Driver death rates by make and model

One way to compare the relative safety of vehicles is to look at driver death rates. From time to time, IIHS researchers compute the rate at which drivers of individual models from a given year are killed in crashes.

Rates are given as the number of driver deaths per million registered vehicle years. (A registered vehicle year is one vehicle registered for one year.) To increase the exposure and thereby improve the accuracy of the calculations, results are included for the previous three model years if the vehicle wasn't substantially redesigned during that time. These calculations take into account only deaths of drivers, not passengers, since every vehicle that crashes has a driver, but not every vehicle has passengers.

Death rates are adjusted for driver age and gender. Results for 2008 vehicles are based on a different statistical model that also takes into account calendar year, vehicle age and vehicle density at the garaging location. Information on deaths is from the National Highway Traffic Safety Administration's Fatality Analysis Reporting System. Data on vehicle registrations come from IHS Markit. Some additional information used to compute the 2008 model year results comes from HLDI's database.

The numbers in parentheses next to the overall death rate are 95% confidence bounds. The rollover death rate represents deaths in single-vehicle rollover crashes and is a subset of the single-vehicle crash death rate. For the model year span, the number of model years included depends on when a particular vehicle was redesigned.

Why can't I find driver death rate data for new models?

It takes considerable time to gather and tabulate the real-world data needed to provide statistically significant results for new models. Complete vehicle registration data for each model year typically are released about two years later, and data on fatalities are first available approximately nine months after the end of the calendar year. For vehicles that have not been fundamentally redesigned, previous model year results are good predictors of the current model's experience.