Crash rates for self-driving cars are lower than the national crash rate of conventional cars, according to a new report out of Virginia Tech Transportation Institute that was commissioned by self-driving car developer Google.

While self-driving cars have a rate of 3.2 crashes per million of miles, traditional driver-operated cars have a rate of 4.2 crashes per million miles, according to the study that adjusted the data for unreported crashes and takes into account accident severity.

The authors note that the data also suggest that conventional vehicles may have higher rates of more severe crashes than self-driving cars, but there is insufficient data to draw this conclusion with strong confidence given the small overall number of crashes for the self-driving cars.

However, there is “statistically-significant data” that suggest less severe events may happen at significantly lower rates for self-driving cars than conventional vehicles, according to the authors.

Additionally, the research report notes of the data used, none of the vehicles operating in autonomous mode were deemed at fault.

Up until now, comparisons based on existing data have been incomplete as requirements in each state for police reported crashes differ, and the majority of severe crashes that go unreported, not the authors. Meanwhile, self-driving cars are required to report every crash, no matter how minor. Thus previous studies have attempted to analyze self-driving car data, which has a full record of all crashes, relative to the current vehicle fleet, which has an incomplete record of crashes. “The comparison is, as the old saying goes, apples to oranges,” the authors state.

To better estimate existing crash rates, the new report, Automated Vehicle Crash Rate Comparison Using Naturalistic Data, examines national crash data as well as data from naturalistic driving studies that closely monitors the on-road experience of 3,300 vehicles driving more than 34 million vehicle miles. It then compares the results to data from Google’s Self-Driving Car program, which includes written reports, video and vehicle kinematic data that deals with the motion of objects.

The study was conducted by the Virginia Tech Transportation Institute but sponsored by Google. The report includes a disclaimer that its final opinions and conclusions are those of the institute and not necessarily those of Google.

A January 2015 study from the University of Michigan and the Sustainable Worldwide Transportation consortium of researchers said that it is not clear that a self-driving vehicle would ever perform more safely than an experienced, middle-aged driver and that during the transition period when conventional and self-driving vehicles would share the road, safety might actually worsen, at least for the conventional vehicles. However, the researchers noted two important caveats to its work: self-driving cars have traveled far few miles (1.2 million) than conventional cars ((3 trillion) and self-driving cars have been in less challenging (no snowy) road conditions, therefore their exposure is not yet representative of that for conventional vehicles.

The Michigan researchers found that accident rates are higher for self-driving cars on the road with regular cars, although driverless vehicles have never been at fault. The driverless cars are usually hit from behind in slow-speed crashes by humans unaccustomed to machine motorists that always follow the rules and proceed with caution. The Michigan study also found that the severity of injuries in self-driving car crashes is lower than it is in crashes involving conventional vehicles.