Drink-driving kids

The blood alcohol limit in New Zealand for those under 20 is 0%. For those over 20, it's 0.08.

It's worth keeping that in mind when you read stories giving the youth and adult arrest numbers:

The Group and traffic officers on Friday and Saturday had checkpoints around Tauranga, Mt Maunganui and Papamoa.

Of the 22 drink drivers caught at the weekend only two were under the age of 20.

''This is an encouraging sign that our youth are now in a position to set a benchmark for the rest of the community".

Ten percent of drivers nailed being youths seems high relative to their proportion of the driving population, but the standard varies across cohorts.

Overall, youth arrest rates for drink driving are substantially down. The reporter here seems to have gotten it, even if the sub-editor went for the inaccurate but eye-grabbing headline "Teens turn blind eye to drink-driving". From the story:

The number of teenagers arrested for drink-driving has halved almost a year after the zero alcohol limit was brought in for under-20s. Police figures show that in the first nine months since the law came into force on August 7 last year, 3091 youths aged 15 to 19 were arrested for drink-driving. The figure for the 12 months before the law change was 6414.

Previously, the drink driving limit for youths under 20 was 0.03. Here's the relevant section from the Land Transport Act 1998. Prior to the law change, consumption of alcohol in the 0.03 to 0.08 range drew a maximum term of 3 months or a fine of up to $2250 and a 3-month license suspension. Those consuming over 0.08, as I understand things, would draw the standard adult range of penalties: maximum jail term of up to 3 months, a fine of up to $4500, and driving disqualification for at least 6 months. The 2011 law change added an infringement offence for consumption in the 0 to 0.03 range: a fine and 50 demerit points.

A prior news story citing the same police figures gave a few more details:

The average numbers of those charged for having 30mg or more had dropped by about 43 per cent since the law change - from two charges per day to 1.1.On average, the numbers of arrests for youth drivers falling within the ‘‘new range’’ of alcohol restrictions dropped from about 18 charges before the law was introduced to 12.6 since.

I'm a bit confused by the stats on the 'new range'; I didn't think you could be arrested for being in the 0-0.03 range prior to the law change.

So:

1. Drink driving arrest numbers are not commensurable pre- and post- the law change; if behaviour had not changed, we would expect arrest rates among youths to have increased substantially.
2. That youth arrest rates dropped says that the bright line rule deterred behaviour.

I would be awfully curious to know more about the actual BAC for those charged before and after the law change.

The best argument for reducing the adult drink driving limit from 0.08 to 0.05 isn't the reduction in accidents among the cohort of drivers who test in the 0.05 to 0.08 range - that's almost certain to be trivially small relative to their representation among the cohort of drivers who were not involved in accidents.* Rather, it's that many people may intend on having only a couple of drinks but, after having gotten to 0.06, decide to have a few more drinks. Since they hadn't planned on getting drunk, they didn't plan on a ride home. And so they may wind up driving home drunk. A lower drink driving limit's benefit might then be in reducing the likelihood of transitioning to heavier drinking when driving home.

The change to the 0% threshold did nothing to affect the penalties faced by 19 year olds (compared to, say, 23 year olds) for consuming amounts of alcohol over the 0.08 threshold for more severe penalties. But it might have affect their likelihood of making plans to have a designated driver or to drink at all. If a substantial proportion of the reduction in under-20s drink-driving arrests comes from reduced numbers of youths consuming past the 0.08 limit (relative to the trend for 23 year olds), then we put more weight on the potential for a 0.05 adult limit to reduce adults' likelihood of driving with BAC > 0.08. If not, then we downweight that potential.

In either case, full cost-benefit analysis would need to weigh benefits from accident reductions against reduced consumption benefits to those who would not have gone on to impose increased accident risk on others: many adults, me included, would substantially curtail consumption to guard against the risk of accidentally going over 0.05 even if the vast majority of the time we never exceeded 0.04. It is a mistake to note that most adults' typical consumption would keep them below 0.05 and that consequently moderate drinkers benefits from moderate consumption would not change. Think of it this way: if we had the death penalty for going over 105 kph if the speed limit were 100, and the police said this shouldn't matter because most people drive the speed limit anyway, it would be a bit nuts. It's easy to accidentally hit 105 even when you're trying hard to stay at 100. It's harder to hit 112 by accident. So the death penalty for driving 105 would likely have most folks stick to 90, just to be safe. Same with a drink driving limit of 0.05: risk-averse people would target 0.03 or below just to be safe. The loss in consumption benefits would be real, and we'd better be sure that the reduction in accident rates is then worth it.

And so I wonder if the Police is collecting data on actual BAC among those breath-tested, and whether time series data on that sorted by age is anywhere available.


* And this, boys and girls, is why the police's emphasis on testing the BAC of those involved in accidents to see what portion fell in the 0.05 to 0.08 range by itself really doesn't help us a ton in deciding whether to cut the limit to 0.05. If 10% of drivers on the road from 8 pm to 4 am would test at .05 to .08, then they'd have to show up in more than 10% of accidents during that time of day to be over-represented. Unless we combine that accident data with data from random police checkpoints to get time-of-day population baselines, we just won't know whether they're over, under, or proportionately represented.