Daily Associations between Affect, Drinking Motives, and Drinking Intensity among U.S. Young Adults

Abstract

Objective:

We investigated the relationships between daily affect, drinking motives, likelihood of drinking, and intensity of drinking, particularly high-intensity drinking (HID), in a sample of young adults. We also explored differences in our outcomes before versus during the early COVID-19 pandemic.

Method:

In the springs of 2019 and 2020, young adult drinkers (N=633) completed 14 consecutive morning surveys (each year) characterizing the prior day’s affect, motives, and alcohol use. We examined between-person and within-person associations of affect and motives with two outcomes: any drinking and drinking intensity on drinking days (1=moderate drinking [1–3 drinks for women, 1–4 drinks for men], 2=binge drinking [4–7 for women, 5–9 for men], and 3=HID [8+ for women, 10+ for men]).

Results:

Young adults reported higher positive affect on drinking days and higher negative affect on non-drinking days. On days when young adults reported greater enhancement motives, positive affect was strongly related to HID. During the early COVID-19 pandemic, young adults were more likely to report drinking, but did not drink more heavily unless they also reported drinking for social motives.

Conclusions:

These results suggest that heightened social, coping, and enhancement motives are risk factors for drinking in young adults. They also suggest that young adults perceive their mood to be better on drinking days, particularly when they were drinking to enhance positive affect. Results are consistent with a positive affect regulation model (i.e., drinking to increase positive affect), but not a negative affect regulation model (i.e., drinking to cope with negative affect).

Introduction

Excessive alcohol use continues to be a public health concern on a global and national scale (Centers for Disease Control and Prevention, 2019; World Health Organization, 2005). Drinking 4 or more drinks on one occasion for females and 5 or more drinks for males (i.e., binge drinking) is most common in young adults (Kanny et al., 2015; Patrick et al., 2019), particularly those under 21 (Esser et al., 2017), and results in more severe consequences than moderate alcohol consumption (Centers for Disease Control and Prevention, 2019). Past studies have found that up to 45% of young adults who binge drink also engage in high-intensity drinking (HID), or episodes during which females consume 8 or more drinks and males consume 10 or more drinks (Patrick, Terry-McElrath, et al., 2016), and that this level of consumption also peaks in young adulthood (Patrick et al., 2017; Patrick & Terry-McElrath, 2019). High-intensity drinkers also engage in binge drinking almost twice as frequently as non-high-intensity drinkers (Patrick, Terry-McElrath, et al., 2016), and there is evidence that HID incurs additional negative consequences above that associated with binge drinking (Patrick & Terry-McElrath, 2021; White et al., 2016). Although it was initially thought that HID characterized a specific subset of young adult drinkers, past studies have found that the tendency to engage in HID varies more within person than it does between people (Patrick, Cronce, et al., 2016; Patrick & Terry-McElrath, 2021). This finding raises questions about which proximal, daily factors are the strongest predictors of HID in young adults.

One of the proximal factors that predicts drinking in a young adult population is affect, frequently studied together with drinking motives (which are often focused on regulating affect). Many studies have examined the relationship between affect and drinking behavior at the between-person level (i.e., a person’s average affect is associated with a person’s overall or average alcohol consumption). Using this measurement level, studies examining positive affect have found either that young adults with generally higher positive affect report lower alcohol consumption (Hussong et al., 2001) or no relationship between a person’s general level of positive affect and alcohol consumption (McCreary & Sadava, 2000; Rankin & Maggs, 2006). In between-person analyses of negative affect, studies have largely found that overall levels of negative affect are not related to alcohol consumption for young adults (Dermody et al., 2013; Hussong et al., 2001; McCreary & Sadava, 2000), but one study found that higher general negative affect was associated with more drinking days (Rankin & Maggs, 2006).

In order to make inferences that affect on a given day might influence drinking, there is a growing body of literature dedicated to studying the within-person, daily relationships between affect and drinking behavior, but these studies have not settled on consistent conclusions. For positive affect, several studies have found that higher positive affect predicted alcohol consumption in the same day (De Leon et al., 2020; Dvorak et al., 2018; Emery & Simons, 2020; Howard et al., 2015; Hussong et al., 2001; Patrick et al., 2015), although some studies have not found a significant relationship between daily positive affect and drinking (Fairlie et al, 2019; Stevenson et al., 2020). For negative affect, the majority of within-person studies have found no association between negative affect and drinking in young adults (Emery & Simons, 2020; Fairlie et al., 2019; Howard et al., 2015; Hussong et al., 2001; Stevenson et al., 2020), although some studies have found an inverse relationship (De Leon et al., 2020; Patrick et al., 2015; Treloar et al., 2015).

