Activist “Researchers” Use Biased Study to Call for Lower BAC LevelsJune 27, 2011
Washington, D.C. — The American Beverage Institute (ABI) criticized a new study, which claims that drinking even trace amounts of alcohol increases a person’s risk of being severely injured in a car crash. The University of California, San Diego demographers behind the study — David Phillips and Kimberly M. Brewer — are using their findings to call for lowering the legal blood alcohol concentration (BAC) limit, which is currently .08 across the United States.
ABI’s managing director Sarah Longwell responded, saying, “Traffic safety officials should take a much closer look at this study before using it as a source for policy prescriptions. The fact is, these researchers have a public policy agenda — seeing BAC limits lowered below .08 — and they use a skewed dataset in order to get results favorable to that agenda. Using Fatality Analysis Reporting System (FARS) data to determine the relationship between small amounts of alcohol and serious injury in car crashes is like surveying for steroid usage by using major league baseball players as your dataset.”
There are two fundamental problems with the study:
1. Cherry-picking the dataset guarantees the worst possible outcome.
The only traffic accidents included in their dataset are accidents where someone died. But it’s absurd to say that this small number of accidents is at all representative.
For instance, there were an estimated 10.2 million motor vehicle accidents in 2008. This study cherry-picks the 34,712 accidents where there was a fatality — that’s about 0.3% of all accidents that year — and attempts to draw conclusions about what happens to a person who has a glass of beer at a ball game and then drives home.
Accidents happen far more frequently to people who are sober. Census Bureau data from 2008 show that 63% of all traffic fatalities happened when the driver’s blood alcohol content was 0.0; another 32% happened when the driver’s BAC was at .08 or higher. Only 6% of traffic fatalities occurred when the driver’s BAC was between .01 and .07.
2. Confusing correlation and causation leads to flawed conclusions
The study claims that there is a significant jump in the severity of driver injuries between a BAC of 0.0 and a BAC of .01. Yet this trend doesn’t continue; for instance, at the legal limit of .08, the severity ratio of driver injuries is lower than at .01.
That’s not the only inconvenient fact the authors don’t highlight. Their signature finding — that injury severity is higher at a BAC of .01 — breaks down when they look at the Western part of the country. In this case, the severity ratio is significantly higher for drivers with a BAC of 0.0 than it is for drivers with .01.
Anomalies like these suggest that the higher injury ratio at low levels of BAC is being driven by other factors that the authors haven’t accounted for. “Correlation is not causation” – it’s a lesson that’s taught in Statistics 101, but these authors appear to have discounted it.