Does it matter who you ask for time-use data?

Time-use statistics are sensitive to measurement error, especially errors that might be introduced based on whether the informant is reporting on herself or reporting on others in the household. In this paper, we use the nationally representative time-use survey in India and propensity score matching methods to investigate how self- and proxy reporting impact the reported time spent on various activities by men and women. Theoretically, we examine the mechanisms underpinning the differences in reporting and empirically test our hypothesis. Proxy informants tend to report higher time use for both men and women on employment activities (14- 26 per cent) and lower time use on production for self-consumption, unpaid domestic work, and care work (5-33 per cent) as compared to self-reports. On average, female proxies differ more from self-reports when reporting about both men and women in their households as compared to male proxies. Investigating the mechanisms, we find that the self-proxy differences are not due to random error but are systemic. Information asymmetry between the self and proxy respondents plays a key role-spouses and self-proxy respondents with similar characteristics have smaller reporting differences than non-spouses and other respondents. Gendered perception of what activities are classified as work also play a key role in the differences.