Research
Is Social Desirability Biasing Survey Responses? Assessing Interviewer Effects in Survey Data (Job Market Paper)
[PAPER]
Social norms play a role in important decisions such as educational and labor supply choices. Could social norms also systematically affect the responses to sensitive questions in surveys? I examine the role of gender norms in two ongoing surveys: the National Longitudinal Survey of Youth 1979 (NLSY79) and the Kenya Life Panel Survey (KLPS). I examine how responses to sensitive topics—such as gender norms and household labor division—vary depending on the gender of the interviewer, which likely affects the perceived norm in the eye of the interviewee. In the NLSY I focus on questions measuring agreement or disagreement with statements such as “Men should share the work around the house with women”. I first document a systematic difference in responses between male and female respondents. Consistent with the role of social norms, when the interviewer is of the opposite gender, these differences shrink by about 30-50%. For example, female respondents are less likely to give the gender-typical response when interviewed by males, and the same pattern holds for males interviewed by females. I examine the parallel patterns in the KLPS, focusing on comparable questions about gender norms. I replicate the key finding that females are less likely to give the gender-typical answer when interviewed by a male, compared to by a female — and conversely for male respondents. I also find the gender-based response patterns in the KLPS do not fit the simplest social desirability model. I conclude the paper by discussing the implications of these findings for the elicitation of attitudes and survey responses on sensitive topics.
Internet Matching Technologies and the Geographic Distribution of Chess Skills (with Sara Ellison , Working Paper)
[PAPER]
There is broad consensus that the internet has the potential to become an important tool for learning. There is less consensus on its likely distributional effects. On the one hand, it could provide previously isolated regions, countries lacking high-quality educational systems, or under-resourced individuals and communities access to skills and knowledge. On the other hand, its offerings might be relatively more accessible to well-resourced communities for many reasons, widening gaps. The COVID pandemic has put these questions into sharp relief, with millions worldwide being moved to online or distance-learning. This paper leverages a historic example of distance-acquisition of one particular skill, chess-playing. We look for evidence of either this democratizing effect or a widening divide on chess skills from internet penetration. We also look broadly at how the geographic distribution of chess skills has changed since the introduction of the internet. Finally, we speculate on what, if anything, can be said about the relevant mechanism given the particular characteristics of how one typically learns to play chess. Preliminary results show evidence of decentralized global chess skill concentration and reveal Internet exposure as a significant predictor of chess skill at the individual and country level.