And further on price elasticity

Seamus reminds us that we need to look at relative elasticities.

Brian Easton provides advice that the Law Commission deems "helpful". His advice is:

It is generally assumed that the demand for alcohol is largely price inelastic. However, it is believed that the main groups whose consumption is sensitive to changes in price are:

• the young;
• binge drinkers; and
• heavy drinkers.

Easton is right that alcohol consumption is relatively inelastic: around -0.44 for beer, which means that a 10% increase in price reduces consumption by 4.4%. But we tend to expect that the folks whose consumption produces greatest harms would also be the folks whose consumption is relatively least elastic. Price elasticity basically tells us how many substitute commodities there are out there for the consumer. If the individual views there as being no substitute for alcohol, his price elasticity will be very low; if he views a Coke as being a not too bad alternative, his price elasticity will be rather high. And, indeed, what numbers I've seen on it show that heavy drinkers are less price elastic than are moderate drinkers.

If we decide to be charitable, then Easton's just reminding us that even heavy drinkers are price sensitive. Which is true. But they're less price sensitive than moderate drinkers. Even if you accept the premise that there are some heavy drinkers who could be made happier by virtue of a tax increase that helps to keep them on the straight and narrow, moderate drinkers' consumption is more strongly affected. Suppose we have a 10% price increase. Moderate drinkers reduce their consumption by about 4.4%; heavy drinkers reduce their consumption by about 2.8%.

Now, how much does a heavy drinker benefit from a 2.8% reduction in consumption? How much is a moderate drinker harmed by a 4.4% reduction? And what are the relative proportions of both in the country? BERL said that 1/6 are "harmful" drinkers, which I still think is a gross overestimate of the true proportion. But let's take it for now. It then has to be the case that a single heavy drinker reducing his consumption by 2.8% is benefited by at least six times as much as a moderate drinker is harmed by a 4.4% reduction in his harmless drinking for the tax increase to potentially be efficiency-enhancing. If the true proportion is 1/10, change it to 10 times. Recall that for the wowsers, consumption of 2 pints per day is harmful; the midpoint of BERL's "high risk" category is 110 grams: 11 standard drinks, or about 5 pints. How big a price increase would be necessary to get the high tail consumers down to "moderate" levels? Well, they'd need to reduce their consumption by about 65%. If your price elasticity of demand is -0.28, then you need a 232% increase in price to reduce consumption by 65%. And of course for that level of a price increase, the moderate drinkers reduce their consumption pretty much to zero. All of these numbers are just back of the envelope, for now. But I don't think they're far out.

The LC cites a WHO study arguing that heavy drinkers are no less elastic than other drinkers. A quick check shows the WHO is looking at 2-3 studies while the above-linked one is a meta-study of more than a hundred, with results that are near identical to another meta-study of similar magnitude. The latter, forthcoming in the Journal of Economic Surveys, doesn't give a breakdown between heavy and light drinkers, but has near identical estimates of average elasticity. I give a heck of a lot more weight to a serious meta-study, and especially when two of them give near-identical results, then to a couple of pieces picked from the potential list. If there are more than a hundred studies, you can always find a few that give you numbers you like.

Looking more closely at the Law Commission's preferred WHO finding, ... wow. Terrible work.

While heavy drinkers are sometimes thought to be likely to be less affected by price, the Committee found that the evidence does not support this belief, with higher prices affecting the amounts consumed by frequent and heavy drinkers. This finding is supported by a large body of evidence which has shown an impact of prices on harms caused by alcohol, also indicating therefore that heavier drinking has been reduced (34). Natural experiments that have occurred recently in Europe as part of changes required as consequences of economic treaties have shown that as alcohol taxes and prices have been lowered, so sales and alcohol consumption have increased (37). In some jurisdictions in Europe, special taxes have been introduced for spirit-based sweet premixed drinks, in response to increases in young people’s drinking (38). These have led to reductions in sales and consumption of the specific drinks.

Nothing in the paragraph supports the WHO's argument. Nothing. Demand curves slope downwards, that's all they're saying. All groups are somewhat elastic in that demand drops when price increases. The question at hand is whether moderate drinkers reduce their consumption by more than do heavy drinkers, and they provide zero evidence supporting their opening claim. I cannot see how the Law Commission dismissed the findings of the meta-study based on this paragraph.

Shocking.

Costs and benefits of alcohol, properly done

I previously took BERL to task for its report on the costs of alcohol use in New Zealand.

My colleague Andrea Menclova points me to a previous Treasury analysis of the costs and benefits of alcohol: Felicity Barker, "Consumption Externalities and the Role of Government: The Case of Alcohol". The important difference isn't that there's a ton of health or other benefits from alcohol use; rather, the important difference is that costs that fall on the individual then are netted out: the individual has to weigh those costs as being less than the benefits. What's the conclusion when we do that?

This paper considers the role of government in the case of externalities and, in particular, in the case of alcohol externalities. The purpose of the paper is to assess whether the current level of the alcohol excise can be justified on externality grounds. The paper assesses various mechanisms to address externalities. These mechanisms are institutional solutions, trade in rights to generate externalities, regulatory measures and Pigouvian taxes. The paper assesses these tools in the case of alcohol and concludes that institutional, trade and regulatory solutions are limited in their ability to address the externalities of alcohol. A specific tax can be justified in the case of alcohol. The externalities are large and there is sufficient information on which to base a tax. Given the information constraints the specific tax must be applied uniformly across a rage of units of consumption, rather than to particular individuals. Where an optimal uniform tax is imposed it is reasonable to assume that the amount of revenue collected by the government would be at least as large as the total externality. In 1999/00 the amount of revenue collected from the tax on alcohol was $580 million. This is near the mid-point of the estimated bound of the external tangible costs of alcohol. Thus the current rate of excise tax can be justified on externality grounds.

