#264 Do Vehicle Safety Inspection Programs Lead to Lower Fatality Rates?

Principal Investigator
H. Scott Matthews
Start Date
Oct. 1, 2018
End Date
June 30, 2020
Research Type
Grant Type
Grant Program
FAST Act - Mobility National (2016 - 2022)
Grant Cycle
2018 Traffic21


States require inspections on vehicle safety components to be performed with varying frequencies and on various subsets of the fleet.  In Pennsylvania, every passenger vehicle is inspected annually. Stakeholders have called for modifications or elimination of safety inspection programs.  However, inspection data have not been available, so efforts to improve programs have been challenging.    In short, discussions of the continued existence of state inspection programs are hampered by a lack of data-driven analysis to show their effects and benefits.

This study will focus at a high level. We will create a data framework that contains historical data on the numbers of fatal crashes related to passenger vehicles for all states, as well as estimates of vehicle miles traveled for all states in urban and non-urban areas. We seek to find fatality rates for all states at the overall, urban and non-urban levels.  With these fatality rates we will do statistical analysis to associate those fatality rates with the existence of safety inspection programs in the states, to see whether there are statistically significant different fatality rates in states with and without safety inspection programs. We will do similar analysis to see whether there are significant differences with and without emissions inspection programs (because even though they dont inspect safety components, they tend to get older cars off of the road).

Our goal is an unbiased study that could help inform stakeholders and the policy process in Harrisburg and other states seeking to make changes to their programs.
Various parties within Pennsylvania and other states have called for significant modifications and/or elimination of their passenger vehicle safety inspection programs.  Such programs call for specified safety inspection tests to be performed on automobiles and light trucks with varying frequencies (e.g., annually) and on various subsets of the fleet (e.g., exempting new cars). But state legislators have been seeking to eliminate these programs because there is a  perception that they are very costly to consumers, and provide little or no benefits to society or to parties other than automobile repair business owners, who earn revenue from fixing vehicles that fail tests. Behind these perceptions are sentiments that "cars have never been safer", yet very little data exists so very few analytical studies have been able to prove this.  

The handful of past studies completed have been very high level analyses of whether states with safety inspection programs have higher crash or fatality rates which provide at best indirect measures of effectiveness. These studies were also done about 20 years ago with data that is almost 30 years old. The datasets used for counting fatalities, VMT, etc., have all been significantly improved over time. That said, these "old studies" are still those that are dominating state-level discussions, and have led to various states eliminating their programs in the past few years.

We seek to do the first study of this issue with the latest generation of publicly available datasets. We will use data from the US DOT’s Fatality Analysis Reporting System (FARS), which has information on every fatal crash in the United States for more than the past 10 years. This dataset alone is a significant improvement alone to what was available previously. Amongst the obvious explanatory variables such as location of the accident, make-model-years of vehicles, and numbers of occupants and fatalities), FARS also has fields that note whether vehicle safety components (such as brakes or tires) were primary or contributing factors in the fatal crashes. We stipulate, though, that preliminary observations of these "component-cause" data fields are quite sparse, since the time and effort to make such determinations may not be possible given the need for police and emergency responders - and thus robust results from this component-level analysis may not be fully realized in the final work.

We thus will focus on the highest level safety metrics, such as estimating fatality rates (e.g., fatalities per vehicle mile traveled) for the overall, urban, and non-urban levels for each state. In line with these estimated rates we will associate states that do and do not have safety inspection programs, so that statistical models can be created that seek to show whether there are statistically significant results that can be shown, i.e., lower fatality rates for states with safety programs. We also will consider whether emissions testing programs show any statistically significant results, as they also tend to remove older cars from the fleet.

Beyond the high-level results, we will attempt to create risk-based measures of expected fatalities as a function of failures in safety components (i.e., accidents caused by brake or tire failures per million vehicle miles driven). 

