Calendar Month Variation in Alcohol and Marijuana Use in a Community Sample of Young Adults

Abstract

Objective:

Alcohol use among adults varies by season, and drinking among 4-year college students has been linked to holidays and the academic calendar. Little research has assessed calendar-related variation in marijuana use. We examined calendar month variation in alcohol and marijuana use among a sample of young adults that was diverse with respect to college status and educational attainment.

Method:

Data were drawn from a community sample of 761 individuals (57% female; ages 18–23 at enrollment). Participants were surveyed monthly for 24 consecutive months. Multilevel models accounted for nesting of monthly data within individuals.

Results:

All measures of substance use showed evidence of variation by calendar month, with the exception of hours high in months when marijuana was used. Drinks per week and peak number of drinks were relatively elevated in summer months and in October and December; the prevalence of heavy episodic drinking was also elevated in October and December. Attending a 4-year college was associated with more drinking than not being enrolled in a 4-year college, regardless of whether a 4-year degree had been obtained. Students in a 4-year college reported more drinks per week than those not in a 4-year college in September and October, whereas the peak number of drinks converged across educational status in the summer months and in December. Across educational statuses, marijuana use was highest in April and December.

Conclusions:

Findings may guide the targeting and timing of substance use prevention programs by focusing efforts before and during higher risk months.

Identifying when young adults are at risk for hazardous substance use and which young adults are at the greatest risk during these periods can inform prevention efforts (Neighbors et al., 2007, 2011). Research suggests that alcohol use may vary with the calendar (Cho et al., 2001; Kushnir & Cunningham, 2014; Lemmens & Knibbe, 1993), with most studies of young adults using 4-year college samples (e.g., Del Boca et al., 2004). Students in 4-year colleges tend to drink more than young adults not attending 4-year colleges (Patrick & Terry-McElrath, 2017; Schulenberg et al., 2020), and it may be that calendar variation in substance use differs by educational status.

Calendar variation in alcohol use

Variation in alcohol consumption by month of the year has been documented in adult samples in both the United States and Europe, suggesting that alcohol use, and notably heavy alcohol use, increases during the winter holidays and summer months relative to other months (Carpenter, 2003; Knudsen & Skogen, 2015; Lemmens & Knibbe, 1993; Uitenbroek, 1996). Among young adults, studies have tied heavy drinking among college students to special events and holidays that offer opportunities for social drinking, such as spring break, orientation week, 21st birthdays, New Year’s Eve, college football games, and St. Patrick’s Day (e.g., Henslee et al., 2015; Lee et al., 2006; Lewis et al., 2009; Neal & Fromme, 2007; Riordan et al., 2015). Del Boca et al. (2004) examined weekly diary data from first-year students at a large university and found that increases in alcohol consumption corresponded with the first 4 weeks of the fall semester, as well as with Halloween, Thanksgiving, New Year’s, and spring break. In a separate study, Clapp and colleagues (2008) found that breath alcohol concentrations among college students walking near or on campus were higher in the spring and winter than in the fall, with seasonal variations accounted for by the timing of holidays and academic events. In a sample of Canadian college students, Tremblay et al. (2010) observed that students drank more at the beginning of each semester and less during portions of the semester when exams were administered.

Because research has often been limited to the academic year among 4-year college samples, we know less about alcohol use in young adult populations during the summer months. White and Swartzwelder (2009) found that heavy drinking in the summer was high among a sample of college-bound high school graduates. It may be, however, that college students do not increase drinking during summer months as older noncollege young adults do (Carpenter, 2003; Knudsen & Skogen, 2015) because college students are likely to be living with parents and away from college peers and associated social events. Miller et al. (2016) found small decreases in alcohol consumption and negative alcohol-related consequences during summer months compared with the academic year among a sample of 4-year college student heavy drinkers.

Calendar variation in alcohol use may differ for young adults who do not attend 4-year colleges. Among non–4-year college students, work schedules, holiday plans, seasonal variation in weather, or hours of daylight may be more salient with respect to drinking behaviors. In one study, Goldman et al. (2011) found that calendar-related variation in drinking did not differ for college and noncollege youth; however, the sample only included individuals ages 18 and 19.