I like the Barker paper. I could quibble about some details (allowing 5% of production losses to be external costs), but it frames the problem correctly and shows an appropriate amount of worry about the limitations of the method. Barker notes not having accounted for some external costs, like the costs of caring for victims of alcohol-related accidents. But the Barker's self-recognized limitations in underestimating things are dwarfed by BERL's inclusion of costs that should not count.

How much would we cull from the BERL report if we were to do things properly? All costs that are internal to the individual have to be dropped immediately. A quick accounting has those at $2.46 billion in the BERL figures (Table 5.5): foregone labour income and costs of producing and distributing alcohol. That should go. BERL counts $1.6 billion as the monetized value of premature death and intangible health costs: the vast majority of those fall on drinkers. Barker suggests that about 1/3 of victims of drunk driving are people not in the drunk driver's car; car accidents contribute about $500 million to the BERL number. We then can count $167 million in real road crash costs falling on other parties; BERL (properly) adds in a small value for mortality costs from homicide. So $169 million in real external costs from a $1.6 billion line item: So we should cut out $1.4 billion from the $1.6 billion as being costs that fall on drinkers themselves as a first cut estimate. So, $3.9 billion are costs that shouldn't count in a cost-benefit analysis: they're automatically offset by benefits as judged by the drinker.

BERL cites total social costs of $5.3 billion. $3.9 billion of that, 74%, cannot honestly be counted as policy-relevant costs from an economic perspective.

Costs of everything, value of nothing

NBR points me to a recent analysis by BERL of the costs of alcohol use in New Zealand. Apparently the Law Society is basing some recommendations towards greater regulation on the report. Kiwiblog also comments. A rather important, but thusfar ignored, point from the report:

This study takes a conventional approach for economic cost studies, which “do not attempt to fully consider the economic benefits of alcohol… and other drugs, and should not be confused with cost-benefit or cost-effectiveness analyses” (Single et al, 2003: 14).

In other words, everything is a cost. Imagine applying this methodology to anything else. What are the costs of car use? Of apple growing? Of coffee? It's very easy to get big numbers on the cost of anything, if you don't offset the corresponding benefits.

If this were a cost-benefit analysis, rather than just a pure cost analysis, things would look very different. First, we'd need to estimate how much folks enjoy the alcohol they drink. We could do that by assuming a linear demand function and extrapolating from the point estimate of the price elasticity of demand. That would be a minor point though. The more important feature would be that many of the costs listed would disappear entirely, washed out by private benefits. Let's work this through. The study counts as costs reduced labour productivity. If you go to work with a hangover, you're less productive. Similarly, if you spend a night out on the town rather than putting in the overtime, you're not producing as much. If we only count costs, then these get included: costs to society via lost output and costs to the government via reduced tax revenues. But if we worry about NET costs rather than gross costs, these have to disappear. Why? Because if I decide to drink and be less productive at work, I'm less likely to get a promotion or a salary increase. My productivity affects my wages. If I decide to be less productive and have a lower expected salary path, that's between me and my employer: I'm bearing the costs. If I decide to do it, that's prima facie evidence that I weigh the benefits as greater than the costs.

The report also counts as costs reduced government tax revenue because of folks being less productive and earning less. By this logic, every time I take a holiday rather than putting in overtime, I'm costing the government money. Think about the underlying assumption here and whether we ought to be viewed as owning ourselves or whether we're cogs for the production of tax revenue for the government.

What other gross costs disappear if we allow for benefits of alcohol consumption? Anything that primarily affects the individual gets dropped. Health, mortality, lost output -- all of those are costs I bear directly. You can make an argument that the health system bears costs, but the report admits that collected excise taxes outweigh the costs to the health system! It's only when they add in taxes on foregone labour income that they can get a negative fiscal effect. Correction 27 April: see below. And again, by that logic, your holiday or OE should be banned because it costs the government money.

I'd previously criticized a similar report on the costs of smoking over in the NZ Med Journal: the O'Dea Report. Lots of similarities between this report and that one. When the desired conclusion drives the method, take care in interpreting the results.

The Law Society should know better than to take this kind of report seriously. I could produce, using the same methodology, a report showing the massive social costs of the existence of BERL, the consulting firm that put out this report. First, assume that their reports have zero value, then add up all the costs of the researchers that waste their time writing them, then add up all the time we waste reading and complaining about them....

Update/correction 27 April I misread the final paragraph of the report. They list excise taxes of $167 million from excessive alcohol use and health costs of $286 million. Note, though, that these are only the proportion of excise taxes paid by individuals defined as consuming "excessively". Actual excise taxes in 2008, by Crown reports, were $573 million. Actual alcohol tax revenue outweighs actual health expenditures on alcohol, but tax revenues from problem drinkers don't cover their health costs. This is always a problem with a linear tax on something that has convex costs. See Felicity Barker's excellent Treasury analysis of 2002 on alcohol taxes.