Monday, 29 July 2013

Sometimes, prices are blunt

Where the social harms from alcohol consumption are nonlinear and excise is constrained to be linear, any excise tax will wind up being a pretty blunt way of trying to internalise harms. An excise tax that raises revenue sufficient to cover the total social cost imposed will necessarily overtax light and moderate drinkers while undertaxing those imposing larger social harms, if we expect that it's the heavy drinkers who impose large social costs on others rather than just large health costs on themselves.

Reasonable policy tries instead to focus on the relevant margins. An alcohol excise tax sufficient to internalise the external harms imposed by heavy drinkers will impose very strong penalty on moderate drinkers. Instead, policy should combine a rather lower excise component with penalties attached directly to the actions that impose external harms: penalties for being drunk and disorderly; penalties for driving while intoxicated; making alcohol use an aggravating factor in sentencing for other crimes.

A new bit of evidence that prices are a blunt instrument in dealing with harms imposed by heavy drinkers. I've typically cited Wagenaar's metastudy showing that heavy drinkers are roughly 60% as price responsive as moderate drinkers: a 10% price increase has drinkers in total reduce consumption by 4.4% but has heavy drinkers reduce consumption only by 2.8%. The University of Maryland's Jon Nelson has a new study out focusing on a smaller proportion of the population of estimates, but the ones with the finer-grained individual data that allows for demographic breakdowns. The abstract summarises things:
Gender differences in drinking patterns are potentially important for public policies, especially policies that rely extensively on higher alcohol taxes and prices. This paper presents a systematic review of alcohol prices and gender differences in drinking and heavy drinking by adults and young adults. Starting with a database of 578 studies of alcohol demand and other outcomes, 15 studies are reviewed of adult drinking including discussion of samples, measurement issues, econometric models, special variables, and key empirical results. A similar discussion is presented for eight studies of drinking by young adults, ages 18–26 years. Four conclusions are obtained from the review. First, adult men have less elastic demands compared with women. Second, there is little or no price response by heavy-drinking adults, regardless of gender. Third, although the sample is small, price might be important for drinking participation by young adults. Fourth, the results strongly suggest that heavy drinking by young adults, regardless of gender, is not easily dissuaded by higher prices. Policy implications, primary study limitations, and suggestions for future research are discussed.
Nelson here provides more of a literature review than a metastudy that would provide a new summary elasticity measure. A couple other interesting findings:
  • Bans on alcohol on university campuses do rather more to deter moderate drinking than to stop heavy drinking;
  • The number of licenced outlets near campus has little effect on drinking or on heavy drinking among young adults.


  1. "policy should combine a rather lower excise component with penalties attached directly to the actions that impose external harms"

    That's a bit too homo oeconomicus for me. I don't have research to point to, but anecdotal evidence suggests that the drunk are less responsive to far-off threats like possible future penalties to possible consequences of their actions, and also, that they are more risk-loving and likely to engage in wishful thinking.

    In other words, I would think that is also a blunt tool.

    I'm not sure about alcohol as a aggravating factor, but I think that at least it should not be a mitigating factor, unless people were forced to get drunk.

  2. Yeah, but the threat of consequences when drunk can reduce the likelihood of getting drunk. Think about how we run drink-driving laws: drunkenness isn't a defence.