Drunk driving

The Undercover Economist makes the following points about drunk driving,

It has always been difficult to test the effect of alcohol on drivers let loose on the roads. The difficulty is this: if half of all crashes involve drunks, that may be because drinking impairs your driving or it may be because there are a lot of drunks on the road – and we can only guess at how many drunk drivers there are.

But the economists Steven Levitt and Jack Porter realised that it was possible to say more, by looking at how often drunk drivers crashed into each other. If 10 per cent of drivers drink, and if drunk drivers are as safe as any other kind of driver and randomly mixed among the sober drivers, then only 1 per cent of two-vehicle crashes should involve two drunks.

Drunk-on-drunk crashes are much more common than one would expect, given the number of drunk-on-sober crashes, allowing Levitt and Porter to reach firm conclusions about the risks of drink driving.

They find a very large effect. Drivers who have been drinking are seven times more likely to cause a fatal crash; those who have drunk over the legal limit (in the US) are 13 times more likely to cause a fatal crash. You might also bear in mind another finding from the paper: "The great majority of alcohol-related driving fatalities occur to the drinking drivers themselves and their passengers." That should be sobering.