What Did They Say? Respondent Identity, Question Framing and the Measurement of Employment

Abstract Using data from a primary survey conducted in rural India, this paper examines how two key survey design features—respondent identity and question framing—affect employment estimates. First, it estimates the causal impact of (a) replacing a single weekly employment question with a set of detailed activity-specific questions, and (b) changing the reference period from a week to individual days. The detailed module yields significantly higher estimates of women’s employment with no corresponding effect for men. Second, using spousal respondent pairs, the paper finds that proxy-reports by men significantly underestimate women’s employment while men's employment estimates do not differ between self- and proxy- reports. Within different types of employment however there are significant deviations for both genders. Intra-household analysis suggests misreporting is driven by asymmetric information and gender norms. Overall, the findings underscore the importance of self-reporting and detailed questions for accurately measuring employment with implications for improving survey design in resource-constrained contexts.