Economists have long employed hedonic wage analysis to estimate income-fatality risk trade-offs, but some scholars have raised concerns about systematic measurement error and omitted variable bias in the empirical applications of this model. Recent studies have employed panel methods to remove time-invariant individual-specific characteristics that could induce bias in estimation. In an analogous manner, this paper proposes to exploit assortative matching on risk attitudes within married couples to control for worker characteristics that are unobserved to the econometrician. I develop and implement a modified hedonic wage estimator based on a within-coupled differenced wage equation for full-time working married couples with the Current Population Survey Merged Outgoing Rotation Group over 1996-2002. The key assumption builds on the findings in the assortative matching literature that individuals often marry those who have common traits across many dimensions, including those that may influence worker wages and are correlated with observed occupational fatality risks. This estimator identifies the compensating differential for occupation fatality risk by using within-couple differencing to remove unobserved determinants of risk attitudes and risk-mitigation ability, on which couples match, from the error term. I find that the value of statistical life (VSL) varies from $9 to $13 million (2016$). The within-couple differenced VSL estimates are stable and more robust to variation in specification of the hedonic wage model than conventional, cross-sectional hedonic wage models. I also find that the value of statistical life takes an inverted-U shape with respect to age.
Read full abstract