Car Safety By the Numbers

My husband was in a serious car crash in 2017. A distracted driver ran a red light and slammed into our passenger-side front fender. Our faithful 2000 Subaru Forester was totaled, but my husband walked away from the crash, unhurt.

Car safety was, not surprisingly, an important factor in our choice of a replacement vehicle. Which car makes and models are most likely to protect their occupants in crashes? And how do we know they’re safer?

The National Highway Traffic Safety Administration (NHTSA) website presents crash test ratings on different makes and models of cars. These are the “star” ratings prominently displayed on the window stickers at new car dealerships. Type in “2015 Kia Optima” and you will see a rating of 5 stars (the highest possible rating) for every category: overall, frontal crash, side crash, and rollover.

A NHTSA brochure explains that “more stars means safer cars”* and describes the controlled tests with crash dummies used to assign the ratings. The frontal crash test, for example, evaluates the seriousness of injuries to an average-size adult male dummy (driver) and a small-size adult female dummy (passenger) when the vehicle crashes into a fixed barrier at 35 mph. This represents a crash with a vehicle of similar weight, and, in fact, the NHTSA star rating for a vehicle can only be compared with other vehicles of similar size and weight (within 250 pounds).

The NHTSA star ratings evaluate injuries for body types similar to the crash test dummies. Female dummies are studied in the driver position for the side pole test, which simulates the effects of crashing into a telephone pole, but not for the other tests (in fact, female dummies were not used at all until 2011). But a small woman may have different frontal crash injuries when in the passenger seat (where she can sit as far back as she likes) than in the driver seat (where she must be far enough forward to reach the pedals). Persons of different genders and ages have different susceptibilities to injuries in a crash.

The Insurance Institute for Highway Safety (IIHS) rates vehicles as Poor, Marginal, Acceptable, or Good, again based on controlled crash tests. The 2019 Kia Optima receives a rating of Good for all six tests: three frontal crash tests, a side crash test, a roof strength test, and a head restraints and seats test; the 2015 Optima is rated Acceptable on one of the frontal crash tests and Good on the other tests.

The NHTSA and IIHS ratings are both based on controlled tests in laboratories. What happens in the real world, where you cannot control what type of vehicle runs into you?

The Auto Professor® rates cars using historical data on crashes involving fatalities instead of performance on crash tests. The 2015 Kia Optima receives a grade of B on the “academic” A-through-F scale: above average for safety, but not in the highest class. Moreover, the website gives different grades for different drivers: for a male or female driver over age 50, the 2015 Kia Optima receives a B-minus; for a female driver between the ages of 23 and 30, it receives an A-minus.

I asked Dr. Norma Hubele, Professor Emerita of Engineering at Arizona State University and the statistician who created The Auto Professor, about the statistical methodology behind the ratings. In brief, here is how it works.

Many fatal crashes involve multiple vehicles, where some of the drivers are killed and some survive. Dr. Hubele predicts the probability that a seat-belt-wearing driver** survives a crash as a function of the make, model, and year of the driver’s vehicle and of the other vehicle(s) involved in the crash. She then uses those probabilities to calculate a grade for each make, model, and year of car, and also separately for male and female drivers in different age groups when enough data are available.

The data-conscious car buyer will want to look at all three sets of ratings. Each measures different aspects of car safety.

The ratings by NHTSA and IIHS:

  • Are from crash tests done under controlled circumstances. Each vehicle is subjected to exactly the same crash tests. This allows direct comparison of how vehicles in a weight class protect the dummies in the crashes studied, without worrying about other factors such as driver skill, road conditions, or variations in the type of crash.

  • Are limited to the factors studied in the tests. One might extrapolate to conjecture that the 5-star Kia Optima will also protect female drivers or non-average-sized male drivers in frontal crashes, but those combinations are not tested.

The Auto Professor ratings:

  • Come from an analysis of crashes dating back to 2001 that involve fatalities. They thus reflect the types of conditions that are most likely to occur in fatal crashes, and include all types of impacts. They also include male and female drivers of all ages, so the ratings can be personalized.

  • Are from historical data instead of experiments conducted under controlled circumstances, so you cannot separate the characteristics of the car from those of the people who are likely to buy that type of car. Does the 2003 Ford Mustang receive a C-minus because of the car or because of the driving habits of the people who have bought the car?

  • Are limited to crashes involving at least one fatality, and thus do not consider crashes such as my husband’s in which no one was injured (which, after all, is the goal).

What did we decide to do? After looking at all the data, we bought a more recent model Forester (5 stars overall from NHTSA, a top safety pick from IIHS, and a grade of B from the Auto Professor) with forward collision warning and automatic emergency braking. And we walk and take public transportation whenever we can.

Copyright (c) 2019 Sharon L. Lohr


*Although the rating scale goes from 1 star to 5 stars, the Center for Auto Safety calculated from NHTSA’s September 2018 crash test spreadsheet that nearly all vehicles receive 4 or 5 stars for frontal and side crashes. NHTSA’s spreadsheets give the vehicle test details and the formulas used to calculate the probabilities of different types of injuries. Five stars are awarded if the probability of injury is less than 0.10 but some vehicles receiving 4 stars have probability of injury close to that threshold; a vehicle with probability of injury 0.101 receives 4 stars but has similar risk to a 5-star-rated vehicle with probability of injury 0.099.

**Why does she study only belted drivers? Non-belted drivers have much higher risks of injury or death in a crash. In 2017, 87 percent of the passenger vehicle occupants who survived a fatal crash were wearing a seat belt or in a child restraint device; only 53 percent of occupants who were killed were using restraints.

The type of analysis done by Dr. Hubele requires a huge amount of data and, sadly, with more than 640,000 fatalities from motor vehicle crashes in the US between 2001 and 2017 (a number approximately equal to the 2017 population of Las Vegas, Nevada), a lot of data exist. More than 37,000 people died during 2017 alone, an average of more than 100 fatalities for each day of the year. Almost 30 percent of those fatalities occurred in a crash involving a driver with blood alcohol level of 0.08 g/dL or higher; more than 10 percent involved a distracted or drowsy driver.

Sharon Lohr