Both positive and negative affect may lead to drinking when young adults also have the goal of enhancing or improving their affect via alcohol use, consistent with the motivational model of alcohol use (Cooper et al., 1995; Cox & Klinger, 1988). In particular, internal drinking motives (enhancement and coping) include a desire to regulate affect and fluctuate meaningfully within-person from day to day (O’Hara et al., 2015). Enhancement motives underlie drinking to enhance one’s positive affect, and coping motives are drinking to reduce one’s negative affect (Cooper, 1994).

Social motives (e.g., to have fun with friends) can also be thought of as a desire to increase positive affect by drinking with peers (Cooper, 1994). Social motives are particularly relevant in a young adult population (Read et al., 2003) and are the most commonly endorsed motives in young adults (Kuntsche et al., 2005; O’Donnell et al., 2019). Studies at the between-person level have found a variety of associations between social motives and alcohol consumption, with findings including no association (Read et al., 2003), a positive association (Lyvers et al., 2010), or that social motives predict only moderate alcohol consumption (Kuntsche et al., 2005). Enhancement motives have often been associated with heavy drinking at the between-person level (Cooper, 1994; Kuntsche et al., 2005). For coping motives, between-subjects literature largely has found that young adults who drink to cope tend to also have more drinking-related problems, but that coping motives are not directly related to alcohol consumption (Kuntsche et al., 2005; Lyvers et al., 2010). However, some studies have found no relationship between coping motives and alcohol outcomes in young adults (Read et al., 2003). Relatively fewer studies have examined HID as a specific outcome, but studies so far have found that coping motives, or drinking to get away from problems, were related to HID at the between-person level in young adults (Patrick, Evans-Polce, et al., 2017; Terry-McElrath et al., 2017).

In studies examining the within-person, daily relationships between motives and drinking in young adults, studies have often found a strong relationship between enhancement motives and increased alcohol consumption that day, similar to between-person findings (Hamilton et al., 2020; O’Donnell et al., 2019; Patrick & Terry-McElrath, 2021). In particular, the relationship between enhancement motives and alcohol consumption is stronger when positive affect is higher (Dvorak et al., 2014; Gautreau et al., 2015; Stevenson et al., 2019). Social motives have not been as consistently included in within-person studies of drinking, but some studies have found that social motives predict social consumption (i.e., drinking with others present; Hamilton et al., 2020) and HID (Patrick & Terry-McElrath, 2021; White et al., 2016), although others have found no association with alcohol use (Blevins et al., 2018; O’Donnell et al., 2019). When previously tested, social motives did not interact with affect to predict alcohol outcomes (Hamilton et al., 2020). Thus, most studies have found a strong relationship between enhancement motives and drinking in young adults, and this relationship is intensified by heightened positive affect. Further, there is preliminary evidence that social motives predict HID, but the question of whether social motives interact with affect has not been frequently investigated.

For coping motives, within-person studies have found mixed results. One study found, for HID specifically, a within-person association between coping motives and HID (White et al., 2016) and there has been some evidence of an interaction such that alcohol use was higher on days characterized by the combination of higher negative affect and coping motives (Armeli et al., 2008; Dvorak et al., 2014). Most studies, however, have found that coping motives on a given day did not have a main effect predicting alcohol use or HID in young adults (O’Donnell et al., 2019; Patrick & Terry-McElrath, 2021), even regardless of negative affect in some instances (Hamilton et al., 2020; Stevenson et al., 2019).

Taken together, these findings suggest that, at the within-person level, positive affect is associated with drinking when enhancement motives are elevated in young adult samples. Findings are mixed on whether negative affect and coping motives (as an interaction) are associated with drinking in young adults. Lastly, social motives consistently predict consumption, but it is not known whether they also interact with affect to predict increased consumption. The current study extends previous work by examining how daily affect and drinking motives interact to predict drinking intensity, with a focus on predictors of HID.

