Estimating the errors in hours of work and wage rates
Labor supply equations are often estimated by regressing hours worked on a wage measure which is itself the quotient of income and hours, leading to downward biases. The author shows that where both income and hours of work are measured with error, the downward bias remains. Using data collected for the RAND Health Insurance Study, this paper presents evidence that hours data are inherently variable, even when reporting error is low. Using data on 350 individuals' wage information for the preceding week and for a "usual week" in the preceding four months, a considerable negative bias is seen in the wage elasticity of both annual and weekly hours with respect to hourly wage, when data are derived from report week as compared with the "usual" week measures.