Estimating Compensating Wage Differentials with Endogenous Job Mobility

Abstract

We evaluate a strategy for using matched employer-employee data to estimate compensating wage differentials. We apply this strategy to study occupational safety using data on the universe of jobs in Brazil from 2003–-2010. In these data, common panel based approaches actually amplify downward bias from nonrandom assignment to jobs. We argue that when firms offer differentiated jobs, the direct effect on wages of moving between jobs with different amenities can be identified by conditioning on unobserved establishment-level heterogeneity. Our preferred model approximates the wage structure well, while relaxing the key assumption that mobility between jobs with different amenities is conditionally exogenous