In addition to affect and motives as important proximal factors influencing alcohol use, the onset of the COVID-19 pandemic caused widespread changes to occupational, educational, and social routines, including alcohol use. Studies in adolescents (Dumas et al., 2020) and adults have reported overall increases in substance and alcohol use as a result of COVID-19 and related stressors (Czeisler et al., 2020; Jacob et al., 2021; Koopmann et al., 2020; Rodriguez et al., 2020). However, a sizeable portion of adults (15–20%) have also decreased their use (Czeisler et al., 2020; Koopmann et al., 2020), particularly younger ones (Chodkiewicz et al., 2020; Neill et al., 2020), and evidence in college students has pointed to an overall reduction in substance use and alcohol consumption (Bonar et al., 2021; Martinez & Nguyen, 2020). Because the current study collected data through March and April 2020, when statewide lockdowns began in response to the COVID-19 pandemic, we have also included exploratory aims to examine changes in drinking intensity and hypothesized relationships between affect and motives and drinking in U.S. young adults after the onset of the pandemic.

The Current Study

Using longitudinal, daily morning reports of retrospective mood and alcohol use in young adult drinkers in the U.S., we predicted that: Hypothesis 1 (H1): Consistent with a positive affect regulation model, reported daily positive affect will be associated with likelihood of drinking and drinking intensity on the same day, and this relationship will be stronger on days for which greater enhancement motives were reported. H2. Consistent with a negative affect regulation model, we hypothesize that reported negative affect will be associated with higher drinking intensity on days when stronger coping motives are also reported. Additionally, we will explore the association between reported negative affect and drinking but do not propose a hypothesis for its direction. H3. Social motives will be associated with likelihood of drinking and drinking intensity in the same day. We will also explore the impact of affect on this relationship but do not have a hypothesis about this interaction. Lastly, due to the onset of the COVID-19 pandemic during previously-planned data collection, we will explore the differences in alcohol use and drinking intensity before the pandemic compared to the early pandemic period (3/11/20 to 4/12/20), as well as whether the hypothesized relationships differed during the pandemic.

Method

Participants

Data came from the first two waves of the Young Adult Daily Life (YADL) Study. YADL (Patrick & Terry-McElrath, 2021) is based on a nationally representative sample of 12^th grade students (N=14,502) in the U.S. who participated in the Monitoring the Future (MTF) study in Spring 2018 (Schulenberg et al., 2019). For the YADL study, participants were eligible if they self-reported past 30-day alcohol use in 12^th grade. A total of 4,240 MTF 12^th grade respondents reported past 30-day drinking and were eligible for the YADL study. A subset (N=828) were excluded because they were randomly selected for participation in the MTF panel study (Schulenberg et al., 2019), and 1,208 were excluded for not providing contact information necessary for follow-up (Patrick and Terry-McElrath, 2021). Out of the 2,204 individuals eligible to participate in YADL, 1048 participated in waves 1 and/or 2 of the study. Respondents could receive up to a $100 incentive for participation each year. This study was approved by a University of Michigan Institutional Review Board.

Procedure

Participants were assessed via online annual surveys approximately one and two years after their 12^th grade surveys. Following each annual survey, participants completed up to 14 consecutive morning surveys reporting alcohol use, motives, and affect for the day prior. Participants had up to three days to complete each daily survey. The current study analyzed these daily morning reports. The total sample included 1,048 participants with 18,509 days of data across waves 1 and 2. The present study included participants who reported alcohol use in at least one of the 28 daily surveys (60.4%, N=633). The analytic sample of 633 individuals was used in all analyses described herein, and after accounting for missing data at the day level, the sample of 633 individuals provided a total of 13,774 days of data with a total of 2,109 drinking days.

Measures

Any drinking.

Each day, respondents were asked whether they consumed alcohol the previous day. Any drinking was coded as 1 and no drinking was coded as 0.

Drinking intensity.

If participants reported drinking the previous day, they were asked how many standard drinks they consumed, from 1 to 25+. We created an ordinal measure for intensity of drinking on drinking days, using sex-specific thresholds (1=moderate drinking coded as 1–3 drinks for females and 1–4 drinks for males, 2=binge drinking coded as 4–7 drinks for females and 5–9 drinks for males, or 3=high-intensity drinking [HID] coded as 8+ drinks for females and 10+ drinks for males [Patrick, 2016]).

Affect.

We assessed positive and negative affect via the 20-item Positive and Negative Affect Schedule (Watson et al., 1988). For each day, the survey questions asked, “On this day, to what extent did you feel…” with response options such as “excited” or “interested” for positive affect, and “upset” or “lonely” for negative affect (range: 1 = very slightly or not at all to 5 = extremely). Scores across 10 items each for positive and negative affect were averaged separately (20 total items were used; Howard et al., 2015). The day-level alpha was .91 for positive affect and .88 for negative affect.

Drinking motives.

