Abstrakt
Recent studies linking household surveys to administrative records reveal high rates of misreporting of program receipt. We use the FoodAPS survey to examine whether the findings of these studies of general household surveys using one or two states generalize to a survey with a narrow focus and across many states.
First, we study how reporting errors differ from other surveys. We find a lower rate of false negatives (failures to report true receipt) in FoodAPS, likely partly due to the shorter recall period of FoodAPS.
Misreporting varies with household characteristics and between interviewers. Second, we examine geographic heterogeneity in survey error to assess whether we can extrapolate from linked data from a few states.
We find systematic differences between states in unconditional error rates but no evidence of substantial differences conditional on common covariates. Thus, extrapolating error rates across states may yield more accurate receipt estimates than uncorrected survey estimates.