![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
In nearly all countries, women are underrepresented and men are overrepresented in national legislatures. This distortion in representation might occur for several reasons. One set of explanations suggests that parties, voters, or both, might discriminate against women. In this analysis we examine potential discrimination by parties and ask if elected officials discriminate against women who are thinking about a career in politics. Evaluating whether discrimination occurs is notoriously difficult with observational studies, so we conduct a field experiment to examine whether elected officials in New Zealand respond differently to potential political aspirants based on their perceived gender. Our results show that elected officials are equally willing to respond to both male and female political aspirants. These findings support the results from recent work conducted in other developed democracies and suggest that parties do not discriminate against female political aspirants at this stage of the recruitment process.
Disclosure statement
No potential conflict of interest was reported by the authors.
Supplementary material
Supplemental data for this article can be accessed here.
Notes
1. We thank Matt Golder, two anonymous reviewers, and the editors for extremely useful feedback. The University of Michigan’s Institutional Review Board approved this study (HUM00153767). We pre-registered the study’s design and our analytical strategy at Open Science Framework (https://osf.io/gwksf/?view_only=b2f06cb2bc284cec8b8c35170d200add). The data that support the findings will be available on the Harvard Dataverse upon publication.
2. In this paper, we talk about gender in terms of men and women. Although gender is distinct from biological sex, and gender as a concept is socially constructed and not dichotomous, the studies we will draw upon generally focus on men and women, and explicitly or implicitly treat gender and sex as collinear for the purposes of their research goals.
3. Interestingly, the average percentage of female legislators in democracies is only slightly higher than the average percentage of female legislators in dictatorships. In only three countries – Rwanda, Cuba, and Bolivia – do women comprise more than half of the legislators.
4. Existing work examining why women are underrepresented in politics highlights multiple stages of the representation process during which discrimination might occur. In this paper, we focus on the role that parties can play in encouraging potential political aspirants. Specifically, we conduct a field experiment in New Zealand to evaluate whether elected officials are equally responsive to female and male political aspirants.
5. Montgomery, Nyhan, and Torres (Citation2018) provide a good introduction to post-treatment bias and how researchers can potentially avoid it.
6. This discussion draws heavily on Gaddis (Citation2018).
7. We did not include members of the various local boards, contacting only members of municipal councils. We excluded national-level politicians who were also members of the government. We provide additional sample details in Appendix A.
8. The emails would appear to be only from students of European descent, rather than from students with Māori, Pasifika, or Asian backgrounds.
9. Another benefit of using a smaller number of names is that it also allowed us to maximize statistical power.
10. Feedback from colleagues at the 2018 New Zealand Political Studies Association annual conference suggested that we should have modified the original language of the email to say ‘course’ rather than ‘class’. Additionally, multiple scholars claimed that the text of the email was, in their experience, unusually polite and formal compared to the correspondence they typically receive from students.
11. We did not collect other information about the officials. We provide more descriptive statistics for these pre-treatment covariates in Appendix A, but in brief, the mean age of the officials in our sample was 53, 32% were women, and 90% were officials holding subnational offices.
12. Like other audit studies, we cannot tell if the officials in our sample receive and reply to our emails or if staff do so instead. As a result, our unit of analysis is the office of the official and not the official herself.
13. To be able to examine the effects of these pre-treatment characteristics as efficiently as possible, we block-randomized on gender, level of office, and age (Moore and Schnakenberg Citation2012). This means that we divided the officials into four groups, each with the same mixture of gender, level of office, and estimated ages. Then two of these groups received emails from women and the other two from men. This approach allows us to minimize the effect of age, level, or gender of the recipients across the treatments on rates of response.
14. We had intended to collected data on response rates as well as quality of response within a two-week period. Due to technical problems with the email provider, we were only able to collect data on responses for up to a week for the earliest messages, and at least three days for the latest messages. Note that evidence from this kind of correspondence study suggests that most people will respond within 48 hours, if they are going to respond at all (Costa Citation2017; Hughes et al. Citation2018).
15. An LPM is preferable to a logit or probit model because it provides an unbiased estimate of the average treatment effect and is easier to interpret (Lin Citation2013; Judkins and Porter Citation2016).
16. Emails were delivered over a three-day period, spanning Wednesday-Friday.
17. We present tabular results in Appendix B.
18. We obtain similar results with a wide range of analyses, such as difference of proportion tests (Appendix C), logit models (Appendix D), and randomization inference tests (Appendix E).
19. To determine this, we re-estimated the LPMs described above but interacted the gender treatment indicator with a measure of (a) official gender and (b) official level of office. The interaction terms from these models were not statistically significant, suggesting that elected officials treated male and female students similarly, regardless of whether the officials were men or women, or occupied national or subnational office.
Additional information
Notes on contributors
Sona N. Golder
Sona N. Golder is Professor in the Department of Political Science at The Pennsylvania State University. She studies political institutions, with a particular interest in coalition formation and representation.
Charles Crabtree
Charles Crabtree is a visiting scholar in the Department of Government at Dartmouth College and a Senior Data Scientist in the Policy Data Lab at the Tokyo Foundation for Policy Research, where he specializes in fairness in politics, with applications to discrimination and repression.
Kostanca Dhima
Kostanca Dhima is a PhD Candidate in the Department of Political Science at Texas A&M University, where she works on gender and politics, in particular women’s political representation in a comparative perspective.