The Effects of Alcohol Warning Labels on Population Alcohol Consumption: An Interrupted Time Series Analysis of Alcohol Sales in Yukon, Canada
There is limited evidence that alcohol warning labels (AWLs) affect population alcohol consumption. New evidence informed AWLs were introduced in the sole government-run liquor store in Whitehorse, Yukon, that included a cancer warning (Ca), low-risk alcohol use guidelines and standard alcoholic drink (SD) messages. These temporarily replaced previous pregnancy warning labels.
The researchers test if the intervention was associated with reduced alcohol consumption.
An interrupted time series study was designed to evaluate the effects of the AWLs on consumption for 28 months before and 14 months after starting the intervention. Neighboring regions of Yukon and Northwest Territories served as control sites.
About 300,000 labels were applied to 98% of alcohol containers sold in Whitehorse during the intervention.
Multilevel regression analyses of per capita alcohol sales data for people age 15 years and older were performed to examine consumption levels in the intervention and control sites before, during, and after the AWLs were introduced.
Models were adjusted for demographic and economic characteristics over time and region.
Total per capita retail alcohol sales in Whitehorse decreased by 6.31% during the intervention.
Per capita sales of labeled products decreased by 6.59%, whereas sales of unlabeled products increased by 6.91%.
There was a still larger reduction occurring after the intervention when pregnancy warning labels were reintroduced (-9.97% and -10.29%).
Applying new AWLs was associated with reduced population alcohol consumption. The results are consistent with an accumulating impact of the addition of varying and highly visible labels with impactful messages.
Reduction in per capita alcohol sales from liquor stores with highly visible alcohol warning labels
An accumulating reduction in per capita alcohol sales from liquor stores was observed in the intervention site of Whitehorse in comparison variously with the baseline period in Whitehorse, with per capita sales in five outlying control regions in Yukon, and also after adjustment for total per capita monthly alcohol sales in neighboring NWT.
These statistically significant reductions were estimated in models that adjusted for a number of economic and demographic predictors of the level of alcohol consumption in different regions. It is noteworthy that in this remote area of Canada, per capita alcohol consumption estimated from sales was significantly higher in the outlying, control regions in Yukon, regions that also had a higher proportion of males, young adults, persons with low income, and non-Indigenous people.
It is also important to note that significant reductions in consumption were observed only in relation to alcohol products that received the manual application of some 300,000 bright yellow and red intervention warning labels and not among products that were not labeled.
In fact, there were significant increases in the consumption of unlabeled products in Whitehorse during the intervention. These products could not be labeled because they were from local or small producers, the containers were too small, or it was otherwise impractical to add labels (e.g., single containers of beer were exempt).
They represented only 3% of sales. Although it is possible that factors other than the absence of labels may account for this finding, the pattern of results is consistent with some customers selecting unlabeled products to avoid seeing the series of stark warning and health messages.
Is it plausible to attribute the observed reductions in per capita alcohol sales to the labeling intervention?
The central question raised by these results is whether it is plausible to attribute the observed reductions in per capita alcohol sales to the labeling intervention.
Against this interpretation is the scant evidence of changes in population consumption as a result of the much-studied introduction of U.S. warning labels in 1989. Furthermore, the greatest reduction in monthly sales was observed after the application of low risk alcohol use guidelines and Standard Alcoholic Drinks labels to product containers was completed at the end of July 2018 (Hobin et al., 2020).
In favor of the hypothesis that the labeling intervention had a causal role, these labels were strikingly different from their U.S. predecessors. They were developed over 4 years, during which the literature on what constitutes effective warning labels was carefully reviewed and both a randomized experiment and a focus group study were conducted to identify effective content and presentation. Thus, the labels presented messages for which there was low awareness at baseline but that both local stakeholders and alcohol users judged to be important information for consumers, that is, warnings of serious health risks for conditions prevalent in Yukon (e.g., colon and breast cancer), low risk alcohol use guidelines, and information about the number of Standard Alcoholic Drinks in alcohol containers to enable consumers to follow the guidelines.
The label design also followed best practices by using multiple colors, adequate size, and inclusion of images as well as text.
Furthermore, a case could be made that the effect size of the reductions in per capita sales reflected the intensity of the intervention. Thus, the smallest effect size (about 3%) occurred at the outset when about 100,000 of the new cancer and low risk alcohol use guidelines labels were applied to most containers for just 30 days.
Over the following 3 months, when the reduction in sales was 5%, there was intense media coverage of the study, which could have served to reinforce the labeling messages and intensify their effect even though no new labels were added.
During the third 4-month phase, approximately 200,000 low risk alcohol use guidelines and Standard Alcoholic Drinks labels were applied, and there was an effect size of approximately 7%.
The post-intervention phase included a change in labeling (i.e., the return of the small BD label that had been placed on alcohol containers for more than 25 years until the beginning of this study in November 2017).
The change in the warning label back to the BD label at this point itself could have created more discussion and attention to health aspects of alcohol consumption.
Last, significant increases in per capita sales were observed in models examining unlabeled products, indicating a measure of specificity for the effect of the intervention warning labels.
No ordinary commodity
Alcohol warning labels warn consumers of the potential dangers and health risks from products. In providing such information, warning labels also deliver a clear message to consumers that alcohol is not an ordinary commodity. After seeing the new label messages, shoppers may have stopped purchasing alcoholic beverages or decided to purchase fewer alcohol products than planned, and therefore the total or some types of products sales could be reduced during the study period.
It is important, however, to acknowledge both the advantages and limitations of the use of sales data to estimate the impacts of a policy intervention. The researchers followed international best practices to estimate local per capita alcohol consumption, estimating total recorded sales from official sources and expressing these as a rate for the proportions of local residents age 15 and older. Because Yukon has a government monopoly on the sale and distribution of alcohol, the monthly data provided on recorded sales provide an excellent and accurate record of off-premise sales across all the regions included. However, these would not include sources of unrecorded alcohol consumption such as homemade and travelers’ imports, although these are likely to be small, especially because Yukon is a fairly remote area.
The effect of the labeling intervention may also be lagged. This study did not examine any lagged effects of the labeling intervention on the consumption because of the short period observed after the labeling was implemented. However, the observed large effect for the post-intervention phase would be consistent with such an interpretation.
Last, the confounding effect of other social policies or factors may exist in Whitehorse. One candidate is the legalization of cannabis that occurred Canada-wide and was implemented on October 17, 2018 (Department of Justice, 2018), when the first government-run online and retail store selling cannabis opened in Whitehorse, midway through the post-intervention period. However, in separate analyses, no differences were observed in alcohol sales from before to after October 17, 2018.
Highly visible alcohol warning labels lead to significantly reduced per capita alcohol consumption
The researchers found that the introduction of new AWLs displayed on the containers of alcohol products sold in a major Yukon liquor store was associated with significantly reduced per capita alcohol consumption.
The accumulating effect size over time can be interpreted as being consistent with a causal effect of the labeling intervention, especially as an opposite change was observed for unlabeled products and no reductions were seen in two separate control regions within and outside Yukon where there were no changes in labeling practices.
The results are also broadly consistent with those from the self-report survey data collected before, during and after the labeling interventions.