Calendar variation in marijuana use

Little attention has been given to calendar-related variation in marijuana use. One potential source of variation is 4/20 events, which occur on April 20 and celebrate marijuana and marijuana culture (Lopez, 2019). Two studies based on data from small samples of 4-year college students positively linked marijuana use to 4/20, as well as spring break and Mardi Gras (Bravo et al., 2017; Walukevich-Dienst & Buckner, 2019).

Current study

We used monthly data from a young adult community sample to examine calendar month variation in alcohol and marijuana use and test whether calendar month variation differed by educational status. With respect to educational status, we considered whether individuals were enrolled in 4-year college and, for those not in 4-year college, we distinguished between those with and without a 4-year college degree, since educational attainment is associated with alcohol and marijuana use (Grucza et al., 2018; Han & Palamar, 2020). We grouped 2-year college students with individuals not attending 4-year college because 2-year colleges are generally commuter schools that do not offer social events or housing options tied to opportunities for social drinking.

Based on prior research, we expected to see calendar month alcohol variation among young adults attending a 4-year college, with increased drinking early in academic terms and during the month of spring break. We expected increases in alcohol use corresponding with December holidays among all young adults, and increases in drinking during summer months, but only for individuals not attending a 4-year college. Our study is exploratory with respect to how calendar month variation in drinking may differ by educational attainment among individuals not attending a 4-year college. Our study is also exploratory with respect to marijuana use, although we expected increases in use in April because of 4/20 celebrations.

Method

Participants and procedures

The sample included 761 young adults from a longitudinal study on substance use and young adult social role transitions (Lee et al., 2017; Patrick et al., 2018). Participants were ages 18–23 years at screening, had consumed alcohol in the prior year, lived within 60 miles of the study office in Seattle, WA, and enrolled in the study between January 2015 and January 2016. Beginning the first month after enrollment and a baseline survey, participants completed 24 consecutive months of online surveys. Participants could complete each monthly survey for 7–10 days at the beginning of each calendar month, and most survey items asked about experiences from the previous calendar month. We emailed Amazon gift card codes as compensation (up to $680 total). The University of Washington’s Institutional Review Board approved all procedures.

Of 778 participants who enrolled in the study and began monthly data collection, 17 were excluded from analyses because they did not complete at least one monthly survey in which both educational status and substance use were reported. The analytic sample’s mean age at baseline was 21.11 years (SD = 1.70), and 57% of the sample reported sex at birth as female. The racial composition of the sample was 59% White, 18% Asian, 5% Black, and 18% other (including Native American, Pacific Islander, and multiracial), and 9% of participants were Hispanic/Latinx. At the beginning of the study, 61% were employed, 10 were married, and 7 were parents. The 761 participants in the analysis sample completed an average of 20.02 (SD = 6.18) monthly surveys, with 77% completing at least 18 and 51% completing all 24. For analyses examining marijuana use, we limited the sample to 499 participants who reported using marijuana in the year before enrollment to match the alcohol use selection criterion for the study. The 499 participants included in the marijuana analyses did not differ significantly (p > .05) from the 262 who were excluded with respect to sex, age, or 4-year college status at baseline, but did by race/ethnicity, χ²(3) = 13.67, p = .003, with the marijuana users more likely to be non-Hispanic White (58% vs. 49%) and less likely to be non-Hispanic Asian (14% vs. 25%).

Measures

Substance use.

For alcohol use in the reference month, we used count measures of typical drinks per week (sum of the typical number of drinks reported for each day of the week; Collins et al., 1985) and the peak number of drinks consumed on any one occasion and a dichotomous measure of whether respondents reported any heavy episodic drinking (HED; threshold of 4+/5+ for females/males in a 2-hour period, depending on sex at birth; National Institute on Alcohol Abuse and Alcoholism, 2003). We also examined a dichotomous measure of any marijuana use. In months in which marijuana use was reported, we used a count measure of the number of hours high based on the number of reported days of marijuana use multiplied by the answer to the question, “On a typical day when you used marijuana, how many hours were you high?” Response options for that item ranged from 0 to 12 with 0 = 0 hours, 1 = 1 hour, and 12 = 12 or more hours.

