Government Data on Computer-Assisted Driving Car Crashes ‘Raise More Questions Than They Answer’

The first-ever data release doesn’t tell us anything new, but may signal a first step towards real oversight for systems manufacturers insist are safe.
Tesla crashed in woods
ROBIN VAN LONKHUIJSEN / Stringer via Getty
Screen Shot 2021-02-24 at 3
Moveable explores the future of transportation, infrastructure, energy, and cities.

For the last year, the National Highway Traffic Safety Administration, an agency within the Department of Transportation, has been collecting data from vehicle manufacturers on car crashes involving computer-assisted driving systems, known in the industry as Automated Driving Assistance Systems (ADAS). This includes Tesla’s Autopilot and Full Self-Driving (which is not self-driving) as well as similar systems available from most manufacturers these days. 


NHTSA released the first set of data on Wednesday, which is a landmark in federal oversight over this still-nascent technology, because until now nobody had any idea how safe these systems are. This data does not answer that question, but it does get us a step closer.

NHTSA received 392 reports of crashes involving ADAS systems, defined as any crash where such a system was active within 30 seconds of the crash, and 273 of them were in Teslas. The crashes were self-reported by the manufacturers themselves. Most of those reports were generated based on vehicle telematics data, but 35 percent were from complaints or claims submitted by motorists. In most crashes (294), NHTSA doesn’t know if anyone was injured as a result, but 11 of the 98 crashes where injury information was available had a serious injury or fatality. And the cars crash into all types of stuff: other cars, “other fixed objects,” animals, poles and trees, buses, trucks, vans, three pedestrians and one cyclist. In 146 cases, NHTSA doesn’t know what the car crashed into.

To put these numbers into perspective, there are approximately 5.2 million car crashes in the U.S. per year, according to the Bureau of Transportation Statistics, although that’s just an estimate because nobody officially tracks that. At least 36,500 drivers crash their cars into buildings every year, but that’s just an estimate by an independent researcher because nobody tracks that either. Even if someone died in each of the 392 ADAS crashes, it would account for less than one percent of the 42,915 estimated road deaths last year in the U.S. 

For now, there are too many issues with the ADAS crash data to draw any conclusions, something NHTSA acknowledges. NHTSA head Steven Cliff told reporters, “The data may raise more questions than they answer.” Among the problems with the data: There is likely some double-counting, there are surely many crashes that never go reported, manufacturers with more centralized data harvesting capabilities (like Tesla) will be more aware of crashes than manufacturers that don’t regularly analyze car data, and none of the data is weighted for how many ADAS-capable vehicles are on the road for each manufacturer, not to mention the percentage of time those systems are actually active.

Nevertheless, the simple act of acknowledging this a problem under NHTSA’s purview is a significant and noteworthy break from NHTSA’s past, where the agency more or less took everything auto manufacturers said about the safety of these systems for granted.