Corruption undermines public trust, distorts how governments spend, and weakens state capacity. As women have gained political ground around the world, a hopeful idea has taken hold: that electing more women might also mean cleaner government. But is it true?

This belief rests on a substantial body of correlational evidence: across and within countries, greater female representation has been associated with lower corruption (Dollar et al. 2001, Swamy et al. 2001, Jha and Sarangi 2018, Bauhr et al. 2019, Decarolis et al. 2023). These correlations are, however, hard to interpret causally: the women who reach office may differ systematically from male politicians, for instance in their political orientation or experience, and the places that elect them may differ too. More broadly, the evidence on whether female politicians govern differently is mixed (Hessami and Lopes da Fonseca 2020): electing more women does not necessarily translate into different policies (Campa and Bagues 2017, Carrer and De Masi 2026).

In a new paper (Peveri and Tricaud 2026), we revisit the corruption question using thousands of close mixed-gender elections and a wide range of corruption measures. Our answer is a clear null: across two decades of Brazilian municipal elections and six different measures of corruption, electing a woman mayor has no detectable effect. And we can explain why earlier work found otherwise.

Measuring the causal effect of female leadership on corruption

Brazil offers a great setting to study this question. Its 5,570 municipalities are autonomous entities responsible for delivering local public services, and mayors hold broad discretion over spending. Corruption in local governments is widespread, with losses estimated at around $550 million per year (Ferraz and Finan 2011), and Brazil’s institutional efforts to detect and sanction it have generated unusually rich data to measure it. Women remain heavily underrepresented in politics: in 2000, the first election in our sample, only 5.6% of elected mayors were women, a share that has plateaued at around 10% since 2012.

To isolate the causal effect of female leadership, we exploit close mixed-gender races: municipal elections where a coin flip determines victory between a female and a male candidate. We compare municipalities where a woman barely beat a man to those where a man barely beat a woman – thousands of close races across the 2000, 2004, 2008, 2012, and 2016 elections. Reassuringly, these close races compare like with like: municipalities that narrowly elect a woman or a man have similar socio-demographic characteristics, and the narrowly elected men and women are similar in political experience and orientation.

We then assemble six corruption measures from three independent sources: predicted corruption scores that are built from municipal budget accounts (extending the machine-learning approach of Ash et al. 2025) and are available for every municipality; irregularities recorded by the federal government’s randomised anti-corruption audits; and newly collected data on legal sanctions.

No effect of female leadership on corruption

The headline result is a precisely estimated zero. Having a female mayor has no significant effect on any of the six outcomes – not on the predicted corruption score, the number of audit irregularities, the probability of legal sanction, suspension from political rights, the financial damage to be reimbursed, or the fines imposed. For our preferred measure, the budget-based score that covers the full sample, we can rule out reductions larger than about 3% and increases larger than about 5%.

This null result holds regardless of mayors’ education, political experience, ideology, or age, and it holds at every point in the four-year electoral cycle. In short, closely elected female and male mayors in Brazil look equally (un)corrupt.

Reconciling with earlier findings: It’s incumbency, not gender

How, then, do we square this with the previous finding that female mayors reduce corruption in Brazil? That result comes from Brollo and Troiano (2016), who also used a close-election design in Brazil but focused on a much smaller subsample: municipalities randomly audited in the two earliest terms (2000 and 2004). When we zoom in on exactly that setting, we too find that female mayors look less corrupt. But this is the only one of our many samples where such an effect appears, and it vanishes once later elections or a broader set of municipalities are added.

Investigating what makes that subsample special, we uncover a confounder: incumbency. In those early-audited municipalities, and only there, women who narrowly won were substantially less likely to be incumbents than the men who narrowly won. That matters because incumbency has a direct effect on corruption. Incumbents have more experience and networks to exploit, and Brazil’s two-term limit means second-term mayors cannot run again, weakening electoral accountability (Ferraz and Finan 2011).

