Saturday, 18 August 2018

Zombies live

Nine years ago, BERL put out its study on the costs of alcohol use.

They're now pushing an inflation-adjusted version of the figure. The figure isn't good, and that they're pushing it now is worse.

The study was riddled with problems and became a laughingstock among economists. Among the methods BERL used to get a gigantic figure on the costs of harmful alcohol use:
  • Including every dollar spent by drinkers on alcohol if they consumed more than a medically-set threshold. You can't do this unless you're assuming that there are zero benefits associated with those drinkers' consumption. The threshold was equivalent to about two pints of strong beer per day. That's more than I drink, but it's odd to assume that folks consuming at that level get zero benefit from it.
    • Oh - that spending included excise tax. They included the excise tax paid as a social cost. So whenever you increase alcohol excise tax, you increase the measured social cost of alcohol unless consumption drops by at least enough to offset the increased expenditure per-unit. They eventually fixed that part when I mocked them for it; dunno if the updated figure includes excise paid by heavier drinkers as a social cost or not. 
  • Double-counting by including lost output among those who die early because of excessive alcohol use, and the value-of-statistical-life measure used by MoT which is inclusive of all costs of death including lost output.
  • In tallying health costs, they took Collins & Lapsley's aetiological tables that give the proportion of the burden of each disorder associated with alcohol use, then zeroed out all the disorders where alcohol reduces health costs. 
  • In tallying crime costs, they used a survey of prisoners who were asked how much alcohol contributed to their offending. Possible answers were "not at all", "a little", "some", "a lot", or "all". If the offender said at least "some", BERL attributed 100% of the costs of that crime to alcohol. 
    • Oh - they also counted the lost output costs of incarceration by assuming that those in jail would have been on the average wage otherwise. That doesn't make any sense unless you're trying to inflate figures. 
I could go on. The full tally is summarised here; the full report is here. Remember Marge listing Homer's failings at Catfish Lake? Anyway, maybe about a fifth of BERL's tallied figure could plausibly count as policy-relevant costs under more normal ways of handling things. 

BERL did not come out well from that study. Nana had to defend it on Jim Mora's show. That did not go well. It was called "Shonky" by a Treasury Deputy Secretary in the National Business Review - but I understand they had to pull back from that because the reporter had characterised it to the Dep Sec as a cost-benefit assessment that had forgotten to run the benefits side rather than as a cost-of-illness study that included a pile of private costs as net social costs by assuming the associated benefits to be zero. 

They presented it at the NZAE meetings; I was discussant. It was standing-room only because Matt Burgess and I had released our review of the report ahead of the meetings.

Geoff Palmer defended the study, and hired Marsden Jacob Associates to back him up on it since he was using it in his Law Commission review. They presented it as an independent review, but note that it's Marsden who was presenting at the anti-alcohol conference this past week. I'd summarised the Marsden-Jacob review here

This past week, BERL provided an updated measure at an anti-alcohol conference. The reporter who called me about it said it was an inflation-adjustment of the old figure rather than new workings; I haven't seen the new figure's workings to check. It's a higher figure than you'd get by inflation alone, so I expect they took the per-capita equivalent of the old figure, inflation adjusted it, then inflated by population increase over the period - but I don't know for sure.  
Eric Crampton, from think tank NZ Initiative, said many of Nana's figures were based on the 2009 study which had been mocked in economic circles for things such as double-counting and counting factors that shouldn't be counted.
Using total cost figures to inform policy was useless in cases such as this. For example, raising excise on alcohol may penalise moderate drinkers but studies showed would only slightly deduce what heavy drinkers drank.
That's not quite right. I'd said that my remarks were based on the 2009 study and would apply to the current one to the extent it relied on the old one. I haven't seen the new one.

But there are two big annoying things.

First, we're again back in the "let's make a big stupid number" world rather than thinking about cost-effectiveness. 

More worryingly, it appears BERL no longer worries about reputation cost associated with that prior work, which was plausibly by now well behind it. Why would you tie your name back to that mess now? None of the plausible answers are good. 

No comments:

Post a Comment