Calendar month.

Calendar month was the month to which substance use questions referred in the monthly surveys. For regression models, dummy codes were used with January as the reference month. Substance use in January was close to average for most of the substance use outcomes examined.

Educational status.

Each month, individuals reported whether they were enrolled in school and, if so, what type. When participants were on break from college, they were instructed to report that they were college students. We defined educational status at each person-month based on three categories: (a) attending a 4-year college, (b) not attending a 4-year college and not having a 4-year college degree, and (c) not attending a 4-year college but having a 4-year college degree. These categories were time varying, with statuses changing during the 2 years of monthly data collection for some participants. When determining whether participants had a 4-year college degree (i.e., at least a bachelor’s degree) at a given time point, we used reports of degree attainment at baseline, Month 12, and Month 24.

There was heterogeneity within the two non–4-year college educational status categories that shifted during the study period as the sample aged. In the first month of the study, 54% of those in the “not 4-year college, no 4-year degree” status were enrolled in a 2-year college program and 14% had an associate or vocational/technical degree; by Month 24, 23% were in a 2-year program and 27% had an associate or vocational/technical degree. In the “not 4-year college, 4-year degree” status, the percentage who had a graduate or professional degree increased from 2% in Month 1 to 11% by Month 24. The distribution of statuses also shifted during the study period. In Month 1, 47% of the sample were attending a 4-year college, 34% were not in a 4-year college and not degreed, and 19% were not in a 4-year college and degreed. At Month 24, 33% were 4-year college students, 26% were not in a 4-year college and not degreed, and 41% were not in 4-year college and degreed.

For college students, we did not have information on the institutions they attended. The Seattle area has three exclusively 4-year colleges, the largest being the University of Washington, and five primarily 2-year colleges, two of which offer 4-year degrees. The status of individuals at these primarily 2-year schools depended on how they identified the program in which they were enrolled. All but one of the colleges in the Seattle area are on a quarter system, with the fall quarter beginning in late September, the winter quarter in early January, a weeklong spring break in March, and the spring quarter ending in June. Based on the total number of undergraduates attending each college in the area, we estimate that more than 90% of students in the study sample were in schools on a quarter system.

Covariates.

Person-level covariates included biological sex (0 = male, 1 = female), age at baseline (grand mean centered), and race/ethnicity (dummy codes for Hispanic, nonHispanic Asian, and non-Hispanic other [including Black, Native American, Pacific Islander, and multiracial], with non-Hispanic White as the reference group). We included time since enrollment in the study as a time-varying covariate, coded as the month of data collection (1–24) divided by 12 and thus in units of years.

Data analysis

Multilevel models with random effects for intercepts accounted for nesting of monthly assessments within individuals. Calendar month and educational status were time-varying predictors. To assess whether substance use outcomes varied by calendar month, we used multivariate Wald tests (Raudenbush, 2004) comparing the fit of models with and without estimates of effects for calendar month dummy variables. Similarly, to test whether calendar month differences varied by educational status, we used multivariate Wald tests to compare the fit of models with and without interaction effects between educational status and calendar month dummy codes.

For count outcomes, over-dispersed Poisson models were used; for dichotomous outcomes (i.e., any HED, any marijuana use), we used logistic models. Estimates of fixed effects in the Poisson models are presented as rate ratios (RRs; sometimes referred to as count ratios or incident rate ratios) that describe the proportional change in the outcome (e.g., number of drinks) associated with a one-unit increase in the covariate (Atkins et al., 2013). Fixed effects for logistic regression models are presented as odds ratios (ORs). Restricted maximum likelihood estimation with HLM 7 (Raudenbush, 2004) was used, generating population average estimates of fixed effects with robust standard errors. For Figure 1, predicted values by educational status and calendar month at the sample mean values for other model covariates were based on model estimates. This figure also shows reference lines for the model-predicted mean counts or prevalence by educational status across all calendar months.

