The Gendered Impact of Corruption Revelations: Unveiling the Role of Parties and Voters in Mexico

Abstract

Studies have found that women’s representation is more likely to spike after corruption scandals. However, the mechanism underlying this increase remains unclear: are parties more likely to nominate women after corruption scandals, are voters more likely to support women candidates, or is it a combination of both? Using an original dataset of audit results and the gender of 47,000 candidates running in over 10,000 mayoral elections in Mexico (2000–2019), we find that voters drive the effect. While political parties are not more likely to nominate women as candidates in municipalities with recent revelations of spending irregularities, women candidates are more likely to win elections after corruption is uncovered. In contrast to previous studies, which expect strategic parties to be behind the increases in women’s representation following corruption scandals, our findings underscore that increases in women’s representation can happen despite parties and not because of parties.

Recent research has found that revelations of corruption can be followed by increases in women’s representation (Diaz and Piazza 2022; Reyes-Housholder and Thomas 2018; Valdini 2019). This relationship is more apparent in high-accountability systems (Armstrong et al. 2022; Esarey and Schwindt-Bayer 2018) and in countries, such as Mexico, that have increasing demands for women’s inclusion and strong stereotypical beliefs about women being less corrupt than men (Guajardo and Schwindt-Bayer 2024b). However, it is not clear empirically who is responsible for the increase in women’s representation after corruption revelations. Are parties using women’s representation strategically to clean their image in the eyes of voters (Armstrong et al. 2022; Valdini 2019)? Do voters prefer women in post-corruption contexts (Barnes and Beaulieu 2014; 2019; Benstead, Jamal, and Lust 2015; Le Foulon and Reyes-Housholder 2021)? Or is it both?

In this study, we answer these questions by leveraging an original dataset of audit results and the gender of 47,000 candidates running in over 10,000 mayoral elections in Mexico (2000–2019). We test whether parties are more likely to nominate women candidates (i.e., the party mechanism) and whether women candidates are more likely to win the election (i.e., the voter mechanism) in municipalities where wrongdoing was recently made public by audits. We find evidence in support of the voter mechanism. While political parties are not more likely to nominate women as candidates in municipalities with recent revelations of spending irregularities, the probability of a woman winning the election in those municipalities increases by 4.4 percentage points—a 65% change. Whereas most research expects strategic parties to be behind the election of women in post-scandal environments, our findings underscore that in some contexts, women can benefit electorally after corruption revelations despite parties, not because of parties.

THE STRATEGIC VALUE OF WOMEN AFTER CORRUPTION SCANDALS

Some studies have found that women are less likely to enter the political arena when corruption is present (Bjarnegard and Zetterberg 2017; Stockemer and Sundström 2019; Sundström and Wängnerud 2014). Others show that women’s representation can spike after corruption scandals. Valdini (2019) found that the percentage of women legislators was more likely to increase in post-scandal environments, and Guajardo and Schwindt-Bayer (2024b) found that, in Mexico, women are more likely to win mayoral offices after audit reports publicly reveal wrongdoing. Armstrong et al. (2022) linked the appointment of female finance ministers to spikes in perceptions of corruption, and Diaz and Piazza (2022) found that corruption revelations inspired female candidates to contest municipal elections in Brazil.

Although an association between the salience of corruption and increases in women’s representation has been uncovered, the mechanism behind the increase remains elusive. On the one hand, parties may use women strategically after legitimacy crises, nominating them as candidates where corruption scandals, accusations, or revelations have occurred with an incumbent mayor. Traditionally, parties have resisted women’s inclusion, exploiting loopholes in parity requirements (Baldez 2004) and marginalizing women even after they are elected (Schwindt-Bayer 2010; Senk 2021). However, the value of women’s inclusion can change after corruption scandals. Once an incumbent party has been tainted with wrongdoing, the party has a strong incentive to “clean” its image and provide a viable alternative in the eyes of voters (Valdini 2019). One low-cost solution for incumbent parties hoping to evade accountability is putting women on the ballot; if voters stereotypically perceive women as less corrupt than men, then a party may get more votes (Barnes and Beaulieu 2014; 2019; Goetz 2007).

