosecuted. This is not to say that governors of other political parties are unimpeachable, nor that they are completely trustworthy. In fact, a 2017 survey reported the trustworthiness of political parties, and the numbers were abysmal: only 17.8 per cent of citizens trust in political parties, and that percentage is even lower in states like Mexico City (around 7%) or Oaxaca (11.8%) (INEGI, 2017). But as discussed above, the PRI’s particular institutional characteristics makes it a more likely case to observe the specific mechanism. Given that an overwhelming proportion of prosecuted governors are from this party (only one non-PRI governor, PAN’S Guillermo Padrés was arrested during this period versus seven PRI governors), the PRI party is more frequently associated with criminal governors, although Padrés’ case also invokes ideas of networks of corruption.9

How arrests (and the PRI) affected its presidential candidate

The previous section argued that recent arrests and prosecution of local executives were used in the media not only as cases or examples of corrupt individuals, but as cases where institutions and political networks, particularly the PRI party, enabled criminal behavior of public servants. I begin this section by proposing an observable implication of the existence of the mechanism set forth in this paper. If indeed the PRI had suffered a reputational loss due to its governors’ criminal prosecution, this loss would not necessarily affect all voters equally, because not all voters pay equal attention to political scandals and the discourses surrounding them. Plausibly, voters in states with a criminal governor would have been exposed to public media reproducing these discourses at a higher rate than voters in states with no criminal governors. This is so because voters are more likely to be interested, or at least exposed, to news and opeds regarding their own governor than other governors. Furthermore, it is also plausible that citizens ruled by a governor who turned out to be corrupt would have suffered the consequences of such actions directly, making the acceptance of the corruption-as-network discourse more likely (Ang, 2019). Therefore, an observable implication of the mechanism proposed here is that voters in states with a criminal governor should be more prone to the depiction of the PRI as an enabler of the grand corruption schemes suffered in their states, and therefore would be the ones that would be more likely to punish the PRI for the malfeasance of its members.

Comparing electoral results across time would not quite provide convincing evidence towards the argument presented here. Showing that there was a decrease in votes cast for the PRI across the years would be insufficient as there are plenty of factors that change from one election to the next. Figure 1, which shows the presidential electoral results of all major parties for presidential candidates since 2000, illustrates the shortcoming of simply comparing across time: the figure shows a steep decrease in the percentage of votes cast for the PRI between 2012 and 2018, a decrease that is comparable to the one suffered between 2000 and 2006. Yet the corruption cases between these two sets of elections are not equal, nor are the characteristics of these elections. Thus, in order to show that the electoral results of 2018 were indeed affected by these corruption scandals, my analysis must show that the latest electoral result is a consequence of the corruption scandal and not due to other reasons, like the quality of the presidential candidate or the opposition.

Source: Own using electoral results reported by ine. Note: The figure shows the main political parties in Mexico. PRD and the PAN formed a coalition and presented a single candidate in the 2018 election. The party Morena was founded after the 2012 elections, but the candidate that competed in 2018, winner Andrés Manuel López Obrador, had already competed as candidate of the PRD in 2006 and 2012.

Figure 1 Presidential electoral results (2000-2018)  

I propose instead to present a research design that exploits both the geographical and temporal variation of corruption prosecution in explaining the presidential vote received by the PRI presidential candidates.10 I do so in two distinct, but complementary, steps. The next subsections explain each of these analysis in depth.

D-i-D estimates of PRI electoral loss

I begin by presenting a difference-indifference design (Lechner, 2011), which compares a set of units at two points in time, some of which will have experienced a treatment between time 1 and time 2. By comparing how the outcome of interest changed across time among treated units to the change across time among untreated units, we can identify an effect of the treatment in the outcome. For this first set of estimations, the outcome of interest is the votes received by the PRI presidential candidates in 2012 and 2018 in each of the electoral districts. Districts in states with PRI governors involved in a criminal corruption are considered units that have received treatment, and they are compared to the districts with no criminal PRI governors. In this way, the hypothesis derived from the discussion above can be rewritten to state that the difference in PRI votes in districts of PRI criminal governors versus districts of non-criminal PRI governors will be larger in 2018 than the difference in PRI votes between these two groups of districts in 2012. I operationalize the outcome of interest using the variable, PRI vote, which reports the ballots issued solely for the PRI party in a given electoral district as a proportion of the effective votes in that district.

The relevant independent variable, or treatment, is whether the governor of a state was involved in a criminal corruption case between 2012 and 2018. Out of the 32 federal entities, I identified 74 governors who had been in power during or after the 2012 election. Since governors in Mexico can only be criminally prosecuted after a political trial,11 or once they have left their office, I removed those who were still in office during the 2018 elections. Out of the 46 governors in this list, I identified seven priistas in six states that had been arrested or issued an arrest warrant due to involvement in corrupt activities.12 Formally, we can write the empirical strategy in model form:

PRI voted,s,y=μd, s,+y*criminal_PRI+δ*Y_2018++ τ*criminal_PRIS*Y_2018++β*Xd,s,y+εd, s, y

Where the main dependent variable, or outcome of interest (PRI vote d, s, y ) is the proportion of votes cast in favor of the presidential PRI candidate in each electoral district (d ) in state s, at time t (either 2012 or 2018). The first set of estimators, µ_d are district-specific intercepts, which capture the average support received by the PRI in each district.13 The variable criminal_PRIs is an indicator variable that takes the value of 1 if at least one governor of state s was from the PRI and involved in a criminal proceeding at some point between 2012 and 2018, and 0 otherwise. Variable Y_2018 also stands for an indicator that takes the value of 1 if the election observed is on 2018, and 0 if it is 2012. Thus, the coefficient δ will capture the average difference on votes received by the PRI presidential candidates across all districts from the election in 2012 to the election of 2018.

The estimator τ is the main estimator of interest, as it is associated to the interaction between the geographic variation, criminal_PRIs, and the time variation, Y_2018. This interaction will take the value of 1 if the observed district is in a state with a corrupt governor after the governor was involved in a criminal proceeding, and 0 otherwise. Following the discussion above, I expect τ to be negative and statistically distinguishable from 0 because the association of a corrupt governor to a corrupt party should be higher in states where the governor was involved in a criminal corruption case.