Figure 1.

Model-predicted substance use by calendar month and current educational status. Error bars are 95% confidence intervals. Flat horizontal lines represent the mean prevalence or count by educational status across all calendar months. Model estimates for alcohol use based on 761 individuals who reported alcohol use in the year before enrollment in the study; estimates for marijuana use based on data from 499 individuals who used marijuana in the year before enrollment; and estimates for hours high from marijuana use based on 441 individuals who reported at least 1 month of marijuana use during the 24 months of monthly data collection.

The analysis approach makes the assumption of data being missing at random after taking model covariates into account (Graham, 2012). In addition to data being missing at certain months because of survey noncompletion, there was a small amount of missing data (<3%) on substance use outcomes because of participants skipping items in the monthly surveys. Sample sizes for models range from 14,793 to 15,188 person-months across 761 individuals for alcohol use outcomes. Any marijuana use was examined in 9,672 person-months for 499 individuals, and the number of hours high was examined in 4,171 person-months in which any marijuana use was reported among 441 individuals.

Results

Descriptive information

Across the 24 months of assessment, only 2% of the analytic sample reported no alcohol use in all months, and 15% reported no months with HED. Across all person-months, average drinks per week was 4.79 (SD = 7.46), and average peak number of drinks was 4.41 (SD = 4.30). HED occurred in 43% of person-months. Among participants who used marijuana in the year before enrollment, 11% reported no months of marijuana use during the 2 years of monthly data collection. Marijuana use was reported in 47% of all personmonths among participants who used marijuana in the year before enrollment. In months in which marijuana use was reported, average number of hours high was 51.67 (SD = 68.84).

Table 1 shows age at baseline, sex, and race/ethnicity by educational status at the first monthly time point. Participants not in a 4-year college with a 4-year degree were older than participants in the two other statuses. Educational status did not differ significantly by biological sex but did by race/ethnicity. Non-Hispanic Asian participants were underrepresented in the “not 4-year college, no 4-year degree” status compared with other racial/ethnic groups, and both non-Hispanic White and non-Hispanic Asian participants were overrepresented in the “not 4-year college, 4-year degree” status.

Table 1.

Age, sex, and race/ethnicity at the first monthly time point split by educational status

Variable	4-year College (n = 341) M (SD)	Not 4-year college, no 4-year degree (n = 250) M (SD)	Not 4-year college, 4-year degree (n = 138) M (SD)	F	p
Age	20.76 (1.51)	20.84 (1.66)	23.00 (0.77)	130.69	<.001
	n (%)	n (%)	n (%)	χ2	p
Female	199 (58)	133 (53)	88 (64)	4.21	.122
Race/ethnicity				19.76	.003
Non-Hispanic White	177 (52)	144 (58)	82 (59)
Non-Hispanic Asian	81 (24)	27 (11)	27 (20)
Non-Hispanic other	59 (17)	52 (21)	20 (15)
Hispanic	24 (7)	27 (11)	9 (7)

Tests of variation by calendar month and calendar-monthby-educational status

Wald tests of variability in substance use by calendar month were statistically significant (p < .05) for all three measures of alcohol use, and tests of differential effects of calendar month by educational status were significant for drinks per week and peak number of drinks (Table 2). Any marijuana use varied significantly by calendar month, but the test of educational-status-by-calendar-month interactions was not significant. For the number of hours high, neither the test of variation by calendar month nor the test of calendar-month-by-educational-status variability were significant. Based on these results, we present both a main effects–only model and a model with educational-status-bycalendar-month interactions for drinks per week and peak number of drinks, models with only calendar month main effects for HED and any marijuana use, and a model with no calendar month effects for hours high. Figure 1 shows model predicted values by calendar month and educational status based on main effects models for HED and any marijuana use and models including interaction terms for drinks per week and peak number of drinks.

