This project evaluates the extent to which the gender of a survey respondent impacts their likeliness to report corruption in the civil service
Findings
Our first hypothesis, regarding the gender of the respondent, was confirmed: the negative coefficient of gender is statistically significant which indicates that women in the public service are more likely to report corruption than men. These results stand in contrast to many of the results of previous literature on whistleblowing, suggesting either that as social ideals and gender roles evolve, so does women’s ability and willingness to report corruption, or that in accordance with the findings of scholars who study business ethics, women report more ethical responses regarding corruption than men. Today, women occupy more managerial positions and operate under different societal expectations, and our results offer confirmation that this could make them more likely to report corruption. This is particularly true with the agencyCulture variable. Of all of the indicators we evaluated, women’s scores varied the most from men’s when asked directly about corruption. This is strong evidence in support of our hypothesis, as this question is the least likely to be misinterpreted and therefore the most robust indicator of corruption. These results do get slightly more nuanced when looking at the question regarding the importance of political connections to obtain a salary increase. Here, women on average actually report that political connections are less important than men do. This could indicate that women tend to have less important political connections and rely more on things like merit and tenure to advance in their careers than men.
Additionally, given that corruption tends to be under-reported, and women seem to report higher levels of corruption, one could logically conclude that women’s assessment of corruption is a more accurate indicator of the level of corruption within a given country. However, given that there is no concrete indicator of how much corruption exists in each country, there is no way to prove this claim. Some authors suggest that having a high perception of corruption in turn causes perceptions of corruption to continue to increase (Grindle 2012). Essentially, the more that some people talk about how corrupt the government is, the more people will begin to also perceive similar levels of corruption. It is important to keep in mind that this survey was administered in 2018, only a couple of years after the biggest corruption scandal in Brazilian history was discovered. This means that in 2018, people could have been hyper-aware and more eager to report corruption. Therefore, we cannot empirically make the claim that women are better indicators of the level of corruption because it is possible that in this case, they over-reported corruption.
Another factor that is worth investigating in the future is whether women are generally just more critical than men. It could be that women’s scores on corruption are indicative of a larger trend that women generally score the bureaucracy lower than men. In order to investigate this, one could compare the average responses of women and men across many different thematic indicators, such as autonomy and capacity, and not only corruption. Our results do indicate a strong relationship, but in the future, we hope to add more complexity to our conclusions by exploring this question.
Our other independent variables also produced some noteworthy results. The relationships observed with tenure largely disprove our initial hypothesis that respondents with longer tenure are more likely to report corruption. In the three models in which tenure has significant results, the relationship is the opposite result of the gender variable, meaning that it appears that the longer someone works in the Brazilian bureaucracy the less likely they are to report corruption. While none of these relationships are particularly large, they also aren’t small enough to be dismissed. One possible explanation for this relationship could be that the longer someone works in the system, the more likely it is for them to be co-opted into it and to benefit from corruption themselves, making them less likely to report. In two of the models, the education variable suggests that those with more education are more likely to report corruption despite our initial hypothesis that education would not affect a meritocracy such as Brazil. The other two models, including the one that directly mentions corruption, do not produce significant results. While not wanting to dismiss the other results, more research needs to be done before drawing conclusive inferences about education’s relationship with corruption. Finally, while age produced some statistically significant results, the results are mixed and of minimal size making any conclusions impossible to make at this time.
Regression Models


Eliza Patterson
Major: Political Science & Hispanic Studies
Hometown: Los Altos, CA
Class: 2022
Project Focus: Gender & Corruption in the Brazilian bureaucracy

Carter Cook
Major: Political Science
Hometown: Jacksonville, FL
Class Year: 2020
Project Focus: Gender & Corruption in the Brazilian bureaucracy