Using a separate close-election analysis comparing male incumbents to male non-incumbents, we indeed show that incumbents are more corrupt across all our measures. Within the Brollo and Troiano setting, second-term mayors are roughly 50% more likely to be corrupt. Moreover, once incumbency is accounted for, the apparent gender effect in early-audited municipalities dissolves: restricting the sample to first-term mayors, or simply controlling for winner characteristics, roughly halves the estimate and renders it statistically insignificant.

A caution for studies of close elections

Beyond gender and corruption, our findings carry a methodological lesson. Comparing narrowly winning and losing candidates is a powerful tool, but it assumes that the candidate’s trait of interest – here, gender – is effectively the only thing that changes. When other characteristics that also affect the outcome change discontinuously at the same point, estimates can conflate the trait of interest with those confounders (Marshall 2024) – here, incumbency.

Conclusion

Electing women to local office in Brazil neither raises nor lowers corruption. This null result should not be read as weakening the case for women’s political representation. Whether or not women govern differently, equity in representation remains a fundamental democratic objective in its own right – and the case for it is only reinforced by how far we still are from achieving it. Women continue to face substantial barriers and discrimination on the path to office (Bertrand and Duflo 2017, Eyméoud and Vertier 2020, Le Barbanchon and Sauvagnat 2022, Fujiwara et al. 2025). The case for bringing more women into politics rests on fairness and on dismantling these persistent barriers – not on the expectation that women will deliver particular policy outcomes such as cleaner government.

References

Ash, E, S Galletta, T Giommoni, and G Pino (2025), “CSI corruption: A machine-learning approach to predicting public-budget fraud”, working paper.

Bauhr, M, N Charron, and L Wängnerud (2019), “Exclusion or interests? Why females in elected office reduce petty and grand corruption”, European Journal of Political Research 58(4): 1043–65.

Bertrand, M, and E Duflo (2017), “Field experiments on discrimination”, in Handbook of Economic Field Experiments, Vol. 1, North-Holland.

Brollo, F, and U Troiano (2016), “What happens when a woman wins an election? Evidence from close races in Brazil”, Journal of Development Economics 122: 28–45.

Campa, P, and M Bagues (2017), “Electoral gender quotas fail to empower women”, VoxEU.org, 9 September.

Carrer, L, and L De Masi (2026), “More women in politics does not always mean more gender equality”, VoxEU.org, 10 April.

Decarolis, F, R Fisman, P Pinotti, S Vannutelli, and Y Wang (2023), “Gender and bureaucratic corruption: Evidence from two countries”, Journal of Law, Economics, and Organization 39(2): 557–85.

Dollar, D, R Fisman, and R Gatti (2001), “Are women really the ‘fairer’ sex? Corruption and women in government”, Journal of Economic Behavior and Organization 46(4): 423–9.

Eyméoud, J-B, and P Vertier (2020), “Gender discrimination in politics: Evidence from a natural experiment in French local elections”, VoxEU.org, 22 May.

Ferraz, C, and F Finan (2011), “Electoral accountability and corruption: Evidence from the audits of local governments”, American Economic Review 101(4): 1274–311.

Fujiwara, T, H Hilbig, and P Raffler (2025), “Biased party nominations as a source of women’s electoral underperformance”, NBER Working Paper 34396.

Hessami, Z, and M Lopes da Fonseca (2020), “Female political representation and substantive effects on policies: A literature review”, European Journal of Political Economy 63: 101896.

Jha, C K, and S Sarangi (2018), “Women and corruption: What positions must they hold to make a difference?”, Journal of Economic Behavior and Organization 151: 219–33.

Le Barbanchon, T, and J Sauvagnat (2022), “Electoral competition, voter bias, and women in politics”, Journal of the European Economic Association 20(1): 352–94.

Marshall, J (2024), “Can close election regression discontinuity designs identify effects of winning politician characteristics?”, American Journal of Political Science, forthcoming.

Peveri, J, and C Tricaud (2026), “Do female leaders reduce corruption? New evidence from Brazil”, working paper.

Swamy, A, S Knack, Y Lee, and O Azfar (2001), “Gender and corruption”, Journal of Development Economics 64(1): 25–55.