Table 2.

Wald tests for variation by calendar months and differential variation by calendar month by educational status

Substance use outcome	Main effects of calendar month χ2 (df = 11)	Interaction effects of calendar month by educational status χ2 (df = 22)
Drinks per week	44.08***	46.18**
Peak number of drinks	95.34***	92.95***
Any HED	48.09***	33.15
Any marijuana use	29.82**	19.98
Hours high in marijuana use months	14.26	13.62

Notes: HED = heavy episodic drinking. For models of alcohol use outcomes, n = 14,793–15,188 personmonths among 761 individuals. Models of any marijuana use based on 9,672 person-months among 499 individuals. Models of hours high based on 4,171 person-months in which marijuana use reported among 441 individuals.

^**p < .01;

^***p < .001.

Multilevel models for alcohol outcomes

The main effects model indicated that drinks per week and peak number of drinks were relatively low in January, February, and November; higher in summer months and into October; and also higher in December (Table 3). October and December had the highest prevalence of HED, with odds 20%–21% higher in these months than in January. Main effects models also showed differences in drinking by educational status, although the pattern of statistically significant contrasts differed by outcome. The “not 4-year college, 4-year degree” status was associated with 16% fewer drinks per week and 7% fewer peak number of drinks compared with attending a 4-year college; the “not 4-year college, no 4-year degree” status was associated with 14% fewer peak number of drinks and 35% lower odds of HED compared with attending a 4-year college.

Table 3.