The final term X d, s, y captures a number of control variables. First, and most importantly, since not all corrupt governors end up being prosecuted, or investigated even, it could be that the variable criminal_PRIs by itself is not only measuring criminal corruption, but rather of corruption itself. If this were true, the estimator of interest (τ) could be “picking up” the effects of corrupt, or even very corrupt governors. I address this by adding a variable that captures a given governor’s involvement in a corruption scandal between 2012 and 2018 (scandal_PRIs). For each PRI governor, I conducted a search of whether he or she had been accused of corruption, and marked 1 when I found evidence of accusations.14

I also include two measures of political performance: economic_growth s and homicides s . The first variable reports the average economic growth in the state the year before the election. Such variable accounts for the possibility that governors that perform poorly might be simultaneously more prone to prosecution and less capable of bringing in votes. The second variable, homicides s is a crucial possible confounder particularly in the case of Mexico, where drug-related violence has increased everywhere, but not at the same rate. In particular, organized crime in Mexico is so closely related to politicians that scholars agree that sometimes it is hard to disentangle structures of drug trafficking and structures of the State (Bataillon, 2015; Illades, 2015). Thus, it could be that governors that allow organized crime to operate freely are both more likely to be arrested and to lose support from voters. I account for this possibility by controlling for the homicides per 100 000 people committed in the state in the calendar year before the election. The final model also includes the population in each district (logged), and Access to Internet as covariates. This latter variable is included because it could be that citizens in states that have more access to information would be both more likely to find and denounce criminal actions of their governors, and less likely to vote for the PRI.

Before presenting the estimations, I discuss a crucial assumption, often referred to as no spillover effects, that underpin the causal inference in the difference-indifference approach. This assumption is necessary for τ to be an unbiased estimate of the effect of criminal trials in electoral outcome. Of course, it is hard to argue that this assumption holds in the data used: after all, criminal governors made national news, and it is more than likely that voters nation-wide followed these cases. However, any spillover would decrease the difference between treated and non-treated districts, biasing τ towards 0. Thus, the spillover effects would make it even harder to find an effect of criminal governors in districts that received the treatment even when there is one. Therefore, the estimations that I present here can be seen as a floor or minimum effect. Since the purpose of this statistical analysis is to present evidence towards the observable implication of my argument, and not to provide an exact point estimate of the effects, violation of the spillover assumption will not irreversibly damage for my argument.

Table 1 presents the results of the estimation of interest. The first line reported stands for the average difference in votes for the presidential candidate in districts with criminal_PRI governors in 2012, that is, before they were prosecuted, and the second row reports a similar comparison, but among politicians that were involved in a scandal. It is important to note that all criminal_PRI governors were involved in a scandal_PRI, so the overall comparison should include these two coefficients. In short, criminal_PRI is only statistically significant when including governors involved in a scandal (models 2, 4, and 5). Overall, what these two estimations jointly suggest is that scandalous governors that were not prosecuted drew an average of 5 per cent more effective votes for the PRI candidate Enrique Peña Nieto, the candidate that would ultimately win the 2012 elections. The candidates that were involved in a prosecution also brought more votes than non-scandalous governors for Peña Nieto, but less than those PRI governors accused of corruption but that were not prosecuted.

Table 1 Change in votes for PRI (districts governed or not by a prosecuted governor) 

Model 1 Model 2 Model 3 Model 4 Model 5
Criminal_PRI -0.002 -0.039*** -0.001 -0.040*** -0.039***
(0.008) (0.009) (0.008) (0.009) (0.009)
Scandal_PRI 0.054***
Y_2018 -0.140***
Criminal_PRI*Y_2018 -0.027***
-0.028*** (
Scandal_PRI*Y_2018 -0.002 -0.003 -0.002
(0.008) (0.008) (0.008)
Access to Internet -0.139***
Economic growth -0.003 -0.006** -0.007**
(0.003) (0.003) (0.003)
Homicides per 100000 0.00003 0.0003 0.0003
(0.0003) (0.0003) (0.0003)
Log(population) 0.026* (0.015)
Constant 0.328*** 0.312*** 0.330*** 0.313*** -0.022
(0.005) (0.005) (0.006) (0.006) (0.194)
District intercepts Y Y Y Y Y
Observations 600 600 600 600 600

Source: Own elaboration. See Appendix for data sources. *p < 0.1; **p < 0.05; ***p < 0.01.

The third row, Y_2018, shows that the PRI presidential candidate lost, per district, an average of 16 per cent of the effective votes between the 2012 and the 2018 elections (statistically significant across all specifications). In perspective, the electoral margin nation-wide between the 2018 winner, Morena’s candidate Andrés Manuel López Obrador and PRI candidate José Antonio Meade was 36.8 per cent. Thus, this 16 per cent loss would have not changed the winner, but it could have placed the PRI in second place, as the PAN candidate Ricardo Anaya obtained 22.3 per cent of the votes (compared to the 16.4 per cent obtained by Meade). Of course, this loss can likely be attributed to a number of factors that changed between 2012 and 2018. For example, it could be that this loss is due to differences in the quality of the PRI presidential candidates, or to the differences in the candidates presented by the opposition. Another possibility, one that is directly related to the topic of criminal governors, is that at least part of this effect is being driven by the corruption scandals. That is, as discussed in section 3, corruption scandals made national headlines, and some of them were so well-known that they shaped the media depiction of the PRI. Of course, since this coefficient is capturing the difference between 2012 and 2018, it cannot be solely attributed to governors: after all, PRI president Peña Nieto was also involved in a well-known corruption case, which could also be impacting votes for PRI across the country.

The main coefficient of interest is reported in the fourth row. This estimation is associated with an indicator variable that takes the value of 1 if the electoral district is in a state with a criminal governor after the governor was arrested or issued an arrest warrant. As explained above, we expect this coefficient to be negative and statistically significant because we would expect the criminal malfeasance to inflict more electoral damage in places that directly experienced that malfeasance. Put differently, these are the districts where we would expect a higher loss of electoral support for PRI because voters in these districts were more exposed to the discourse of PRI institutionalized corruption. The estimations show a statistically significant and negative, albeit somewhat small (around 2.5 per cent), effect of criminal governors in votes for the PRI.

Importantly, this effect is robust to the inclusion of the control variable scandal_ PRI, included as treatment effects in models 2, 4, and 5. Taken together, these models show that the electoral weakening of the PRI in states with criminal governors is not only attributed to the corruption itself, but its criminalization. Furthermore, models 2, 4 and 5 show no evidence to suggest that scandalous governors cost more electorally to the PRI than non-scandalous governors: in both of these models I find no statistically significant difference between these two groups. Of course, this lack of effect could also be due to the sample size. Nevertheless, the inclusion of this variable enables me to show that prosecution of governors affects vote choice distinctly than just corrupt governors.

