Principal investigator: Adriano Valladão
Article title: WHAT HAPPENS WHEN A WOMAN WINS AN ELECTION? EVIDENCE FROM CLOSE RACES IN BRAZIL
Article authors: Fernanda Brollo and Ugo Troiano
Location of the intervention: Brazil
Sample size: 161 municipalities
Sector: Gender
Type of interventionThe effect of gender on corruption.
Main variable of interestCorruption
Evaluation method: Discontinuous Regression
Policy Problem
There is a view that the gender of a politician is important in the allocation of public spending, linking women to choices more focused on social issues such as health, childcare, and education. Similarly, there is also an association between gender and corruption, where female leaders are associated with less corruption and bribery compared to men. It is possible to study the effect of a politician's gender on corruption in Brazilian municipalities.
Evaluation Context
Using information about electoral disputes, administrative irregularities, and patronage, it is possible to analyze whether the chance of being involved in corruption cases is higher or lower for women. The focus will be on the municipal administration of mayors, highlighting that they are directly elected by a majority vote for 4-year terms with the possibility of one reelection. It is also important to note that elections in municipalities with fewer than 200 inhabitants have only a single round and may have a plurality of candidates, a potential difficulty in comparing the effect of gender. Furthermore, corruption is objectively measured by irregularities committed by the municipal government; that is, it does not depend on subjective indices such as the perception of corruption.
Intervention Details
Brazil, during the research period, had 5567 municipalities, and the focus was on places where the 2000 and 2004 elections were marked by a contest between candidates of the opposite gender. Given the problem of party plurality, only municipalities with two candidates were selected, totaling 723 electoral contests or 7% of the total. Then, some characteristics of the elected mayor were collected, such as gender, education level, party affiliation, and political experience.
The information on corruption comes from audits of municipal governments randomly selected by the federal government since 2003 as part of an anti-corruption program. The program consists of sending auditors to municipalities to analyze how local administrations are spending money transferred by the federal government since 2001. The audit reports allow for the classification of irregularities into: (i) illegal procurement processes (such as limiting competition, manipulating bids, bids won by irregular firms, among others); (ii) fraud; (iii) overpricing; and (iv) embezzlement. After filtering, corruption data was obtained for 161 municipalities.
Public positions distributed throughout a mayor's term can be a means of gaining political support. Thus, patronage can be measured by the temporary jobs offered by the municipal public administration, including appointed positions that do not clearly define the criteria and requirements for the position. Temporary positions contrast with permanent and stable jobs obtained through competitive public examinations. Data on patronage is only available for the term lasting between 2005 and 2008 and totals 381 municipalities.
Methodology Details
Measuring how a politician's gender affects potential corruption cases is not simple. First, this effect would be influenced by unobservable characteristics of the municipality, such as its stance on acceptance of women, social preferences, and demographic characteristics. Second, it would be necessary to compare the administrations of a female mayor with those of a male mayor under the same conditions to attribute the effect to gender, and in practice, only one of them is observed. In other words, it is not possible to know exactly whether certain decisions would be adopted or not if the mayor's gender changed; therefore, some hypothesis is necessary to compare the data and reach a conclusion.
It is assumed, then, that cities where women won the election by a small margin when running against a man are a good counterpoint to cities where the opposite happened (men winning against women by a small margin). This argument is justified since, in close races, the probability of victory would be the same for men and women, and elections would be decided by other random factors of the voters and not by the candidate's gender. Therefore, the treatment group can be defined as the municipalities where a woman was elected in a contest against someone of the opposite sex by a small margin, and the control group as those where a man was elected under the same circumstances. Finally, it is highlighted again that the effect of the politician's gender on corruption is obtained when the margin of victory in the election was small.
Results
Using electoral data, corruption measurements, and the comparison logic proposed above, it is possible to compare the probabilities of corruption between men and women and attribute this value strictly to gender. The results show that the probability of women being involved in corruption cases is between 29 and 35 percentage points lower than that of men. The calculation is credible since the audits carried out by the federal government are random and do not depend on the gender of the mayor; that is, the difference in probability above is not due to a bias on the part of the federal government in potentially choosing more cities with male mayors than with female mayors.
Another way to measure corruption would be through temporary positions. It was found that female mayors hire 52% of the temporary jobs in mayoral administrations, on average over the 4-year term. Furthermore, the proportion of temporary to tenured positions is 9% lower in municipalities with a female mayor than with a male mayor. The dynamics of appointments to temporary positions in the last year of a term were also investigated, as this could more clearly indicate a trade-off to secure political support in the following election. While temporary hiring is similar in the first year of a term between municipalities led by men and women, in the last year female mayors hire 10 to 13 percentage points fewer temporary positions than male mayors.
Lessons in Public Policy
Just as the gender of politicians is associated with the allocation decisions of their administration, the same is true for corruption. Analyzing the effects of the gender of mayors in Brazilian municipalities shows that women are less likely to be involved in corruption cases. Furthermore, another dimension that signals misconduct occurs through the hiring of temporary positions, happening on a larger scale for men in the last year of their term compared to women.
Reference
Brollo, Fernanda; Troiano, Ugo. “What happens when a woman wins an election? Evidence from close races in Brazil.” Journal of Development Economics, vol. 122, p. 28-45, 2016.
