Effects of within- and between-person assessments of alcohol expectancies and valuations on use and consequences moderated by sex

Abstract

Background.

Alcohol expectancies (AE; beliefs about likelihood of outcomes) and valuations (beliefs about desirability of outcomes) may help explain young adult alcohol use. However, it remains unclear how variability in AE and valuations over time are related to alcohol-related outcomes, and whether these associations are moderated by sex. The current study addressed these gaps by examining within-person variability between positive and negative AEs, valuations, and alcohol-related outcomes over a 12-month period.

Methods.

Data were collected from 433 college students (M_age=20.06; 59.81% women) who completed surveys at four timepoints: baseline, 3-month, 6-month, and 12-month follow-ups.

Results.

First, we found substantial within-person variability in both AE and valuations (ICCs ranged from 50% to 66%), and differences in variability by sex, with women showing more variability than men. Multilevel models revealed that weekly drinking was significantly higher at timepoints in which participants held relatively greater AE for sociability, sexuality, and risk/aggression, but lower when participants expected greater effects on self-perception. Weekly drinking was also higher when participants reported more favorable valuation of risk/aggression. Participants experienced significantly more negative consequences at timepoints in which participants held relatively greater AE for sexuality and self-perception. No AE were associated with reduced likelihood of negative consequences. Participants experienced more negative consequences at timepoints in which they reported more favorable valuation of self-perception; no valuations were associated with fewer consequences. Several between- and within-person associations were moderated by sex.

Conclusions.

These findings suggest that AE and valuations are dynamic; young adults’ beliefs about the effects of alcohol varied over time, and that both negative and positive AE and valuations may be important correlates of use and consequences. These findings have clinical implications for interventions designed to challenge expectancies and valuations with the goal of reducing alcohol use and associated consequences.

Introduction

Alcohol misuse among young adults is a major health concern. Among young adults aged 18 to 25, nearly 55% reported alcohol use within the last month, with nearly 35% reporting binge drinking in the last month. Of particular concern, approximately 10% of young adults in this age group also met criteria for Alcohol Use Disorders (AUD) in the past year (Lipari and Park-Lee, 2019). Alcohol misuse is also associated with increased risk of preventable death, as well as a host of psychosocial consequences such as academic difficulties, risk of injury, and risk of physical and sexual assault (Hingson et al., 2005; Wechsler et al., 2010). Understanding the risk factors for young adult alcohol use is important for reducing risky alcohol use and related consequences.

Alcohol-related outcome expectancies (AE) represent an individual’s belief about the likelihood of experiencing positive and negative effects when consuming alcohol, which in turn influences decisions to drink (Fromme et al., 1993). Among young adults, positive AE (i.e., expecting positive or favorable effects from drinking) have been associated with increased alcohol consumption cross-sectionally (Fromme et al., 1993; Fromme and D’Amico, 2000; Zamboanga, 2006; Zamboanga and Ham, 2008) and longitudinally (Aas et al., 1998; Zamboanga et al., 2006), as well as experiencing a greater number of negative consequences (Blume and Guttu, 2015; Neighbors et al., 2003; Park and Grant, 2005). Overall, negative AE (i.e., expecting negative or adverse effects from drinking) have received less attention than positive AE (Jones et al., 2001), and findings for the role of negative AE in relation to alcohol use are mixed (Neighbors et al., 2007; Patel and Fromme, 2010; Zamboanga et al., 2006). While some studies suggest that negative AE are related to less alcohol consumption (Fromme et al., 1993; Fromme and D’Amico, 2000; Nicolai et al., 2010), others have found that negative AE are related to increased alcohol consumption (Mann et al., 1987; McMahon et al., 2009) or not related at all (Ham and Hope, 2005; Neighbors et al., 2007; Valdivia and Stewart, 2005; Young et al., 2006). Similarly, studies have found mixed effects for the role of negative AE in experiencing negative consequences; while some studies show that greater negative AE are related to experiencing more negative consequences (Lac and Luk, 2019), others have found no relation (Young et al., 2006), or that the relation may be expectancy and consequence dependent (i.e., subjective versus physical; (Lee et al., 2020). In contrast, previous studies have shown that greater negative expectancies may be protective from negative consequences, such as relapse (Jones and McMahon, 1994). These mixed findings may reflect variability in the extent that young adults view ‘negative’ expectancies as actually being adverse or less favorable (Patrick and Maggs, 2011). Further, these findings demonstrate the need to utilize methodologies that assess the unique contributions of each specific AE domain (as opposed to a broad categorization of positive versus negative), as well as examination of the evaluative judgement of each AE.