Party hypothesis (H1): Incumbent parties will be more likely to nominate women as candidates after recent corruption revelations.

Yet, a recent empirical study in Latin America finds that, where corruption perceptions are higher, parties do not have more women on their ballots (Funk, Hinojosa, and Piscopo 2021). This finding suggests that another mechanism may be at work—voters. Voters have several reasons to prefer women in the aftermath of corruption scandals. First, voters may subscribe to gender stereotypes and view women as more honest, more trustworthy, and less corrupt than men (Barnes and Beaulieu 2014; 2019; Benstead, Jamal, and Lust 2015; Goetz 2007; Le Foulon and Reyes-Housholder 2021).¹ Second, voters might view women as political outsiders and less likely to have access to the networks necessary to engage in corruption (Goetz 2007; Reyes-Housholder and Thomas 2018; Sundström and Wängnerud 2014). Third, voters might also expect women to eschew opportunities to engage in corruption if they perceive them as more risk averse than men (Barnes and Beaulieu 2019). Fourth, voters may respond to the party’s selection of a candidate representing a fresh anticorruption direction or being an anticorruption crusader. For these reasons, voters may be more likely to support women in a post-corruption election.

Voter hypothesis (H2): Women candidates will be more likely to win after recent corruption revelations than men candidates.

So far, evidence of parties and voters preferring women candidates after corruption revelations has been mixed. Observational studies have found that women are more likely to win after recent corruption revelations (Diaz and Piazza 2022; Guajardo and Schwindt-Bayer 2024b; Valdini 2019) but have not tested the mechanisms behind this relationship. Survey experiments have found null or small effects for voters preferring women after corruption (Batista Pereira 2020; Elia 2024; Le Foulon and Reyes-Housholder 2021; Schwindt-Bayer, Esarey, and Schumacher 2018) but have focused on the voter mechanism only and just in a hypothetical way. In this study, we offer a novel, replicable observational design that causally tests the voter and party mechanisms in one specific context, Mexico. We expect results to apply to other highly corrupt, party-centered systems where women’s representation is increasing.

DATA AND METHODS

We use data on 47,141 candidates for mayor running in 10,119 municipal elections and audit results in Mexican municipalities from 2000 to 2019 to assess whether recent corruption revelations affect where women are nominated as candidates and their likelihood of winning the election.² Mexico is a country where party leaders have strong control over candidate selection (Kerevel 2013; Motolinia 2021) and quotas require them to nominate women to all legislative and executive offices, including mayors (Piscopo 2017).³ Moreover, a large proportion of Mexican voters view women as less corrupt than men (Batista Pereira 2020).

We use data on 2,967 audits of municipal finances from Mexico’s Supreme Audit Institution (ASF) to create a measure for recent corruption revelations in the municipality—our main explanatory variable. Auditors publish audit results a year after they conduct audits. Local news outlets often publicize audit results (Larreguy, Marshall, and Snyder 2020; Stanig 2014) and discuss them on social media.⁴ We code as “1” cases where an audit report that found wrongdoing was made public in the year before the election and as “0” cases where no audit occurred or no spending irregularities were found.⁵ Wrongdoing implies that auditors found a non-zero amount of irregularities in the social infrastructure fund (FISM), which is exclusively used for improving basic infrastructure and tackling poverty in the municipality.⁶ Following previous studies, we interpret FISM spending irregularities as corruption since they imply deviations from spending guidelines and are directly controlled by the mayor (Chong et al. 2015; Guajardo and Schwindt-Bayer 2024b; Larreguy, Marshall, and Snyder 2020).

To test H1 (the party hypothesis), we assess whether parties are more likely to nominate women in municipalities with recently revealed spending irregularities. The unit of analysis is the party candidate for mayor in a municipal election. We interact our indicator of recent corruption revelations with an indicator for whether the candidate belongs to the incumbent party (1) or opposition (0) and predict whether the party nominates a woman (1) or a man (0). To test H2 (the voter hypothesis), we assess whether a woman candidate is more likely to win the mayoral election after recent spending irregularities are revealed. We interact our indicator of corruption revelations with the candidate’s gender to predict whether a candidate won the election (1) or not (0). Winning, of course, results from getting the most votes from voters; thus, this modeling strategy allows us to test whether women candidates are more likely to win the most votes and get elected in municipalities with recent corruption revelations.

