Firm-level evidence on the formation and durability of employment relationships.

Abstract

This dissertation uses firm-level data to investigate how employment matches are formed and how sensitive these matches are to aggregate fluctuations. The first essay investigates how the matching process between workers and firms affects empirical estimates of wage growth. Using data from the Ford Motor Co. and the A. M. Byers Co. the dissertation shows that short-career workers enjoy similar early-career wage growth as workers with longer careers. This finding provides evidence against the hypothesis that poorly matched workers leave firms sooner because their wage growth reflects substandard productivity. The second and third essays concern the durability of employment relationships over the business cycle. Previous research has contended that job destruction is more cyclically sensitive than job creation. The second essay demonstrates, however, that the sensitivity of destruction is overstated by heavy reliance of previous work on manufacturing data. Evidence from Michigan during the 1980s shows that in most other sectors, job creation has the higher variance. Furthermore, a predictor of how much creation varies relative to destruction is the trend in sectoral employment growth. This finding is consistent with a simple (S,s) model of employment demand augmented with a non-zero trend growth rate in desired employment. The large time-series variance in manufacturing job destruction probably reflects the asymmetric response of manufacturing employment to aggregate shocks, brought about by the bunching of manufacturing firms close to the job-destruction border of the (S,s) region (and far from the job-creation border) during periods of declining manufacturing employment. The third essay estimates a micro-level model of employment adjustment to determine whether firms prefer a few large adjustments to continuous small ones (as an (S,s) model would imply), and whether idiosyncratic dispersion in desired growth rates across firms changes in recessions. Evidence is found that large adjustments are preferred to small ones, but the essay finds no evidence that idiosyncratic dispersion in desired growth rates either rises or falls over the business cycle.