Expectancy valuations refer to an individual’s belief about how good or bad certain outcomes would be if they occurred while drinking. Utility Theory (Edwards, 1954) and Social-Learning Theory (Bandura, 1977) suggest that decisions to consume alcohol are not only influenced by the expected effects of alcohol, but the extent to which an individual values that outcome. Similar to the effects of AE on decisions to drink, outcomes that are rated more positively are associated with increased alcohol use (Fromme et al., 1993; Gaher and Simons, 2007; Wood et al., 1996). However, like findings from AE studies, the role of negative valuations is less clear, with some studies finding outcomes rated more negatively are associated with decreased alcohol use (Dunne et al., 2013), whereas others have found negative evaluations associated with increased alcohol use and alcohol-related consequences (Werner et al., 1993) Clearly, the relations between valuations and alcohol-related outcomes are complex, with studies finding discrepant results (Jones et al., 2001; Werner et al., 1993), further highlighting the need for nuanced assessment of valuations in addition to their corresponding AE.

Previous research on AE and valuations demonstrates that both AE and valuations are important predictors of alcohol consumption, and several individual factors may contribute to discrepant findings between AE and valuations. Specifically, positive versus negative expectancies are context dependent (Ham et al., 2013; Monk and Heim, 2013; Wall et al., 2001). It is well-documented that specific physical environments (e.g., bars, parties), social contexts, and emotional states are predictors of quantity and frequency of alcohol consumption (Clapp et al., 2009; Monk and Heim, 2013; Paschall and Saltz, 2007; Treno et al., 2000). Such environments may alter AE and valuations, which in turn can affect alcohol use behavior. Further, AE and valuations are person dependent, and may vary by factors such as group membership, age, drinking history, or sex (Lewis and O’Neill, 2000; Monk and Heim, 2013; Nicolai et al., 2010). For example, previous research has shown that college students may endorse negative consequences as either positive or neutral (Mallett et al., 2008; Patrick and Maggs, 2011), suggesting that negative AE may not be perceived as negative by all individuals. The discrepant findings for AE may also be attributable to measurement issues, as measures typically utilize a-priori determined ‘positive’ versus ‘negative’ AE without consideration of the participants’ subjective value of a given outcome. Given that the effect of AE and valuations are influenced by external factors (e.g., setting), and that valuations may not directly correspond to AE, it is important to consider these related but distinct beliefs independently.

One of the more complex individual factors to disentangle in AE research is potential sex differences. Previous research has shown that men and women vary in their AE (Jones et al., 2001; Monk and Heim, 2013; Read et al., 2004; Williams and Ricciardelli, 1996). For example, while women endorsed social enhancement more strongly than men, the relationship between social enhancement and alcohol consumption was only evident for men (Read et al., 2004). These findings suggest that the translation from AE to alcohol consumption behavior may be gender specific. However, as noted by Monk and Heim (2013), the relationships between AE and drinking outcomes are complex and multi-faceted, as they may depend on other factors such as quantity and frequency of alcohol use, cultural differences, context, and AE type (i.e., positive versus negative). Notably, there is a considerable lack of research examining the role of sex on valuations, and to our knowledge, no studies have examined the moderating role of sex on valuations at the within-person level, which would provide valuable information about the tractability of AE, valuations, and outcomes for men and women. Such information has important implications for prevention and intervention, as approaches may need to be tailored to men and women. Together, these studies highlight the importance of nuanced assessment of the role of sex in the relationships between AE, valuations, and alcohol-related outcomes.

