Unemployment Insurance and Job Search Behavior*

Abstract

Unemployment insurance (UI) can affect unemployment duration and re-employment wages, through various dimensions of unemployed workers’ job search behavior. We shed light on the effects of UI on job search behavior using new longitudinal data: we link administrative registers to data from a major online search platform, and track the job applications sent over the unemployment spells of about 500,000 French workers. We identify changes in individual search behavior caused by UI around benefits exhaustion, after accounting for changes in the sample composition—i.e. dynamic selection—and for the effect of the time spent unemployed—i.e. duration dependence. We show that search effort (the count of job applications) increases by at least 50% during the year preceding benefits exhaustion and remains relatively high thereafter. The target monthly wage decreases by at least 2.4% during the year preceding benefits exhaustion, and remains relatively low thereafter. We document particularly large dynamic selection around benefits exhaustion, as some workers increase their search effort more before exhaustion and find a job faster. We also show evidence for duration dependence: workers decrease their target wage by 1.5% over each year of unemployment, irrespective of their UI status. Overall, the effect of UI on individual search behavior is consistent with the predictions of search models and constitutes evidence against search-free models where UI merely subsidizes leisure.