The estimation (model 5) shows that the population of a district is statistically significant and positive, suggesting that more populated districts voted more frequently for the PRI on average. Finally, the two variables that measured the gubernatorial performance, Economic growth and Homicides per 100 000 show some interesting results. On the one hand, Economic growth is both negative and statistically significant, suggesting that indeed governors that performed poorly received less votes, on average, for the PRI. Interestingly, however, violence in a state is not statistically associated with the votes received by the PRI candidate, a finding that is interesting in light of existing literature linking electoral behavior and violence.15

Table 1 also included, for all specifications, the variable Access to Internet. This covariate was included as a way to control the possible confounder that states with more informed citizens are states are both more likely to prosecute governors for malfeasance as well as less likely to vote for the PRI candidate. If this were the case, then adding this control variable would affect our main findings and the coefficient of interest would disappear. Yet, the main result of interest is quite robust to the inclusion of the covariates. Furthermore, the coefficient associated to Access to Internet is negative and statistically significant across all models, which suggests that more informed districts indeed cast less votes for the PRI. Of course, we must be cautious not to overinterpret this particular coefficient as it is merely added as a control variable in these estimations.

Cross-sectional estimation of vote for PRI

I complement the evidence presented above with a cross-sectional estimation of votes for José Antonio Meade in 2018, explained as a function of Access to Internet. The logic behind this second set of estimations is as follows: if indeed the discourse of PRI prosecutions as evidence of institutional corruption affects vote choice because voters are more exposed to it, then voters who have more access to internet should be less likely to cast a vote for the PRI presidential candidate. Again, given the lack of individual-level data on vote choice, I rely on aggregate district-level results, and use the variable Access to Internet which reports the percentage of houses with internet at the district level (INEGI, 2015). The estimated model can be formalized as follows:16

PRI voted,s=μs,'+γ'*criminal_PRIs+ δ'*InternetAccessd,s++τ'*criminal_PRIs* InternetAccessd,s++β'*Xd,s+εd,s'

Where the PRI vote in a given district in the year 2018 is a function of whether the state had a governor that ended up being prosecuted (coefficient γ’), and the proportion of households that have access to internet (coefficient δ’). According to the previous discussion, this last estimate should be negative and statistically significant: voters in districts where internet access is higher, I argue, will be more exposed to the news of malfeasance as well as the idea that such malfeasance is evidence of larger corruption networks. In addition, models 3 through 4 add an interaction effect between these two variables. The coefficient τ’ associated to this interaction effect measures the additional effect of internet access in the states whose governors were prosecuted for malfeasance.

I include a similar set of control variables to account for possible confounders. First, I add the variable Scandal_PRI which, as discussed, might be the driver behind the overall loss of votes for PRI, as opposed to the criminalization of corruption. I also include performance covariates (Economic growth and Homicides per 100000) to account for the possibility that corrupt governors are bad at governing and more likely to get caught for mismanagement. Finally, models 3 and 4 include the variable Phone Access, which measures the proportion of households that have landlines. This variable is included as a way to account for the possibility that the coefficient of interest, δ’ is capturing not only the effect of higher access to public media, but also of voters with higher income. Put differently, it could be the case that districts with higher internet access are also districts with wealthier citizens, which have distinct voting patterns. Since information on median income or socioeconomic status is not readily available at the district-level, I opt to proxy it using Phone Access, a measure that should be correlated with socioeconomic characteristics but is not a direct determinant of exposure to media and discourses.

The first line in Table 2 reports the average difference in votes obtained in states with a criminal governor versus those with no criminal governor. The table shows that all average estimations are negative, which is to be expected, but they are only statistically significant (p < 0.05) in model 4, where we include the variable Scandal_PRI. This means that in three out of the four models, we cannot reject the hypothesis that voters governed with a prosecuted PRI governor voted less, on average, than voters who were not governed by a prosecuted PRI governor. Although this is a surPRIsing result, it bears noting that this analysis contains a relatively low number of observations (only 300), which would make it harder to find small effects. The fact that all estimates are negative suggest that this could be the case.

Table 2 Votes for PRI in 2018 (districts with varying access to internet) 

Model 1 Model 2 Model 3 Model 4
Criminal_PRI -0.031* -0.027 -0.028 -0.063***
(0.018) (0.021) (0.021) (0.020)
Scandal_PRI 0.054***
Internet access -0.138***
Internet -0.011 -0.011 -0.008
Access*Criminal_PRI (0.032) (0.032) (0.032)
Economic growth 0.007 0.007 0.007 0.012
(0.009) (0.009) (0.009) (0.008)
Homicides per 100000 -0.001* -0.001* -0.001* -0.001*
(0.001) (0.001) (0.001) (0.001)
Phone access -0.010 -0.011
(0.035) (0.034)
Constant 0.199*** 0.199*** 0.199*** 0.177***
(0.014) (0.014) (0.014) (0.013)
State intercepts Y Y Y Y
Observations 300 300 300 300

Source: Own elaboration. See Appendix for data sources. *p < 0.1; **p < 0.05; ***p < 0.01.

Importantly, however, the third line displays the coefficients associated with Internet Access, which are the main coefficients of interest. As can be seen, these results support the existence of the mechanism proposed: in all estimations, higher internet access is associated with lower votes for the PRI. This result is quite similar across specifications: around 0.12 per cent average less votes for PRI when comparing districts where no houses have internet access to all households having internet access. Importantly, this result is robust to the inclusion of the variable Phone Access (model 4), which can be taken as evidence that the effect of Internet Access is not purely picking up socioeconomic characteristics of the electorate in each district.

Finally, the Table 2 shows that the coefficient associated with the interaction effect (τ’) is consistently negative, which is in accordance to the mechanism discussed here, but it is not statistically significant in any of the models. This lack of results suggests that voters with more exposure to the discourses referenced above in states with criminal governors were not necessarily less likely to vote for the PRI when compared to voters with more exposure in states with no criminal governors. This absence of evidence should not be confused with evidence of absence. First, interaction effects necessitate higher number of observations to be able to be estimated.17 In addition, it could be that the variable Internet Access is not appropriately capturing the underlying concept of exposure to corruption scandals. Therefore, Table 2 does not provide evidence to support the mechanism explained above, at least when operationalized in this way.

Importantly, the results presented in Table 1 are robust to the inclusion of two additional possible confounders, as well as its estimation without the state of Veracruz. First, it could be that in states where the PRI performs really well electorally, governors feel more comfortable abusing public office and they are more likely to suffer bigger losses. To test if this is the case, model 3 of Table A2 in the includes the variable of Past_Legislative_Vote_PRI, which captures the PRI’S electoral performance in the previous (legislative) federal election. This variable is not statistically significant, nor does it change the substantive results reported above. Second, the model 4 of Table A2 includes the variable Proportion_Urban_Sections, which captures the proportion of electoral sections that are in urban areas for each district. This variable is included to control for the possibility that governors in states with higher urbanization rates are more likely to suffer electoral consequences for their criminal behavior, and although the variable is indeed negative, it does not change the main results of interest. Finally, Table A4 in the Appendix reports the same set of estimations but removing Veracruz from the sample. I include these estimations because Duarte’s case is paradigmatic in terms of the amount of funds embezzled, which could be driving the main findings. Again, I find no evidence that this is the case: even removing Veracruz from the sample I find that governors that were arrested indeed lower the votes obtained by their party in their states.

Votes for PRI and allies

Before offering some concluding remarks, I discuss some issues with the operationalization of the dependent variable, votes for the PRI. The Mexican electoral system allows citizens to mark several parties in the presidential ballot, as long as these parties are postulating the same candidate. The previous set of results were estimated using only the ballots that were marked for the PRI, thus getting rid of other votes that selected the PRI and other parties. The reason behind only looking at PRI votes is precisely that the mechanism I set forth here was PRI-specific. That is, there is no reason to think that parties that presented the same partisan candidate as the PRI also suffered electorally as a consequence of the gubernatorial prosecutions.

I take advantage of this feature, and present the same set of estimations than Table 1 using the variable PRI+ vote, a variable that reports the proportion of effective ballots that were marked for PRI and other parties of its electoral coalition.18 I take these estimations to be a plausibility probe of my mechanism: since there is no reason to think that other parties would suffer electoral losses as intensely as the PRI, we should not observe such a sudden change in electoral preferences that were expressed for these PRI allies.

The results presented in Table 3 supports my interpretation. The main coefficient of interest, associated to the interaction term Criminal_PRI*Y_2018 is negative, but only statistically significant in models 1 and 3, none of which include the variable Scandal_PRI. Once we add this covariate, the effects disappear. This suggests, once again, that the mechanism by which voters electorally punished criminal corruption was, at least partly, via the PRI party. Based on these estimations, there is no reason to suggest that the parties that ran in alliance with the PRI suffered because of the latter’s problematic governors. Thus, the reputational loss discussed in section 3 seems to have affected only the PRI.

Table 3 Change in votes for PRI and allies (districts governed or not by a prosecuted governor) 

Modelo 1 Modelo 2 Modelo 3 Modelo 4 Modelo 5
Criminal_PRI -0.012 -0.056*** -0.012 -0.057*** -0.056***
(0.008) (0.009) (0.009) (0.009) (0.009)
Scandal_PRI 0.065***
Y_2018 -0.212***
Criminal_PRI*Y_2018 -0.019** -0.009 -0.020** -0.011 -0.012
(0.009) (0.011) (0.009) (0.011) (0.011)
Scandal_PRI*Y_2018 -0.014 -0.015* -0.014
(0.009) (0.009) (0.009)
Access to Internet -0.142***
Economic growth -0.003 -0.008** -0.008**
(0.004) (0.003) (0.003)
Homicides per 100000 0.0001 0.0003 0.0003
(0.0003) (0.0003) (0.0003)
Log(población) 0.031*
Constant 0.406*** 0.387*** 0.407*** 0.388*** -0.010
(0.005) (0.005) (0.006) (0.006) (0.211)
State intercepts Y Y Y Y Y
Observations 600 600 600 600 600

Source: Own elaboration. See Appendix for data sources. *p < 0.1; **p < 0.05; ***p < 0.01.

Discussion and implications

This paper began with the assertion that the recent wave of detained governors had political consequences beyond those suffered by the individuals involved in those cases. Because of the prevalence and importance of these cases, public media reproduced depictions of these detentions as evidence not only of the malfeasance revealed, but also of networks of impunity. These networks were embodied by the PRI, the party that contributed the most to the pool of criminal governors. As a consequence, the PRI lost its appeal as a viable party, and this was reflected in its electoral results of 2018, where its candidate suffered a resounding loss.

This argument has important theoretical and political implications. Theoretically, this paper identifies an effect that corruption prosecution has on political parties. This effect contributes to the growing literature on the relationship between corruption and trust in political systems and institutions. Furthermore, it identifies the electoral implications of this effect: by associating corrupt governors with their parties, parties suffer electorally even in non-gubernatorial races (Ang, 2019).

Politically, the Mexican experience can inform us of potential ways of improving accountability of political parties beyond simply voting them out of office. After the reactions described here, where media and pundits blamed the PRI for the voraciousness of their governors, PRI militants and directives attempted to assuage the concerns raised by the corruption scandals and their ensuing prosecutions. Perhaps noting that electoral support for his campaign was dwindling, PRI presidential candidate José Antonio Meade attempted to address the corruption scandals head-on, and during a political rally in Veracruz claimed that “It hurts that Javier Duarte injured our prestige. It hurts because it does not define us, and it does not describe us. The priismo of Veracruz is not defined by corruption or impunity” (Gómez, 2018). Similar statements were made by PRI national leader Enrique Ochoa Reza, who attempted to convince voters that the PRI party was free of the “sins of corruption” after calling Duarte a “national embarrassment”.19 Attempts to distance the PRI from these governors continued even after the 2018 elections. A document allegedly written by a priísta group to the PRI leadership blamed the then president Peña Nieto and the “shameful generation” of politicians for the disastrous electoral results (El Financiero, 2018).

But the attempts to solve the crisis brought upon by the criminal governors went beyond vague statements. Notably, Javier Duarte and César Duarte were expelled from the PRI ranks when they escaped their arrests warrants, an extreme action considering that other governors involved in corruption scandals were never publicly rejected by PRI.20 The willingness of the PRI to expel and publicly reject any association with criminally prosecuted governors marks a change from its previous treatment of corruption scandals. This suggests that the 2018 electoral PRI crisis fundamentally shaped what was considered acceptable within the PRI. This, in turn, opens up the possibility that voters, when punishing candidates for what their copartisans did, are forcing parties to improve internal discipline and a better selection of politicians and candidates.


This paper aimed at understanding the consequences of the recent wave of corruption scandals for the PRI party, both in terms of its reputation as well as its 2018 electoral results. I argued that the criminal prosecutions were not merely understood as cases involving individuals, but rather as instances that evidenced networks and institutions that enabled criminal practices. Specifically, the networks evidences were often embodied in the PRI, the party that most frequently contributed to the pool of criminal governors.

This paper also showed that this association of criminal behavior with a specific institution, the PRI, affected the electoral results of candidates associated with such party. I argued that, if this association were indeed affecting the appeal of the PRI as a political option, the loss of votes should be greater in places where voters were more likely, on average, to have been exposed to public media that reproduced these associations. I argued that two groups of voters that would be more exposed would be (1) those in states that had a prosecuted governor, and (2) those who had greater internet access. The evidence presented here shows that indeed districts in states with governors that were prosecuted lost more votes for the PRI from 2012 to 2018 than districts in states with no prosecuted governors. This effect, although small, is consistent with the implications of my argument. Similarly, I found that districts higher proportion of households with internet access were indeed associated with less votes for PRI, even after including Phone Access as a way to control for possible socioeconomic lurking variables.

I finalized this paper by arguing that these dynamics are somewhat promising for the possibility of strengthening partisan accountability and responsiveness to the voters. The findings presented here, I argue, are a somewhat hopeful interpretation of the electoral effects of criminalizing corruption. Particularly, it is worth noting that if politicians and entire political parties realize that they pay the consequences of what members in their group do, they might be incentivized to monitor and punish corruption within their ranks. Should this trend continue, Mexico’s prospects of developing a responsive party system that can be held accountable might not seem so elusive.


Adsera, Alícia, Carles Boix y Mark Payne (2003), “Are You Being Served? Political Accountability and Quality of Government”, Journal of Law, Economics, and Organization, 19(2), pp. 445-490. [ Links ]

Agren, David (2018), “Mexico: ‘Worst Governor in History’ Sentenced to Nine Years for Corruption”, The Guardian, 27 de septiembre, secc. World News, disponible en: [fecha de consulta: 15 de enero de 2019]. [ Links ]

Ang, Milena (2017), “Corrupting Accountability: Elite Control and Corruption Prosecution in Comparative Perspective”, tesis doctoral, University of Chicago-Department of Political Science. [ Links ]

Ang, Milena (2019), “Guilty by Association? The Spillover Effects of Corruption-Related Prosecutions”, preparado para el Seminario en IC3JM, 22 de marzo. [ Links ]

Angel, Arturo (2017), “Opinión: Del nuevo PRI a la corrupción de Javier Duarte”, The New York Times, 8 de mayo, secc. Opinión, disponible en: [fecha de consulta: 18 de abril de 2019]. [ Links ]

Animal Político (2012), “Retienen avión de Javier Duarte con 25 mdp en efectivo”, Animal Político, 30 de enero, disponible en: [fecha de consulta: 15 de enero de 2019]. [ Links ]

Animal Político (2016), “Empresas fantasma de Veracruz”, 24 de mayo, disponible en: [fecha de consulta: 15 de enero de 2019]. [ Links ]

Armesto, María Alejandra (2015), “Competencia política y gasto particularista de los gobiernos subnacionales en México y Argentina”, Latin American Research Review, 50(1), pp. 160-183. [ Links ]

Barrow, Lynda K. (2007), “Party On? Politicians and Party Switching in Mexico”, Politics, 27(3), pp. 165-173, doi: [ Links ]

Bataillon, Gilles (2015), “Narcotráfico y corrupción: Las formas de la violencia en México en el siglo XXI”, Nueva Sociedad, 255, pp. 54-68. [ Links ]

Cantú, Francisco y Scott Desposato (2012), “The New Federalism of Mexico’s Party System”, Journal of Politics in Latin America, 4(2), pp. 3-38, doi: https://doi. org/10.1177/1866802X1200400201. [ Links ]

Carey, John M. y Matthew S. Shugart (1995), “Incentives to Cultivate a Personal Vote: A Rank Ordering of Electoral Formulas”, Electoral Studies, 14(4), pp. 417-439. [ Links ]

Chandler, David (2006), “Building Trust in Public Institutions? Good Governance and AntiCorruption in Bosnia-Herzegovina”, Ethnopolitics, 5(1), pp. 85-99, doi: https://doi. org/10.1080/17449050600558272. [ Links ]

Chang, Eric C.C., Miriam A. Golden y Seth J Hill (2010), “Legislative Malfeasance and Political Accountability”, World Politics, 62(2), pp. 177-220. [ Links ]

Cheibub, José Antonio y Adam Przeworski (1999), “Democracy, Elections, and Accountability for Economic Outcomes”, en Adam Przeworski, Susan Stokes y Bernard Manin (eds.), Democracy, Accountability and Representation, Nueva York: Cambridge University Press, pp. 221-250. [ Links ]

Costas-Pérez, Elena, Albert Solé-Ollé y Pilar Sorribas-Navarro (2012), “Corruption Scandals, Voter Information, and Accountability”, European Journal of Political Economy, 28(4), pp. 469-484, doi: [ Links ]

Downs, A. (1957), “An Economic Theory of Political Action in a Democracy”, The Journal of Political Economy, 65(2), pp. 135-150. [ Links ]

El Financiero (2017), “El PRI está limpio de pecados de corrupción: Ochoa Reza”, 18 de mayo, disponible en: [fecha de consulta: 15 de enero de 2019]. [ Links ]

El Financiero (2018), “Priistas acusan a Peña y a ‘la generación de la vergüenza’ por derrota electoral”, 3 de julio, disponible en: [fecha de consulta: 15 de enero de 2019]. [ Links ]

Excélsior (2017), “Duarte sabía que aplicaban agua por ‘quimios’ a niños con cáncer: Yunes”, 17 de enero, disponible en: [fecha de consulta: 15 de abril de 2019]. [ Links ]

Expansión (2016), “Los 12 gobernadores salientes dejan una herencia millonaria... en deuda”, 9 de mayo, secc. Política, disponible en: [fecha de consulta: 14 de junio de 2019]. [ Links ]

Ferejohn, John (1990), “Information and the Electoral Process”, en John Ferejohn y James Kuklinski (eds.), Information and Democratic Processes, Champaign: University of Illinois Press. [ Links ]

Fernández-Vázquez, Pablo, Pablo Barberá y Gonzalo Rivero (2016), “Rooting Out Corruption or Rooting for Corruption? The Heterogeneous Electoral Consequences of Scandals”, Political Science Research and Methods, 4(2), pp. 379-397. [ Links ]

Ferraz, Claudio y Frederico Finan (2005), “Exposing Corrupt Politicians: The Effect of Brazil’s Anticorruption Program on Electoral Outcomes”, manuscrito inédito, Berkeley: University of California. [ Links ]

Ferri, Pablo (2018), “Ex gobernadores corruptos ante el juez, la otra contienda de México”, El País, 31 d enero, secc. México, disponible en: . [fecha de consulta: 18 de abril de 2019]. [ Links ]

García, Dennis A. (2016), “Padrés: Tejedor de redes de corrupción”, 7 de diciembre, disponible en: [fecha de consulta: 15 de enero de 2019]. [ Links ]