In addition to limitations associated with contextual and individual factors, the bulk of our understanding of the role of AE and valuations is limited to cross-sectional findings, which has several limitations. First, given the bidirectional nature of alcohol use and expectancies (i.e., alcohol use influences expectancies, expectancies influence alcohol use; (Aas et al., 1998; Patrick and Maggs, 2011), it is important to evaluate AE and valuations longitudinally to better understand their dynamic nature. Second, cross-sectional data limit our ability to understand positive and negative AEs’ unique role in alcohol use and associated consequences, as analyses rely on between-person comparisons. Thus, cross-sectional studies do not account for within-person fluctuations of AE and valuations and may fail to capture nuances associated with the complex relations between negative AE, valuations, and alcohol-related outcomes.

To address these limitations, daily studies have compared within-and between-person associations between AE, valuations, and alcohol-related outcomes, which show variability in these relationships across individuals over short periods of time (Butler et al., 2010; Lee et al., 2015; Ramirez et al., 2020) Such studies have the ability to parse out within- and between-person effects of AE and valuations on drinking outcomes. For example, studies have shown that at the within-person level, negative expectancies are associated with increased likelihood of alcohol use on a given day (Patrick et al., 2016); whereas between-person data suggested that negative AE were associated with a decreased likelihood of alcohol use on a given day (Ramirez et al., 2020). Additionally, daily studies have shown that high-intensity drinking is positively associated with positive valuations of alcohol-related consequences (Patrick et al., 2016), and that event-specific drinking is higher when AE are more positively valued (Armeli et al., 2005; Patrick et al., 2016). The few studies examining sex differences have found that female participants demonstrated a stronger association between positive expectancies and drinking on a given day (Ramirez et al., 2020), but sex did not moderate the relationship between daily AE and alcohol-related consequences (Lee et al., 2020). While these findings highlight the importance of within-person analyses, further exploration of within-person variability over greater lengths of time, the role of valuations, and the role of sex is warranted given the dearth of studies examining these variables simultaneously. By examining within- and between-person data over longer periods of time, it may be possible to identify notable shifts in beliefs that are important antecedents to fluctuations in alcohol use and the associated consequences that are not captured in daily studies.

Current Study

The aims of the study were to (1) examine the extent that AE and alcohol valuations fluctuate within-person over time and whether there are sex differences in these fluctuations, (2) examine whether within-person variability of AE and alcohol valuations predict alcohol-related outcomes (weekly drinking, negative consequences), and (3) examine if the relationship between AE, alcohol valuations, and alcohol-related outcomes is moderated by sex at the between- and within-person levels.

Regarding Aim 1, we expected that there would be significant within-person variability in positive and negative AE and valuations across timepoints. Given the limited and generally mixed findings on sex differences across AE and valuations and time, we considered the latter half of Aim 1 exploratory, but anticipated that men and women would significantly differ in variability of AE and alcohol valuations across timepoints. For Aim 2, we expected that between- and within-person AE and alcohol valuations would predict weekly drinking and negative consequences. Consistent with previous findings, we generally expected that positive AE and valuations would be associated with greater weekly drinking and experiencing more negative consequences, whereas negative AE and valuations would be associated with less weekly drinking and fewer negative consequences, but we did not make specific predictions about each AE and valuation independently. Further, consistent with previous studies examining multilevel data, we anticipated that different associations would emerge at the within-person level compared to the between-person level. However, given the limited studies in this area, combined with the fact that previous studies have utilized different AE and valuation measures or grouped items into positive or negative, we did not make specific predictions for each AE and valuation at each level. For Aim 3, we expected that the association between AE, alcohol valuations, and alcohol-related outcomes would be moderated by sex. Consistent with the exploratory nature of sex differences in Aim 1, we expected different associations between men and women would emerge. Further, based on findings from previous studies using multilevel data, we expected different associations at the within-person level compared to the between-person level but did not make specific predictions about individual AE and valuations due to limitations in the extant literature.

Materials and Methods

Participants and Procedures

The current study entails secondary analysis of college students recruited to participate in two randomized control trials (RCT) that tested the efficacy of a personalized normative feedback interventions. Participants were sampled at baseline, 3-month, 6-month, and 12-month follow-ups. Specifically, the current sample was screened into the respective RCT studies but randomized to a control condition and therefore received no active treatment components (LaBrie et al., 2013; Larimer et al., 2021). Aggregating the control samples from two RCTs for secondary analysis is justified given that both RCTs sampled students from the same two U.S. universities, employed identical measures within the same follow-up timelines, and procedures for control conditions were nearly identical.