Effect estimates for models predicting monthly alcohol use outcomes

	Drinks per week				Peak number of drinks				Any HED
	Main effects		Model with interactions		Main effects		Model with interactions		Main effects
Predictor	RR	[95% CI]	RR	[95% CI]	RR	[95% CI]	RR	[95% CI]	OR	[95% CI]
Between-person Intercept	7.37	[6.38, 8.53]	7.15	[6.14, 8.32]	6.31	[5.72, 6.97]	6.45	[5.83, 7.15]	1.22	[0.98, 1.51]
Age (grand mean centered)	1.01	[0.96, 1.06]	1.01	[0.96, 1.06]	1.00	[0.97, 1.03]	1.00	[0.97, 1.03]	0.95	[0.89, 1.02]
Female	0.65	[0.56, 0.75]	0.65	[0.56, 0.75]	0.76	[0.68, 0.84]	0.76	[0.68, 0.84]	1.00	[0.81, 1.22]
Race/ethnicity (ref. = non-Hispanic White)
Non-Hispanic Asian	0.56	[0.46, 0.69]	0.56	[0.46, 0.69]	0.62	[0.53, 0.73]	0.62	[0.53, 0.73]	0.50	[0.38, 0.66]
Non-Hispanic other	0.81	[0.66, 1.00]	0.81	[0.66, 1.00]	0.80	[0.70, 0.91]	0.80	[0.70, 0.92]	0.77	[0.60, 1.00]
Hispanic	0.97	[0.75, 1.27]	0.97	[0.75, 1.26]	0.96	[0.81, 1.14]	0.96	[0.81, 1.14]	0.99	[0.68, 1.43]
Within-person (time varying)
Time since enrollment in years	0.97	[0.93, 1.02]	0.97	[0.93, 1.02]	0.97	[0.94, 1.00]	0.97	[0.94, 1.00]	0.85	[0.80, 0.91]
Educational status (ref. = 4-year college)
Not 4-year college, no 4-year degree	0.91	[0.82, 1.01]	0.94	[0.80, 1.10]	0.86	[0.81, 0.92]	0.80	[0.73, 0.89]	0.65	[0.55, 0.77]
Not 4-year college, 4-year degree	0.84	[0.77, 0.93]	0.92	[0.80, 1.05]	0.93	[0.87, 0.99]	0.91	[0.84, 1.00]	0.94	[0.80, 1.09]
Calendar month (ref. = January)
February	1.02	[0.96, 1.08]	1.06	[0.96, 1.16]	0.97	[0.93, 1.01]	0.98	[0.92, 1.05]	0.90	[0.81, 1.00]
March	1.04	[0.99, 1.10]	1.08	[0.99, 1.18]	1.01	[0.97, 1.05]	0.99	[0.93, 1.06]	0.94	[0.83, 1.06]
April	1.05	[0.98, 1.12]	1.08	[0.98, 1.18]	1.02	[0.98, 1.06]	1.02	[0.95, 1.08]	1.01	[0.90, 1.14]
May	1.09	[1.03, 1.16]	1.14	[1.03, 1.26]	1.06	[1.02, 1.10]	1.08	[1.01, 1.16]	1.05	[0.94, 1.18]
June	1.12	[1.05, 1.19]	1.15	[1.04, 1.27]	1.04	[1.00, 1.09]	0.96	[0.89, 1.03]	1.03	[0.91, 1.16]
July	1.13	[1.05, 1.21]	1.17	[1.03, 1.33]	1.06	[1.01, 1.10]	1.01	[0.93, 1.10]	1.02	[0.90, 1.15]
August	1.15	[1.07, 1.23]	1.13	[0.97, 1.30]	1.04	[1.00, 1.09]	0.97	[0.90, 1.05]	1.01	[0.90, 1.15]
September	1.16	[1.09, 1.23]	1.27	[1.15, 1.42]	1.02	[0.98, 1.07]	1.03	[0.96, 1.10]	1.08	[0.96, 1.21]
October	1.11	[1.05, 1.18]	1.23	[1.12, 1.35]	1.07	[1.03, 1.12]	1.10	[1.03, 1.17]	1.20	[1.08, 1.34]
November	1.03	[0.98, 1.09]	1.05	[0.97, 1.14]	0.96	[0.92, 1.00]	0.95	[0.90, 1.01]	0.96	[0.86, 1.07]
December	1.14	[1.08, 1.20]	1.10	[1.01, 1.20]	1.12	[1.08, 1.16]	1.03	[0.96, 1.10]	1.21	[1.08, 1.35]
Not 4-year college, no 4-year by
February			0.92	[0.79, 1.07]			0.94	[0.85, 1.04]
March			0.98	[0.86, 1.11]			1.12	[1.02, 1.23]
April			0.98	[0.82, 1.18]			1.06	[0.95, 1.18]
May			0.97	[0.81, 1.15]			0.98	[0.88, 1.09]
June			1.03	[0.86, 1.22]			1.27	[1.13, 1.42]
July			0.94	[0.78, 1.14]			1.12	[1.00, 1.26]
August			1.06	[0.87, 1.29]			1.17	[1.05, 1.31]
September			0.83	[0.71, 0.98]			1.00	[0.89, 1.12]
October			0.86	[0.72, 1.02]			0.94	[0.84, 1.06]
November			1.04	[0.90, 1.20]			0.99	[0.90, 1.10]
December			1.12	[0.97, 1.30]			1.25	[1.13, 1.38]
Not 4-year college, 4-year degree by
February			0.94	[0.84, 1.06]			0.99	[0.91, 1.08]
March			0.90	[0.80, 1.00]			0.98	[0.90, 1.07]
April			0.92	[0.82, 1.05]			0.98	[0.90, 1.07]
May			0.89	[0.79, 1.01]			0.94	[0.85, 1.03]
June			0.90	[0.79, 1.03]			1.09	[0.98, 1.21]
July			0.93	[0.79, 1.08]			1.05	[0.94, 1.16]
August			0.99	[0.84, 1.17]			1.10	[0.99, 1.22]
September			0.85	[0.75, 0.96]			0.99	[0.91, 1.09]
October			0.82	[0.73, 0.93]			0.97	[0.88, 1.06]
November			0.91	[0.81, 1.01]			1.01	[0.92, 1.10]
December			1.01	[0.91, 1.12]			1.11	[1.02, 1.20]

Notes: Bold typeface indicates p < .05. HED = heavy episodic drinking; OR = odds ratio; RR = rate ratio; CI = confidence interval; ref.= reference category. n = 14,793-15,188 person-months among 761 individuals.

