Measuring Gender Bias in U.S. House Elections: An Outcome Test Approach

• In Event: Women and Politics Research & Federalism and Intergovernmental Relations Virtual Poster Session
  In Virtual Poster Session: Women and Politics Research Virtual Poster Session

Abstract

Becoming a representative of the Unites States House requires selection through three sequential stages: choosing to run, winning the primary nomination and winning the general election. While there is the potential for gender bias at all of these stages, empirically, we have observed few differences between the rates of men and women winning conditional on running. While much of the focus has shifted to looking at gender differences when candidates are choosing to run, we still lack a comprehensive picture of the amount of bias, if any, through all the stages. Differential selection mechanisms for women may result in only the most qualified women being elected suggesting women should actually be winning at higher -- not at equivalent rates, to men. Selection can be affected by two different mechanisms: 1) election aversion (or political ambition) where women perceive there to be bias or high costs so as to self-select and only choose to run if they are highly qualified and/or 2) voter discrimination. It is likely that both of these mechanisms, election aversion and voter discrimination, are consistent with the empirical observations that conditional on winning, women are more effective legislators and that conditional on running, women and men win at equal rates. How then do we explain the continued low-level of representation by women in Congress? To answer this question, I combine federal outlay data on U.S. House members and data on money raised by candidates with primary and general election results. This comprehensive dataset gives a richer picture of the hurdles to descriptive representation than data that focuses on any singular stage. Additionally, embracing methodological pluralism, I use a method more commonly found in economics and criminal justice for studying discrimination that has yet to be used in evaluating bias in elections. The outcome test is a methodological approach specifically used for studying bias in selection processes. Using an outcome test approach on both a candidate outcome and legislator outcome I am able to evaluate gender bias comprehensively throughout the entire selection process. As gender is only one important dimension of descriptive representation this approach may be extended to better understand the intersectionality of selection bias in elections.

Author

• Elisha Cohen, Emory University