More nonsense on blood alcohol limits

The latest press beat-up: the government has ignored expert advice from its officials, obtained by OIA, on the benefits of reducing the drink driving limit.

As I noted a few months ago, we simply can't tell whether the number of accidents involving drivers in the 0.05-0.08 range is very high, about what we'd expect, or very low, unless we know how many drivers on the roads at different times of day are in that range.

It's a little galling. Suppose that Expert Advice noted that women drivers were responsible for 50 deaths in 2009, followed by bleating about banning women from driving. We might expect a competent press to ask whether this is a particularly high number given the proportion of female drivers on the road. Since the 2009 road toll was about 300 deaths, and women would be a bit fewer than half of all drivers (since driving couples tend, for whatever reason, to mostly have the man driving), the proper conclusion from a finding of 50 deaths would be that women are safer than average. But you can't know that unless you know the accident rate rather than just the number of accidents.

Other bits of particular nonsense in The Press's coverage:

Last year, road crashes involving alcohol killed 137 people and caused 565 serious injuries at a social cost of $875m.

The social cost figure will be overstated to the extent it includes costs that drunk drivers impose on people inside their own vehicle. And the figure does include those costs. By memory, somewhere around 17% of injuries accrue to people outside of the drink driver's vehicle. The social cost figure has to be reduced to be a number that economists would consider a reasonable measure of actual external costs. And, you'd want the number of accidents where alcohol was responsible for the crash, not just where the driver had had alcohol. Sober drivers have accidents too: a proper accounting would include only the costs of excess accidents.

The Government was told that, based on data in about 300 international studies, a lower limit would save up to 33 lives and prevent up to 686 injuries each year. Aside from social cost savings of between $111m and $238m a year, ACC expected additional savings of up to $94.5m on claims.

I'd need to have a look at those studies to be able to comment on them. They're almost certainly expecting that reducing the limit will have reasonably large effects on drivers who would otherwise be driving at alcohol levels well in excess of the current limit. There's some evidence of such effects, but I'd want to look carefully to look at whether they controlled for things like enforcement blitzes that tend to accompany changes in drink driving laws and endogeneity issues where areas in which drink driving has attracted greater stigma and areas experiencing spikes in drink driving accidents are the ones most likely both to see reductions in drink driving AND reductions in the legal limit. I'd suspect both to be fairly major issues, and I'd be very surprised if Transport officials evaluating the lit were competent to know how to read this stuff. I know I've seen one study showing that reductions in blood alcohol limits were associated with crash reductions a year or two prior to the law change, suggesting that endogeneity issues are going to be just a bit important.

But assuming for the moment that their numbers on deaths and injuries are correct, we'd need to compare a corrected measure of social cost savings against reductions in consumer surplus that would result from the lower blood alcohol limit. And, from memory, these social cost figures tend to include things like health costs; adding claims savings then constitutes double counting, unless they've been far more subtle than I'd expect in how they've run those figures.

At least Transport Minister Stephen Joyce seems to be standing up to the wowsers for now.