Testing the Stationary Search Model

Abstract

This paper uses tight parametric assumptions to model individual unemployment histories in a structural form. Starting from a simple search model it is shown that individual spells are distributed exponentially with different hazard rates for different individuals. In contrast to the usual reduced form approach the choice of the reservation wage is directly incorporated into the analysis. In effect, a log—linear simultaneous equations system is derived that explains both reservation wages and elapsed spell durations of currently unemployed individuals in terms of exogeneous variables. Moreover, it is demonstrated that apart from constant terms an identical structural form should be valid for accepted wages and completed spell durations for individuals who are re-employed at the time of the survey. The empirical part of the paper uses data from the Federal Republic of Germany to estimate the parameters of this structural form both for a subsample of unemployed and a subsample of re-employed individuals. Evidence on the performance of structural search models was previously provided only for currently unemployed individuals by the studies by Kiefer/Neumann (1979), Lancaster (1985), and Jones (1988). This study is therefore the first to provide evidence for re-employed individuals. Furthermore it is the first of its kind to investigate German data. In this paper the main emphasis is given to the robustness of the estimation results across subsamples, not to the formulation of formal tests for the validity of cross—subsample restrictions. The resulting evidence is mainly favorable, and the paper concludes with an extensive discussion of merits and remaining deficiencies of the proposed model.