Estimating Compensating Wage Differentials with Endogenous Job Mobility

Abstract

We estimate compensating wage differentials for occupational fatality risk using administrative longitudinal matched employer-employee data from Brazil. Our method documents, and corrects for, the presence of bias from endogenous job mobility and nonrandom assignment of workers to firms that may be correlated with unobserved job characteristics. We find that changes in risk across jobs are correlated with changes in residual wages, so estimates that only control for unobserved worker heterogeneity are biased downward. Controlling for unobserved plant and job-match effects, while allowing for correlation between worker effects, plant effects, and risk, implies compensating differentials that are about 8 times larger than within-worker estimates, and lie between cross-sectional and within-worker estimates. The implied value of a statistical life (VSL) for prime-age male workers, after correction for endogenous mobility, is estimated to be 330,000 reais – equivalent to 42 years employed at the average wage. In addition, our data allow us to measure fatality risk within very detailed industry-occupation cells, alleviating concern about measurement error and aggregation bias that has been highlighted in recent research.