Drinking motives were assessed for drinking days only. We measured motives via items from the Drinking Motives Questionnaire (Cooper, 1994), adapted for use at the daily level (Patrick & Terry-McElrath, 2021). Drinking motives were 13 items based on the stem, “Why did you drink on [this day]?” with response options of no (0), somewhat (1), or definitely (2). Subscales included enhancement motives (2 items, day-level α=0.62: because I liked the feeling; to have fun), social motives (2 items, day-level α=0.86: to improve a party/gathering; to make a party/gathering more fun), and coping motives (7 items, day-level α=0.83: to avoid dealing with my ongoing problems; to feel less depressed; to cheer up; to forget my ongoing problems/worries; to feel more confident/sure of myself; to feel less nervous; because I was angry). A fourth subscale assessing conformity motives was excluded because both relevant items had very low endorsement in this sample of young adult drinkers.

Covariates.

We included two covariates at the day level: Weekend days (1=Thursday, Friday, or Saturday) vs. weekdays (0=other days), and annual survey timing. Annual survey timing was coded as 2019 (February to April 2019), pre-COVID 2020 (2/11/20–3/10/20), and during-COVID 2020 (3/11/20–4/12/20). Pre-COVID 2020 was used as the reference group in regression analyses.

At the person level, covariates included biological sex, race/ethnicity, and college student status. Sex was coded as female or male (1=female, 0=male). Race/ethnicity was measured via ethnicity (Hispanic or Latino/a) and race (Asian or South Asian; Native Hawaiian or other Pacific Islander; Black or African American; American Indian or Alaska Native; White; Arab, Middle Eastern or North African; or Other). Race/ethnicity was subsequently recoded into three categories: non-Hispanic White (67.4%), Hispanic (20.8%), and non-Hispanic Other (11.8% total; 5.2% multiracial, 3.1% Black, 3.0% Asian, and 0.5% all other) due to sample sizes. College student status was based on whether respondents attended a 4-year college or university full-time in 2019 and/or 2020 (1=full-time student, 0=other).

Data Analysis Plan

Due to the nested structure of the data, two-level hierarchical regression models were used. The nesting of days within individuals can cause correlated residuals, and hierarchical regression accounts for this form of dependence by including random effects and analyzing days and individuals as separate levels of data (Raudenbush & Bryk, 2002). There are two equations in these hierarchical regression models: a day-level equation (within-person) and an individual-level equation (between-person). Within-person, time-varying measures are included at the day level, and between-person, time-invariant measures are included at the individual level.

We first examined the relationship between affect and likelihood of drinking on a given day using a two-level logistic hierarchical regression (all days=13,774; persons=633). We did not examine the relationship between motives and likelihood of drinking because motives were only measured for drinking days. We then used two-level ordinal logistic hierarchical regression to examine how affect, drinking motives, and their interactions predict drinking intensity on drinking days only (drinking days=2,109; persons=633). Therefore, we ran three models: modeling affect only for any drinking (Model 1), modeling affect only for drinking intensity (Model 2), and then adding in motives and interactions of motives with affect for drinking intensity (Model 3). To explore how these relationships varied after the onset of the COVID-19 pandemic, we also examined main effects and interactions between the pandemic and affect and motives (see Supplemental Table 1 for any drinking and Supplemental Table 2 for drinking intensity).

In order to estimate within-person relationships, affect and motives were group-mean centered at the day level and the person-means of affect and motives were included at the individual level (Raudenbush and Bryk, 2002). For Model 3, we examined the interaction between day-level affect and drinking motives by including interaction terms for each motive with both positive and negative affect. In order to ascertain practical significance of our results, we specified a smallest effect size of interest. The authors determined that the odds of drinking or odds of drinking more heavily (from moderate to binge, or from binge to HID) would need to increase or decrease by at least 10% for each unit change in affect (range: 1–5) and at least 30% for each unit change in motives (range: 0–2) for the effect to be practically significant. We also conducted two sets of sensitivity analyses that (1) used negative binomial multilevel regression to examine a count measure of the number of drinks for drinking intensity, and (2) an analysis testing whether the multilevel ordinal logistic regression analysis for drinking intensity violated the proportional odds assumption. The first set of sensitivity analyses was conducted to determine if any results changed when using a count measure of the number of drinks compared to the ordinal measure. The second set of sensitivity analyses tested whether an ordinal measure could be used, as an ordinal measure is not recommended if the proportional odds assumption is violated. Attrition weights were used in analyses to account for attrition and nonresponse. We used Stata v.16.1, specifically using the melogit (with weights) and the meologit (with weights) commands for the multilevel logistic and ordinal logistic models, respectively.

