Data and Car Safety Debates

Figure 1 . Number of deaths from motor vehicle crashes, 1975 to 2021. Source: Fatality Analysis Reporting System, National Highway Traffic Safety Administration.

Approximately 43,000 people were killed in motor vehicle crashes on U.S. roadways in 2021, the highest level of fatalities since 2005 (Figure 1).

Why did fatality counts decrease so steeply between 1975 and 2010, why are they rising now, and what can you do to reduce your and your family’s risk of being injured in a crash? Dr. Norma Hubele’s new book Backseat Driver: The Role of Data in Great Car Safety Debates investigates what the data have to say about these questions.*

Hubele traces the history of the use of data to improve car safety, starting with a World War I pilot’s investigation of airplane and automobile design changes that could reduce injuries, and ending with a discussion of self-driving vehicles.

The book looks at major milestones in car safety improvement from a data’s-eye view. Hubele presents these milestones as debates about appropriate uses of data. For example, Chapter 5, on the roof crush resistance debate, discusses the datasets that different investigators used to study injuries in rollover crashes (the most lethal type of crash). All of the investigators used data originating from police accident reports, but reached different conclusions. How could that happen? Hubele explains that different investigators studied different subsets of the data and different sets of explanatory variables. She writes that while all the groups “started with police accident reports, each group analyzed very different sets of data in the end. It is not surprising that they arrived at contrary results.” Investigators from the National Highway Traffic Safety Administration used data from a probability sample of police accident reports in which specialized crash investigators documented the circumstances and injuries relating to the crashes. That data set included information on the number of quarter turns the vehicle made, and thus crashes in which the vehicle made just a quarter turn (and hence had no roof-to-ground contact) could be excluded from the analysis. Other investigators used larger samples of police accident reports, but these samples had no information on the number of rollover turns and hence those investigators could not distinguish single-quarter-turn from multiple-quarter-turn rollovers.

Other data debates in the book include effects of fuel-efficiency standards on car safety, how car safety ratings are determined (I wrote earlier about Hubele’s car safety rating system TheAutoProfessor.com, which gives car models a grade of A through F based on data from the Fatality Analysis Reporting System), and data used in car recall decisions. The last two chapters discuss a different use of data, where car manufacturers employ algorithms and massive databases to enable cars to make their own safety decisions through automated driver-assistance systems (Chapter 10) and self-driving cars (Chapter 11). Each chapter concludes with comments from the author, giving her personal opinion on the debate discussed in the chapter.

Backseat Driver: The Role of Data in Great Car Safety Debates tells the stories of the data debates that have helped address the systems problems of automotive safety. It also highlights the importance of ongoing statistical investigations and research to help reduce the tragic losses from motor vehicle crashes.

Footnotes

Figure 2. Number of deaths from motor vehicle crashes per 100 million vehicle miles traveled, 1975 to 2021. Source: Fatality Analysis Reporting System, National Highway Traffic Safety Administration.

*Chapter 4 of Backseat Driver: The Role of Data in Great Car Safety Debates discusses the importance of looking at crash fatality data in multiple ways. For example, in Figure 1, I displayed the total number of fatalities from motor vehicle crashes from 1975 to 2022, but that graph does not account for changes in population or traffic over the time period. The number of deaths per 100 million vehicle miles traveled dropped slightly from 2020 to 2021 (Figure 2; the rate dropped even though the count increased because people drove more in 2021 than in 2020), but the 2021 rate, like the 2021 count, is close to levels last seen in 2007. Hubele also discusses additional ways of calculating fatality rates from motor vehicle crashes, and how each gives a different perspective on the data.