Food security dynamics and measurement error

Abstract

We examine intra- and intergenerational food security dynamics in the United States using longitudinal data from the Panel Study of Income Dynamics (PSID) while accounting for measurement error. We apply recently developed methods on the partial identification of transition matrices and show that accounting for measurement error is crucial as even modest errors can dwarf the information contained in the data. Nonetheless, we find that much can be learned under fairly weak assumptions; the strongest and most informative assumption being that measurement errors are serially uncorrelated. In particular, although the evidence—both intragenerational and intergenerational—is consistent with significant mobility, we also find food security status to be persistent for at least some households in the tails of the distribution. We further document some heterogeneities in dynamics across households differentiated by race and education. Finally, the impact of measurement error in the context of underlying dynamics is widely applicable to other areas of applied microeconomics generally as well as to food security dynamics in less developed countries specifically.