Social scientists rely heavily on self-reported data. But can respondents be trusted to report the truth? In this paper, the authors compared survey self reports with administrative data and found that nearly 50% of recent borrowers did not report their high-interest consumer loans. Under-reporting appeared to be correlated with several characteristics, in particular gender. Relying strictly on self-reported data may lead to biased inference, and the authors outline some methodological implications for identifying impacts of credit access on borrower behavior and outcomes. Matching female surveyors to female respondents appears to be one low-cost mitigation strategy. The best strategy, however, is to avoid reliance on self-reported data by using lenders’ administrative data or the credit bureau, when feasible.
Dean KarlanJonathan Zinman
Publication type: 
November 01, 2007
Program area: