“It has for some time past been the policy of Great Britain to discourage the consumption of spirituous liquors, on account of their supposed tendency to ruin the health and to corrupt the morals of the common people.”
– Adam Smith (1776), The Wealth of Nations
Governments have long used taxation to correct for the socially costly overconsumption of alcohol. Pigou (1920) noted that by setting a tax equal to the marginal external costs of consumption, it is possible to fully correct for the externality. The idea is deceptively simple, but potentially difficult to implement in reality. Since his seminal work, much research has been dedicated to the design of corrective, or Pigouvian, taxes. This includes work on the measurement of externalities (e.g. Luca et al. 2015), corrective tax design in the presence of other distortionary taxes (Bovenberg and Goulder 1996), and general equilibrium considerations (e.g. Golosov et al. 2014).
Diamond (1973) studied optimal corrective taxes when there is heterogeneity in consumption patterns and the marginal externality of consumption. Ideally, we want each consumer to face a price increase equal to the marginal external costs associated with his consumption. In the classic Pigouvian setting there is one representative consumer, which means that a single tax rate can perfectly correct for the externality. However, when the marginal externality of consumption varies across people, a single tax rate can no longer achieve the first best. The price increase induced by the tax will be too high for some consumers (those with a low marginal externality) and too low for others (those with a high marginal externality).
There is much evidence to suggest that the externalities from alcohol consumption do vary considerably across drinkers. A heavy drinker on his tenth beer of the evening is more likely to have a larger negative effect on other people than a light drinker on his first. In the US, only 7% of the population are frequent binge drinkers, but they account for around 75% of the costs of excessive alcohol use (Centers for Disease Control and Prevention 2016). This means that we are in Diamond’s world and unable, assuming we can’t set consumer specific tax rates, to achieve the first best.
Varying alcohol taxes across products can improve welfare
In this setting, Diamond’s second best prescription is to set an alcohol tax rate equal to the average marginal externality across drinkers. However, in a recent paper, we show that it is possible to improve on this optimal single rate by varying tax rates across alcohol products (Griffith et al. 2017). Although the externality is generated by ethanol (pure alcohol), ethanol is bundled with other product characteristics that consumers may differentially value. It is correlation between demands for different alcohol products and the marginal externality of ethanol consumption that allows product tax rates to improve welfare, relative to a single rate applied to all ethanol.
To see why, consider a simple example in which there are two types of consumer, heavy drinkers and light drinkers, and two products, beer and wine. Heavy drinkers have a high marginal external cost associated with their alcohol consumption and drink only wine. Light drinkers have zero marginal external cost associated with their drinking and drink only beer. Ideally both types would face a tax equal to their marginal external cost. The optimal single tax rate that is applied to both beer and wine is equal to the average marginal external cost across the two types of drinker. However, we can do better than this by setting a high tax rate on wine, and a zero tax rate on beer – this targets the excess alcohol consumption of the heavy drinkers, while leaving unaffected the alcohol consumption of the lightest drinkers.
We show that this intuition generalises to a more realistic setup in which there is a set of differentiated alcohol products bought by consumers with different preferences. This logic is similar in spirit to that used in the tagging literature – Akerlof (1978) noted the potential value of using observable characteristics to identify (or ‘tag’) the potentially needy in order to improve the efficiency of the benefit system. We use the correlation in consumers' preferences for different alcohol products and their marginal externality to tag consumption that is likely to have high marginal external costs. We also show that the optimal tax rates depend on the correlation between consumers’ price responsiveness and their marginal externality.
Heavier drinkers buy different products and are more willing to switch between products in response to price changes
We implement our method using data on the UK alcohol market to show that our ideas have empirical relevance. In order to recover the optimal tax rates, we require estimates of consumers’ demands for different alcohol products, as well as an estimate of the externality of alcohol consumption.
Heavy drinkers buy a different mix of products to lighter drinkers. For instance, heavy drinkers prefer stronger alcohol products – i.e. those with more ethanol per 100ml, measured as alcohol by volume (ABV). Figure 1 shows that the average ABV content of products bought by households that consistently purchase fewer than 7 portions of ethanol per adult per week is around 13%, but this rises to over 18% for households that consistently buy in excess of 35 portions per adult per week. In part, this is due to the fact that heavy drinkers get a relatively high share of their ethanol from spirits and less from cider and beer, but also because they prefer stronger products within each of these broad segments of the market.
