A large Canadian survey spanning nearly two years of the COVID-19 pandemic found that higher household income, not lower, was linked to heavier drinking, and that racial and ethnic minority individuals at middle income levels had elevated odds of heavy episodic drinking compared to White individuals at the lowest income bracket. These findings underscore how income, race, gender, and age interact in complex ways to shape substance use risk during a public health crisis.

The COVID-19 pandemic altered substance use patterns in ways that public health systems are still grappling with. Most prior studies examined income or race as independent risk factors, but real-world vulnerability rarely operates along a single axis. If income, race, gender, and age interact to create pockets of elevated risk that standard screening tools miss, then clinical and public health responses built on simple socio-economic targeting may be directing resources to the wrong places. Understanding these intersections is essential for any clinician who screens for alcohol or cannabis misuse in a post-pandemic population.

The COVID-19 pandemic created a natural experiment in which economic disruption, social isolation, and psychological distress converged across socio-economic lines. This study, led by Somé and colleagues at the Centre for Addiction and Mental Health in Toronto, recruited approximately 1,000 English-speaking Canadian adults per wave across nine consecutive web-based surveys from May 2020 to January 2022. The research team applied an intersectional analytical framework, testing not just whether income, education, and employment predicted alcohol and cannabis outcomes, but whether those associations differed by gender, ethnoracial background, and age. Backward stepwise logistic regression with Bonferroni correction was used to model four outcomes: heavy episodic drinking, perceived increase in alcohol use, past-week cannabis use, and perceived increase in cannabis use.

The core findings challenge simple narratives. Moderate and high household income were associated with greater odds of heavy episodic drinking and self-reported increases in alcohol use, as was unemployment. Racial and ethnic minority individuals with household incomes of $40,000 to $79,999 had greater odds of heavy episodic drinking than White individuals at the lowest income tier. University education appeared protective, particularly for women and older adults, for both alcohol and cannabis outcomes. The authors note critical limitations: response rates ranged from only 10 to 18 percent, all measures were self-reported and retrospective, and the binary White versus racial minority grouping conceals important heterogeneity. The authors call for longitudinal, representative studies to determine whether these intersectional patterns reflect causal processes.

Beyond Income Alone: How Race, Gender, and Age Reshaped Alcohol and Cannabis Risk During COVID-19

We tend to assume that financial hardship drives drinking, that the lower you are on the economic ladder, the more you reach for a drink to cope. But a Canadian study spanning nearly two years of the pandemic found a more complicated picture: higher income was associated with heavier drinking, mid-income racial minority individuals drank heavily at rates exceeding low-income White individuals, and a university education appeared to protect women in ways it did not protect men. The rules of pandemic drinking turned out to depend heavily on who you were, not just how much money you had. What this study genuinely contributes is its intersectional architecture. By testing interaction terms rather than simply stacking variables, the authors reveal that income does not operate uniformly across demographic groups. A mid-income Black or Indigenous Canadian faced different odds of heavy episodic drinking than a mid-income White Canadian at the same household income level. That is a finding that standard models, which treat each predictor as independent, would never surface. Before any criticism, this matters. It validates what many of us in clinical practice have long observed: patients do not arrive with a single socio-economic label, and the confluence of identity markers shapes their risk in ways that simple screening questions about income or employment status cannot capture.

The central methodological problem, however, is that all of these findings rest on self-reported, retrospective assessments of substance use collected from a commercial web panel with response rates of 10 to 18 percent. In epidemiological terms, asking people whether they drank more during the pandemic than before is a perceived-change measure, not a prospective behavioral observation. It is the difference between measuring someone’s sleep with a polysomnogram and asking them over breakfast whether they think they slept well. Memory, mood, framing effects, and social desirability all color the answer, and those distortions are not random across income and racial groups. Similarly, if you surveyed only the people who answered the phone during dinner, you might conclude that most people are home at six in the evening, but you would be missing everyone who was not. The 10 to 18 percent who completed these surveys may systematically differ in their substance use habits, health concerns, and free time from the 82 to 90 percent who did not. This means the intersectional patterns the authors describe could be real population phenomena or artifacts of differential non-response, and we cannot currently distinguish between the two.

The aggregation of diverse racial and ethnic groups into a single binary variable further limits the precision of the race-income interaction finding. Grouping Indigenous, Black, Asian, Arab, and Latinx communities under “racial minority” for a health analysis is a bit like diagnosing all respiratory symptoms as “lung problems”: technically defensible when you have too few cases to do better, but obscuring clinically meaningful differences between populations with very different histories, stressors, and relationships to alcohol and cannabis. To a patient concerned about their pandemic drinking, I would say that this study reminds us stress and substance use do not follow a simple rulebook, and their patterns are worth discussing regardless of economic situation. To a colleague, I would emphasize the screening signal: mid-income racial minority patients may not self-identify as high-risk and may not be flagged by standard tools, so a more thorough substance use history is warranted. To a policymaker, I would urge caution: the direction of these findings is plausible and clinically resonant, but the data quality does not yet support major resource allocation decisions. Intersectional analysis is not methodological decoration. It can reveal risk concentrations at specific combinations of socio-economic and demographic characteristics that are invisible to main-effects models, and those concentrations may matter more for health equity than average group differences.

