A data-driven assessment of the efficacy of vehicle safety inspection programs.
Author(s): Acharya, Matthews
To ensure adherence to safety requirements, many jurisdictions administer vehicle safety inspection and maintenance (I/M) programs. In the United States, I/M programs (managed at the state-level) have received significant political opposition over the last half-century, leading to the abolition of nearly half the 32 I/M programs that were present in the mid-1970’s.
Much of this opposition posits that inspection programs are not an effective means of preventing road fatalities and are therefore a waste of time and money for vehicle-owners and administrative authorities. As jurisdictions continue to debate the need for I/M programs, it is imperative that statistical analyses be conducted to assess the effectiveness of safety inspection programs, in achieving their stated aim of preventing road fatalities.
The primary aim of this study is to use publicly-available historical data to quantitatively analyze the correlation between the presence or absence of vehicle safety inspection programs, with motor accident fatality rates, while controlling for improved vehicle safety over time, as well as such factors as urban vs. rural driving, accidents due to inclement weather, driving under the influence, etc.
The U.S. National Highway Traffic Safety Administration (NHTSA) administers a database through the Fatality Analysis Reporting System (FARS), which has, since 1976, collected data on motor-vehicle accidents anywhere in the United States, which resulted in at least one fatality. FARS contains information on each accident (location, weather and road conditions, number of vehicles involved, etc.,), in addition to information on individual vehicles involved in these accidents (make and age, vehicle-related contributing factors to the accident, speed, etc.,), as well as demographic information of drivers, passengers and pedestrians. This study will use FARS data from 1976-2018, to develop a time-series panel data model (fixed or mixed effects) for the population adjusted fatality rate (i.e., fatalities per 100,000 population in each state), with at least one regressor corresponding to the presence or absence of inspection programs.
Initial iterations of a pooled-data time-series regression show, with high confidence, a significant negative correlation between the dependent variable (i.e., fatalities per 100,000 population in each state) and a binary variable corresponding to whether or not the state had an I/M program two years prior to the year being assessed (i.e., a lagged variable).
The preliminary findings point to a measurable and statistically significant reduction in fatality rates from the implementation of safety inspection programs in the U.S. This data-driven finding can inform discussions and policymaking at a time when I/M programs are being scrutinized in statehouses across the country. By using publicly available data and providing free access to all the methods and code used to conduct these analyses, the authors intend to make this study fully reproducible.
We wish to show that data-driven analyses of historical trends can prove useful to inform policy decisions, and may even be used to model the impact of specific policy changes, such as the implementation or repeal of an inspection program. Furthermore, this study's findings support a hypothesis that safety inspections are effective at their stated aim of reducing road fatalities.