Our goal is an unbiased study that could help inform stakeholders and the policy process in Harrisburg and other states seeking to make changes to their programs.

We note that we began a project similar to this in 2016, but had to stop partway through because the funded post-doc unexpectedly left campus. We had created a preliminary data framework and thesis chapter which were submitted for publication, but it was not accepted. We received greatly beneficial comments on changes we will need to make this time around in improving the model - and thus plan to engage a PhD students in Statistics working on their "first year research project" to complete the project. We have assembled a team of PIs with past experience working together, and who can help to recruit and attract the right student for this project. We have no reason to doubt whether the project could be completed by the end of 2019 (even if it does not show the results we expect).

Compuspections, LLC–We have been working with the CEO and Director of Compuspections LLC for nearly 5 years.  Compuspections is a small PA business that provides inspection record management and reporting software for inspection stations (ranging from small garages to large dealerships). Records are extremely comprehensive, including information such as all four actual tire tread measurements (in units of 1/32 of an inch) at time of inspection, all maintenance work done to meet state inspection program requirements, labor and material costs, final pass/fail status, etc.  In short, this data can fill the gap identified above in terms of clearly noting what happens in the workflow of a safety inspection from time of entry to point of exit from the station and not simply whether the vehicle passes when leaving.  Compuspections will continue to provide us with a large amount of in-kind expertise (their CEO has owned an inspection station for 20 years).  Past efforts to justify or modify the inspection program by the various stakeholders involved have been limited since their motivations have been questioned (e.g., by stations to maintain the current system and associated revenues).  A primary goal of this effort is to have a study with multiple data sources but authored by researchers from CMU, a neutral third party. Our partnership with Compuspections has already provided us with 10 million inspection records over 10 years and helped us to build relationships with other interested parties such as PennDOT, various state legislators, AAA, Pennsylvania Automotive Association (PAA), etc.  
October 1, 2018 - Preliminary Discussions with prospective students in Statistics ADA PhD program
January 1, 2019 - Beginning of Project with chosen student(s)
May 15, 2019 - Beginning of "full time" effort by students (in summer, without classes)
August 15, 2019 - Interim results available from student effort
Oct 1, 2019 - Revision of interim results
December 31 2019 - Final report submitted (and journal paper)    
Deployment Plan
Expected Accomplishments and Metrics
Fatality Rates for All States

Component-Level Rates

Final Report to Mobility 21

Journal Paper (modified from initial version) submitted to an outlet such as "Accident Analysis and Prevention" (in progress, July 2020).  

We are also working on a two-page "policy brief" document of this project to be used to communicate the results to diverse stakeholder audiences.    

Individuals Involved

Email Name Affiliation Role Position
pf12@andrew.cmu.edu Fischbeck, Paul SDS/EPP Co-PI Faculty - Tenured
hsm@cmu.edu Matthews, H. Scott CEE/EPP PI Faculty - Tenured
cschafer@cmu.edu Schafer, Chad Statistics and Data Science Co-PI Faculty - Tenured


Amount of UTC Funds Awarded
Total Project Budget (from all funding sources)


Type Name Uploaded
Presentation 2018_-_Template_for_Fatality_Rates_Mobility_21_Project.pptx May 6, 2019, 4:18 a.m.
Data Management Plan Data_management_Plan_for_Mobility21_Fatality_Rates.docx Sept. 9, 2018, 8:21 a.m.
Publication HKSTS19_64_6Oct2019.doc April 6, 2020, 2:45 p.m.
Presentation hksts-safety-inspection-statistics-2019.pptx April 6, 2020, 2:45 p.m.
Progress Report 264_Progress_Report_2020-03-30 April 6, 2020, 2:45 p.m.
Final Report TSET_264_Fatality_Rates_Final_Report_with_Appendix.pdf July 16, 2020, 4:22 a.m.

Match Sources

No match sources!


Name Type
CITA Deployment Partner Deployment Partner