Young Adult Birthday Celebrations as Windows of Risk for Alcohol and Cannabis Use: 21^st Birthdays Compared to Other Young Adult Birthdays

Abstract

Objective:

Alcohol use among college students during 21^st birthday celebrations constitutes a well-known example of event-specific drinking when alcohol use is both pervasive and heavy. Less is known about how 21^st birthday alcohol use compares to other birthday celebrations during young adulthood, whether similar increases occur for cannabis use on 21^st birthdays, and whether the “21^st birthday effect” is similar for non-college young adults. Alcohol and cannabis use during 19^th to 25^th birthday celebrations were explored among college and non-college students.

Method:

Participants were 720 young adults ages 18 to 23 (M = 21.1, SD = 1.7) at enrollment who completed 24 monthly surveys and 204 reported on a 21^st birthday. Participants resided in a state where cannabis was legal and were asked the month following their birthday whether they engaged in alcohol and cannabis use as part of their birthday celebration.

Results:

Multilevel models found a 21^st birthday effect for alcohol use as individuals consumed over twice as many drinks on their 21^st birthday than would have been expected given age trends in birthday drinking, and this effect held for college and non-college students. A 21^st birthday effect for cannabis was not found.

Conclusions:

21^st birthday celebrations represent a high-risk drinking event for young adults in general, and the current findings suggest event-specific prevention programs targeting all young adults turning 21 is warranted. Unlike alcohol where turning 21 is associated with socially and culturally normative use, a similar 21^st birthday effect was not found for cannabis.

Introduction

Identification of particular time periods or life course events that carry heightened risks for heavy alcohol use creates windows of opportunities for targeted interventions (Neighbors, Foster, Fossos & Lewis, 2012). Among young adult college students, heavy drinking has been tied to weekends compared to midweek (Kushnir & Cunningham, 2014), holidays (e.g., New Year’s Eve, St. Patrick’s Day and 4^th of July) (Neighbors et al., 2011; Kushnir & Cunningham, 2014), Spring Break (Lee, Maggs, & Rankin, 2006), and tailgating and sporting events (Glassman, Dodd, Sheu, Rienzo, Wagenaar, 2010; Neighbors, Oster-Aaland, Bergstrom, & Lewis, 2006; Cadigan, Martens, Dworkin & Sher, 2018). 21^st birthday celebrations represent a well-established example of event-specific drinking (e.g., Lewis, Lindgren, Fossos, Neighbors & Oster-Aaland, 2009; Neighbors, Spieker, Oster-Aaland, Lewis & Bergstrom, 2005; Rutledge, Park, & Sher, 2008). In the United States, alcohol use becomes legal at age 21 and celebrations among many young adults commonly include activities such as “drink your age” where celebrants drink 21 shots of alcohol in a relatively brief period of time (Rutledge, Park & Sher, 2008). Drinking games and having drinks purchased for the celebrant are common features of 21^st birthday celebrations (Neighbors et al., 2014), and 21^st birthdays are often associated with exceeding celebrant’s prior maximum number of drinks and an elevated risk of negative consequences (Brister, Sher & Fromme, 2011; Lewis et al., 2009; Rutledge, Park & Sher, 2008). A majority of research has focused on 21^st birthday drinking among college student populations, and less is known regarding drinking that occurs on other birthdays during young adulthood and the extent that 21^st birthdays are events associated with elevated risk compared to other surrounding birthdays in young adulthood. Further, 21^st birthday alcohol use among those who do not attend 4-year colleges remains largely unexamined.

Little is known about the relationship between 21^st birthdays and cannabis use. Some research has suggested that cannabis use is higher on specific events such as April 20^th or “4/20”, Mardi Gras, and St. Patrick’s Day (Buckner, Walukevich & Henslee, 2018; Bravo, Pearson, Conner & James, 2017; Walukevich-Dienst & Buckner, 2019). Cannabis is currently legally sold for non-medical use in 11 states, and four additional states (Arizona, Montana, New Jersey, and South Dakota) approved legalization of non-medical cannabis via ballot measures during the November 2020 election. Every state that has legalized cannabis has set the legal age for purchasing, possessing, and using cannabis at 21 years. Thus, 21^st birthdays represent the first opportunity for individuals to legally engage in cannabis use in these states. To date, no research has examined the association between birthday celebrations including 21^st birthdays and cannabis use. Previous research has found that event-specific cannabis use was linked with celebration motives (Walukevich-Dienst & Buckner, 2019), but cannabis use has not been studied in relationship to birthdays during young adulthood. Understanding whether a similar phenomenon occurs for cannabis use during birthday celebrations, particularly for 21^st birthdays, is important because it would identify the need for targeted event-specific interventions around the birthday celebrations and also because of potential negative health consequences.

