Estimating Dynamic Models from Repeated Cross-Sections

A major attraction of panel data is the ability to estimate dynamic models on an individual level. Moffitt (1993) and Collado (1998) have argued that such models can also be identified from repeated cross-section data. In this paper we reconsider this issue. We review the identification conditions underlying alternative estimators and present an instrumental variables type estimator that is consistent under a relatively weak set of conditions. These conditions, however, are not trivially satisfied in applied work.