Compensating Wage Differentials and Unobserved Productivity

It is well known that the inability to observe workers' full labor market productivity can bias estimates of compensating wage differentials. This paper attempts to determine how serious this bias is likely to be. It adopts a stochastic framework of workers' tastes over job attributes and models their equilibrium wage-job attribute choices. Workers' productivity is assumed to consist of observed and unobserved components. Applying the standard estimation methodology, we find that the degree of bias can be surprisingly large. On the basis of our analysis, we conclude that contemporary labor market studies are likely to severely underestimate workers' willingness to pay for job attributes. This has implications for a number of applications of compensating wage differentials, including value of life studies.