The present study utilizes a sample from Washington State where non-medical cannabis use has been legalized for adults age 21 and over since 2012 (and available in retail outlets since 2014) and examines alcohol and cannabis use on birthdays from age 19 to age 25. We hypothesize that, consistent with previous research, alcohol use will be highest for 21^st birthday celebrations compared to other birthdays. We also hypothesize that cannabis use will be highest for 21^st birthday celebrations as there may be similar cultural norms and pressures for using cannabis on this birthday when a celebrant is no longer legally restricted from use. Examination of a potentially differential effect of 21^st birthday by college status is exploratory. Prior research has found that college students drink more than non-college young adults and that, while past-year cannabis use did not differ by college status, past-month and daily cannabis use was higher among non-college students (Schulenberg et al., 2020). It is unclear as to how these patterns of use will translate to birthday celebration behavior.

Method

Participants and Procedures

Data were collected from young adults participating in a longitudinal study on alcohol use and young adult social role transitions (Patrick et al., 2018). A sample of 779 participants were recruited from the community using a variety of methods such as social media and advertisements (e.g., online and print). Eligible participants were 18 to 23 years of age at screening, had consumed alcohol in the prior year, and lived within 60 miles of the study office in Seattle. All participants were enrolled in the study between January of 2015 and 2016. Participants completed a baseline survey, followed by 24 consecutive months of online surveys. The present analyses utilize data from the month following the participant’s birthday, which included specific items about their alcohol and cannabis use during their birthday celebrations the prior month. The study procedures received approval from the University of Washington’s Institutional Review Board.

Of the participants enrolled in the project, 720 had data from at least one monthly survey pertaining to birthday alcohol or cannabis use, and these participants comprised the sample for the current study. Compared to those enrolled in the parent project but not included in present analyses (n=59), the present analytic sample was more likely to be female and Non-Hispanic Asian and less likely to be Hispanic and to have reported cannabis use in the month prior to enrollment (see Table A in the online supplement). These participants reported on 1,287 birthdays that occurred during the two years of monthly data collection and spanned 19^th to 25^th birthdays. Of these, 204 were 21^st birthdays. Compared to participants who only reported on one birthday (n=153), those who reported on two birthdays (n=567) were older and less likely to report heavy episodic drinking (HED; 4/5+ drinks during a two-hour period for females/males) and cannabis use at baseline (see supplement Table B). The mean age for the analysis sample at enrollment was 21.11 years (SD = 1.70), and 58% of the sample reported sex at birth as female. The sample was 55% Non-Hispanic white, 19% non-Hispanic Asian, 9% Hispanic, and 18% non-Hispanic other (of which, 24% were Black/African American, 3% American Indian/Alaska Native, 3% Native Hawaiian or other Pacific Islander, and 57% multiracial). At the beginning of the study, 47% were enrolled in a 4-year college and 61% were employed part or full-time. Only 10 participants were married, and only 7 were parents. At baseline, 88% of participants reported alcohol use during the past 30 days, 48% reported HED during the past 30 days, and 41% reported past 30-day cannabis use.

Measures

Birthday celebration substance use.

Participants were asked retrospectively about their alcohol and cannabis use during their birthday celebration in the month following their birthday. The primary measure of birthday alcohol use was a count of number of drinks based on the survey item: “How many drinks containing alcohol did you drink when celebrating your birthday?” Responses could range from 0 to 25 or more drinks. All alcohol use questions were asked in terms of number of standard drinks. For secondary models to illustrate birthday effects on drinking, we created dichotomous measures at different margins of drinking: any use, HED, and high-intensity drinking (8/10+ for females/males). The primary measure of birthday cannabis use was dichotomous and based upon the item: “Did you use marijuana to celebrate your birthday?” (coded 1 = yes and 0 = no). Participants who reported using cannabis on their birthday were asked how many hours they were high. Hours high, ranging from 0 to 24, was examined in supplemental analyses.

Covariates.

In addition to age, other covariates used in the analysis models included sex at birth (male = 0, female = 1) and race/ethnicity (mutually exclusive dummy-codes for non-Hispanic White, non-Hispanic Asian, Hispanic, and non-Hispanic other with non-Hispanic White serving as the reference group). For analyses of whether birthday substance use differed by 4-year college status, we included a variable for whether participants reported that they were enrolled in a 4-year college program at the time of the birthday month (0 = not in 4-year college, 1 = 4-year college student). Participants were instructed to report their current status as being in a 4-year college if they were on a scheduled break from school.

