SangEun Kim and Kristin Michelitch
Do politicians tend to be more responsive to their own social groups, such as their own gender or ethnic group? In particular, when politicians from historically-marginalized groups attain elected office, do they tend to be more responsive to their fellow group members? Many experimental studies in Global North countries find evidence of politicians’ greater responsiveness to in-group members, especially politicians from historically-marginalized groups (e.g., Costa’s (2017), Magni & Ponce de Leon (2020)).
In contrast, we find that neither women nor men politicians are any more responsive to their own gender in an experiment on gender bias with subnational politicians in Uganda, where one third of legislative seats are reserved for women. We leverage a novel in-person survey experiment with politicians as part of a larger field experiment on politician accountability (Grossman & Michelitch (2018)). Respondents include 333 subnational Ugandan politicians across 20 subnational legislatures, one tier below the central government. Politicians’ response rate was 92% due to sustained local fieldwork and a partnership with ACODE, a Ugandan public policy NGO.
Politicians were asked to allocate a fixed hypothetical budget (100,000 UG shillings) between making service delivery improvements after hearing two audio recordings of citizens stating grievances and requests for improvements by constituents. Specifically, constituents state common complaints about how staff absenteeism – an absent teacher versus an absent health clinic worker – negatively affected their child’s education or health, respectively. The voices were randomly assigned as (a) male voice-school, female voice-health clinic, or (b) female voice – school, male voice-health clinic.
By examining politicians’ allocation decisions toward improvements in the two, we find that politicians are no more (or less) responsive to requests from same-gender citizens, nor to any particular gender. We further find no significant differences between men and women politicians. The study is well-powered and well-vetted to have a high degree of causal validity and fit the local context.
Rather than respondent identity, we find in exploratory analysis that politicians’ perceptions about the relative visibility of improvements to garner votes through improvements in the education versus health sector is the leading determinant of politician allocation decisions. These findings support the idea that perceived electoral incentives, rather than constituent gender identity, may play a more significant role in politician responsiveness toward individual constituents in the study area.
Of course, the experiment does not rule out gender discrimination in Ugandan politics. First, men, the historically-priviledged and currently more politically active group in this context, are more likely to contact politicians with requests — thus, their priorities may be better represented. Further, our experiment regards two service delivery sectors that are equally prioritized by men and women, but sectors classically prioritized by women versus men may lead to different results. Indeed, elsewhere we show that water — disproportionately women’s priority — is given extremely little attention compared to the other top service delivery sectors in legislative activity. Lastly, politicians have a variety of job duties in which gender discrimination may occur, negatively affecting other aspects of women’s representation. Garcia-Hernandez, Grossman, & Michelitch (2018) show that women are more peripheral in informal networks, which drives a gender gap in legislative activity favoring men in these same legislatures.
Nonetheless, this study raises important academic and policy questions about the conditions under which politicians are more responsive to citizens with whom they share a salient identity group, especially politicians from historically-marginalized groups. Most research finding that politicians are more responsive to in-group members tends to be conducted in the Global North, especially the United States. By contrast, our study is one of few conducted in the Global South (e.g., McClendon (2016), Distelhorst & Hou (2018), or Gaikwad & Nellis (2020)) where discrimination toward historically-marginalized groups, such as women, can be particularly stark but where affirmative action in the form of reserved seats can be more common. More research is necessary to investigate what societal or institutional features might lead to differences in politicians’ in-group responsiveness.
In addition to underscoring a need in additional studies to gain cross-national variation, our study also challenges the assumption that descriptive representation of a historically-marginalized group automatically leads to substantive representation. On the one hand, lack of gender bias is heartening as a result – politicians did not systematically ignore women as the historically-marginalized group. Of course, the mere presence of women politicians due to the affirmative action could raise awareness and lead to such an across-the-board result. On the other hand, others may be disappointed that women politicians did not disproportionately respond to women citizens. Identity-based motivations, should they exist, may be crowded out by accountability incentives to the voters (or parties) to which politicians owe their (re)election. In some contexts, removal of other barriers (e.g., Grossman, Michelitch, & Santamaria (2017)), in addition to affirmative action for office-holding, may be necessary to facilitate the substantive representation of citizens from historically-marginalized groups.