On the Measurement of Job Risk in Hedonic Wage Models

Preprint
Publisher Website
Google Scholar
  • 1 January 2003
    • preprint
      • Published in RePEc
Abstract
We examine the incidence, form, and research consequences of measurement error in measures of fatal injury risk in U.S. workplaces using both BLS and NIOSH data. These data are commonly used in hedonic wage studies. Despite the fact that each of our measures of job risk purport to measure the same thing – the risk of a fatality while on the job – the various measures of job risk are not highly correlated, with the maximum correlation being 0.53. Indeed, many of the estimated value of statistical life estimates are negative. We find that the National Institute of Safety and Health’s industry risk measure produces implicit value of life estimates most in line with both economic theory and the mode result for the existing literature than other risk measures examined. Because we find non-classical measurement error that differs across risk measures and is not independent of other regressors, innovative statistical procedures need be applied to obtain statistically improved estimates of wage-fatality risk tradeoffs.

This publication has 0 references indexed in Scilit: