A Forest Fire Theory of Recessions and Unemployment

Abstract

We show that a class of models of matching from firms' perspectives (that nest most standard search models) necessarily exhibit specific dynamic patterns of employment. Despite the prevalent use of models within this class, these patterns have not been noted before. Focusing on labor search as a leading example, in presence of search frictions, firms prefer to hire low productivity workers in high-demand states to take advantage of current profit opportunities. This implies that bad matches accumulate during expansions, leading to drops in employment, and other inputs, that are positively related to the length of the expansion when a negative shock eventually hits: the longer since the last contraction, the larger the absolute and percentage drop in the employment of labor and capital inputs.We also show that this phenomenon is present in observed dynamics, using US micro-data. We find that five additional years of expansion lead to an additional drop of thirty percent in employment growth at the onset of a downturn.