Results

Descriptive statistics

For Model 1 (all days among participants who ever reported drinking in waves 1 or 2, N=13,774), drinking was reported on 15.7% of days. For Models 2 and 3 (drinking days only, N=2,109), 58.1% of drinking days involved moderate drinking, 30.6% involved binge drinking, and 11.4% involved high-intensity drinking (HID). The average levels of positive affect, negative affect, and motives are presented in Table 1.

Table 1.

Descriptive Statistics for Analytic Samples (Person N=633)

	All Days		Drinking Days
	%/Mean	(SE)	%/Mean	(SE)
Outcomes
Any drinking (% of days)	15.74	(0.40)	~	~
Ordinal drinking intensity (% of days)
Moderate alcohol use (1–3/1–4)	~	~	58.05	(1.35)
Binge drinking (4–7/5–9)	~	~	30.59	(1.24)
High-intensity drinking (8+/10+)	~	~	11.36	(0.88)
Affect scales
Positive (mean; range=1 to 5)	2.35	(0.01)	2.41	(0.02)
Negative (mean; range=1 to 5)	1.47	(0.01)	1.40	(0.01)
Motive scales
Enhancement (mean; range=0 to 2)	~	~	1.17	(0.02)
Social (mean; range=0 to 2)	~	~	0.68	(0.02)
Coping (mean; range=0 to 2)	~	~	0.18	(0.01)
Covariates: Level 1 (day)
Weekend (Thurs/Fri/Sat) (% of days)
No	56.97	(0.53)	35.4	(1.33)
Yes	43.03	(0.53)	64.6	(1.33)
Year (% of days)
2019	51.04	(0.54)	49.68	(1.38)
2020 (pre-COVID)	41.98	(0.54)	40.60	(1.36)
2020 (early-COVID)	6.98	(0.26)	9.72	(0.81)
Covariates: Level 2 (person)
Race/ethnicity (% of person sample)
Non-Hispanic White		67.41	(2.44)
Hispanic		20.76	(2.25)
Other (including multiracial)		11.83	(1.52)
Sex (% of person sample)
Female		43.41	(2.35)
Male		56.59	(2.35)
College status (% of person sample)
Full-time at 4-year college		64.62	(2.59)
Other		35.38	(2.59)

Notes. All days N = 13,774, drinking days N = 2109.

Hierarchical regression results

Any drinking.

Table 2 shows hierarchical logistic regressions for Model 1. For significant effects, all odd ratios and even the lower value of each confidence interval exceeded the authors’ criteria for the smallest effect size of interest. Greater positive affect on a given day was positively associated with drinking on that day (see Figure 1; AOR=1.37; 95% CI=1.19, 1.59). Each step up the positive affect scale (on a scale from 1 to 5) corresponded with a 37% increase in the odds that the day was a drinking day, adjusting for covariates. For each unit increase in reported negative affect, the odds of a drinking day were 31% lower (see Figure 2; AOR=0.69; 95% CI=0.57, 0.84). Between-person levels of affect were not related to drinking. For between-person covariates, males reported higher odds of drinking compared to females (AOR=1.27; 95% CI=1.04, 1.55), and full-time college students reported higher odds of drinking compared to young adults who were not full-time college students (AOR=1.40; 95% CI=1.11, 1.78). Analyses for time period revealed a main effect for odds of drinking, but no interaction with daily affect. Compared to pre-COVID 2020, the odds of drinking were higher on days that occurred during early stages of the pandemic (AOR=1.78; 95% CI=1.26, 2.51). Supplemental Table 1 shows that for any drinking, there were no significant interactions between time period and affect.

Table 2.

Multilevel Logistic Regression Results for Any Drinking and Affect

Sample: All Days among Young Adults who Report Any Drinking
	Model 1
	AOR	(95% CI)	p-value
Level 1: Day (Difference from personal affect mean)
Positive affect	1.37	(1.19, 1.59)	<0.001
Negative affect	0.69	(0.57, 0.84)	<0.001
Weekend (ref.=weekday)	3.09	(2.72, 3.52)	<0.001
Year (ref=2020 pre-COVID)
2019	0.97	(0.83, 1.13)	0.677
2020 early-COVID	1.78	(1.26, 2.51)	0.001
Level 2: Person (Personal motive mean)
Positive affect	0.94	(0.75, 1.17)	0.562
Negative affect	0.81	(0.64, 1.03)	0.081
Race/ethnicity (ref=Non-Hispanic White)
Hispanic	0.83	(0.62, 1.10)	0.186
Other (Non-Hispanic)	1.10	(0.82, 1.47)	0.531
Male (ref=female)	1.27	(1.04, 1.55)	0.021
College (ref=non-college)	1.40	(1.11, 1.78)	0.005
Variance component
Level-2 variance	0.57	(0.42, 0.76)

Notes. Person N = 633, Days N = 13,774; college=full time student in 4-year college

Variance component at level 2 was statistically significant

Figure 1.

