Measurement error and the estimation of gross flows from longitudinal economic data

Abstract

The use of longitudinal survey data in economic research is considered with special reference to the estimation of gross flows. Alternative models of measurement error for discrete variables are reviewed. A common assumption is that errors on different occasions are independent. Two reasons for departures from this assumption are unit heterogeneity, where different units have different misclassification rates, and state heterogeneity, where different rates apply to individuals in different underlying states which partition the observed states. The effect of assuming independent errors in the presence of heterogeneity is considered. Approaches to adjusting for measurement error in gross flow estimation are reviewed.