This study sits at the hypothesis-generating stage of the research arc. It does not replace prior single-wave pandemic substance use surveys but extends them by adding temporal breadth across four pandemic phases and, more importantly, by demonstrating that main-effects models of socio-economic risk are insufficient. The intersectional income-by-race finding for heavy episodic drinking aligns with broader structural vulnerability frameworks in public health but has not yet been replicated in longitudinal or probability-sampled populations, meaning it should be treated as a directional signal rather than an established fact.

From a pharmacological and safety standpoint, the study’s cannabis findings are particularly relevant for clinicians managing patients on medications with CYP450 interactions or those using cannabis alongside alcohol. The protective association between higher education and reduced cannabis use among women and older adults may reflect differential access to information about drug interactions and health literacy rather than a direct causal effect of education itself. Clinicians should consider incorporating brief, non-judgmental substance use screening into routine encounters with all patients, not just those who fit traditional socio-economic risk profiles, particularly during and after periods of widespread social disruption.

This is a repeated cross-sectional observational study using nine consecutive web-based surveys with pooled multivariable logistic regression. In the evidence hierarchy, it sits below cohort studies and randomized trials, generating associations rather than causal estimates. The single most important inference constraint is that different individuals were sampled at each wave, so apparent trends across pandemic phases may reflect changes in who responded, not changes in population-level behavior.

Prior pandemic substance use studies, including early Canadian and international cross-sectional surveys, generally reported increased alcohol consumption associated with psychological distress and economic disruption but rarely tested intersectional interactions. This study’s finding that higher income predicted heavier drinking is consistent with pre-pandemic epidemiological literature showing a positive income-alcohol consumption gradient in high-income countries, though the interaction with race and ethnicity adds a novel dimension. The protective effect of university education for women partially aligns with findings from European pandemic surveys showing that higher-educated women moderated their drinking more successfully during lockdowns.

The study challenges simpler accounts from single-wave pandemic surveys that treated socio-economic status as a monolithic predictor and extends the literature by demonstrating that intersectional modelling can surface risk patterns invisible to additive frameworks.

The most consequential analytic choice was the use of backward stepwise variable selection for the logistic regression models. Stepwise methods are known to be sensitive to multicollinearity and sampling variability, meaning that retained interaction terms may not replicate in an independent dataset. A pre-specified model based on prior theory, or a penalized regression approach such as LASSO, could have produced different retained terms and potentially altered which intersectional associations emerged as statistically significant.

Additionally, disaggregating the binary racial and ethnic minority category into its component groups, even at the cost of wider confidence intervals, could have revealed whether the income-race interaction is driven by one specific community or is consistent across all aggregated groups. This distinction has direct implications for intervention targeting.

The most likely overinterpretation is concluding that higher income causes heavier drinking during a pandemic or that mid-income racial minority Canadians are definitively at elevated risk. The cross-sectional design with self-reported retrospective perceived-change measures cannot establish causation regardless of statistical adjustment. The income-drinking association may reflect pre-existing consumption patterns, social drinking norms in higher-income groups, or differential non-response rather than a causal effect of income on pandemic-era drinking behavior.

A related misreading involves treating the binary racial and ethnic minority grouping as a meaningful clinical category. The aggregated finding masks substantial heterogeneity across Indigenous, Black, Asian, Arab, and Latinx communities, and should not be applied to any single group without disaggregated data.

This study contributes a multi-wave, intersectional analysis showing that income, race, gender, and age interact in non-additive ways to shape pandemic-era alcohol and cannabis use in Canada. It does not establish causation, cannot be generalized to all Canadians, and relies on self-reported data from a low-response web panel. For practice now, it is best understood as a directional signal: standard income-based risk stratification may miss clinically meaningful vulnerability concentrations, and broader, more nuanced screening approaches are warranted.

Does this study prove that higher income leads to heavier drinking?

No. The study found an association between higher household income and heavy episodic drinking, but because it is a cross-sectional survey, it cannot determine the direction of causation. Higher-income individuals may have had greater access to alcohol, different social norms around drinking, or pre-existing consumption patterns that the study could not measure.

Should people in racial minority groups be more worried about alcohol use?

The study found that racial and ethnic minority individuals at middle income levels had statistically elevated odds of heavy episodic drinking compared to White individuals at the lowest income level, but this finding comes from a broadly aggregated category that groups very different communities together. It is best understood as a signal that warrants further investigation, not a definitive risk label for any individual or community.

Does education protect against heavy drinking or cannabis use?

The study found that university education was associated with lower odds of heavy drinking and cannabis use, but primarily for women and older adults. Education may serve as a proxy for health literacy, social networks, or coping resources rather than being directly protective. These findings are associations and should not be interpreted as guarantees.

Can I apply these findings to myself or my patients?

These results describe population-level patterns from a Canadian web panel and cannot be applied to any specific individual. However, they do suggest that clinicians should consider substance use screening broadly across socio-economic groups and not assume that only lower-income patients are at risk for pandemic-related increases in drinking or cannabis use.

Have thoughts on this? Share it: