Brief 54: The Effect of Incumbent Malfeasance Revelations: Evidence from Mexico

The authors develop a simple Bayesian model, which predicts that voters will electorally punish (reward) the incumbent party when they receive information that the incumbent is engaged in significantly more (less) malfeasance than expected. The model further predicts that substantial updating in either direction will increase turnout. These predictions are supported by evidence from a field experiment in Mexico, in which leaflets that provided information about incumbent misuse of funds prior to the municipal elections of June 2015 were delivered to voters. On average, voters had sufficiently unfavorable priors that the intervention – which often reported substantial levels of malfeasance – did not significantly alter beliefs about incumbent malfeasance. The incumbent party’s vote share increase by three points, although the increase was greatest where the incumbent engaged in the least malfeasance and voters updated their beliefs most favorably. As predicted, turnout increased most when voters learned of high or low malfeasance levels.

Read Full Study

Category: Elections

Tags: Mexico, voter behavior, information, malfeasance

Date of Publication: Wednesday, August 1, 2018

EGAP Researcher: Eric Arias, Horacio Larreguy , John Marshall, Pablo Querubin

Geographical Region: North America

Research Question:

Under what conditions do voters punish or reward parties upon receiving information about malfeasance of incumbents?

Preparer: Tanu Kumar



This study was conducted in the context of the 2015 municipal elections in Mexico, which were generally characterized by competition between the centrist Institutional Revolutionary Party (PRI) and either the right-wing National Action Party (PAN) or the left-wing Party of the Democratic Revolution (PRD). Mexican municipalities are responsible for roughly 20% of government spending and are governed by mayors elected to three-year terms that were, at the time of the study, non-renewable. Roughly one-quarter of the municipal budget comes from the FISM program, which consists of direct federal transfers mandated for infrastructure projects benefiting those living in marginalized localities. 

In 1999, the Federal Auditor’s office (ASF) began conducting independent audits of the use of FISM funds. The results of these audits note: (1) the share of funds spent on projects that do not directly benefit the poor; and (2) the share of funds spent on unauthorized projects including those diverted for personal expenses and electoral campaigns. In the study’s sample, these shares ranged from 0% to 58%. According to surveys of untreated households, the average voter in the study believed that the incumbent party engaged in medium to very high levels of such forms of misallocated spending.

Figure 1. Example of local information leaflet in Ecatepec de Morelos, México (where 45% of the
146.3m pesos received were spent on unauthorized projects)

Research Design:

The study was conducted in 26 municipalities in the states of Guanajuato, México, San Luis Potosí, and Querétaro. The municipalities and states were chosen based on holding municipal elections in 2015, municipalities receiving ASF audit results in 2015, security and logistical considerations, and ensuring that incumbents from different parties were proportionately represented within these states. Within each municipality, up to one-third of the electoral precincts were included in the study, with a focus on accessible rural precincts and those with fewer registered voters. In total, 678 precincts were included in the sample. Precincts were then stratified to create blocks of six or seven similar precincts within municipalities. Within these blocks, the authors randomly assigned whether a precinct received leaflets detailing either non-poor or unauthorized spending in the municipality (Figure 1). (Several variants, which did not significantly affect voting behavior, are explored in other papers.) Leaflets were delivered directly to up to 200 households within treated precincts on behalf of the non-partisan Mexican partnering NGO. One set of outcomes, namely precinct-level turnout and incumbent party vote share, was measured using official electoral returns. Voter beliefs and uncertainty about incumbent malfeasance were elicited through a post-election survey of 10 voters from each treated precinct and 10 voters from a randomly selected control precinct within each block. Voters’ prior beliefs in a given municipality are inferred from the post-election beliefs of respondents in control precincts.

This study was part of EGAP’s Metaketa Initiative (, in which teams of researchers run coordinated evaluations of interventions across contexts that aim to activate similar causal mechanisms. This Metaketa round focused on Information & Accountability ( by implementing a series of experimental projects that assessed the role of information in promoting political accountability in developing countries.


On average, the intervention did not affect voters’ beliefs about incumbent party misuse and misallocation of funds, indicating that voters possessed relatively pessimistic but accurate prior beliefs about incumbent party malfeasance. However, the provision of information actually increased the incumbent party’s vote share by 3 percentage points, most likely because information reduced voter uncertainty about incumbent party malfeasance and incumbent parties responded more effectively to the intervention. Moreover, the delivery of unsurprising information – moderate levels of malfeasance – depressed turnout by about 1 percentage point, whereas the delivery of surprising information – extremely low or high levels of malfeasance – increased turnout by about 1 percentage point. Finally, in treated precincts, incumbent and challenger party organizations respectively often attempted to discredit or incorporate the information revealed into their campaigns, particularly when reported levels of malfeasance were high. They did not, however, condition their behavior on voters’ prior beliefs or the extent of voters’ belief updating.

Policy Implications:

Overall, the study finds that voters’ responses to information interventions depend on how the information delivered relates to their prior beliefs. In particular, where voters possess unfavorable or imprecise beliefs about incumbent performance, interventions may actually reveal the incumbent to be less corrupt than originally believed. These findings can help explain the mixed evidence from prior studies on the effects of information on electoral sanctioning and political participation in developing democracies. They also reinterpret studies that have found that negative campaigning and reporting of corruption causes citizens to disengage from the political system by highlighting a nonlinear relationship between information content and turnout. Finally, the findings suggest that an important way to increase sanctioning behavior is to increase voter expectations of politicians, perhaps through civic education, a more critical media, or by inducing higher quality candidates to run for office.