Main effect of positive affect on odds of drinking

Figure 2.

Main effect of negative affect on odds of drinking

Drinking intensity.

Table 3 presents Models 2 and 3, ordinal hierarchical logistic regressions predicting drinking intensity on drinking days only. For significant effects, all odds ratios and even the lower value of each confidence interval exceeded the authors’ criteria for the smallest effect size of interest. Model 2 shows that participants reported higher levels of negative affect on days with higher drinking intensity (negative affect: AOR=1.67; 95% CI=1.12, 2.48). In contrast, although participants reported higher positive affect on drinking days compared to non-drinking days, positive affect was not related to heavier drinking intensity. At the between-person level, positive affect was related to higher intensity of drinking, but negative affect was not related. For covariates, Non-Hispanic Other/multiracial young adults endorsed lower drinking intensity compared to non-Hispanic White young adults (AOR=0.42; 95% CI=0.25, 0.71). Being a full-time college student was associated with higher drinking intensity relative to not being a full-time college student (AOR=2.15; 95% CI=1.30, 3.56). Drinking days in 2019 were associated with lower drinking intensity as compared to pre-COVID 2020 (AOR=0.63; 95% CI=0.48, 0.82).

Table 3.

Multilevel Ordinal Logistic Regression Results for Drinking Intensity, Affect, and Drinking Motives

Sample: Drinking Days Only
	Model 2			Model 3
	AOR	(95% CI)	p-value	AOR	(95% CI)	p-value
Level 1: Day (within-person affect and motive)
Affect
Positive affect	1.10	(0.86, 1.42)	0.457	0.61	(0.44, 0.85)	0.003
Negative affect	1.67	(1.12, 2.48)	0.012	1.00	(0.64, 1.56)	0.997
Drinking motives
Enhancement motive	~	~	~	5.55	(3.65, 8.43)	<0.001
Social motive	~	~	~	2.91	(2.19, 3.86)	<0.001
Coping motive	~	~	~	5.09	(2.36, 10.98)	<0.001
Affect and motive interactions
Positive affect × enhancement motive	~	~	~	2.46	(1.31, 4.62)	0.005
Positive affect × social motive	~	~	~	0.91	(0.52, 1.62)	0.756
Positive affect × coping motive	~	~	~	1.15	(0.53, 2.47)	0.723
Negative affect × enhancement motive	~	~	~	2.14	(0.71, 6.51)	0.178
Negative affect × social motive	~	~	~	1.10	(0.52, 2.33)	0.804
Negative affect × coping motive	~	~	~	0.26	(0.10, 0.71)	0.008
Weekend (ref.=weekday)	3.00	(2.20, 4.08)	<0.001	2.07	(1.53, 2.80)	<0.001
Year (ref=2020 pre-COVID)
2019	0.63	(0.48, 0.82)	0.001	0.79	(0.58, 1.07)	0.127
2020 during-COVID	1.17	(0.69, 1.98)	0.571	1.32	(0.68, 2.56)	0.411
Level 2: Person (between-person affect and motive)
Affect
Positive affect	1.59	(1.21, 2.11)	0.001	1.14	(0.78, 1.66)	0.508
Negative affect	0.94	(0.66, 1.34)	0.727	1.09	(0.64, 1.87)	0.741
Drinking motives
Enhancement motive	~	~	~	2.96	(1.70, 5.13)	<0.001
Social motive	~	~	~	3.90	(2.40, 6.34)	<0.001
Coping motive	~	~	~	0.60	(0.21, 1.73)	0.347
Race/ethnicity (ref=Non-Hispanic White)
Hispanic	0.60	(0.31, 1.16)	0.13	0.73	(0.35, 1.54)	0.407
Other (Non-Hispanic)	0.42	(0.25, 0.71)	0.001	0.42	(0.22, 0.78)	0.006
Male (ref=female)	1.00	(0.70, 1.42)	0.995	0.81	(0.54, 1.21)	0.301
College (ref=non-college)	2.15	(1.30, 3.56)	0.003	1.76	(0.94, 3.29)	0.077
Variance component
Level-2 variance	1.62	(1.10, 2.39)		1.98	(1.27, 3.10)

