Evaluation of the Population Distribution of VSLs by Combining New Vehicle Safety Estimates with a Model of Vehicle Choice

Many of the studies estimating the value of statistical life (VSL) use labor market estimates that may apply best to a subset of the population that is relatively homogeneous in terms of income and other characteristics.  Since many households choose to own automobiles, the safety implications of this choice can be used to estimate the distribution of VSLs and its correlation with income and other demographic variables using broad support in the explanatory variables.  A recently developed model of automotive safety (Economic Inquiry 53(3): 1606-29) uses Fatality Analysis Reporting System (FARS) data to estimate the level of safety of vehicles at the model by model year level.  Specifically, the probability of someone dying in a particular vehicle over the course of a year is calculated.  This probability is a complex function of vehicle weight, class, manufacturer, vehicle age, mean number of vehicle occupants, age of vehicle occupants, and number of miles driven in a year.  The safety estimates are combined with information on vehicle costs and data from the National Household Travel Survey (NHTS) to calculate the marginal cost of saving a statistical life for each vehicle type for a particular household.  The NHTS provides data from a random selection of households about household characteristics, the vehicle(s) owned by those households, and the way in which these vehicles are used.  The safety cost variable is included in a regression model of vehicle choice with a rich set of control variables to impute a VSL for each household in the NHTS.  The resulting distribution of VSLs provides evidence about the variation in risk preferences within the population and the correlation of these preferences with income and other demographic variables of interest.