About VSL

Overview

Society often faces choices that involve tradeoffs between physical risk and pecuniary returns, and our decisions reveal our willingness to trade money for the risk of physical harm (Ashenfelter 2006). "The ratio of the wealth we are willing to accept in exchange for a small change in the probability of a fatality is expressed in units of 'dollars per death,' or the dollar value of a fatality.  It is for this reason that this tradeoff is often called the value of a 'statistical' life" or VSL (Ashenfelter 2006, pC10).  VSL estimates are used for understanding and informing public policies where risk-reward tradeoffs are important (Kniesner, Viscusi, and Ziliak, 2006).

As pointed out by Fuchs and Zeckhauser (1987) valuing health and life as a commodity can be troubling because, "we cannot produce life years directly or buy them on any market. Goods that cannot be produced or sold often become priceless through a natural process that leads to interdependent utility considerations, great concerns for equity, probing discussions of obligations to the future, and assertions that these goods shall not be allocated through the market" (Fuchs and Zeckhauser 1987; pg. 267).  Therefore, the only way the value of life can be estimated is via indirect methods.  It is for this reason that the empirical literature calls it the value of a 'statistical' life (Ashenfelter 2006). 

Estimation Approaches

Surveys of the large VSL literature can be found in Aldy and Viscusi (2007), Krupnick (2007), Viscusi (1993), Viscusi and Aldy (2003). Viscusi (1993) outlines three indirect methods that can be used to estimate the VSL: a) survey methods b) observed non-labor market risk tradeoffs, and c) observed non-labor market risk tradeoffs.  The last two are the preferred methods and the focus of this website.  Both preferred methods usually hinge on estimation of a hedonic regression.

Survey Methods

Survey methods allow researchers to consider a broad range of risks, not just those directly observed on the job or within the environment. Survey methods have been used to investigate the mortality risks associated with pesticides (Evans and Viscusi 1991; Viscusi et al. 1987), risk of respiratory ailments (Viscusi et al. 1991; Krupnick and Cropper 1992), reductions in risk fatality rates (Jones-Lee et al. 1985), altruism (Viscusi et al. 1988), the affect of age on the VSL (Krupnick 2007), the benefits of landmine clearance in Thailand (Gibson et al. 2007), and Sweden's "Vision Zero" target of traffic safety (Hultkrantz et al. 2006) to cite a few.

Non-labor Market Studies

Measuring the VSL outside of the labor market uses non-labor market behavior to investigate their risk-reward tradeoffs.  The fundamental premise is to use an agent's decision to take a given action or not (a discrete choice) to reveal a lower bound on the implied VSL resulting from their actions.  This method has been used to investigate the VSL from highway speeding (Ghosh et al. 1975), the use of seatbelts (Blomquist 1979), willingness to pay to avoid hazardous waste sites and pollution (Gayer et al. 2000; Portney 1981), automobile safety (Atkinson and Halverson 1990), fuel economy standards (Dreyfus and Viscusi 1995), and the purchase of bicycle helmets (Jenkins et al. 2001) to cite a few.  An excellent study is Ashenfelter and Greenstone (2004), who use speed limit increases and their effect on fatalities and "time saved" to identify a VSL estimate of $1.6M. A similar methodology is used in Schnier Horrace and Felthoven (2008) on the discrete choice of a captain's decision to fish or not on a given day in the Alaskan crab fishery. The estimated VSL is $5M.  These methods rely on hedonic regression (link to hedonic regression page).

Labor Market Studies

Estimation of the VSL from labor market data uses the hedonic wage decomposition model (Rosen 1986; Thaler and Rosen 1976).  The model regresses observed wage rate on the occupational fatality rate and additional control variables to account for job- and agent-specific characteristics.  Additional variables have been added to the models, such as the job related non-fatal injury rate (which often is not statistically significant), workers compensation (Arnould and Nichols 1983), and annuity benefits (Moore and Viscusi 1988).  An excellent study is Kniesner, Viscusi, Woock and Ziliak (2006), who solve a variety of econometric problems with disaggregate panel data and estimate a VSL of $5.5-7.5M. In-depth surveys of wage decomposition models are Viscusi (1993) and Viscusi and Aldy (2003).  The estimated coefficient on the job related fatality rates is used to estimate the wage workers marginal rate of substitution between wealth and the laborer's occupational risk (Viscusi 1993; Viscusi and Aldy 2003). These methods rely on hedonic regression.

Estimation Issues

Empirical investigations of the VSL must control for omitted variable bias, endogeneity of fatality risks and heterogeneous preferences, to name a few. Omitted variable bias arises when the empirical model does not control for worker or safety-related productivity (Ashenfelter and Greenstone 2004). 'Heterogeneous preferences' refers to differing risk preferences of the agent being studied. Kniesner, Viscusi, Woock and Ziliak, (2006) demonstrate that many of these concerns can be mitigated with panel data.