Politicians, Publicly-Released Audits of Corruption, and Electoral Outcomes in Brazil
Researchers evaluated how exposing information about corrupt politicians impacted election results in Brazil. Taking advantage of a federal anticorruption program that randomly assigned municipalities to be audited, the researchers compared the electoral outcomes for mayors in two randomly selected groups of municipalities: those that were audited before and those that were audited after the 2004 election. This provided an opportunity to observe whether voters’ access to information prior to the election about politicians’ corruption levels affected vote share and reelection rates for incumbent mayors. Publicly released corruption audits reduced reelection rates of corrupt incumbent municipal mayors.
In a well-functioning democracy, citizens hold politicians accountable for their actions by choosing to reelect candidates who have best represented their interests in the past. However, when citizens lack information about politicians’ performance, they cannot monitor them and hold them accountable for their actions. Efforts to audit politicians and publicly release this information may allow citizens to identify corrupt politicians and punish them at the polls, which could force them to act in the best public interest. However, it is not clear whether making information available causes electoral accountability, or whether additional related factors (such as the level of political competition or voter efforts to collect information) are the true cause.
In 2003, the Government of Brazil initiated a pioneering anticorruption program where municipal governments were randomly selected to be audited by the Controladoria-Geral da União (CGU), a federal agency responsible for overseeing public resources. The goal of the program was to reduce the misuse of public funds and to enhance civic participation in monitoring public expenditures.
From a sample of the municipalities in Brazil with less than 450,000 inhabitants, which represent about 73 percent of the total population, the program randomly selected a group of municipalities to be audited. The random selection of municipalities was held on a monthly basis and performed in conjunction with the national lotteries. To guarantee the fairness of the process, representatives of the press, political parties, and members of the public were invited to witness the lotteries.
By July 2005, a total of 373 municipalities with a mayor eligible for reelection had been audited. For 205 municipalities, the results of the audits were publicly released prior to the 2004 election, making the information potentially capable of influencing voters’ decisions. The remaining 168 municipalities were audited after the election and thus formed the comparison group.
Once a municipality was randomly chosen, the audit program consisted of the following:
- The CGU collected information on all federal funds transferred to the municipal government from 2001 to 2003.
- A group of 10–15 CGU auditors was sent to the municipality to examine accounts and documents and to inspect the existence and the quality of the goods and services provided by the municipality. Auditors also met with local community members and municipal councils to get direct complaints about any wrongdoing.
- After approximately 10 days of inspection, a detailed report was submitted to the CGU, a public prosecutor, and the municipal legislative branch.
For each municipality audited, a summary of the main findings was posted on the internet and disclosed to major media sources. Although the researchers do not have direct evidence showing that voters learned about the audit reports, anecdotal evidence documented by the local media suggests that information from the audits not only reached voters, but was also used widely during the election.
For the purposes of this evaluation, the researchers noted whether a local radio station was present in each municipality. About a quarter of municipalities have at least one station, and 79 percent of households own a radio.
To differentiate corruption levels, the researchers counted the number of violations associated with fraud in procurements, diversion of public funds, or over-invoicing found during each mayor’s tenure.
The audit policy and the release of information prior to the election decreased the share of votes and the probability of reelection for corrupt incumbent mayors.
Among municipalities where two corruption violations were reported, the audit policy reduced the likelihood of reelection by 7 percentage points (or 17 percent). This is in comparison to mayors with the same number of violations in the comparison group of municipalities, where information from the audits was released after the election. Additionally, incumbents with reported violations who were reelected had a smaller margin of victory.
Informed voters more severely punished higher levels of corruption.
The impact of the audit policy increased with the number of violations reported. In municipalities audited before the election, every additional corruption violation reported reduced the likelihood of an incumbent’s reelection by 7 percentage points.
Would voters have punished these politicians even in the absence of the audits?
In comparison group municipalities, where audits had not been conducted and released before the elections, reelection rates remained steady across corruption levels. This result suggests that voters dislike corruption, but lacked the information they needed to punish corrupt politicians.
Local media enhanced political accountability by disseminating information to voters.
Given the low level of newspaper circulation in Brazil, radio stations usually play an important role in politics. In cases where three violations were reported, incumbents in municipalities with a local radio station were 16 percentage points less likely to be reelected. Incumbents with the same level of corruption in municipalities without a local station were only 3.7 percentage points less likely to be reelected.
While the availability of radio stations was not a part of the randomized program implementation, local radio stations seem to be an important channel through which the policy disseminated information, influenced voters, and affected the electoral performance of corrupt politicians.
Information channeled through local media favored noncorrupt incumbents.
While the presence of the local media magnifies the negative effect of the policy on the electoral performance of corrupt incumbents, it also helps to promote noncorrupt incumbents. In cases where no corruption was found, the policy increased the likelihood of reelection by 17 percentage points in municipalities where a local station was present.
Voters care about corruption but may not have the information necessary to take action.
In the absence of the audits, the consistent reelection rates across corruption levels indicate that voters lacked information about their mayor’s level of corruption.
In response to credible information, voters update their beliefs about politicians and change their voting behavior.
Voters punished incumbents with corruption levels higher than the median corruption level of one violation. They also rewarded incumbents when no cases of corruption were found.
Local media can enhance political accountability and candidate selection.
By widely disseminating the results of the audits, local radio stations increased the effect of the audit policy. Voters in municipalities with local stations more severely punished corrupt politicians (and more strongly rewarded honest ones) than in areas lacking a local station.
A higher level of accountability is achieved when relevant information is both produced and disseminated.
A policy that produces reliable information on politicians’ performance, such as anticorruption audits, combined with the public release of this information through the right channels is an effective way to hold politicians accountable.
Ferraz, Claudio and Frederico Finan. 2008. "Exposing Corrupt Politicians: The Effects of Brazil's Publicly Released Audits on Electoral Outcomes." The Quarterly Journal of Economics. 123 (2): 703-745