Gómez, Eirinet (2018), “La corrupción de Javier Duarte no define ni describe a los priistas, asegura Meade”, La Jornada, 8 de enero, disponible en: [fecha de consulta: 15 de enero de 2019]. [ Links ]

Grzymala-Busse, A. (2007), “Encouraging Effective Democratic Competition”, East European Politics & Societies, 21(1), pp. 91-110. [ Links ]

Guardado Rodríguez, Jenny (2009), “La identidad partidista en México: Las dimensiones políticas de la competencia en las elecciones presidenciales de 2000 y 2006”, Política y Gobierno, volumen temático, pp. 137-175. [ Links ]

Huckfeldt, Robert y John Sprague (1990), “Social Order and Political Chaos: The Structural Setting of Political Information”, en John Ferejohn y James Kuklinski (eds.), Information and Democratic Processes, Champaign: University of Illinois Press. [ Links ]

Ibarra, Mariel y Silber Meza (2016), “Los piratas de Borge: El saqueo de bienes institucionalizado en Quintana Roo”, Expansión, 6 de julio, disponible en: [fecha de consulta: 15 de enero de 2019]. [ Links ]

Illades, Esteban (2015), La noche más triste: La desaparición de los 43 estudiantes de Ayotzinapa, Ciudad de México: Grijalbo. [ Links ]

INEGI (Instituto Nacional de Geografía y Estadística) (2015), Estadísticas Intercensales a escalas geoelectorales, Ciudad de México: INEGI, disponible en: [fecha de consulta: 1 de noviembre de 2018]. [ Links ]

INEGI (Instituto Nacional de Geografía y Estadística) (2017), Encuesta Nacional de Calidad e Impacto Gubernamental (ENCIG), Ciudad de México: INEGI, disponible en: [fecha de consulta: 1 de abril de 2019]. [ Links ]

Iyengar, Shanto (1990), “Shortcuts to Political Knowledge: The Role of Selective Attention and Accessibility”, en John Ferejohn y James Kuklinski (eds.), Information and Democratic Processes, Champaign: University of Illinois Press. [ Links ]

Kim, Myunghee y Mychal Voorhees (2011), “Government Effectiveness and Institutional Trust in Japan, South Korea, and China”, Asian Politics & Policy, 3(3), pp. 413-432, doi: [ Links ]

Klašnja, Marko y Joshua A. Tucker (2013), “The Economy, Corruption, and the Vote: Evidence from Experiments in Sweden and Moldova”, Electoral Studies, 32(3), pp. 536-543, doi: [ Links ]

Kumlin, Staffan y Peter Esaiasson (2012), “Scandal Fatigue? Scandal Elections and Satis- faction with Democracy in Western Europe, 1977-2007”, British Journal of Political Science, 42(2), pp. 263-282. [ Links ]

Langston, J. (2010), “Governors and Their Deputies: New Legislative Principals in Mexico”, Legislative Studies Quarterly, 35(2), pp. 235-258. [ Links ]

Lastra Guerrero, Gilberto (2016), “Niegan amparo a ex gobernador de Durango”, Milenio, 12 de enero, disponible en: [fecha de consulta: 17 de abril de 2019]. [ Links ]

Lechner, Michael (2011), “The Estimation of Causal Effects by Difference-in-Difference Methods”, Foundations and Trends in Econometrics, 4(3), pp. 165-224, doi: [ Links ]

Ley, Sandra (2018), “To Vote or Not to Vote: How Criminal Violence Shapes Electoral Participation”, Journal of Conflict Resolution, 62(9), pp. 1963-1990, doi: https://doi. org/10.1177/0022002717708600. [ Links ]

Linz, Juan J. (1985), “Democracy: Presidential or Parliamentary, Does It Make a Differen- ce?”, manuscrito inédito. [ Links ]

Lupu, Noam (2016), Party Brands in Crisis: Partisanship, Brand Dilution, and the Breakdown of Political Parties in Latin America, Nueva York: Cambridge University Press. [ Links ]

Maas, Cora J.M. y Joop J. Hox (2005), “Sufficient Sample Sizes for Multilevel Modeling”, Methodology, 1(3), pp. 86-92, doi: [ Links ]

Magaloni, B. (2006), Voting for Autocracy: Hegemonic Party Survival and Its Demise in Mexico, Nueva York: Cambridge University Press. [ Links ]

Mainwaring, Scott y Matthew S. Shugart (1997), “Juan Linz, Presidentialism, and Democracy: A Critical Appraisal”, Comparative Politics, 29(4), pp. 449-471, doi: https://doi. org/10.2307/422014. [ Links ]

Manin, Bernard, Adam Przeworski y Susan Stokes (1999a), “Elections and Representation”, en Adam Przeworski, Susan Stokes y Bernard Manin (eds.), Democracy, Accountability and Representation, Nueva York: Cambridge University Press, pp. 29-54. [ Links ]

Manin, Bernard, Adam Przeworski y Susan Stokes (1999b), “Introduction”, en Adam Przeworski, Susan Stokes y Bernard Manin (eds.), Democracy, Accountability and Representation, Nueva York: Cambridge University Press, pp. 1-26. [ Links ]

Maravall, José María (1999), “Accountability and Manipulation”, Democracy, Accountability and Representation, Nueva York: Cambridge University Press, pp. 154-196. [ Links ]

Merolla, Jennifer L., Laura B. Stephenson y Elizabeth J. Zechmeister (2007), “La aplicación de los métodos experimentales en el estudio de los atajos informativos en México”, Política y Gobierno, XIV(1), pp. 117-142. [ Links ]

Meza, Silber (2009), “‘Gobernadores, nuevos virreyes’: Macario Schettino”, Noroeste News- paper, 21 de noviembre, disponible en: [fecha de consulta: 8 de noviembre de 2015]. [ Links ]

Morgenstern, S. y E. Zechmeister (2001), “Better the Devil You Know than the Saint You Don’t? Risk Propensity and Vote Choice in Mexico”, The Journal of Politics, 63(1), pp. 93-119. [ Links ]

Peschard, Jacqueline (2017), “Javier Duarte: icono de la corrupción”, El Universal, 17 de abril, disponible en: [fecha de consulta: 18 de abril de 2019]. [ Links ]

Ramírez Rodríguez, Roberto y Alfredo Erquizio Espinal (2012), “Análisis del ciclo político electoral a partir de variables de gasto público por entidad federativa en México, 1993- 2009”, Paradigma Económico, 4(2), pp. 5-27. [ Links ]

Rennó, Lúcio Remuzat (2008), “Rewarding the Corrupt?: Reelection and Scandal Involvement in the Brazilian 2006 Legislative Elections”, Colombia Internacional, 68, pp. 98-106. [ Links ]

Samuels, David (1999), “Incentives to Cultivate a Party Vote in Candidate-Centric Electoral Systems: Evidence from Brazil”, Comparative Political Studies, 32(4), pp. 487-518, doi: [ Links ]

Schaffner, B.F. y M.J. Streb (2002), “The Partisan Heuristic in Low-Information Elections”, Public Opinion Quarterly, 66(4), pp. 559-581. [ Links ]

Scheiner, E. (2006), Democracy without Competition in Japan: Opposition Failure in a One- Party Dominant State, Nueva York: Cambridge University Press. [ Links ]

Tercer Grado (2012), “Enrique Peña Nieto-Parte 1”, 23 de mayo, Ciudad de México: Televisa, disponible en: [fecha de consulta: 30 de marzo de 2020]. [ Links ]

Torres Cofiño, Marcelo (2017), “El PRI y la corrupción”, Milenio, 16 de septiembre, disponible en: [fecha de consulta: 15 de enero de 2019]. [ Links ]

Trelles, Alejandro y Miguel Carreras (2012), “Bullets and Votes: Violence and Electoral Participation in Mexico”, Journal of Politics in Latin America, 4(2), pp. 89-123, doi: https:// [ Links ]

Velázquez Guadarrama, César (2006), “Determinantes del gasto estatal en México”, Gestión y Política Pública, XV(1), pp. 83-109. [ Links ]

Zuckermann, Leo (2003), “Los nuevos virreyes (elecciones 2003)”, Proceso, 13 de julio. [ Links ]

Zuckermann, Leo (2018), “Juegos de poder: La corrupción, el PRI y Meade”, Excélsior, 9 de enero, disponible en: [fecha de consulta: 15 de enero de 2019]. [ Links ]

1A paper studying the Spanish housing boom and the ensuing rise in scandals found that voters are willing to reward corruption if it is beneficial for them (Fernández-Vázquez et al., 2016).

2Since the opposition cannot gain government expertise, its quality decreases (Morgenstern and Zechmeister, 2001).

3For example, research on corruption scandals and corruption prosecution among mayors in Spain found that prosecuting had a much larger impact in vote choice than corruption (Costas-Pérez et al., 2012), and a study of Mexican legislators found that criminalizing governors affected co-partisan electoral performance across parties (Ang, 2019).

4Many of the recent gubernatorial corruption scandals involve local executives that acquired astronomical amounts of public debt. For example, a report on the 2016 elections pointed out that among the states with gubernatorial elections, the highest indebted states were Veracruz, Quintana Roo, and Chihuahua. The governors of these three states have since been prosecuted for corruption (Expansión, 2016). Existing analyses on local finances and political cycles agree that governors strategically plan public debt and expenditures with political purposes, although the extent and institutional determinants of these practices is still in question (Velázquez Guadarrama, 2006; Ramírez Rodríguez and Erquizio Espinal, 2012; Armesto, 2015).

5I am grateful to an anonymous Política y Gobierno reviewer who pointed out this distinction.

6 For example, in 2012 two of his collaborators were arrested while travelling on a state-owned air- craft and carrying suitcases full of cash (Animal Político, 2012).

7In January of 2017, Miguel Ángel Yunes, Veracruz governor after Duarte left power, denounced that Duarte’s administration had allowed the purchase of fraudulent medicine, intended to treat children with cancer (Excélsior, 2017).

8“Yes, the capture of Javier Duarte is a step, but it is only the first. The former officials and also the front men and accomplices who made the great possible robbery must appear before the courts” (Angel, 2017).

9For example, see García (2016). Some institutional characteristics suggest that prosecutions among PRI politicians would be more likely to trigger the mechanism set forth here than the PAN prosecutions. For example, whereas the PRI candidate is designated by its leadership, the PAN decides its candidate by holding internal elections.

10I choose to focus on presidential candidates only so as to minimize potential confounders when it comes to quality of candidates. Furthermore, as was discussed in the previous section, José Antonio Meade was clearly identified as a person suffering from the party’s corrupt reputation, but he was not accused of corruption himself.

11Although legally possible, no governor in modern Mexico has ever been prosecuted after being stripped from immunity through a political trial (juicio de desafuero).

12The identified governors are César Duarte (Chihuahua), Jesús Reyna (Michoacán), Rodrigo Medina (Nuevo León), Beto Borge (Quintana Roo, still at large), Andrés Granier (Tabasco), and Javier Duarte and Flavino Ríos (Veracruz).

13Since districts are perfectly nested within states, adding state’s specific intercepts would only aggregate average district votes per state. However, table A3 in the Appendix presents the same set of estimators but adding state-specific intercepts, and the substantive results remain unchanged.

14Almost all governors have, at one point or another, been accused of corruption. To tease out the “quality” of these accusations, I only coded as one the accusations that were either picked up by three or more national newspapers, reports of the governors being investigated for malfeasance, or accusations that were tied to specific governmental programs. In the 32 states, 15 had been governed by executives that had been involved in a scandal as defined here, including the 6 states that had governors who were formally charged. The variable Criminal, then, measures the additional effect of being involved in a judicial process when you have also been involved in a scandal.

15Although most of this literature has addressed the issue of violence and political participation and not necessarily vote choice (Ley, 2018; Trelles and Carreras, 2012).

16Note that the analysis of this implication does not require comparison across time, as is the case with the previous analysis. There is no reason to think that the decrease in votes should be more contingent upon internet coverage in 2018 than the decrease in votes in 2012.

17See Maas and Hox (2005) for sample sizes in multilevel modelling.

18See the Codebook for a list on parties of each election.

19See El Financiero (2017). In 2016, the PRI also expelled Tomás Yarrington, former governor of Tabasco, and suspended Roberto Borge, former governor of Quintana Roo.

20For example, Mario Marín, the Puebla governor who was accused of torturing a journalist as retaliation for an investigation involving a child abuse ring was never expelled from the PRI.



For every state, I identified the governors that were in power on June 30th of 2012, and left power before June 30th 2018.

  • vote_PRI: Proportion of votes cast only in favor in favor of the PRI, as a proportion of effective number of votes.

  • vote_PRI_allies: Proportion of votes cast in favor of the PRI and parties that endorsed the PRI candidate, as a proportion of effective number of votes. In 2012, these are votes for PRI and pvem. In 20128, these are votes for PRI, pvem and na.

  • criminal_PRI: 1 if at least one of the governors associated with that state was arrested and s/he was from the PRI

  • scandal_PRI: 1 if at least one of the governors associated with that state was from the PRI reported to be either investigated by corruption, or accused for corruption in the media with evidence or for a specific program of government.

