Mismatch Cycles∗

Using a recently developed worker-occupation mismatch measure for the US labor market, we document that mismatch is pro-cyclical. However, there is substantial heterogeneity by previous employment status: mismatch is pro-cyclical for workers in ongoing job relationships and job-tojob transitions, consistent with the cleansing effect of recessions; while mismatch is counter-cyclical for new hires from unemployment, consistent with the sullying effect of recessions. Our empirical findings show that the cleansing effect dominates. We then provide new evidence that conditional on mismatch, business cycle conditions at the start of the match are important in explaining variation in job duration. We explain the observed pattern through a model of learning about workers’ mismatch, in which the learning technology varies over the business cycle.