This dissertation addresses two important issues in the literature estimating the Value of a Statistical Life. The first issue is the potential endogeneity bias in cross-section hedonic wage models. The second issue is the transferability of the VSL between different policy contexts. To address the first issue, we estimate cross-section and panel hedonic wage models to identify the bias due to the time-invariant worker heterogeneity. We also consider potential endogeneity bias due to measurement error associated with risk variable, time-variant worker heterogeneity and simultaneity between wage and risk in panel models. We obtain labor market data from the 1996 Survey of Income and Program Participation panel, and occupational fatal risk data from Scotton (2000). We find that the cross-section hedonic wage model is significantly biased upward due to unobserved time-invariant worker heterogeneity, but not from time-variant worker heterogeneity or simultaneity between wage and risk. Our results are sensitive to the inclusion of industry variables, but not sensitive to the sample of workers used in estimation. To address the second issue, we examine whether or not workers and firms differentiate heterogeneous risks to determine the risk-wage compensation levels. We focus on two very different fatal risks in terms of the degree of workers’ control over the risk and the degree of dread associated with risk: violent assaults and risks related to non-violent events. We use occupational drivers to mitigate potential unobserved heterogeneity of job characteristics and measurement error associated with risk variables. The labor market data comes from the basic CPS, and the occupation-geographic specific risk rates for each cause of death are created from the non-public Census of Fatal Occupational Injuries. We find that occupational drivers require larger compensation to accept a marginal increase of violent risk as compared to non-violent risk. This is true for both fatal and non-fatal risks. Our results are quite robust. This study suggests that current direct use of VSL obtained from hedonic wage studies in benefit estimation of various governmental programs should be reconsidered.
Available at: http://works.bepress.com/ikuho_kochi/2/