Whom to ask? Testing respondent effects in household surveys

Household questionnaires typically survey the most knowledgeable household member, but this can lead to inaccurate data if they have limited information. Using data from survey experiments with 4,100 households in Paraguay and Uganda, we investigate whether there are discrepancies in intra-household reporting when multiple household members are interviewed. We randomly vary who responds to a survey on household income and food consumption using common approaches to respondent selection. We find that the mean and distribution of these variables is insensitive to respondent selection. However, there are discrepancies between spouses of the same household in Uganda that depend on the gender of the respondent or recipient of the income. Taken together, our results indicate that respondent selection does not markedly affect the aggregate analysis of households, but that it may matter for obtaining accurate information on income and consumption for a given household or by gender.