Data Analysis

We used multilevel models with random intercepts to account for clustering of up to two birthdays within participants (i.e., 1,287 time points nested within 720 individuals). For analysis of number of drinks consumed and supplemental analysis of hours high, a negative binomial distributional model was used, and the regression coefficients were exponentiated so that associations between covariates and count outcome are expressed as rate ratios (RRs). The difference in RRs relative to 1 can be interpreted as the percentage increase or decrease in the number of drinks associated with a one unit increase in the given covariate (Atkins, Baldwin, Zheng, Gallop, & Neighbors, 2013). Logistic models were used for the secondary drinking outcome measures of any drinking, HED, and high-intensity drinking, and for any cannabis use. Initial models included terms for sex, race/ethnicity, birthday age (centered at age 21), birthday age squared, and a binary variable for whether it was the 21^st birthday (coded 1 = yes, 0 = no). The inclusion of birthday age and birthday age squared captures the age trajectory of substance use across young adult birthdays. Birthday age squared (a second order polynomial coding of birthday age) allows for a quadratic trajectory that can account for an upward trend until birthday age 21 followed by a downward trend, given that substance use commonly peaks between the ages of 21 and 23. Most importantly, the model captures the deflection off this curve due to the birthday being the 21^st birthday.

To examine differences by college status, a second set of models was estimated restricting the birthday age range to birthdays between 19 and 22 to better reflect the age range of typical undergraduate students, and the trajectories of birthday substance use for participants in 4-year colleges were compared with those not in 4-year college. These models used 722 time points of data on 467 individuals. Since only four birthday years were being modeled, the age-squared term was dropped from these models as the potential downward trend after age 21 that would be captured by the quadratic term is not applicable in the restricted sample, and the models thus specified a linear association between substance use outcomes age. A binary variable for 4-year college status was added, as well as a 4-year college status by 21^st birthday interaction term to test if the effect of the 21^st birthday differed by 4-year college status.

Models were run with Stata 15.1 (StataCorp, 2017). Maximum likelihood estimation was used for the multilevel models, and unit-specific estimates (i.e., conditional on random effects) are provided. The “margins” command was used to produce model-predicted estimates for the generalized linear models and convert these estimates to prevalence and counts used in the figures. A small number of cases (<1%) in the analysis sample had data at a particular birthday time point on one type of substance use but not the other, and 21% of individuals in the analysis sample had data on only one birthday. For those with data on only one birthday, the multilevel modeling approach yields unbiased estimates for the sample population under the assumption that data are missing at random after accounting for measured covariates (Graham, 2012). Although 21^st birthday effects were the primary focus of the study, in order to estimate trajectories of birthday substance use across a broader age span, we included data from individuals who were too young or too old to experience a 21^st birthday during the study period. The models assume that age cohorts were drawn from the same population and that age cohort effects, independent of the effects of developmental age, are minimal. Sensitivity analyses were run using data only on individuals who experienced a 21^st birthday during the two years of monthly data collection.

Results

The average number of drinks participants reported drinking on their birthdays was 3.55 (SD = 4.58). Participants reported drinking at least one drink of alcohol on 65% of the birthdays and using cannabis on 17%. On their 21^st birthdays, nearly all (89%) drank and, on average, they drank in greater quantities (mean number of drinks = 5.88, SD = 5.55). (Observed means and prevalence for substance use outcomes by age and by educational status are reported in supplemental Table C.) Unconditional multilevel models indicated between-individual variance were statistically significant for both outcomes, with intra-class correlations (ICCs) of .40 for number of drinks consumed and .84 for any cannabis use.

Table 1 shows the estimates for the model predicting primary outcomes of number of drinks consumed and any cannabis use, and Figure 1 shows model-predicted values for drinking and cannabis use outcomes by birthday age at the sample mean of other model covariates. Age was positively associated with number of drinks, with respondents drinking less on pre-21^st birthdays (19^th and 20^th) compared to post-21 birthdays (22^nd through 25^th), although the estimated effect of age squared reflects a leveling off of this trend with later birthdays. The 21^st birthday itself was a large deflection off of this trajectory. Individuals consumed over twice as many drinks on their 21^st birthday than would have been expected given the overall age trajectory of birthday drinking (Rate Ratio [RR] = 2.32, 95% CI = 1.88-2.86, p < .001). A similar effect estimate for 21^st birthday was found in models limited to individuals age 19-20 at enrollment (RR = 2.31, 95% CI = 1.86-2.86, p < .001; see Table D in online supplement). The first two panels of Figure 1 show model-predicted number of drinks and prevalence of any drinking, HED, and high-intensity drinking by birthday age. Complete model estimates for dichotomous drinking outcomes are shown in Table E in the online supplement. These supplementary models indicated that the odds of any drinking, HED, and high-intensity drinking were all over three times greater for the 21^st birthday compared to other birthdays.

