Friday 3 July 2009

Of externalities, elbows, and knowing one from the other

Economists tend to think that murder is a bad thing. Why? Well, despite the murderer presumably enjoying the act, his gain comes at a cost that he doesn't personally bear: the death of his victim. That's the kind of cost that economists tend to call an externality. And so economists tend to support laws against murder. We similarly tend to support laws against theft: while the thief tends to think taking other folks' stuff is a good idea, the thief's victims tend to be hurt by it and the thief won't weigh those folks' losses against his gains. In these kinds of cases, individuals' rational calculation of their own costs and benefits lead to socially bad outcomes because of the substantial external costs.

No, this isn't heading towards commentary on the Clayton Weatherston trial. Rather, consider Adrian Slack's latest response to our critique of BERL's analysis of the costs of alcohol.
“So for example someone who murders someone, from the individual’s point of view, Eric would be, I presume, quite comfortable with that. The person who decides to murder someone else makes an evaluation of what are the benefits and costs to me of this action? Society says ‘well some people do murder other people’, but society says ‘that’s not good.’”
Is this close enough to Godwin's Law for BERL to be deemed as having conceded defeat?

For the record, I do not support murder: I care about the external costs. One of our biggest critiques of the BERL report is that they conflate internal and external costs. I don't think that BERL has a ... conventional ... understanding of the difference between the two. As we reminded BERL in our report:
Externalities can be imposed on private citizens and, conversely, costs to a business of employing an unproductive worker are not externalities. Perhaps this misunderstanding contributed to BERL’s decision to count private costs as social costs. A basic staple of principles-level economics is that costs or benefits are not external if the agents are linked through a contractual nexus: the baby crying next to me on the long-haul flight does not impose an externality on me because I have chosen to buy a ticket that includes the risk of such unpleasantness and was accordingly charged less for that ticket. Similarly, a worker who slacks off on the job or takes inordinate numbers of sick days does not impose externalities on his employer. It is worrying that BERL does not seem to recognize this basic definitional issue.
Adrian, what colour is the sky in your world?


  1. My jaw dropped when I read the quote on murder.

    Giving Slack the benefit of the doubt, maybe he was saying that you're likely to count the utility of the murderer as a benefit partially offsetting the cost to the murdered, whereas the BERL authors are not because society deems that sort of utility as inadmissible.

    In that case, he's saying that economists should ignore the benefits heavy drinkers get from their drinking. Even though people might enjoy getting drunk, this pleasure is too uncivilized to be accepted by society as legitimate.

    That's the only way I can make sense of his comment.

  2. I made a comment on Bard's comment in the Update to this.

  3. Brad (or is it Bard?): David Friedman has an apposite comment in Law's Order about how an economic case has to weigh the criminal's utility. Of course, the (correct) presumption in a forced transaction is that it's efficiency reducing.

  4. It's not a great comment by Slack, it indicates a basic misunderstanding of a number of things, it sounds desperate, and it may be aimed at smearing Eric as well, I don’t know. But the weakness of BERL’s overall response runs much deeper than this.

    I haven't counted but there might be 30 or 40 specific concerns we raise about the BERL report in our critique. In the three weeks they’ve had the report now BERL has responded to perhaps 3 or 4 points, and we don’t like much of what they've said.

    BERL's most repeated response is to say we assume perfect rationality/perfect information in our critique. This response achieves two things. One, it is an attempt to discredit our analysis without attacking any part of it. Two, BERL can say they didn’t make that crazy assumption – "we're the reasonable ones here".

    It is of course a red herring. Let me say it again: our critique does not depend in any way on perfect rationality or information. We simply think that on the whole people drink for a reason and a better approximation to the zero benefits BERL assumes, particularly given the low threshold for harm that BERL uses, is that private benefits will roughly equal private costs. For every person who wakes up with regrets there are others who had a good time out. Show me the perfect rationality/perfect info assumption in that. Rational addiction finds strong support in the empirical literature.

    In any case, BERL's argument is a non-sequitur. It does not follow from deciding consumers are sometimes irrational or imperfectly informed that all consumers drinking more than what BERL is says is harmful are irrational and that none enjoy any gross (not net) benefit whatsoever. We are still waiting for BERL’s explanation for these $2.2 billion assumptions. And still waiting for something concrete on the other $1.9 billion of problems.

    The irony of Slack’s murderer comment is that he stood up at the NZAE conference and said he's not prepared to make radical value judgments.

  5. Is it time that we moved on from the question of accuracy to that of robustness? I haven't read all of the debate on this, so forgive me if you've already dealt with robustness elsewhere. If not, I'd be interested to read something about the degree to which BERL's conclusions are robust to the poor assumptions you've identified, and particularly about how serious a violation would really be for their identification and location of the various (non-offset) costs.

  6. BERL does a lot of robustness checks at the end of their paper; just not against any of the bits that seem to matter much. BERL's numbers are highly non-robust to:
    - realistic multiplier specification
    - degree to which you want to assume crime disappears with a reduction in alcohol use
    - consistent application of counterfactuals for resource costs (ie reduction to threshold rather than reduction to zero)
    - assumptions regarding counterfactual labour force characteristics for prisoners
    - whether "harmful" drinkers are allowed to enjoy any benefits that offset their costs of drinking, or whether they're assumed to have only gross costs and no gross benefits.

  7. I just really hope that the government, and Geoffrey Palmer are following all this commentary. They're so quick to spend our money, not so quick to ensure they're received value for it. BERL are a waste of paper, dollars, and oxygen.