  • y_2018: 1 if electoral year is 2018, 0 if 2012.

  • Internet Access: Proportion of households in district with access to Internet

  • Population (logged): Population per district, logged.

  • Economic growth: Average quarterly economic growth in the year before the election.

  • Homicides per 100 000: Homicide count in the six months PRIor to the elections (January to June). Source for 2012: INEGI, Source for 2018: Semáforo delictivo.

  • Telephone Access: Proportion of households in district with landlines

Summary statistics

Table A1 Descriptive statistics 

Min. 1stQ. Med. Mean 3rdQ. Max. Obs.
Vote PRI 0.04 0.13 0.20 0.22 0.30 0.55 600
Vote PRI+ 0.05 0.14 0.24 0.26 0.38 0.63 600
Y_2018 0.00 0.00 0.50 0.50 1.00 1.00 600
Criminal_PRI 0.00 0.00 0.00 0.21 0.00 1.00 600
Scandal_PRI 0.00 0.00 0.00 0.47 1.00 1.00 600
Internet Access 0.01 0.13 0.25 0.27 0.38 0.82 600
Population (log) 12.34 12.79 12.87 12.86 12.94 13.54 600
GDP growth -2.16 0.51 0.76 0.74 1.13 3.14 600
Homicides 0.91 5.42 7.59 10.55 13.02 43.02 600
Phone Access 0.02 0.25 0.38 0.40 0.54 0.87 600

Source: Own elaboration.

Alternative estimations

Table A2 Votes for PRI (additional controls) 

Model 1 Model 2 Model 3 Model 4
Criminal_PRI -0.039*** -0.040*** -0.039*** -0.047***
(0.009) (0.009) (0.009) (0.008)
Scandal_PRI 0.053*** 0.055*** 0.055*** 0.063***
(0.007) (0.007) (0.007) (0.007)
Y_2018 -0.140*** -0.140*** -0.140*** -0.156***
(0.005) (0.005) (0.005) (0.005)
Criminal_PRI*Y_2018 -0.026*** -0.027*** -0.027*** -0.024***
(0.010) (0.010) (0.010) (0.009)
Scandal_PRI*Y_2018 -0.001 -0.002 -0.002 -0.003
(0.008) (0.008) (0.008) (0.008)
Access to Internet -0.145*** -0.144*** -0.144*** 0.005
(0.015) (0.015) (0.015) (0.024)
Economic growth -0.006* -0.006* -0.005*
(0.003) (0.003) (0.003)
Homicides per 100 000 0.0003 0.0003 0.0003
(0.0003) (0.0003) (0.0002)
Past_Legislative_Vote_PRI 0.016
Proportion_Urban_Sections -0.098***
Constant 0.313*** 0.314*** 0.313*** 0.343***
(0.005) (0.006) (0.006) (0.007)
District intercepts Y Y Y Y
Observations 586 586 586 586

Source: Own elaboration. *p < 0.1; **p < 0.05; ***p < 0.01.

Table A3 Votes for PRI presidential candidate (state intercepts) 

Model 1 Model 2 Model 3 Model 4
Criminal_PRI -0.011 -0.046** -0.007 -0.045**
(0.019) (0.019) (0.019) (0.020)
Scandal_PRI 0.052*** 0.059***
(0.015) (0.016)
Y_2018 -0.157*** -0.156*** -0.154*** -0.150***
(0.004) (0.005) (0.005) (0.006)
Criminal_PRI*Y_2018 -0.027***
Scandal_PRI*Y_2018 -0.004 -0.012
(0.009) (0.010)
Population (logged) 0.007 0.007 0.007 0.006
(0.015) (0.015) (0.015) (0.015)
Economic growth -0.006 -0.008*
(0.004) (0.004)
Homicides per 100 000 -0.001 -0.001*
(0.0004) (0.0004)
Constant 0.213 0.197 0.225 0.217
(0.194) (0.193) (0.194) (0.193)
State intercepts Y Y Y Y
Observations 600 600 600 600

Source: Own elaboration. *p < 0.1; **p < 0.05; ***p < 0.01.

Table A4 Votes for PRI presidential candidate (without Veracruz) 

Model 1 Model 2 Model 3 Model 4
Criminal_PRI 0.012 -0.023** 0.016 -0.023**
(0.010) (0.011) (0.010) (0.011)
Scandal_PRI 0.052*** 0.057***
(0.008) (0.008)
Y_2018 -0.159*** -0.158*** -0.158*** -0.155***
(0.004) (0.005) (0.004) (0.005)
Criminal_PRI*Y_2018 -0.034*** -0.032*** -0.040*** -0.035***
(0.009) (0.010) (0.009) (0.010)
Scandal_PRI*Y_2018 -0.003 -0.008
(0.008) (0.008)
Population (logged) 0.047*** 0.040** 0.047*** 0.042***
(0.016) (0.016) (0.016) (0.016)
Economic growth -0.006*
Homicides per 100 000 -0.0003 -0.0002
(0.0003) (0.0003)
Constant -0.299 -0.237 -0.303 -0.245
(0.211) (0.206) (0.211) (0.205)
District intercepts Y Y Y Y
Observations 559 559 559 559

Source: Own elaboration. *p < 0.1; **p < 0.05; ***p < 0.01.

Table A5 Votes for PRI presidential candidate (PRI governors only) 

Model 1 Model 2 Model 3 Model 4
Criminal_PRI -0.041*** -0.046*** -0.047*** -0.050***
(0.010) (0.010) (0.010) (0.010)
Scandal_PRI 0.020 0.013
(0.013) (0.013)
Y_2018 -0.161***
Criminal_PRI*Y_2018 -0.018* -0.018* -0.012 -0.014
(0.009) (0.009) (0.009) (0.010)
Scandal_PRI*Y_2018 0.003 0.006
(0.012) (0.012)
Population (logged) 0.030 0.031 0.031 0.032
(0.019) (0.019) (0.019) (0.019)
Economic growth 0.013*** 0.012**
(0.005) (0.005)
Homicides per 100 000 0.0004 0.0003
(0.0004) (0.0004)
Constant -0.050 -0.085 -0.081 -0.099
(0.250) (0.251) (0.250) (0.250)
District intercepts Y Y Y Y
Observations 303 303 303 303

Source: Own elaboration. *p < 0.1; **p < 0.05; ***p < 0.01.

Received: June 15, 2019; Accepted: April 03, 2020

Special thanks to Joy Langston and the attendants of the Política y Gobierno workshop held at CIDE in Mexico City, April 2019, for their thoughtful comments on earlier empirical results. All errors remain my own.

Creative Commons License Este es un artículo publicado en acceso abierto bajo una licencia Creative Commons