Notes. Person N = 633; Days N = 2109

Model 2 includes only affect

Model 3 includes affect, motives, and affect/motives interaction

Variance component at level 2 was statistically significant in all models

College=full time student in 4-year college

In Model 3, at the between-person level, positive and negative affect were not related to drinking intensity. At the within-person level, the conditional effects of enhancement, social, and coping motives strongly predicted drinking intensity (AOR=5.55, 95% CI:3.65, 8.43; AOR=2.91, 95% CI=2.19, 3.86; and AOR=5.09, 95% CI=2.36, 10.98, respectively). The conditional effect of positive affect was negatively related to drinking intensity when motives and motive interactions were in the model (AOR=0.61, 95% CI=0.44, 0.85). There was an interaction between positive affect and enhancement motives (AOR=2.46; 95% CI=1.31, 4.62), as well as between negative affect and coping motives in predicting drinking intensity (AOR=0.26; 95% CI=0.10, 0.71). We calculated predicted log-odds for drinking intensity and examined the predicted scores for each predictor in the interaction term (at minimum and maximum values) to demonstrate the nature of these interactions (see Figure 3 for positive affect and enhancement motives; Figure 4 for negative affect and coping motives). At high levels of enhancement motives, drinking intensity and positive affect were positively related; and at low levels of enhancement motives, drinking intensity and positive affect were negatively related. For the negative affect-coping motives interaction, at low levels of coping motives, drinking intensity and negative affect were positively related; and at high levels of coping motives, drinking intensity and negative affect were negatively related.

Figure 3.

Predicted Scores for Drinking Intensity by Positive Affect and Enhancement Motive Interaction

Figure 4.

Predicted Scores for Drinking Intensity by Negative Affect and Coping Motive Interaction

Lastly, in exploratory analyses, there was a significant interaction between social drinking motives and time period (AOR=3.37; 95% CI=1.27, 8.98), such that higher social motives were associated with higher drinking intensity in the same day (i.e., binge or HID rather than moderate drinking) during the early pandemic. There were no other significant interactions between time period and drinking motives or affect.

Sensitivity analyses showed that no conclusions changed (and results exactly mirrored main results) when a count measure and negative binomial regression were used. We also found that no relationship in the ordinal logistic regression model violated the proportional odds assumption, supporting the use of an ordinal drinking outcome.

Discussion

The current study examined the relationships between affect, drinking motives, and two outcomes: likelihood of alcohol use and drinking intensity (on drinking days). Participants reported these variables for the day prior in daily morning reports, allowing for day-level analyses of affect, motives, and drinking. These relationships were examined in a young adult sample during springs of 2019 and 2020, and differences in these relationships before versus during the early COVID-19 pandemic were explored.

We hypothesized that positive mood would be positively associated with likelihood of drinking and drinking intensity, consistent with a positive affect regulation model. Results mostly supported our hypotheses--young adults reported higher positive affect on drinking days compared to non-drinking days, though positive affect was not related to higher intensity of drinking on drinking days. This result could be interpreted as a daily association between positive affect and drinking (versus non-drinking), consistent with previous literature (De Leon et al., 2020; Dvorak et al., 2018; Emery & Simons, 2020; Howard et al., 2015; Hussong et al., 2001; Patrick et al., 2015). However, many prior studies examining the affect regulation theory of drinking have examined affect that was rated before drinking, generally allowing studies to draw conclusions about affective antecedents to drinking. Because participants retrospectively reported their affect and drinking in this study, the current study may be better equipped to draw conclusions about young adults’ judgments of their mood on days with and without drinking events. That is, for positive affect, young adults reported their affect to be more positive on days when they also reported drinking. This interpretation is consistent with the drinking intensity models as well: positive affect was not associated with intensity of drinking as a main effect, but when participants reported that they drank to enhance their positive affect (i.e., enhancement motives, and thus perceived their positive affect and drinking to be related), there was a strong relationship between greater positive affect and higher drinking intensity. Nonetheless, finding that the interaction between enhancement motives and positive affect was associated with drinking is consistent with hypotheses and also with studies that measured affect prior to drinking (Dvorak et al., 2014; Stevenson et al., 2019). Taken together, these results suggest that positive affect is associated with heavier alcohol consumption when young adults are motivated to enhance positive affect by drinking, consistent with a positive affect regulation model.

