VoxEU Column Health Economics

Tobacco and alcohol: Interdependence in consumption?

Recent results based on a large German micro data set show that tobacco and alcohol are complements. Smoking bans are thus likely to reduce alcohol consumption too, but not by much. One cigarette less per day reduces drinking by 1% of a half-pint of beer. Smoke-free pubs are not in danger of becoming alcohol-free too.

Tobacco kills. No surprise then that smoking faces increasingly severe restrictions. Even in Germany, one of the last rich countries to introduce such restrictions, smoking was recently banned from public sector buildings and public transport, and smoking bans in bars and restaurants are being introduced state by state.

Recent evidence suggests that smoking bans can successfully reduce tobacco consumption but there is a popular belief that restrictions on tobacco merely encourage smokers to turn to other substances, such as alcohol. Since excessive drinking is also harmful, a full evaluation of the merits of restricting tobacco consumption needs to determine whether alcohol and tobacco are complements or substitutes.

A new empirical approach

Economists usually determine whether goods are complements or substitutes by estimating a system of demand equations and checking the cross-price elasticity. Jimenez and Labeaga (1994) and Picone et al. (2004) have done this for alcohol and tobacco, but their approach is hindered by insufficient variation in the two prices in a single market. The problem is that sin taxes typically move in tandem, so consumers in the same market usual face similar alcohol and tobacco price changes. To get around this, economists have relied on price variation across regions and across time, but this generates misleading results since the observed consumption variations are likely to reflect many unobserved developments such as changes in tastes.

In recent research, we suggest an alternative study design that does not rely on price data but statistically mimics a controlled experiment (Tauchmann et al 2007). Specifically, our approach aims at directly answering the question: “Imagine one could deprive individuals of tobacco, how would this typically affect the consumption of alcohol?” While it would be impossible to conduct such an experiment in practice, the tools of modern econometrics allow us to statistically mimic it in a non-experimental setting. In particular, we utilise instrumental variables that exclusively determine either the consumption of alcohol or the consumption of tobacco, but not both.

The instrument relies on the fact that the connection between parents’ and children’s consumption of tobacco is very strong, but there is no direct causal link when it comes to drinking. Hence, we can treat the fraction of variation in consumers’ tobacco consumption that is due to parental influence as if an experimenter had exogenously and intentionally varied it. This allows us to estimate the connection between alcohol and tobacco consumption.

We apply this new approach to data from the “Population Survey on the Consumption of Psychoactive Substances in Germany”, a large survey, begun in 1980, that addresses various drug related issues. We use information at the individual level for more than 25,000 respondents aged 16 to 39 years.

Less smoking, less drinking: results for Germany

Our empirical analysis suggests that there is a positive and statistically significant link between smoking and alcohol consumption, i.e. alcohol and tobacco are complements. Restricting access to tobacco, our results suggest, is also likely to reduce the consumption of alcohol. Yet, in qualitative terms, the effect seems to be very small. Our results suggest that one cigarette less per day results in roughly a tenth of a gram of alcohol less per day. This represents merely one-hundredth of a half-pint of beer.

Another interesting question is whether the pattern of interdependence varies across gender, as women and men on average exhibit quite different smoking and drinking patterns. Separate analyses for both genders, in fact, reveal a pronounced difference. For males, the effect of reduced smoking is substantially larger than for females, for whom this link is apparently insignificant.


Jimenez, S. and Labeaga, J. M. (1994). “It is possible to reduce tobacco consumption via alcohol taxation”, Health Economics 3: 231–241.

Picone, G. A., Sloan, F. and Trogdon, J. G. (2004). “The effect of the tobacco settlement and smoking bans on alcohol consumption”, Health Economics 13: 1063–1080.

Tauchmann, H., S. Göhlmann, T. Requate and Ch.M. Schmidt (2007). “Tobacco and Alcohol: Complements or Substitutes? - A Structural Model Approach”, CEPR Discussion Paper 6780.

105 Reads