You're here: Home / What we do / Publication archive / The Globe / Issue 1 2009 / Price and Alcohol Consumption

Price and Alcohol Consumption

Russell BennettsRussell Bennetts, Economic Research Officer of the Institute of Alcohol Studies, reviews a meta-analysis of the relationship of the price of alcohol and levels of drinking.

As more studies are published relating to the relationship between the price of alcohol and levels of drinking, a new meta-analysis can provide a useful means with which to survey the field. Alexander Wagenaar, Matthew Salois and Kelli Komro have provided such a systematic review that is thorough, well researched and up to date. While other narrative reviews and meta-analyses on this topic have previously been published, this paper’s originality stems from the epidemiological method used and the inclusion of papers too recent to have been included in other such reviews. Wagenaar et al. conclude that they ‘know of no other preventative intervention to reduce drinking that has the numbers of studies and consistency of effects seen in the literature on alcohol taxes and prices.’

A meta-analysis differs from a purely narrative review by providing a numerical estimate on the strength of a relationship and the significance of this. For this review, Wagenaar et al. looked at studies that examined the relationships between measures of alcoholic beverage price or tax levels and sales of alcohol or self-reported drinking. This provided them with quantitative estimates for the strength of the relationships between these variables, along with estimates of the variability or error of them. The authors were thus able to illustrate, using their fresh empirical method, the large body of evidence showing that beverage alcohol prices and taxes are inversely related to drinking.

Data collection for this review consisted of an extensive literature search; nine databases were searched for relevant papers. Theoretically this would have allowed the authors to find any paper dating back as far as 1823, but in the final review the oldest paper included came from 1970. In fact, the majority of the papers included in the final computations came from the past 20 years. Some studies were excluded based on a number of criteria, such as duplicate publications of a study/ data set, empirical studies that did not provide sufficient amounts of data, reviews that reported no new data and those not written in English. This procedure resulted in the selection of 112 appropriate papers for inclusion, containing 1003 empirical estimates of the relevant relationship under study.

The data collected by the authors came from papers that, while conceptually similar, used a wide range of statistical methods and, as such, this data was not initially easily comparable. For this reason, a coding method was utilised whereby the data was standardised into a comparable metric. For each estimate in the review, the measure of effect, its standard error, the effect’s significance level and the analysis sample size were coded. The statistical analysis then focused on estimating a standardised effects size r for each separate estimate of the underlying relationship of interest and calculating the standard errors of these estimates.

The r estimates can be interpreted as the standardised slope of the relationship between price/tax and consumption. This means essentially that the larger the r value, the more of an effect price/tax has for that beverage or drinking behaviour. The equivalent is true for the quoted mean elasticity estimates, which tend to be larger due to the averaging-out of individual differences when aggregating the population into larger statistical units. As the results show, the effect of price/tax on consumption differs greatly by beverage. Interestingly, the ranking of beverages by how much price/tax affects their consumption differed by whether the simple mean or the standardised effect size was used. Looking at the simple means, spirits, followed by wine then beer are most affected by price/tax. Whereas when looking at the r values, wine, then spirits, followed by beer, are most affected. Overall the findings are statistically strong and of a noteworthy magnitude.

The authors of this review are forthcoming with the weaknesses in their method. They admit that the exclusion of non-English language papers is likely to mean that their meta-estimates are inflated, noting that this is likely to be in range of 2%. Publication bias is also likely to be responsible for the final results being upwardly biased. Papers that find statistically significant findings are far more likely to have been published and thus included in this meta-analysis. However, while these two factors will have biased the results somewhat, the effect is unlikely to be large enough to have biased the conclusions of the review. It should be noted that such limitations to the model were surely an ex ante result of the chosen systematic review methodology and no evidence exists that efforts were made ex post, to choose for inclusion papers that would help bias the results in favour of a desired conclusion.

Unlike previous meta-analysis papers on this topic, Wagenaar et al. did not follow the econometric tradition as they explicitly cumulated the evidence by ‘weighting each estimated effect by the inverse of its variance’ and used random-effects models in assessing the ‘precision of the cumulative estimates.’ Naturally many of the papers included in their analysis used an econometric approach, as that is the predominant evidence available. This may be somewhat logically inconsistent in some eyes, but an epidemiological/ social science approach to this topic and data does present an illuminating addition to the evidence base.

Moreover, any issues a purist may have with their meta-analytical method, regarding the combination of differing sets of covariates, apply equally to other such reviews. The use here of random-effects models, rather than fixed-effects, to combine different studies allowed them to relax the assumption that the same underlying effect is being estimated by all studies and helped them take into account study-level variability.

In their conclusion, the authors focus on both relevant work in progress and the areas that require further research. They are keen to emphasise that the price elasticities presented in this paper, and in much of the economic literature on this topic, are not inherent properties of the drinks studied. Indeed, ‘results across studies suggest that the magnitude of price effects varies across groups, situations and times.’ Future research is required to ascertain more clearly how the effects of price/tax differ between communities or societies that have very high or very low consumption levels.

Social scientists will hopefully be able to shed further light on the complex interactions between the price of, and tax levels on, alcohol and the web of ‘individual, community and societal influences on drinking behaviour’ that differ across diverse social and cultural environments. More research is also required regarding how much of a tax rise is passed on in a price rise, as the latest research suggests this differs quite significantly across beverages.

An inquisitive reader of this review may be left pondering whether the results have direct relevance to reducing alcohol-related harm. There can be little doubt that much statistical evidence is presented to support the notion that increasing the price or tax levels on alcohol reduces its consumption. Intuitively, one may think that this will lead to a reduction in harm, but it would be useful to have empirical evidence to confirm or negate this notion. Wagenaar et al. have also considered this and are currently conducting a similar meta-analysis of price/tax effects on morbidity and mortality outcomes. Alongside this current review, their ongoing work should form an important basis for policy on alcohol taxation, the primary means by which most public bodies can affect the pricing of alcoholic beverages sold privately.

Wagenaar et al. have produced a review that is a highly valuable addition to the evidence base on the effects of price and tax on alcohol consumption. An important finding is the significant but low magnitude of effect of price/tax on heavy drinking when compared to overall drinking. This contradicts some other studies that have previously emphasised the high sensitivity of heavy drinkers to price/tax. Clearly the consideration of factors outside the economic remains imperative to the study of alcohol consumption. Nonetheless, this review has undoubtedly spun concretely an essential part of the ‘whole web’ of factors Wagenaar et al. describe as influencing drinking behaviour. This paper deserves a place on the desk of all researchers, both as a source of new insights and as an invaluable compendium of prior relevant research. Their future work is much anticipated.

Price and Alcohol chart