Empirical implications of alternative models of firm dynamics

This paper considers two models of firm behavior that allow for heterogeneity among firms, idiosyncratic (or firm-specific) sources of uncertainty, and discrete outcomes (exit and/or entry). Both the characteristics of firm level panel data sets and the nature of the questions we characteristically ask of them imply that models with these features are likely to underlie most detailed empirical work on microdata sets. The models we consider are a Bayesian learning model due to Jovanovic [20], and a model of research and exploration due to Ericson and Fakes [11]. The payer provides a simple set of empirical procedures to help determine which, if either, of these two models might be appropriate for alternative data sets. We conclude by checking for the implications of the two models on an eight-year panel of Wisconsin firms. The results indicate that one model is consistent with the data on manufacturing firms while the other is consistent with the data on retail trade. (C) 1998 Academic Press.