Stability of Estimates of the Compensation for Danger

Estimates of the extra earnings for jobs with higher risks of death can be used in cost-benefit studies involving risk changes. Because of this use, the magnitude and stability of the estimated coefficient are important. Part of the current study closely reproduces a widely quoted 1982 study by Marin & Psacharopoulos to check on the stability. We also examine the robustness of the estimate to the inclusion/exclusion of non-fatal risks and other relevant characteristics. While the magnitude of the co-efficient has increased threefold from the earlier study, the coefficient is robust to other changes in the specification. There could be selectivity bias in the estimates of the extra return because people can select their occupation on the basis of its riskiness. Our findings suggest that one common technique to deal with selectivity bias in a continuous variable can give unreliable results in practice.