Figure 1. Average strength of alcohol products purchased
To estimate consumers’ demands for different alcohol products, we use a discrete choice demand model, allowing for rich variation in consumer preferences. We show that heavier drinkers respond differently to price changes, compared with lighter drinkers. Following an increase in the price of an alcohol product, the heaviest drinkers are more than three times as likely to switch to another alcohol product, rather than to not buying alcohol at all, than the lightest drinking households.
This willingness to switch between alcohol products means that the heaviest drinking households’ demand for ethanol is less price responsive than lighter drinking households. Figure 2 shows how different groups of households respond to a 1% increase in the price of all alcohol products. Households that consistently buy fewer than 7 portions of ethanol per adult per week reduce their ethanol demand by more than 2%, but households that consistently buy more than 35 portions per adult per week reduce their ethanol demand by only 0.7%.
Figure 2. Own price elasticity for ethanol
Welfare can be improved by setting higher rates on strong spirits
This variation in preferences for different products and price responsiveness across heavy and light drinkers provide scope to improve welfare by varying tax rates across products. To recover the optimal rates, we assume that the external costs of drinking are a convex function of ethanol consumption. We use evidence from the medical literature to calibrate this function. Although the magnitudes of the optimal tax rates depend on the precise calibration, as long as the marginal externality is higher for heavier drinkers, the qualitative pattern of optimal rates across products holds.
By moving to a system of optimal taxes that vary across 18 different types of alcohol (and that involves higher taxes on high-strength spirits, and bringing the taxes on cider in line with those on beer), the government can improve welfare by more than a billion pounds per year, relative to the current UK system. This reduces the gap between the current UK system and the first best (which would be achieved by setting consumer specific tax rates) by around a half. This system of taxation also improves on the optimal single rate applied to ethanol (a la Diamond) by around half a billion per year.
In this column, we have argued that we can exploit variation in consumers’ preferences for different products to improve the design of corrective taxes. We focus on an application to alcohol taxation, but the framework that we develop is well suited to other applications in which there are external costs of consumption that vary across people. For example, concern about obesity and the excess consumption of sugar has led to growing interest in sugar taxes. In this case, it is likely the marginal external costs of consumption vary across people. If obese people, or people who consume too much sugar, buy a different mix of products to lower risk individuals, then setting higher tax rates on those products is likely to be a more effective means of reducing socially costly sugar consumption.
We have focused on how to design such taxes when there is variation in the marginal externality of consumption. In order to do this, we have abstracted both from re-distributional concerns and supply side considerations. Kaplow (2012) shows that under certain assumptions, the income tax system can be adjusted to perfectly offset the re-distributional effects of corrective taxes. Nonetheless, in order to know what adjustments to the income tax system are necessary, we would need to work out the distributional effects of the optimal taxes we recover. We assume that taxes are fully passed on to consumer prices. If the patterns of pass-through vary across products, then this could change the optimal rates across products. We leave for future research the incorporation of imperfect competition between firms into the optimal corrective tax model described here.
Finally, we note that taxation is only one instrument available to governments to correct for the problems of excess alcohol consumption. Regulation is another popular tool, for example, governments can prohibit the sale of alcohol at certain times, or impose restrictions on the upstream market for alcohol. The interaction of these various policies is another topic for future research.
Bovenberg, A L, and L H Goulder (1996), “Optimal Environmental Taxation in the Presence of Other Taxes: General- Equilibrium Analyses”, American Economic Review 86 (4), 985-1000.
Centers for Disease Control and Prevention (2016), “CDC - Data and Maps – Alcohol”.
Diamond, P A (1973), “Consumption externalities and corrective imperfect pricing”, The Bell Journal of Economics 4 (2), 526-538.
Golosov, M, J Hassler, P Krusell, and A Tsyvinski (2014), “Optimal Taxes on Fossil Fuel in General Equilibrium”, Econometrica 82 (1), 41-88.
Griffith R, M O’Connell, and K Smith (2017), “Design of optimal corrective taxes in the alcohol market”, CEPR Discussion Paper 11820.
Jackson, C K, and E G Owens (2011), “One for the road: Public transportation, alcohol consumption, and intoxicated driving” Journal of Public Economics 95, 106-121.
Luca, D L, E Owens, and G Sharma (2015), “Can Alcohol Prohibition Reduce Violence Against Women?”, American Economic Review 105 (5), 625-629.
Pigou, A C (1920), The Economics of Welfare, McMillan & Co., London.
Train, K E (2003), Discrete Choice Methods with Simulation, Cambridge University Press.
US Department of Justice (2005), “Drinking in America: Myths, Realities, and Prevention Policy Report by Office of Juvenile Justice and Delinquency Prevention”.