We use linear probability models with clustered standard errors on municipality election and state election year fixed effects to account for unmeasured factors associated with specific states and election years.⁷ Under this setup, candidates in a municipality election are compared to candidates in other municipality elections in the same state and year based on whether they experienced a recent revelation of corruption.

We expect audit assignment to be unrelated to women’s representation in mayoral offices.⁸ However, criteria for audit assignment are not random and can be based on performance indicators and signs of institutional weakness (Chong et al. 2015; Larreguy, Marshall, and Snyder 2020).⁹ For that reason, we control for factors that could affect both the emergence of women and audit assignment. One factor is an index of human development and population in the municipality since some studies of women’s representation expect a positive relationship between women’s emergence and levels of development (Hughes 2011, but see and Hinojosa 2012 and Reyes-Housholder and Thomas 2018), and these factors affect audit assignment in our data (see Supplementary Material, section 7). We also include indicators of electoral competition, specifically the margin of victory in the last election and the Petersen index of electoral volatility in the municipality, since women could be sent to losing districts and municipalities with more recurrent audits could be more volatile and competitive for audited parties. Summary statistics for all variables are presented in Supplementary Material, section 8. For replication data, see Guajardo and Schwindt-Bayer (2024a).

FINDINGS

Overall, we find that women are more likely to be elected after recent revelations of wrongdoing in Mexican municipalities because of voters, not parties. Political parties are no more likely to nominate women as candidates in municipalities with recent revelations of spending irregularities, compared to municipalities with no revelations. However, a woman candidate is more likely to win an election in a municipality where wrongdoing was recently revealed, an effect that is not found among men.

Figure 1 presents the results for the party hypothesis (H1). Contrary to our expectations, we find that incumbent parties are not more likely to nominate women as candidates to municipalities with recently revealed spending irregularities; neither are opposition parties (more details in Supplementary Material, section 9). On average, the probability of a candidate being female is approximately 0.45, regardless of whether the candidate ran for the incumbent or opposition party or whether corruption was recently revealed. In Supplementary Material, section 10, we analyze whether any party is more likely to nominate women to municipalities where corruption was recently revealed. We find that smaller parties (PANAL, PES, PVEM), regional parties (inside the “Other” category), and leftist parties (MORENA, MC, and PT) are overall more likely to nominate women as candidates compared to parties such as PRI and PAN. However, a recent revelation of wrongdoing does not appear to motivate parties to strategically run a woman in a municipality. If anything, large mainstream parties like MORENA and the PAN-PRD coalition are less likely to nominate women after recent revelations.

Figure 1. Parties Are Not More Likely to Nominate Women after a Recent Revelation of Corruption

Note: Predictions from a linear probability model with 95% confidence intervals, all other variables at their means. Recent revelation of corruption in the past year (1) or not (0). Full model results can be found in Supplementary Table A.6 (column 2).

We conducted a series of additonal analyses and robustness checks. We explored whether strategic action by parties is contingent on the gender of the audited mayor since we might expect the strategic use of women to only be valid if revealed wrongdoing was tied to a man. Analyses reported in Supplementary Material, section 11, reveal that results do not differ depending on the gender of the audited mayor. Additionally, it is possible that revelations in the previous year do not provide enough time for party leaders to react strategically with candidate selection. Results in Supplementary Material, section 12, show that extending the time window of recent corruption revelations to three years does not lead to different results. Analyses presented in Supplementary Material also test whether the null relationships hold when we only consider audited municipalities (Supplementary Material, section 13); they do. Another analysis examines whether different relationships existed before and after the national parity law started to require quotas at the subnational level in 2014 (Supplementary Material, section 14). The effect of corruption revelations is consistently null. Finally, models in Supplementary Material (see section 15) include tests that address spillover concerns among neighboring municipalities. The null finding for corruption revelations persists after excluding neighboring (non-treated) municipalities from the sample that have been potentially affected by treatment.

