A Two-Stage Theory of Women’s Representation: Evidence from Mayoral Elections

• In Event: Representation and Electoral Systems & Normative Theory Virtual Poster Session
  In Virtual Poster Session: Representation and Electoral Systems Virtual Poster Session

Abstract

Which electoral institutions are most conducive to fostering descriptive political representation of women? This paper aims to address this old question in a novel two-stage theoretical framework and test it using evidence from mayoral elections in Latin America. It argues that descriptive representation of women is a result of two distinct sets of processes that occur at two stages. First, women must present themselves as candidates running for political office. Second, those women that run for office must be elected. One does not logically imply the other because large numbers of women may run for office without being elected. Nor are the two processes necessarily empirically related through the same causes. Women’s political presence as candidates or as elected politicians may be enhanced or hindered by very different causal processes at either level.

Different variables may affect whether women run for office, and different variables affect whether they are elected or not. The first stage determinants decide whether women are selected as candidates. The access to viable candidacies may be restricted to women by their lack of resources compared to their male counterparts. The selection of candidates often hinges on the party elite’s perception of their abilities, qualifications and experience (Jalalzai, Krook 2010: 17). The role of such party gatekeepers in candidate selection and the specific institutional setting in which this selection operates is often crucial. Both formal institutions such as the type of electoral systems faced by the parties and informal practices, such as clientelism and corruption (Stockemer 2011, Sundström, Wängnerud, 2016) are critical at this stage. Even if women overcome the first hurdle of becoming viable candidates, this is no guarantee that they will overcome the second hurdle and get elected. The second stage presents a wholly new glass ceiling. Even with dramatic increases in the proportion of female candidates, the actual number of elected female politicians may not increase at all or increase at a much lower rate (Miguel 2008, Górecki, Kukołowicz 2014).

This situation requires (1) that theories are formulated for both levels separately, and (2) that empirical strategies exploring the determinants of women’s representation accommodate this two-stage process. Existing research is sensitive to this two-stage process in terms of theorizing (theorized determinants are usually specifically related to one and/or the other stage), but this is usually unaccompanied by research designs that would similarly be sensitive to this, which could lead to biased estimates. In fact, the most common mode of analysis of the determinants of women’s descriptive representation is one in which the outcome variable is based on the share of women elected to the representative body (i.e. the exclusively the second stage process) and which estimates a regression model of the effects of theorized explanatory variables. But, some of such variables may affect women’s representation in opposing directions. This was argued with respect to electoral-institutional variables such as party system fragmentation: “high party fragmentation may well increase the number of women nominated as candidates, but to actually win seats the fragmentation needs to be lower.” (Reynolds 1999: 553). Including such variables in models which only measure the second stage variable (share of elected women) may conceal its powerful effects – but in opposite direction – at either the first or the second stage.

The paper argues that this disjunction between theorizing and empirical testing presents serious inferential problems including biased estimates of effects. The paper illustrates this on the case of selecting and electing female mayors in Latin American countries and shows that both these processes have some fundamentally different determinants and should be distinguished in any analysis of women’s descriptive representation. The proposed methodological solution (Heckman selection model) solves some of the inferential problems faced by existing analyses. The analysis is based on the municipal-level data of over 9,000 municipalities in 15 Latin American countries with competitive elections. Using a hierarchical framework, this allows for testing of the effect of variables both at the municipal-level (e.g. party competition, voter turnout etc.) and at the country level (compulsory voting, simultaneity of elections etc.). It is hypothesized that such individual variables will impact differently the probability of female candidates competing in the elections (the first stage) and actually winning (the second stage).

Authors

• Karel Kouba, University of Hradec Kralove

• Tomas Dosek