Do Vehicle Crash Tests Save Lives? Impacts on Market Decisions and Accident Mortality

Motor vehicles are tested for crashworthiness and the results made available to consumers by government and private organizations.  These programs could be welfare improving if they correct an informational market failure in which consumers would otherwise be unable to obtain adequate information about vehicle safety when choosing a vehicle.  In the absence of such a market failure, vehicle safety would be optimally provided even without crash test programs and there should be no response of automakers or consumers to the information provided by these tests.  This project investigates the response of automakers and consumers to the implementation of a frontal crash testing program by the Insurance Institute for Highway Safety (IIHS) in 1995 and evaluates the degree to which this program impacted vehicle safety and motor vehicle accident mortality.  Significantly more vehicles underwent substantial redesigns in 1995 than was usual before or after, suggesting manufacturers were preparing for the previously announced testing program.  In addition, there is no correlation between vehicle redesign year and crash test results for vehicles designed before 1995, but a significant trend toward improved results following implementation of the IIHS tests that year.  Data from the National Household Travel Survey suggests that consumers also responded to the testing program, with cars receiving a “Good” rating increasing in prevalence subsequently.  The impact of the testing program on motor vehicle mortality was evaluated by refining an estimation model developed in earlier work.  The model uses data from the Fatality Analysis Reporting System on all accidents in the United States from 1991 to 2011 in which at least one driver died.  Since only one of the drivers died in many of the accidents, it is possible to estimate the impact of vehicle and driver characteristics on the probability of driver death conditional on experiencing a severe accident.  The vehicle characteristic of interest, redesign year, is likely to be endogenous since manufacturers knew the timing of program implementation.  To correct for this, redesign year is instrumented with “expected redesign year,” which was determined by regressing time between redesigns on vehicle model fixed effects and using the regression results to predict redesign year.  The final model is estimated using maximum likelihood techniques and provides an estimate of the impact of vehicle redesign year on probability of driver death in a head-on collision after controlling for driver demographics; vehicle weight and class; airbag and seatbelt use; drunk driving; and vehicle make.  The relationship between driver death probability and redesign year mimics the crash test results closely.  There is no trend in death rates for vehicles redesigned before 1995 and a significant downward trend for those redesigned afterwards.  This trend predicts a 10% reduction in driver mortality in head-on collisions from 1991 to 2011, which is 20% of the reduction actually observed during that time.  Using the value of statistical life approach, this improvement would be valued at more than $500 per vehicle.