Figure 2 presents the results for the voter hypothesis (H2). It shows that whereas male candidates are not more likely to win after recent revelations of irregularities, women are (full table in Supplementary Material, section 16). On average, spending irregularities published in the past year increase the likelihood of a candidate winning the election from 0.067 to 0.111 among women (a 65% increase). The effect of corruption revelations among women is most potent in the first year of the revelation of corruption (as shown in Figure 2). The effect size decreases after two and three years but remains positive and significant (see Supplementary Material, section 16). This effect is not found among men, with the probability of a male candidate winning the election remaining unmoved regardless of whether there was a recent revelation of spending irregularities. Figure 2 also shows that, on average, women candidates are less likely to have voter support than male candidates both when corruption has been revealed and when not. Models in Supplementary Material, section 16, include additional controls, such as whether the candidate ran with a coalition, whether the municipality neighbors a treated municipality, the proportion of women that ran as candidates in the election, and party dummy variables.

Figure 2. Women Are More Likely to Win after a Recent Revelation of Corruption

Note: Predictions from a linear probability model with 95% confidence intervals, all other variables at their means. Recent revelation of corruption in the past year (1) or not (0). Full model results can be found in Supplementary Table A.8 (column 2).

Supplementary Material includes additional analyses that test extensions of our hypothesis and underscore the robustness of the results. First, we explore whether results vary depending on the gender of the audited mayor. The voter hypothesis (H2) expects voters to prefer women because they view them as less corrupt. However, if the revelation of corruption was tied to a woman, we would expect stereotypes about women being less corrupt to be weakened. Figures in Supplementary Material (section 17) show that the effect of recent corruption revelations among women is only found when the audited mayor was a man. Second, we conduct placebo tests with the year the audit was announced (before results became publicly available) as the explanatory variable instead, in order to assess whether it is in fact spending irregularities becoming public that is driving the effect and not anything related to audit selection. Models in Supplementary Material (section 18) find no effect if an audit was conducted that same year. We also explore whether results hold before and after quotas started to be mandated for subnational offices. We find evidence of an effect both before and after quotas (see Supplementary Material, section 19). Effects are slightly larger in the pre-quota period, when women were less common in elections and stereotypes about women being less corrupt were stronger. As we did for the party hypothesis models, we also tested whether the effect holds when we only consider audited municipalities. The effect remains positive but loses significance (Supplementary Material, section 20, provides additional details). Finally, models in Supplementary Material, section 21, explore whether results vary depending on the size of the revelations of corruption, finding consistent results. Women are more likely to win when irregularities are greater than 0, but differences in the amount of irregularities produce similar increases in women’s probability of winning.

CONCLUSION

Who is behind the apparent increase in women’s representation after corruption scandals—parties, voters, or both? Using an original dataset that exploits corruption revelations at the subnational level and candidate-level data on close to 20 years of mayoral elections in Mexico, we find evidence of voters being behind the increase of women in office. While parties are not more likely to nominate women as candidates in municipalities where spending irregularities have been revealed, the probability of a woman winning the election increases by 4.4 percentage points (65%) if audit results reveal wrongdoing in the municipality. The same effect is not found among men.

This study makes several contributions. It offers an observational design that could be replicated in other countries to study the relationship between women’s representation and corruption revelations. To our knowledge, no previous studies have explored these dynamics with fine-grained candidate-level data, allowing us to distinguish the probability of parties nominating women as candidates in corrupt contexts from the probability of women candidates winning when corruption has been revealed. Additionally, our findings differ from those of studies that expect strategic parties to be behind the increases in women’s representation after corruption scandals. We find that even for a party-centered system like Mexico, women benefit electorally after corruption revelations despite parties, not because of parties. This suggests that the mechanisms driving the increases in women’s representation after corruption scandals may be context dependent. Future research is needed to determine how generalizable these results are, whether parties learn and adapt to voter preferences, and to explore which contextual factors determine whether parties matter more than voters or vice versa, such as variation in stereotypes of women being less corrupt or the degree to which the system is party-centered.