How Drunk Is Too Drunk to Drive?

Our podcast “The Suicide Paradox” featured sociologist David Phillips, who spoke about his research on copycat suicides (a phenomenon he calls “the Werther Effect”). More recently, Philips has been studying drunk driving. Particularly, he’s been looking at drivers who are merely “buzzed” — with 0.01 percent blood alcohol concentrations (BACs) — and has found that the severity of life-threatening motor vehicle accidents increases significantly at BACs far lower than the current U.S. limit of 0.08 percent. In an email, Philips describes his latest research on buzzed drivers:

My current research, just published in Injury Prevention, shows that even minimally buzzed drivers (with BAC=0.01%) are 46% more likely to be blamed for an accident than are the sober drivers they collide with. This indicates that there is no safe level of alcohol for drivers: any amount of alcohol markedly increases the risk to drivers and their passengers. We reached this conclusion after examining an official, U.S. dataset of more than 570,000 car crashes. The findings have implications for drivers, passengers, police, judges, lawyers, insurance companies, advocacy organizations (like MADD) and regulatory agencies.

The paper, “Official blame for drivers with very low blood alcohol content: there is no safe combination of drinking and driving” (abstract; PDF), is co-authored with Ana Luiza Sousa and Rebecca Moshfegh. The researchers used an official nationwide U.S. database (Fatality Analysis Reporting System) for the years 1994–2011. They found that there’s “no sudden transition from blameless to blamed drivers at BAC=0.08% (the U.S. legal limit). Instead, sole official blame increases smoothly and strongly with BAC (r=0.98 (0.96–0.99) for male drivers, p<0.000001; r=0.99 (0.97–0.99) for female drivers, p<0.000001). This near-linear SOB-to-BAC relationship begins at BAC=0.01% and ends around BAC=0.24%.”

Andrew B

I enjoy your site and podcasts but am frequently amazed at you bringing in articles with such a weak scientific basis. Correlation does not mean causation. Maybe, for example, people who drink even a little are worse drivers than people who do not drink at all. Maybe Mormons or even non-religious non-drinkers are less risk taking in driving than others. Again, correlation is not causation.

neil wilson

Japan punishes drivers with a BAC of .01%.

BUT Japan also punishes EVERY driver in the car. If you are in the car with a driver who has been drinking then you are punished too.

Extremely strict.

I have no idea how effective it is in general but it sure kept people from drinking and driving when I was there.

Voice of Reason

I feel like Japan is more metropolitan though. There are many more options for alternative forms of transportation than we have here. There are only a handful of places to live in America that allow you to get everywhere you need to go efficiently without owning a car.

Voice of Reason

So you keep lowering the DUI threshold, but what does it do? Most people who are very against drunk driving are astonished when they find out how little it takes to get a DUI. It really comes down to how lucky you're feeling that night. People are still going to drink, and if you live in the suburbs, your only chance to get home from a bar is going to be to drive. To you lower the thresholds, and you'll just get more and more innocent people thrown in jail for the night, paying $10,000 of their hard earned dollars in lawyers, and not being able to drive to work.


Apart from the bias mentioned by many other posters, and which the authors acknowledge and try to address, there is also the flaw that this analysis is based only on a subset of accidents where someone died. As has been shown in other analyses from this database (e.g. air bag effectiveness) this can lead to false conclusions.

As an analogy, suppose we wanted to assess the effect of radiation on human beings. If we used a database of cancer patients to make the assessment our analysis might show that radiation had a beneficial effect. However, radiation is harmful and can cause cancer. It doesn't seem likely that low alcohol levels would actually reduce accidents, but this flawed analysis may have grossly exaggerated the harmful effect.

Josh Miner

I'd be curious as to the relative risk of being involved in a SOB accident with various complicating factors (e.g., BAC, texting, eating, drowsiness, etc.). A sort of scale of dangerousness. What is the BAC that is equivalent to texting, in terms of relative risk?

Alaska Pete

I wish my insurance company had a teetotaler discount! Could be stamped across my ID and bar me from purchasing alcohol. I'd jump at it if it gave me a sizeable discount. : - )