Table 1.

Estimates for Multilevel Models Predicting Number of Drinks Consumed and Any Cannabis Use on Birthday

	Number of drinks		Any cannabis use
Fixed Effect	RR	95% CI	OR	95% CI
Intercept	2.65	(2.16-3.25)	0.06	(0.02-0.21)
Female (male = 0, female = 1)	0.84	(0.69-1.01)	0.60	(0.31-1.18)
Race/ethnicity (Ref. = Non-Hispanic White)
Non-Hispanic Asian	0.64	(0.50-0.83)	0.10	(0.03-0.35)
Hispanic	0.94	(0.67-1.32)	1.11	(0.35-3.49)
Non-Hispanic other	0.89	(0.70-1.14)	0.72	(0.30-1.70)
Birthday age (centered at age 21)	1.16	(1.07-1.25)	0.86	(0.67-1.11)
Birthday age squared	0.97	(0.94-1.00)	1.01	(0.91-1.11)
21st birthday	2.32	(1.88-2.86)	0.76	(0.37-1.57)
Random Effect
Intercept standard deviation	0.85		3.39

Note. Birthday Ages: 19-25 Years; for number of drinks, n = 1,278 birthdays for 718 individuals; for cannabis use n = 1,286 Birthdays for 719 Individuals. RR = rate ratio, CI = confidence interval, OR = odds ratio, Ref. = reference group. Bold typeface indicates p < .05.

Figure 1.

Model-predicted Alcohol and Cannabis Use During Birthday Celebrations by Birthday Year

In contrast to increases in drinking with age, the age trajectory for birthday cannabis use bent slightly downwards, although both age and age-squared effects were not statistically significant. The effect of the 21^st birthday on cannabis use was also nonsignificant (Adjusted Odds Ratio [AOR] = 0.76, 95% CI = 0.37-1.57, p = .462). This pattern is illustrated in the “Any Cannabis Use” panel of Figure 1. Sensitivity analyses using individuals age 19-20 at enrollment yielded a similar estimate for the 21^st birthday effect (AOR = 0.61, 95% CI = 0.30-1.25, p = .180; see Table D in the online supplement). Supplemental analyses of hours high among those reporting cannabis use also found no 21^st birthday effects for number of hours high among those who reported cannabis use (RR=0.81, 95% CI = 0.62-1.05, p=.112; see Table F in online supplement).

The model-based estimates for tests of differential effects of 21^st birthday by 4-year college status are shown in Figure 2. (Full model estimates provided in Table G in the online supplement). The percentage of young adults in 4-year college on their birthdays ranged from 68% when turning 19 to 42% when turning 22. For the model predicting number of drinks, there was a positive main effect of 4-year college status (RR = 1.87, 95% CI = 1.42-2.46, p < .001), indicating that 4-year college students drank 84% more drinks on their non-21^st birthday celebrations compared to individuals not in 4-year college. The interaction between 4-year college status and 21^st birthday (RR = .64, 95% CI = 0.41-1.01, p = .053) was negative but not statistically significant. For cannabis use, the main effect for 4-year college status on birthday cannabis use was not statistically significant (AOR = 0.59, 95% CI = 0.27-1.27, p = .173), and the interaction term in this model was also negative but not significant (AOR = 0.27, 95% CI = 0.07-1.06, p = .061). As shown in Figure 2, for number of drinks the positive deflection of drinking off of the trajectory of birthday drinking from age 19 to 22 was slightly larger for those not in 4-year college than for college students; for cannabis use, there was a small positive deflection for individuals not in 4-year college and a small negative deflection for 4-year college students.

Figure 2.

Model-predicted Number of Drinks and Any Cannabis Use for 19^th through 22^nd Birthdays by 4-Year College Status

Discussion

The presence of a 21^st birthday effect for alcohol was consistent with previous studies finding that 21^st birthday celebrations are associated with increased alcohol use (e.g., Lewis et al., 2009). The present study found that alcohol use as part of 21^st birthday celebrations was quite common with nearly 90% of our sample reporting that they drank alcohol, over half of participants (53%) engaging in heavy episodic drinking, and nearly 1 in 4 (23%) engaging in high-intensity alcohol use. This study found that the average number of drinks consumed during 21^st birthdays was more than twice that reported during 20^th birthday celebrations and also substantially more than was reported during subsequent birthday celebrations.