As illustrated in Figure 1, models with interaction terms pointed to spikes in drinks per week unique to 4-year college students in the months of September and October. The interactions model for peak number of drinks indicated significant interactions between both non–4-year college statuses and the month of December, with the peak number of drinks converging across educational status in this month. There was also some convergence between the “4-year college” and “not 4-year college, no 4-year degree” statuses in the summer months.

Multilevel models for marijuana outcomes

As shown in Table 4 and Figure 1, the prevalence of any marijuana use peaked in April and December, with odds of any marijuana use in those months 17% and 13% higher, respectively, compared with January. Estimates for effects of educational status indicated nonsignificant differences by educational status for both any marijuana use and hours high, although the “not 4-year college, no 4-year degree” status was associated with 22% more hours high in marijuana use months compared with attending a 4-year college.

Table 4.

Effect estimates for models predicting marijuana use outcomes

	Any marijuana		Hours high
Predictor	OR	[95 % CI]	RR	[95 % CI]
Between-person Intercept	1.53	[1.17, 2.01]	49.14	[40.13, 60.17]
Age (grand mean centered)	0.95	[0.88, 1.04]	0.96	[0.90, 1.03]
Female	0.81	[0.63, 1.05]	0.56	[0.45, 0.70]
Race/ethnicity (ref. = non-Hispanic White)
Non-Hispanic Asian	0.46	[0.32, 0.66]	0.38	[0.27, 0.55]
Non-Hispanic other	0.72	[0.51, 1.02]	0.90	[0.69, 1.19]
Hispanic	0.75	[0.46, 1.21]	0.89	[0.51, 1.55]
Within-person (time varying)
Time since enrollment in years	0.88	[0.80, 0.97]	1.07	[0.96, 1.19]
Educational status (ref. = 4-year college)
Not 4-year college, no 4-year degree	0.96	[0.79, 1.17]	1.22	[0.99, 1.51]
Not 4-year college, 4-year degree	0.85	[0.67, 1.09]	0.97	[0.83, 1.13]
Calendar month (ref. = January)
February	1.01	[0.88, 1.14]
March	1.13	[0.99, 1.30]
April	1.17	[1.02, 1.34]
May	1.09	[0.96, 1.25]
June	0.95	[0.82, 1.10]
July	1.05	[0.90, 1.24]
August	0.96	[0.82, 1.12]
September	0.94	[0.82, 1.08]
October	0.97	[0.85, 1.11]
November	0.91	[0.80, 1.04]
December	1.13	[1.00, 1.29]

Notes: Bold typeface indicates p < .05. OR = odds ratio; CI = confidence interval; ref. = reference category. For model of any marijuana use, n = 9,672 person-months among 499 individuals; for model of hours high, n = 4,171 person-months among 441 individuals.

Discussion

The findings indicate that both alcohol use and whether individuals used marijuana varied by calendar month. Drinks consumed per week and the peak number of drinks consumed on an occasion increased in the summer months, October, and December, although the October increase in drinks per week was specific to 4-year college students and summer month increases in the peak number of drinks was specific to not being in a 4-year college without having obtained a degree. The findings that these measures of drinking were elevated in the summer and the holiday month of December are similar to findings from adult samples not limited to young adults (Carpenter, 2003; Cho et al., 2001; Knudsen & Skogen, 2015; Lemmens & Knibbe, 1993). The lack of summer increases in the peak number of drinks for those attending a 4-year college matches findings from Miller et al. (2016) and suggests that the summer effect on heavy drinking is dampened for students when they may be away from the 4-year college environment. HED was more prevalent in October and December than other months, pointing to the effect of holidays and other events that occur in these months.

