Wednesday, 28 April 2010

Marsden Jacob review - continued

In the last post, I noted some problems with the Marsden Jacob review. Most annoyingly, from a personal perspective, they strongly critique the Crampton and Burgess work on two points that severely mischaracterise our work. Contra Marsden:
  • In actuality, none of our results depend on the cardioprotective effects of alcohol for serious drinkers;
  • their critique of our narrow focus highlights a small robustness check at the end of our work rather than the vast bulk of what we actually did - they take the robustness check as constituting the entire method rather than cursory treatment of an alternate method.
They then argue at paragraph 58 that we take a flawed conception of cost internalisation. Because some costs of drinkers' behaviour is borne by others, including friends, families and government, drinkers drink more than they would were they to bear the full costs. We did note the costs to government as a cost of harmful drinking, including costs to the public health system and policing costs. Since they seem to believe that our method discounted all fiscal externalities, presumably they missed that part.

On cost shifting to family and friends, I can't help but wonder whether their argument proves too much: all kinds of consumption decisions we make have effects on friends and family. I may spend too many late nights reading shonky alcohol policy reports and be grumbly the next day; if I had to bear all of the costs thereby imposed on my family, I might do less of that. But, of course, families have ways of internalising those effects: my wife rightly ensures as much. Marsden Jacob here are effectively launching a full assault on notions of consumer sovereignty: if we can't rely on intrafamily mechanisms for solving these problems, then there exists no consumption decision that isn't subject to their critique. This may well be a good argument to have in an economics journal as a critique of existing orthodoxy, but it has no place in a commissioned report that ought to be providing a mainstream economic approach.

This isn't the first time though that Marsden Jacob use their commissioned policy report to challenge the standing orthodoxy though. In their critique of the positive health effects of drinking, they give full weight to a few critiques of the "previously near unanimous orthodoxy" rather than to that orthodoxy. Again, sound technique would have had them accepting the best consensus of the literature in which they are not expert rather than the overblown critiques of a few anti-alcohol folks.

Let's now move on to the rest of the report.

They argue that the deadweight costs of alcohol taxation are 3% or less for a 50% increase in alcohol taxes. Since alcohol demand is fairly inelastic, the Harberger triangle is relatively small. They then invoke the inverse-elasticity rule. But as Seamus noted here and I noted here, the inverse-elasticity rule isn't as simple as the undergraduate version makes out. If supply is not perfectly elastic, then we need also consider reductions in producer surplus, not just consumer surplus. If there are cross-price effects between taxed goods - if inelastically demanded alcohol is a strong complement to elastically demanded restaurant meals, for example, then Ramsey doesn't reduce to inverse-elasticity. And, if taxed goods and leisure have cross-price effects, then we get the Corlett and Hague result. Higher alcohol excise tax doesn't flow automatically from Ramsey considerations. Neither does it necessarily flow from the inverse-elasticity rule: the McLeod report having argued that existing (lower) levels of excise taxation could not be justified on Ramsey grounds.

I'll leave it to Seamus to come in later with more comprehensive discussion of Ramsey taxation.

But, whatever efficiency claims - debatable though they may be - might be made for moving increasing alcohol taxes and reducing marginal income tax rates disappear when those alcohol tax revenues are instead earmarked for anti-alcohol advocacy groups like ALAC. It's consequently disingenuous at best, or blindingly stupid at worst, for Palmer to cite Marsden Jacob's numbers on efficiency gains while simultaneously advocating that some of the raised tax revenue be earmarked for anti-alcohol programmes. To re-iterate: the efficiency claims rely entirely on the consumer surplus that's turned into tax revenue either being rebated to consumers directly or being used to offset other taxes that impose higher deadweight costs. Palmer simply can't pair the efficiency claims with a call for some of those revenues to be earmarked.

Marsden Jacob is entirely correct however that proper Pigovean analysis seeks to equate the private marginal cost of consumption with the social marginal cost of consumption rather than equating total tax revenue with total external cost. To my mind, if total tax revenue roughly approximates total external cost, then things probably aren't far out of whack as a first cut. Their approach in simulation work from page 34 onwards requires that all excise taxes raised are rebated to consumers. Ideally, all collected alcohol tax revenues would be lump-sum rebated to all drinkers. Such a scheme could have pretty desirable characteristics: it effectively would induce progressivity into the alcohol tax schedule. Heavy drinkers would pay net taxes; moderate drinkers would receive net subsidy. The linearity of existing taxes is a pretty big problem: moderate drinkers are overtaxed while heavy drinkers are undertaxed relative to external cost imposed. The tax plus lump sum rebate to every drinker mechanism, noted at paragraph 64, is a pretty neat idea. I'd go so far as to say I could be convinced to endorse it if I believed that all raised revenue would be so rebated and wouldn't just wind up being earmarked for anti-alcohol advocacy. Of course, I'd only endorse it to the extent that the taxes reflected actual external harms rather than things like lost wages that are largely internalised: they're already part of private marginal cost, so putting a Pigovean tax on that part is double counting.

