Even "minimally buzzed" drivers are more often to blame for fatal car crashes than the sober drivers they collide with, reports a University of California, San Diego study of accidents in the United States .

Led by UC San Diego sociologist David Phillips and published in the British Medical Journal group's Injury Prevention, the study examined 570,731 fatal collisions, from 1994 to 2011.

The researchers used the official U.S. Fatality Analysis Reporting System (FARS) database for the study, because it is nationally comprehensive and because it reports on blood alcohol content (BAC) in increments of 0.01 percent. They focus particularly on "buzzed drivers," with BAC of 0.01 to 0.07 percent, and, within this group, the "minimally buzzed" (or BAC 0.01 percent).

Phillips and his co-authors find that drivers with BAC 0.01 percent - well below the U.S. legal limit of 0.08 - are 46 percent more likely to be officially and solely blamed by accident investigators than are the sober drivers they collide with.

The authors also find no "threshold effect" - "no sudden transition from blameless to blamed" at the legal limit for drunk driving. Instead, blame increases steadily and smoothly from BAC 0.01to 0.24 percent.

Despite this evidence, "buzzed" drivers are often not punished more severely than their sober counterparts. In practice, Phillips said, police, judges and the public at large treat BAC 0.08 percent as "a sharp, definitive, meaningful boundary," and do not impose severe penalties on those below the legal limit. That needs to change, Phillips said. "The law should reflect what official accident investigators are seeing."

The researchers measured blame by looking at more than 50 driver factors coded in the FARS database, including such "unambiguous" factors as driving through a red light or driving on the wrong side of the road.

Many of the study's analyses take advantage of what the authors call "a natural experiment": two-vehicle collisions between a sober and a drinking driver. "Because the two drivers collide in exactly the same circumstances and at exactly the same time," they write, "this natural experiment automatically standardizes many potentially confounding variables," including weather and roadway conditions.

The findings are unequivocal, Phillips said. "We find no safe combination of drinking and driving - no point at which it is harmless to consume alcohol and get behind the wheel of a car," Phillips said. "Our data support both the National Highway Traffic Safety Administration's campaign that 'Buzzed driving is drunk driving' and the recommendation made by the National Transportation Safety Board, to reduce the legal limit to BAC 0.05 percent. In fact, our data provide support for yet greater reductions in the legal BAC."

Phillips noted that, although federal agencies recommend reducing the legal BAC limit below 0.08 percent, there has been very little research on the dangers of driving at very low levels of BAC. "We appear to be the first researchers to have provided nationwide evidence on traffic accidents caused by minimally buzzed drivers," he said.

More than 100 countries around the world have limits set at BAC 0.05 percent or below.

In calling on all 50 U.S. states to follow suit, NTSB Chairman Deborah A.P. Hersman said in a statement last spring: "Alcohol-impaired crashes are not accidents. They are crimes. They can - and should - be prevented. The tools exist. What is needed is the will."