The October effect on alcohol use found in this study is new, although it may be tied to the Halloween effect found by Del Boca et al. (2004) and the beginning-of-the-term effect noted by Tremblay et al. (2010). For most college students in the sample, the academic year began the last week of September, and September and October are months in which fraternity/sorority systems typically have “rush” week, during which they recruit and screen applicants at a series of parties. In addition to Halloween, October usually includes homecoming weekend. Most of those in a 4-year college likely attended the University of Washington, and many in the “not 4-year college, 4-year degree” status were recent graduates of that university. Drinking often accompanies fall football games (Neal & Fromme, 2007), which are watched by both students and recent graduates.

Unlike other studies of college students (e.g., Clapp et al., 2008; Del Boca et al., 2004), we did not see evidence of a spring break increase in alcohol use. Most of the colleges represented in our sample had spring break in March. The break, however, is 1 week, and we assessed drinking monthly. College calendar effects may differ between specific schools or schools in specific regions, which may account for differences in seasonal patterns found in our study compared with prior research.

The present study is one of the first to examine calendar variation in marijuana use. The prevalence of any marijuana use was highest in April, which may be tied to 4/20 celebrations (Bravo et al., 2017; Walukevich-Dienst & Buckner, 2019). Marijuana use prevalence was also elevated in December, perhaps linked to fewer work or school obligations during this month. Hours high in marijuana use months did not significantly vary across calendar months. Although the two measures of marijuana use capture important dimensions of marijuana use, a sample of more frequent marijuana users or measures that capture frequency of use, consequences, or marijuana potency (Cuttler & Spradlin, 2017; Pearson, 2019; Prince et al., 2018) may be more sensitive to seasonal variation.

Limitations

We found differences in alcohol use by educational status similar to findings from other research (Schulenberg et al., 2020), but estimates of effects of educational status are limited by heterogeneity in the non–4-year-college statuses. In secondary analyses, we used multiple strategies for coding non–4-year-college statuses (e.g., treating 2-year college as a separate category and including associate and vocational/ technical degrees in the criteria for degreed). Results were similar across coding schemes, with the primary contrast in substance use patterns being with attending a 4-year college. Importantly, our study does not disentangle how factors such as employment, living situation, and romantic relationship status overlap with educational status and influence calendar month variation in substance use. Also, our analytic models mix within-person variation in educational status with between-person variation in longer-term educational paths. We were primarily interested in calendar month variation and how the current 4-year college environment might moderate this variation. Therefore, we focused on differences in educational status in any given month rather than on longer-term educational paths that may be relevant for substance use (Fleming et al., 2012).

Despite the study’s longitudinal design, monthly measures included retrospective self-reported data for the previous month, which could have introduced retrospective memory bias. The sample came from the Seattle metropolitan area, which has a high cost of living, one large public university, and a young adult population largely composed of 4-year college students and 4-year college graduates. A larger sample of noncollege and nondegreed individuals might reveal different findings. Generalizability may be further limited by the sample living in a state that legalized recreational marijuana for adults. An additional limitation is that we focused on monthly variation in substance use, which does not capture variation attributable to day of the week and only imperfectly captures effects of events such as NewYear’s Eve, Halloween, and 4/20 that occur on particular days of the year.

Implications and future research

Despite these limitations, our study has strengths, including the examination of monthly data from a community sample of young adults. Findings can help inform the targeting and timing of substance use prevention. First, our findings support the use of preventive interventions timed to the beginning of the fall quarter or semester in order to decrease drinking among 4-year college students. This time of the year is a period of heightened drinking for this population, and students can be provided with alcohol protective behavioral strategies to avoid high-risk social events and reduce their alcohol consumption if they do drink (see Pearson, 2013, for a review). Second, for noncollege young adults, prevention efforts could focus on the month of December and summer months, which appear to offer increased opportunities for heavy alcohol use outside of the college environment. Such efforts should, of course, consider other important contexts such as employment, living, and family situation. Regarding marijuana use, prevention efforts tailored to the month of April and specifically targeting 4/20 could be useful. Future research examining if contexts and social roles beyond educational status (e.g., employment status, living situation) influence calendar variation in alcohol and marijuana use is needed to better understand how prevention strategies could be most effectively applied.

Conflict-of-Interest Statement

The authors report no conflicts of interest.