In both RCTs, college students were randomly selected from university Registrar lists to be invited via e-mail to participate in the respective studies. Students were invited from two west coast universities: Campus 1, a midsized private university, and Campus 2, a large public university. Screening criteria for enrollment into the RCT studies required at least one episode of heavy drinking within the past month (i.e., 4+/5+ drinks in a single occasion for women/men, respectively).

The first RCT (i.e., RCT-1; (LaBrie et al., 2013) invited 11,069 students with 4,818 (43.5%) completing the preliminary screener (Campus 1 n = 1,784; Campus 2 n = 3,034). Following the screening survey, 2,034 (42%) met inclusion criteria and were randomized to one of 11 conditions (i.e., n = 184 randomized into the control condition). The second RCT (i.e., RCT-2; Larimer et al., 2021) invited 5,998 students, and 2,688 (44.8%) completed the screener (Campus 1 n = 1,212; Campus 2 n = 1,476). Of these, 1,494 (55.5%) met inclusion criteria and were randomized into one of six conditions (i.e., n = 249 randomized into the control condition). Combining the control groups from both RCTs, the total sample for the present study was 433 students (59.81% female; M_age = 20.06 years at baseline; 68.6% White, 15.7% Asian, 8.2% Hispanic/Latino, 3.8% multiracial, 3.7% other race/ethnicity).

Participants who were eligible for the RCT studies were immediately redirected to an online baseline survey that included additional items pertaining to alcohol use, AE, and valuations (as well as other related constructs). Follow-up surveys (3-, 6-, and 12-months after baseline) asked participants to respond to the same set of items at each timepoint.¹ In total, there were 1,478 unique responses across timepoints with minimal attrition at each follow-up; responses were received from 359 participants at 3-month (82.9%), 346 at 6-month (79.9%), and 340 at 12-month (78.5%).

Participants were compensated with $15 for the screening survey and $25 for the baseline survey. RCT-1 offered $30 for completing 3- and 6-month follow-ups and $40 for completing the 12-month follow-up. RCT-2 offered $25 for the 3-month, $30 for the 6-month, and $35 for the 12-month follow-ups. Participants could also earn a bonus for completing all four waves ($30 for RCT-1 and $25 for RCT-2). Institutional Review Board approval was obtained from both universities where participants were recruited, and a Federal Certificate of Confidentiality was obtained to further protect participants.

Measures

Items pertaining to participant demographics were asked within the initial screening and baseline surveys. The remaining items regarding alcohol use, AE, and valuations were asked at all four timepoints.

Alcohol Expectancies and Valuations.

The 38-item Comprehensive Effects of Alcohol Questionnaire (i.e., CEOA; Fromme et al., 1993) was used to assess both expectancies and valuations of alcohol use. The CEOA is an empirically validated measure that assesses both positive and negative AE, as well as positive and negative valuations, which were expanded from previous studies that typically only included measures of positive AE. Specifically, the original CEOA consisted of 139 items measuring discrete effects of alcohol across several domains; confirmatory factor analyses resulted in two, 38-item scales measuring positive and negative AE and valuations. The same items are used to measure both AE and valuations, where AE are assessed in terms of the participants’ perception of how likely each outcome is (1 = Disagree to 4 = Agree) and valuations are assessed in terms of participants’ subjective feelings about how good or bad each outcome is (1=Bad to 5=Good, with a scale midpoint of 3=Neutral). Initial psychometric research has identified seven subscales for both expectancies and valuations. Positive AE and valuations include sociability (e.g., “I would be outgoing”), tension reduction (e.g., “I would feel peaceful”), liquid courage (e.g., “I would feel courageous”), and sexuality (e.g., “I would feel sexy”). Negative AE and valuations include cognitive and behavioral impairment (e.g., “I would be clumsy”), risk and aggression (e.g., “I would act aggressively”), and self-perception (e.g., “I would feel moody”). A subsequent psychometric study argued that a six-factor structure may be more ideal for valuations (Valdivia and Stewart, 2005), but a preliminary confirmatory factor analysis of the baseline data (Time 1) from the current sample indicated that the proposed six-factor model of valuations had inferior fit compared to originally theorized seven-factor model (i.e., Δχ²= 304.78, p<.001). However, the original seven-factor model of valuations still showed suboptimal fit to the data: CFI=0.81, TLI=0.79, RMSEA=0.08, SRMR=0.09. McNeish et al. (2018) have argued that these fit indices should be considered alongside factor loadings for each item, which were all above the commonly referenced .40 threshold.² Nevertheless, a deeper investigation of potential sources of model misfit is described in the supplemental materials. Internal reliability estimates (Cronbach’s α) indicated adequate subscale reliability for all seven expectancy and valuation scales across the four timepoints. AE subscale reliabilities ranged from α=0.66 to α=0.89 and the valuation subscale reliabilities ranged from α=0.74 to α=0.94.