If on the other hand raised revenues just go into the general pool - or, worse, to projects like anti-alcohol advocacy that many drinkers find positively offensive - the MJ results hold only if the consumer derives the same enjoyment from a dollar's worth of tax-paying as he does from a dollar's worth of alcohol. That seems unlikely.

I continue to confess unfamiliarity with the Sheffield simulation approach here used. But I wonder about some of those results. I've not seen any good evidence that young drinkers are more responsive to prices than adults: the price elasticity of demand for young drinkers is not different from that of adults. Grossman, Chaloupka et al, 1998 Economic Inquiry, put it at -0.41: basically the same as for adults. And, we have good evidence that heavy drinkers' consumption is far less elastic than moderate drinkers' consumption. The simulation results show strongest effect of tax on young drinkers; they note at paragraph 95 that Sheffield estimates higher price elasticity for harmful and hazardous drinkers. Both of those make me more than a bit worried about using the Sheffield results.

Honestly, I'm a bit confused here. At paragraph 95 they claim Sheffield finds higher price elasticity for harmful and hazardous drinkers; at paragraph 82, they assume a uniform price elasticity of demand for all drinkers "for consistency with the University of Sheffield study". But, they're at best assuming uniform elasticity; at worst, they're assuming heavy drinkers are more price elastic. Neither one sits well with the best empirical results that heavy drinkers are far less elastic than moderate drinkers.

Everything following in paragraph 96 assumes at best uniform price elasticities for moderate and heavy drinkers; at worst, that heavy drinkers are more price responsive. They also note their results may be sensitive to those assumptions. Writes MJ at paragraph 96.v:
Unless the price elasticities of demand for alcohol for heavy drinkers are substantially lower than for other drinkers, the higher price increases for heavy drinkers means that harmful and excessive (and likely) younger drinkers will face the largest reductions in consumption and therefore the biggest changes in harms to themselves – and to other parties.
Ok. So they have some results that depend on an assumption of at least equal elasticity and that, they admit, are sensitive to whether there are large differences in price elasticity of demand. The best numbers I've seen show youth drinkers' elasticity on par with averages, and heavy drinkers much less elastic (-0.28 vs -0.44). The Law Commission repeats those numbers throughout their report as well, so I'm not going out on a limb in relying on them.

At Paragraph 97, MJ cites the Sheffield results as strongly diminishing the risk highlighted in the Burgess Crampton submission to the Law Commission that "moderate drinkers respond to price increases by more than heavy drinkers." They quote us from our submission, then say:
Indeed, for England and Wales and for Scotland, the policy simulation models suggest exactly the opposite.
I can only laugh and shake my head at this point. In our submission, we were citing evidence that elasticity varies greatly between moderate and heavy drinkers (with heavy drinkers being far less elastic); their evidence against that is a policy simulation that assumes there to be no difference in demand elasticity between moderate and heavy drinkers and possibly assumes heavy drinkers to be more elastic!

And, if they had access to our submission to the Law Commission, they would have seen our extensive discussion there of how our results do not hinge on a narrow conception of rationality. Pages 7-9 discuss that in depth. But MJ simply asserts that our results are sensitive to assumptions about rationality. Again, I'd be happy for them to provide some reasoned argument about how our results were sensitive in a way we hadn't considered. I'd thought yesterday that they'd simply not seen our extensive discussions on the blog and in our submission to the Law Commission on rationality. But they're here citing that submission.

In summary: the only substantive result from the Marsden Jacob commissioned analysis is that an excise tax increase on an inelastic good, if the tax revenue is rebated to drinkers lump sum or offsets more distortionary forms of taxation, may be desirable. Even that result is questionable absent more explicit consideration of cross-price elasticities between alcohol and other consumption goods, and between alcohol and untaxed leisure. Moreover, the McLeod Report specifically rejected alcohol excise tax increases on Ramsey grounds.

Fortunately, Simon Power appears to have entirely rejected alcohol excise tax increases. iPredict's still giving a 16% chance of a 10% or more increase in the excise tax, but no real chance of anything higher.

No comments:

Post a Comment