Layoffs and duration dependence in a model of turnover (Q1069571)

This paper addresses the problem of turnover in the labor market, and develops a semi-Markov model in this context. This relaxes the traditional Markov model assumption adopted in the most recent literature on the subject, in the sense that while the transition rate from unemployment (state 1) to employment still does not depend on the duration of unemployment (as in the Markov model), the reverse transition rate does depend on the duration of the current spell of employment. The authors argue that the basic reason for this dependence is that layoffs are often generated on a ''first in, last out'' principle, and they show that an individual worker's reaction to this constraint generates duration dependence in employment transition. Following some theoretical analysis on this two-state semi-Markov model, the authors develop some predictions and present some preliminary results based on the Denver income maintenance experiment data set. These results indicate that the particular semi-Markov model used in the paper is closer to reality than the Markov model used heretofore.