BERL redux

Adrian Slack, principal author of the BERL report of which I've been rather critical, provides a couple of responses over at The Visible Hand.

First in response to “BERL should have made the LC very aware of…”. We didn’t prepare the report for their consumption. As with anything that enters the public domain, it is the consumer’s right to interpret it as they see fit and for them to take responsibility for their reaction to it, not for the author to manage their response to it.

In general, I would agree. But not if the response seems likely to turn into a real world policy implementation unwarranted by the paper. If someone took my paper on political knowledge, one finding of which is that Maori on average and correcting for everything else have less political knowledge than do others, and went on to make a policy argument for restricting Maori from voting, I would be doing a lot of jumping up and down about how that conclusion is not warranted from the evidence presented. Ideas have consequences. If folks want to misinterpret my ideas, that's up to them; if folks want to turn a misinterpretation of my ideas into policies, that's something I'd want to do something about. I'd have thought the same of others.

“Eric Crampton takes to the second point of the report thoroughly” - from a position underpinned by strong rationalist assumptions and a non-verifiable argument (e.g. if someone consumes a good the claim that the private benefits must equate or exceed the costs is not verifiable).

All we really have in economics is revealed preference. Preference is revealed by action. I can't look into people's heads to verify that folks choosing A over B really prefer A over B, all costs included, but neither can Adrian. However strong my rationalist assumptions are in figuring that preferences are revealed by action, it takes stronger ones to assume the reverse. It would be a strange strange world if a preference for Holdens was revealed by buying Fords, or a preference for the Canterbury Crusaders were revealed by wearing a Wellington Hurricanes jersey.

Counting the costs (or benefits) of non-harmful consumption were irrelevant as they do not impose a social cost, and benefits were out of scope. But including either of these components would have no net impact on the estimates for non-harmful use.
Valuing the benefits to harmful users is a complicated area. Consumption decisions by addicted, intoxicated or problematic users are likely to violate some of the rationalist assumptions underpinning conventional consumer choice theory. It is unlikely that many harmful drinkers, for example, rationally decide (or can be modelled as deciding) about their drinking behaviour and career prospects. Labour-leisure choices are typically taught that way at an undergraduate level, and provide useful insights, but the assumptions underpinning these models do not hold universally.

As Friedman taught us, "as-if" is all we need. Predictions from the Becker-Murphy model of addiction tend to bear out in the real world (long-term price elasticity of demand greater than short-run elasticity, for example) regardless of underpinning assumptions about rationality. The upshot of the Becker-Murphy model is that we cannot simply point to the existence of addiction and conclude irrationality: their model generates addictive behaviour within a purely rational, full information environment where folks choose between an addictive good and a composite commodity that includes everything else in the world. Why move to assuming irrationality when outcomes are consistent with a pure rationality model?

And even if you want to bring in some bad decision-making, why assume that the benefits are zero? That's torturing the method far too much. Drive consumer surplus down to epsilon if you want, but saying that the benefits to the drinker are less than the cost of the alcohol consumed really looks to be nothing more than a way to inflate costs. If you define benefits as zero, you get to count all of the resources that go into producing the good as costs; if benefits are at least as much as what folks pay for the product, then you don't. There is no basis for assuming benefits to be equal to zero. What basis do I have for saying the benefits are at least as much as folks pay for product? Evidence that they do in fact pay for product!

Low risk drinkers, for example, are affected at work on average one day in four years. This is more likely to reflect a poor, impulsive (and time inconsistent) decision than a calculated choice. Moderate drinkers are affected only about a day per year, but it is unlikely that an employer will employee for having a hangover the day after the end of year party. Employers do not willingly contract in to this kind of behaviour, but bear the costs.
For high risk drinkers, employers have difficultly identifying, let alone punishing, such behaviour. In practice, there is a myriad of reasons that the costs of such behaviour represent externalities rather than internally and rationally borne costs. These might include the relatively low importance of reputation (and therefore reputation risk) in industries that harmful alcohol users are employed in or the difficulty for employers in identifying and documenting alcohol-related problems.

All we need is that the probability of retention or promotion is an increasing function of job performance. If excessive alcohol consumption negatively affects job performance, it reduces the probability of good retention or promotion outcomes.

To paraphrase Dr Cullen, it takes a peculiarly warped sense of values to equate the output lost due to premature mortality with rational consumption choices. Equivalently, people whose drinking makes them unemployable (resulting indirectly in output losses) are more likely to be trapped by addiction than making a rational career choice. The professional who might be modelled using the equivalent variation argument who tosses up between preparing for work the next day or having another beer is unlikely to fall into this category.

Then call me warped. I consciously and deliberately choose a diet that includes richly-marbled steaks, even though such choices may well take a year off my life. Does that make me irrational? I'm maximizing the product of years of life by the quality of each year rather than just years of life. Who's to tell me that my utility function is wrong?

Moreover, if the underlying model is that folks are helplessly trapped by addiction and are completely unable to make choices, I completely fail to understand how a tax increase on alcohol is supposed to help them. If the argument is that there's a vertical demand curve, a price increase doesn't affect Q. We do have decent evidence that consumption responds to prices, but that gives lie to the argument that folks can't make choices.

You can argue your case for the assumptions and perspective you would use, and the emphasis that you would put on Homo Economicus. The research sought to answer a set of questions, with no restrictions from the client on the method beyond it being internationally recognised and no conclusion in mind. The report was upfront that it was commissioned research, what its aims were, that it was conducted independently, and that it was externally peer reviewed by leading academics in this field.

The conclusion was driven entirely by the choice to weigh the benefits of alcohol consumption as being zero for the five hundred thousand (Table 4.1) consumers reckoned to be consuming half of the alcohol drunk in New Zealand. I'm pretty sure that I'm interpreting the report correctly here. Table 4.1 lists 513,000 harmful drinkers in 2005/2006. Half a million in a country of about 4,000,000 - one in eight - is defined as getting zero benefits from the alcohol they consume. Is this plausible?

Turning to the appendix, we find that folks in the "hazardous" and "high risk" groups count as getting zero benefits from their drinking. You're in the hazardous group if you consume 40 grams of alcohol per day. That's opaque. There's a reason that's opaque. Let's pull back the sheet. 40 grams of alcohol. One ounce of draught beer by volume weighs one ounce. A pint is 16 ounces then, either by weight or by volume. A beer that's 5% alcohol by volume is 4% by weight. 40 grams is 1.41 ounces. So hazardous drinking is 1.41 ounces of alcohol per day, which is about 35 ounces of normal beer. So two pints of regular beer, 32 ounces, puts you just below the limit for "hazardous drinking". A beer with lunch and a wine with dinner: BERL says you get benefits from alcohol. Add in one evening drink, and you get zero benefit. Not just zero net benefit: zero gross benefit. You'd have done better by putting your money through a shredder. According to BERL. Without that assumption, about three quarters of the listed costs melt away.

That's what really bugs me about this report. As David Friedman pointed out a long time back, you really oughtn't have your thumb on the scales when weighing up costs and benefits. Adrian's saying that they had no remit to consider benefits, but that's obfuscation. They defined benefits as being zero for 12.5% of New Zealanders who they say consume half of the produced product, so they're taking a position on the benefits while pretending not to. I hate it when paternalism is disguised as being economic analysis.

Rauparaha over at TVHE seems to be feeling a little bit conciliatory. I'm not.