Farrar reports that Cabinet will soon be deciding whether to drop the drink driving limit from 0.08 to 0.05 or to do a bit more research, having police test all drivers involved in an accident to get a picture of how many accidents involve folks in the 0.05-0.08 range.
I like the call for more research, but just knowing how many drivers are in the 0.05 to 0.08 range isn't enough. You have to know what proportion of all drivers on the road are in that bracket if you want to know whether the accident rate is substantially higher for that class of drivers. This could be done by keeping all the data from roadside breath tests when the police set up their alcohol checkpoints. Or, under a few assumptions about the underlying distribution, you could back it out from the accident rate. How? Well, the proportion of accidents involving one driver in the 0.05-0.08 range rises with the proportion of such drivers on the road, but the proportion of accidents involving TWO drivers in that range rises with the square of the proportion of those drivers on the road. Steve Levitt used that a while back to estimate the number of drunks on the road in the States; it wouldn't be terribly difficult to replicate that analysis here if the underlying data were available. I'd be curious to see what the results here would look like.
I fear, though, that research wouldn't matter at all for the folks wanting to reduce the limit to 0.05. Their counterargument, which can't easily be falsified, will be that folks who get to 0.06 are more likely then to decide to get to 0.12; with a limit at 0.05, they'll never get to the point at which the craziness sets in and the next three drinks become unavoidable. And should the limit drop to 0.05 with no effect on the number of accidents in the 0.1 and up range, that will just mean that the evil alcohol marketers have become more crafty in getting their message across, not that the wowsers' argument was wrong in the first place. Or that the limit should really have been 0.03, or 0.01, because clearly 0.05 was too high.
Meanwhile we still hear reports of folks who've finished a temporary licence suspension after their nth drink driving conviction getting their licence back without having undergone any kind of rehab and without any kind of conditions like ignition interlocks. Because it isn't about stopping the repeat drink drivers; it's about stopping social drinkers from drinking socially.
iPredict's saying there's an 82% chance of a reduction in the drink driving limit. I'm short, having sold around $0.86. But I'd be buying at $0.75.
"Their counterargument, which can't easily be falsified, will be that folks who get to 0.06 are more likely then to decide to get to 0.12"
ReplyDeleteIt's not easily falsified, yet that doesn't mean it's wrong. Indeed, anecdotal evidence on the influence of alcohol on myopia and risk-taking suggests there's something to it.
A different point: A drawback to using the methodology employed in the Levitt paper to base policies on would be that it's hard to understand and consequently you'd get counterarguments along the lines of "lies, damned lies. . ."
@Lemmus: I'd of course prefer data from the checkstops; not sure whether that's available though.
ReplyDeleteIt's certainly plausible that that mechanism holds, but oughtn't we weigh the costs to sensible moderate drinkers from the reduction to 0.05 against the costs of an alternative regime - actually getting serious about the folks who repeatedly drive beyond 0.1....
I certainly agree with the last paragraph.
ReplyDeleteHow good the checkstop data would be (compared to alternatives) depends on the details. I'm envisioning some sort of weighted random selection of time-places.