Impact on wine sales of removing the largest serving size by the glass

Findings in context

Removing the largest serving size of wine by the glass in licensed premises had the hypothesised effect of reducing the volume of wine sold. This is in keeping with the wealth of literature assessing the impact of smaller serving sizes on food consumption [24] and the very limited literature on alcohol consumption, comprising 2 studies conducted in semi-naturalistic contexts [29].

The results of the current study suggest that when the largest serving size of wine by the glass (typically 250 ml) was not available, people shifted towards the smaller options (125 ml and 175 ml) and neither drank the equivalent amount of wine nor more, for example, by opting to buy wine by the carafe or bottle. The increase in sales was slightly larger for 125 ml servings compared to 175 ml serving. Given that in most premises the largest serving was 250 ml, this implies that people did not automatically choose the next available size (175 ml). It is not clear why this was the case. One possibility is that when customers who had planned to drink a large glass of wine (250 ml) were told it was not available, they planned to drink two 125 ml services but stopped after one. People have the tendency to consume a specific number of “units” (e.g., number of glasses or bottles, number of cookies or slices of cake), regardless of portion or package size [35]. This helps explain why smaller serving sizes reduce alcohol consumption: people tend to order a pre-set number of glasses, and with less alcohol in each glass they drink less overall.

The current study found no evidence that the intervention affected beer and cider sales, suggesting people did not compensate for their reduced wine consumption by drinking more of these alcoholic drinks. Importantly, there was also no evidence that the intervention affected total daily revenues, implying that participating licensed premises did not lose money as a result of removing the largest serving size for glasses of wine. This might reflect the pricing of glasses of wine, with 125 ml servings usually having a higher profit margin than 250 ml glasses [36]. Important to note is that the study was not powered to provide statistically meaningful data on secondary outcomes.

Strengths and limitations

To our knowledge, this is the first study to estimate the impact on sales—a proxy for consumption—of removing the largest serving of sizes of alcoholic drinks in pubs, bars, and restaurants. Further strengths include features of the design that reduce the risk of bias, including the use of objective measures to assess the primary and secondary outcomes, i.e., electronic records of sales, as well as the large number of participating premises and the high retention rate.

The study has several limitations. First, due to the complexity of the sales reports provided by participating premises, it was not possible to assess the sales of all alcoholic drinks. While we could reliably assess sales of wines, beers, and ciders—estimated to make up more than 70% of alcoholic drinks in licensed premises in the UK [32]—we were unable to assess sales of spirits or cocktails. It is not known, therefore, whether people compensated for reduced wine consumption by drinking more of these other alcoholic drinks. It is also not known whether customers compensated for the reduced serving sizes by drinking wines higher in % alcohol by volume (ABV). This possibility, however, is minimal given that there were very small variations in the % ABV of wines sold by premises, ranging from 12% to 14%, information very rarely mentioned in wine lists anyway. Also, wine lists did not change during the study and none of the premise managers interviewed at the end of the study mentioned any changes in the drinking patterns of their customers during the intervention period relating to wine strength. Second, we were also unable to control for the number of patrons visiting the premises during each study period, to assess whether differences in sales could be attributed to differences in how busy premises were. We attempted to take this into consideration by using total revenue as a proxy measure of busyness. Third, the majority of premises were in London and constituted a small proportion of those approached, which potentially restricts the generalisability of the findings. Fourth, the study used an A-B-A reversal design, which has a higher risk of bias than an experimental design, although the analyses accounted for notable events including public holidays and temperature fluctuations, which may have confounded the effects. Fifth, although outcomes were assessed using objective measures, they concerned sales rather than actual consumption. Measuring consumption directly at scale in these kinds of real-world settings is not feasible. Sales are, however, a valid—as well as practicable—proxy for consumption [37] and are commonly used in behavioural research [38–40]. Finally, the impact of removing the largest serving size of wine was assessed only during a four-week period. Whether observed effects are sustained over time remains to be assessed in future research.

Implications for research and policy

Although the intervention resulted in a relatively small reduction in the volume of wine sold by each premises (7.6% or 421 ml per day, equivalent to approximately 4 standard drinks or 5 units per premises), given that no level of alcohol consumption is currently considered safe for health with even light and moderate consumption contributing to the development of many cancers [41], such a reduction could meaningfully contribute to population health. In England, legal serving sizes of wine by the glass are 125 ml, 175 ml, and multiple of these sizes [42]. It is a requirement that the smallest size—125 ml—is offered to customers but there is no restriction on the largest serving size that can be offered. Restricting the sale of the largest serving of wine by the glass (250 ml) in licensed premises—similarly to the ban proposed by the mayor of New York City in 2012 on serving sizes of sugary drinks larger than 16 ounces [43]—could contribute to policies for reducing alcohol consumption at the population level and merits consideration as part of alcohol licensing regulations.

This is the first study to assess the impact of this intervention. The findings, therefore, require replication in future studies. If effects are replicated, it would strengthen the case for a such consideration.

It is unknown whether similar effects would result from removing the largest serving size for other types of alcoholic drinks including beer, as no real-world studies assessing this exist. We attempted to conduct such a study to estimate the impact of removing the largest serving size for beer—usually a pint (568 ml)—and replacing it with a two-thirds measure, but were unable to find any pubs, bars, or restaurants willing to do this from almost 2,000 contacted. This likely reflects that the pint has been the customary serving size for beer in the UK for centuries [44]. In contrast, it was not until relatively recently that licensed premises started serving 250 ml glasses of wine. Smaller glasses containing 125 ml wine were once considered the standard size for serving wine by the glass [45]. In the UK, this default has now been replaced with the 175 ml measure. The size of wine glasses has also increased in recent years, almost doubling in the last thirty years [46], and likely contributing to an increase in wine consumption [26].

Regulating serving sizes in licensed premises could help shift social norms for what constitutes an appropriate serving size [47], both for consumption out of the home, such as in pubs and bars, and for consumption at homes where most drinking occurs [48]. This possible indirect effect of the intervention awaits study. Whether regulating serving sizes interacts with existing alcohol controls concerning the pricing, marketing, and licensing regulations [22,23]—also awaits study.

Interventions that reduce serving or package sizes are generally less supported by the public than information-based interventions, such as health warning labels [49–51]. In the current study, managers reported receiving few complaints from customers when the largest serving size was removed (4/21 premises reported receiving some complaints). Although the intervention would potentially be acceptable to premises’ managers given there is no evidence that it can result in a loss in sales, were it to be implemented as part of alcohol control policies it would likely lead to opposition from the alcohol industry, given its potential to reduce sales of targeted drinks [52]. The impact of such opposition on policy-makers will in part be modified by the level of public support for the intervention [53,54]. Public support for this and indeed a range of policies in health and other domains is sensitive to evidence of the policy’s effectiveness [55,56]. Communicating the effectiveness of a policy to achieve a valued outcome increases its public support. While we are unaware of any studies examining this in the context of alcohol serving sizes, a policy that increased the price of alcohol by introducing a minimum unit price of £1 was supported by 63% of a representative sample of the English population when informed of its effectiveness at reducing crime and hospital admissions, compared with 43% when not given this information [57]. Research is needed to explore the acceptability of sizing interventions for reducing alcohol consumption, as well as methods for increasing low levels of public support.

Conclusion

Removing the largest serving of wine by the glass from the range of options available in licensed premises reduced the volume of wine sold. This suggests this is a promising intervention for decreasing alcohol consumption across populations, which merits consideration as part of alcohol licensing regulations.