ATTRITION IN PANEL SURVEY DATA AND THE ESTIMATION OF MULTISTATE LABOR-MARKET MODELS

In the analysis of labor market transitions. spells that are incomplet e due to panel survey attrition are treated as spell that are subject to independent right-censoring. It is likely, however, that the unobse rved characteristics affecting individual transitions on the labor mar ket are related to those affecting the attitude toward survey particip ation. In that case, the transition rate estimates are inconsistent. I n this paper we analyze the relation between durations spent in labor market states and the duration of panel survey participation, by expli citly modelling and estimating both stochastic processes. We use multi -state multi-spell models which allow for stochastically related unobs erved determinants.