Gender discrimination in politics: Evidence from French elections
Gender discrimination in politics: Evidence from a natural experiment in French local elections
The underrepresentation of women in politics is a widely shared observation, but the reasons for this underrepresentation are still imperfectly understood and are the subject of much debate.
Several explanations have been put forward by the economic literature: women may be less willing to engage in politics because of a lack of self-confidence or a steeper trade-off between family balance and competitive environment (Bertrand et al. 2010); they may face higher obstacles within parties once they engage in politics (Casas-Arce and Saiz 2015, Esteve-Volart and Bagues 2012); and they may suffer from voter discrimination bias (Beaman et al. 2009).
In a recent paper, we test the gender bias from voters’ hypothesis (Eyméoud and Vertier 2017).
A natural experiment: The 2015 departmental elections
Historically, French departmental councillors were elected for a period of six years according to a two-round, uninominal, majoritarian voting system: each councillor ran accompanied by a substitute. The elections were held every three years, thus renewing departmental councils by half in each election.
The law of 17 May 2013 reformed voting arrangements. The election now follows a two-round, mixed-pair, majoritarian voting system: the candidates are now required to run as a gender-mixed pair, composed of a man and a woman, accompanied by two substitutes with corresponding genders. This method achieves one of the reform’s stated objectives: to ensure strict gender parity after the departmental election.
On 22 and 29 March 2015, the first and second rounds of the departmental councillor election were held in 2,054 cantons following the new electoral system. There were 9,097 pairs of candidates in this election.
Appearance on ballot is in alphabetical order
Article 3 of the law provides that the order of candidates on the ballot is in alphabetical order. To ensure compliance, the law also stipulates that ballots that violate this rule are considered invalid.
Figure 1 Names on ballot must be in alphabetical order
Notes: If Rebecca Vannier forms a pair with Peter Dupont, she appears second on the ballot (since the letter D is before the letter V in the alphabet). On the other hand, if Rebecca Bannier forms a pair with Peter Dupont, she appears first on the ballot (the letter D being after the letter B in the alphabet). (The names in the smaller font are the candidates’ substitutes.)
In our study, we show that the fact that a woman is in the first position on the ballot is random: it occurs in half of the cases and none of the candidates’ observed characteristics explains their position on the ballot.
Figure 2 Cantons by gender of first name on ballot, March 2015 election
Notes: The cantons coloured in red are those in which the gender of the first candidate appearing on the ballot of the first pair (in alphabetical order) is male. Cantons in blue are those in which the gender of the first candidate is female.
We then compare the electoral performances of pairs that belong to the same political group but are located in different cantons. Specifically, we look at whether pairs with a woman in the first position received on average a different number of votes compared to pairs with a man in the first position.
Identifying a difference in vote between the two types of pairs means that we simultaneously observe two things:
Imperfect understanding of election rules by voters
Discriminatory behaviour of voters regarding women
Given that the elected candidates will be granted the same powers regardless of their position on the ballot, voters have no logical reason to change their vote depending on the candidate’s order of appearance. If they do, this indicates that voters imperfectly understood the rules of the election and likely thought that the person appearing first on the ballot would be granted greater powers.
Consequently, any difference in votes between pairs whose first candidates differ in gender is necessarily explained by discriminatory behaviour by voters regarding women.
Comparing the electoral performances of candidate pairs running in the departmental election, we show that only right-wing pairs suffered from discrimination.
In fact, unlike candidates from other political parties, right-wing pairs with a woman in the first position lost between 1 and 2 percentage points of the vote compared to pairs of the same political stripe but with a man in the first position (that is, a decrease in the share of votes received of around 5%). They also saw their probability of going to the second round (and being elected) reduced by about 5%.
The missing votes
The reasons why right-wing voters may discriminate are complex, and we cannot observe what actually happened in the voting booth. However, the data allow us to identify possible explanations for their choices.
First, decisions seem to have been made in the voting booth, since we find no differences in abstention rates or the share of invalid and blank votes. Voters transferred their votes to other pairs; on average, other parties symmetrically received more votes when opposing a right-wing pair in which the woman appeared first.
Second, discrimination appears to be lower when voters are better informed about candidates. We used the SciencesPo electoral archives to gather a subsample of ballot papers (about 12% of the total) from which we extracted additional information sometimes supplied by the candidates (age, photographs, political experience, socio-professional category).
Analysis of this additional information suggests that when information about candidates is available on the ballot, the effect of discrimination disappears. It also seems that pairs who decide to add information to the newsletter also receive more votes.
Public policy implications
Several elements of interest for public policy emerge from the results of our study:
1. The information conveyed by the ballot is not neutral.
Our results show that electoral performance and the degree of discrimination against candidates vary greatly depending on the amount of information available in the voting booth. Experimental economics has recently shown that small informational differences can have substantial effects on the behaviour of individuals: under these conditions, it is worth asking whether we need to harmonise – upwards or downwards – the information contained in ballots.
2. There is a need to reflect on the implications of quota policies.
Although the reform achieved its objective of forming mixed assemblies, it did not guarantee real parity at the highest level of departmental assemblies: since 2015, almost 90% of assembly presidents have been men.
3. The issue of women’s place in politics is linked to numerous perspectives on gender discrimination.
It has been established that discrimination in the political sphere is directly connected to discrimination in the labour market (Le Barbanchon and Sauvagnat 2018) and that this, in turn, partly results from gender bias in educational institutions (Breda et al. 2018).
The fight against gender discrimination will only be effective if we connect these different public policies, and reflect on their interactions.
Beaman, Lori, Raghabendra Chattopadhyay, Esther Duflo, Rohini Pande and Petia Topalova (2009). “Powerful women: Does exposure reduce bias?”, The Quarterly Journal of Economics 124(4): 1497–540.
Bertrand, Marianne, Claudia Goldin and Lawrence F Katz (2010). “Dynamics of the gender gap for young professionals in the financial and corporate sectors”, American Economic Journal: Applied Economics 2(3): 228–55.
Breda, Thomas, Julien Grenet, Marion Monnet and Clémentine Van Effenterre (2018). “Can female role models reduce the gender gap in science? Evidence from classroom interventions in French high schools”.
Casas-Arce, Pablo, and Albert Saiz (2015). “Women and power: Unpopular, unwilling, or held back?”, Journal of Political Economy 123(3): 641–69.
Esteve-Volart, Berta, and Manuel Bagues (2012). “Are women pawns in the political game? Evidence from elections to the Spanish Senate”, Journal of Public Economics 96(3): 387–99.
Eyméoud, Jean-Benoît, and Vertier, Paul (2017), “Gender biases: Evidence from a natural experiment in French local elections”, mimeo, SSRN 2913453[EN3] .