Given the growing international interest in cannabis legalization, the possibility that increased cannabis use might lead to either decreased or increased alcohol use has important implications for public health, economic growth, and government policy (e.g., [15]). However, past medical and economic studies have offered conflicting results: while many found that increased cannabis use was associated with decreased alcohol use, others reported increased alcohol use or no apparent effect.

This study contributed to this debate by analyzing how Canada's per capita monthly alcohol sales changed as legal medical cannabis sales expanded. Unlike most previous research, this study analyzed sales in a country where the federal government had legalized and regulated medical cannabis, rather than one where cannabis was federally illegal; and the study used cannabis sales levels, rather than cannabis law implementation, as the explanatory variable. The results should interest researchers, industry regulators, and policymakers in Canada, as well as in other countries that are considering or pursuing legalization.

Research on the potential relationship between cannabis use and alcohol use has been underway for decades (e.g., [14]) but with mixed results. Reviewers of the medical [28] and economic [3, 9] literatures have noted that while many studies suggest increased cannabis use is associated with decreased alcohol use (i.e., the two substances are economic substitutes), other studies suggest that it is instead associated with increased alcohol use (i.e., the two are economic complements), or that there is no aggregate impact on average. The paragraphs below highlight recent examples of this research.

Only one previous U.S. study [23] directly compared alcohol sales to cannabis sales. It estimated a 15% decrease in Washington state's 2015 alcohol demand due to state-licensed recreational cannabis sales. By contrast, most research compared states that implemented medical cannabis laws (MCL) and/or recreational cannabis laws (RCL) to states without such laws. For example, Baggio et al. [6] analyzed grocery store scanner data and estimated that alcohol sales decreased 12.4% after MCL implementation. Wang et al. [40] found that online searches for alcohol-related terms fell 10.9% after RCL, while cannabis-related searches rose 16.5%. And respondents to several surveys have reported less binge drinking in MCL states [30] and/or RCL states [1, 13].

On the other hand, Veligatia et al. [38] used state tax revenues to estimate alcohol sales and found no significant change after MCL or RCL. Conyers and Ayres [11] found no significant association between medical cannabis dispensary locations and alcohol-related emergency department visits. Calvert and Erickson [7] analyzed alcohol sales as estimated from consumer surveys and found increases in some product categories but decreases in others after RCL. Wen et al. [41] found adults in MCL states were 10% more likely to report binge drinking. And Kim et al. [21] found respondents in RCL states were 24.6% more likely to report using both cannabis and alcohol in the past month, versus only 6.7% less likely to report using only alcohol; the MCL figures were 6.4% and 4.9%, respectively.

This ambiguity of results might be due to cannabis-alcohol relationships depending partly on the purpose, product, or context. For example, O'Hara et al. [26] found that college students using drugs to cope tended to choose cannabis as an alcohol substitute, whereas other students chose it as a complement. And some observers believe alcohol users are more likely to consider drinking cannabis beverages than smoking dry cannabis [20].

Another potential explanation for the ambiguity is that most research has employed cannabis law implementation as the explanatory variable, rather than measuring subsequent changes (if any) in cannabis sales. Apart from Miller and Seo [23], even studies that estimated changes in cannabis use after MCL or RCL (e.g., [21]) modeled those changes as legalization outcomes, rather than as alcohol predictors. This is an important consideration, as cannabis laws vary considerably in their details [27], implying their impacts on cannabis use likely vary too [38]. This indirect approach also meant that any observed alcohol changes were attributable to the cannabis law changes, and only by implication to presumed changes in legal or illegal cannabis sales. Finally, the U.S. federal illegality of cannabis, even in MCL and RCL states, might have affected consumers’ willingness to use cannabis and/or to disclose such use. For example, U.S. military veterans faced great difficulties accessing medical cannabis [25], whereas Canadian veterans had their medical cannabis expenses reimbursed by the government [39].

Aside from being an open theoretical question, the extent to which alcohol use increases or decreases with cannabis use also has important implications for public health, economic growth, and policymaking. If cannabis is complementary to alcohol, then cannabis legalization might benefit established alcohol producers as well as new cannabis firms; the amplified gains in employment and tax revenues could therefore increase legalization's financial appeals to policymakers. However, such synergy would also aggravate the public health harms of both substances. By contrast, if cannabis is a substitute, its legalization might be less problematic for public health but also less attractive economically, as cannabis sales could reduce alcohol-related employment and taxes. For example, Miller and Seo [23] estimated 40% of Washington state's cannabis taxes in 2015 were cannibalized from alcohol and tobacco taxes, an effect that Irvine and Light [19] suggested might impact Canada too.

The Canadian government first legalized medical use of cannabis in 2001, but sales only began growing meaningfully after the Marijuana for Medical Purposes Regulations (MMPR) were fully implemented in March 2015 [18]. Purchases totaled just $16 million in 2015’s second quarter but hit $161 million in 2017’s fourth quarter [17]. Despite that rapid growth, sales remained small relative to those of alcohol. For example, from October 2017 to September 2018, medical cannabis sales totaled just $595 million, versus $23.5 billion for liquor, wine, and beer stores [31, 32]. Furthermore, although MMPR applied nationwide, patient participation varied widely across Canada's 10 provincially regulated health care systems: for example, product shipments in April 2017 were 22 times higher per capita in Alberta than in Quebec [17].

On 17 October 2018, Canada's new Cannabis Act [8] took effect. This law largely retained the existing medical cannabis system but created a new recreational cannabis market to operate in parallel [10]. Recreational sales were restrained by production shortfalls for the first six months [4]; but they grew rapidly from spring 2019 onward as more licensed stores opened [5], and quickly surpassed medical sales.

Canada's cannabis legalization was, in effect, a grand policy experiment. It consequently provided a novel opportunity to estimate cannabis-alcohol relationships. Although federal legalization meant there were no all-or-nothing “treatment” and “control” groups as such, the wide provincial variations in patient participation provided ample contrasts for statistical analysis. Federal legalization also meant government agencies began collecting official cannabis sales data, much as they were already doing for alcohol.

Given this opportunity, this study estimated the relationship between legal sales of medical cannabis and alcohol in Canada, thereby addressing several interesting questions.

To what extent did alcohol sales change after legal medical cannabis sales began?

Were alcohol sales negatively associated with legal cannabis sales, suggesting the substances were substitutes; or positively associated, suggesting they were complements?