Study design and data sources
We used an interrupted time series (ITS) analysis to investigate if the dollar value of per capita alcohol sales in Canada changed during the first 9 months (March – November 2020) of the pandemic relative to prior years. We also examined changes in retail sales of essential and non-essential goods during the same time period. We used monthly sales data starting in January 2010 to November 2020 (latest available data at the time of analysis) from Statistics Canada’s Monthly Retail Trade Survey (MRTS). The MRTS estimates monthly sales (including “brick and mortar” and online) through repeated, cross-sectional, mandatory (response rate in 2018: 94.2%) surveys of retailers identified through the Canadian Business Registry [19]. We included sales from Quebec (QC), Ontario (ON), Manitoba (MB), Saskatchewan (SK), Alberta (AB) and British Columbia (BC), the six provinces for which Statistics Canada reports monthly alcohol sales. These provinces contain 93% of the Canadian population, and reported 93% of total alcohol retail sales (by value) for 2018/2019 [20].
We obtained the following data for provinces from Statistics Canada to calculate per capita sales and control for demographic confounders in our analyses; total population aged (15+), percent of population aged 20-29, and percent of population male [21]. We used assigned values from July 1st for each corresponding year of observation. We also obtained the monthly Consumer Price Index (CPI) for each jurisdiction from Statistics Canada to adjust for inflation and standardized all dollar values to January 2020 dollars [22].
Primary outcome
Our primary outcomes were the inflation adjusted monthly dollar value (in CAD) of per capita retail sales (alcohol, and essential/non-essential goods) for individuals aged 15 + .
Retailers were classified into different industry groups according to the 4 digit-level North American Industry Classification System (NAICS 2017) [19]. Alcohol retail sales were estimated from retailers classified under “Beer, Wine and Liquor Stores” (NAICS code 4453 – a retailer whose primary business is selling packaged alcoholic beverages including beer, wine and liquor). The dollar value of monthly sales is estimated through a monthly sample of retailers (n = 1865 retailers on average per month in 2018) and sales reports from provincial liquor authorities. Off-premise alcohol sales from establishments that sell products other than alcohol, such as grocery stores and convenience stores, are only partially captured in the MRTS. Among jurisdictions in this study, alcohol is sold in convenience stores in Quebec and in select grocery stores in Ontario, Quebec, and British Columbia. Lastly, the MRTS indirectly captures on-premise (e.g. bars and restaurants) sales as provincial liquor authorities and sampled retailers report sales to establishments licensed to sell alcohol for on-premise consumption [23].
Prior to the COVID-19 pandemic, the majority of research on alcohol sales has used detailed data to estimate changes in the volume of pure ethanol purchased/consumed per capita over time. However, in this study we assessed changes in the per capita dollar value of alcohol sold. This data was used for two reasons: 1) examining the dollar value of alcohol sold facilitates comparisons to other sectors of Canada’s economy, including services deemed essential and non-essential during the pandemic and, 2) detailed alcohol sales data are not publicly available in many countries including Canada, and require special requests to individual provincial and territorial alcohol regulators. Several other reports have similarly used the value of alcohol sold to monitor changes in population-level use in a timely fashion during the pandemic [12, 14, 24].
We defined industry groups as “essential retail sales” if they were deemed integral to preserving life, health and basic social functioning and non-essential sales as all other industries. In effect, essential industries were permitted to continually operate with in-person customers during periods of lockdown while non-essential retail sales could only operate online or with curbside delivery during lockdowns. The essential and nonessential designations were determined using provincial policy documents [25]. The following industry groups were declared essential in all six provinces: grocery stores, convenience stores, automotive repair and maintenance, gas stations, health and personal care, building materials and garden stores and general merchandise stores. Non-essential industry groups in all provinces included: furniture sales, electronics and appliances, clothing and accessories, and sporting goods, hobby, book and music retailers [19]. New and used car sales were considered non-essential in all jurisdictions except Saskatchewan. Alcohol retailers were not included in the essential sales category total in our analyses.
Statistical analyses
We calculated the inflation adjusted per capita (aged 15+) value of monthly sales (in dollars) for alcohol, essential and non-essential sales for each of the six provinces. We used the CPI to standardize sales to January 2020 using the following formula (monthly per capita sales * CPI score for month/ CPI score for January 2020). For alcohol sales we used the CPI values for on and off-premise alcohol retailers and for essential and non-essential sales we used the overall CPI values. We visualized the average retail sales from each of the six provinces to create a summary value for Canada. For each type of retail sales we calculated the relative percent change in the value of sales for the current year (March to November 2020) compared to the prior three-year average (March – November 2017 – 2019). We used segmented linear regression models to estimate the effect of the COVID-19 pandemic on changes in the value of alcohol, essential and non-essential sales. Our dependent variable for the three respective models was the monthly value of sales per capita (aged 15+). In order to generate stable time trends, we used all available sales data dating back to January 2010. The models included a binary pandemic indicator, (January 2010 – February 2020 compared to March 2020 – November 2020). The models included two continuous parameters for time; one for the time period before the pandemic and one for the time period during the pandemic. We included categorical variables for both month (12 categories) and province (6 categories) in the models to capture seasonality and differences among the provinces. We also included the percent of the population that is male and percent aged 20-29 as continuous covariates in convention with prior studies examining changes over time in alcohol sales [3]. Durbin–Watson tests for autocorrelation where significant (p < 0.001) for all three models so we used a first order monthly autoregressive structure.
We were primarily interested in three parameters in the model: the binary pandemic indicator, interpreted as the average monthly change in per capita spending during the pandemic; the pre- pandemic time parameter, interpreted as the monthly change in per capita sales before the pandemic; and the time during the pandemic parameter, interpreted as the monthly change in per capita sales during the pandemic. In order to understand how changes in the three outcomes (alcohol, essential and non-essential sales) differed, an additional regression model was fit with interaction terms between all variables in the model and the three outcomes. The interaction terms and the three key parameters were interpreted as the difference among outcomes. All tests of significance were two sided and we report 95% confidence intervals from the models. All analyses were conducted in R Version 4.2.