Consistent with prior research (e.g., Schulenberg et al., 2020), the present study found more alcohol use among 4-year college students compared to non-4-year college young adults. Adding to the literature, the 21^st birthday effect on alcohol use was present, regardless of college status, with the increase being greater among those not in 4-year college. Clearly, 21^st birthday celebrations represent an event that is associated with increased risk for heavy alcohol use and, although this study did not look at negative consequences, we might reasonably expect that this increase in consumption is associated with an increase in negative consequences among young adults, regardless of college status. These findings reinforce the importance of continued prevention efforts targeting alcohol use during 21^st birthday celebrations (e.g., Neighbors et al., 2009; Neighbors et al., 2012b) among both college students as well as expanding these efforts to young adults who are not attending university.

Unlike alcohol, there was no evidence of a 21^st birthday effect for cannabis. Instead, cannabis use was quite stable across birthdays, participants who used on a birthday were likely to do so on their next and those who did not use on one birthday were unlikely to do so on their next. Several possibilities could explain the absence of a 21^st birthday effect for cannabis. First, cannabis becoming legal at age 21 is still novel, particularly compared to being able to use alcohol legally at that age. Cultural norms established around excessive alcohol use during 21^st birthday celebrations that are reinforced in popular culture may not, as yet, be established and depicted around cannabis use. Additionally, motivations for alcohol use may include greater celebratory and social aspects than cannabis. Lastly, cannabis, unlike alcohol, lacks public venues for consumption. At age 21 bars are legally accessible for the first time and represent an attractive setting for alcohol consumption. There is not an equivalent opportunity for onsite consumption of cannabis. As it is expected that more states will legalize cannabis use, it will be of interest to see if any states adopt a more permissive attitude toward public consumption as this could create the condition for event-specific cannabis use, including 21^st birthday celebrations.

Limitations and Future Directions

The present study was conducted in one of the first states to legalize non-medical cannabis use, and the community sample consisted of young adults who had used alcohol in the year prior to enrollment but had not necessarily used cannabis during that time period. Relative to the nationwide population of young adults in the United States, a relatively large proportion of the sample were either in college or had graduated from college. In addition, those included in the analytic sample differed on some sociodemographic characteristics compared to those who enrolled in the parent project but were excluded from the analysis sample because they did not provide data on birthday substance use. These selection issues limit generalizability, although the sample and its data did provide heterogeneity on primary variables of interest (i.e., birthday substance use, birthday year, and college status) and thus provide the basis for assessing associations among these variables. That some individuals enrolled in the study provided data on only one birthday and differed from those who reported on two with respect to substance use prior to enrollment in the project suggests the possibility that data were not missing at random and might bias results. For instance, it is possible that amount or type of birthday substance use affected whether individuals completed surveys in the month after their birthday and, more importantly, whether this type of nonresponse was differentially related to substance use by whether it was the 21^st birthday or whether the individuals was enrolled in college. Additional study limitations include data being self-report and potentially subject to errors of retrospective recall (e.g., hours high). Also, only 204 of the sample experienced a 21^st birthday during the course of the study, and models assume that age cohort effects were minimal.

Although we conducted supplementary analyses on hours high among those who reported cannabis use during birthday celebrations, prevalence of birthday use was fairly low (<20%), which prevented more fully examining intensity of use. Lower rates of birthday cannabis use might be expected given that we did not screen out abstainers, so caution should be exercised in comparing rates of alcohol and cannabis use. Our analyses are largely limited to describing prevalence of cannabis use during birthday celebrations and do not focus on different types of cannabis users (e.g., Pearson et al., 2017), consequences of cannabis use, or the potential effects of co-use or simultaneous use of both alcohol and cannabis, all suggested areas for future research on event-specific birthday substance use. Future research could examine within-person trajectories of alcohol and cannabis use over time to see whether increases in use, particularly cannabis use, occur in the months or years following the transition to legal age. Daily diary questions about the context of birthday celebrations, such as whether the celebration occurs in public venues like parties or bars (Lewis et al., 2008) could prove useful in further identifying factors contributing to higher risk alcohol use.

Public Health Statement:

A “21^st birthday effect” where alcohol use prevalence and quantity increased as part of 21^st birthday celebrations was found for all young adults; however, there were limited findings for a “21^st birthday effect” for cannabis in the current sample. Findings highlight the need to expand event-specific prevention efforts targeting 21^st birthday drinking to include non-college young adults.