For negative affect, we found that young adults judged their affect to be more negative on days they did not drink. If young adults do indeed drink to regulate negative affect, we would have found the opposite—participants would report drinking on days high in negative affect. Finding that mood was worse on abstinent days indicates non-support for the affect regulation hypothesis of drinking, which had mixed support in prior work as well (De Leon et al., 2020; Emery & Simons, 2020; Fairlie et al., 2019; Howard et al., 2015; Hussong et al., 2001; Patrick et al., 2015; Stevenson et al., 2020; Treloar et al., 2015). Further, inconsistent with hypotheses, we found a counterintuitive interaction between negative affect and coping motives. Rather than drinking more on days with high negative affect and high coping motives, young adults reported drinking more on days with low negative affect and high coping motives. It is possible that drinking to cope effectively alleviated negative affect, leading to reports of low negative affect on drinking days and on high-coping days. However, studies that can test this question by measuring affect prior to and after drinking episodes have also not consistently found that negative affect prompts drinking (positive results: Dvorak et al., 2014; Mohr et al., 2005; negative results: Hamilton et al., 2020; Stevenson et al., 2019) or is improved following drinking to cope (mixed results: Piasecki et al., 2014; Treloar et al., 2015). Hence, the current study does not find support for the negative affect regulation model, and this result is not inconsistent with past studies.

Young adults reported drinking more heavily when they were drinking for social reasons, consistent with hypotheses and previous research (Patrick, Fairlie et al., 2019; White et al., 2016). Social motives did not interact with affect to predict drinking, indicating that the effect of social motives on drinking functions independently of affect, consistent with a previous study (Hamilton et al., 2020). It is noteworthy that social motives still had a strong relationship with drinking intensity even with other strong relationships in the same model (i.e., affect, coping motives, and enhancement motives). This result highlights the importance of studying social motives as they relate to alcohol consumption in young adults.

Lastly, some past research has found an overall decrease in the amount of alcohol young adults have consumed during the COVID-19 pandemic (Chodkiewicz et al., 2020; Martinez & Nguyen, 2020; Neill et al., 2020), but the present study examined any drinking separately from the amount consumed, finding that drinking was more likely in the initial weeks of the pandemic (3/11/20 to 4/12/20) but intensity of drinking on drinking days was not. Further, we found that social motives were more strongly associated with drinking intensity during the early pandemic, compared to before. It is possible that these increases in drinking were a result of increased opportunities to drink due to class cancellations (for students) and businesses closing (for working young adults) in the initial weeks of lockdown, and social motives were a more important motivator of high-intensity drinking during that time due to pandemic-related restrictions and isolation (Kathirvel, 2020). It is also possible that young adults were simply more likely to emphasize the social importance of their drinking events (and therefore report higher social motives) while reflecting on drinking episodes during the pandemic, due to increased awareness of the risks of social gatherings. Whether these associations remained throughout the pandemic is an area for future research.

Limitations

The current sample was selected based on reports of past-30-day alcohol use in 12^th grade, limiting generalizability to this particular population in the US; young adults who began drinking after 12^th grade were not included. In addition, the present study collected all data retrospectively in morning reports. This is a strength in that this approach is capable of capturing the overall mood of each day, rather than only capturing isolated moments, but it is also possible that this approach introduces retrospective recall biases. In addition, it is not possible to make conclusions about temporal ordering of affect, motives, and drinking within a day, as would be necessary to test a mediational pathway from affect to motives to drinking. Further, although exploratory analyses examined differences during the early COVID-19 pandemic period, fewer than 10% of our daily surveys were completed during that time and data collection finished in April 2020. Therefore, results pertaining to changes in alcohol consumption during the COVID-19 pandemic do not reflect longer term patterns throughout the pandemic and may be underpowered.

Conclusion

Overall, young adults reported higher positive affect on drinking days than non-drinking days, which they characterized as higher in negative affect. Young adults reported the heaviest drinking when they reported high positive affect and desire to enhance positive affect (enhancement motives). Participants were more likely to drink during the early COVID-19 pandemic (as compared to pre-pandemic days in 2019 or 2020) but did not report higher drinking intensity. However, the association of social motives and drinking intensity on a given day was stronger during the early pandemic than it had been prior. Results in this young adult sample support a positive affect regulation model for drinking (i.e., drinking to increase positive affect), but not a coping/negative affect regulation model).

Public Significance Statement.

Young adults were more likely to drink on days when they also reported high positive affect, and drank heavily when also drinking to enhance their positive affect. Young adults were less likely to drink on days with high negative affect, but if they did drink on those days, they drank more heavily. Drinking to have fun with friends was also associated with heavier drinking, particularly after the onset of the COVID-19 pandemic.