Weekly Number of Drinks.

An index of past month alcohol use was calculated using the Daily Drinking Questionnaire (i.e., DDQ; Collins et al., 1985). The DDQ asks participants to report the average number of drinks they typically consumed on each day of a typical week in the past month, and the drinks on each day are summed to compute an estimate of weekly number of drinks. Extreme values were recoded to three standard deviations above the mean to reduce the effect of possibly spurious outliers (Tabachnick and Fidell, 2019).

Negative Consequences.

The 23-item Rutgers Alcohol Problem Index (i.e., RAPI; White & Labouvie, 1989) was used to assess negative consequences related to drinking in the past three months. Regarding a list of potential consequences, participants reported the number of times each negative consequence happened on a 5-point scale (0=Never to 4=More than 10 times). Example items include “Neglected your responsibilities?” and “Had a fight, argument, or bad feelings with a friend?” As done in previous studies, frequency counts were dichotomized into 0 = did not occur and 1 = did occur and summed to create a total consequences score, which is an approach that has been validated by Martens and colleagues (2007). Internal reliability for the dichotomized RAPI scale was strong across the four timepoints ranging from α=0.85 at baseline to α=0.92 at 12-month follow-up.

Analyses

The first analytic step entailed estimating whether mean scores on study variables differed between men and women, which was accomplished using Welch’s two-sample t-tests. We then estimated the extent that variability in study variables was due to between-person differences vs. within-person fluctuations across study timepoints. This entailed estimating intraclass correlation coefficients (ICC) that range from 0 to 1, indicating the percentage of variability that is attributed to between-person differences, relative to variance at the within-person level. For example, an ICC value of 0.60 would indicate that 60% of the variability is due to between-person differences while the remaining 40% is due to within-person variability across timepoints. To further estimate within-person fluctuations, we calculated within-person standard deviations that show the level of variability, on average, the sample had for each study variable across the four timepoints. Furthermore, we used a Welch’s two-sample t-test to examine differences in within-person standard deviations between men and women.

Primary analyses estimated between- and within-person associations between alcohol AE and valuations and weekly number of drinks and negative consequences of alcohol use. This entailed multilevel modeling using maximum likelihood estimation whereby within-person covariates were person-mean-centered to enable interpretation of these expectancies/valuations as deviations from each participants’ average level of AE and valuations (Enders and Tofighi, 2007). That is, analyses at the within-person level estimated the extent that an individual’s fluctuation in AE and valuations relate to the outcome variables (weekly drinks and negative consequences) after accounting for stable between-person effects.

Given that alcohol use and negative consequences are inherently positively-skewed count variables, we employed a Poisson count regression approach to appropriately handle this distribution (Atkins et al., 2013). In count regression, effect coefficients are exponentiated to yield rate ratios (RR) that estimate the proportional change in the outcome variable associated with a one-unit increase in the independent variable (conditional on random effects). In addition to expectancies and valuations, models include cohort (RCT-1 or RCT-2), campus (public or private university), birth sex, age at baseline, and Greek status as theoretically relevant control variables. Although there was minor attrition across the four timepoints, multilevel modeling is a flexible approach that is able to estimate between- and within-person effects for participants who have missing timepoints (Kwok et al., 2008).

A second step pertaining to the multilevel models entailed estimating the extent that associations with AE and valuations were moderated by sex at both the between- and within-person levels. Expectancies × sex and valuations × sex interactions were entered in a separate second model and, when there was evidence for a significant interaction, simple-slopes were estimated to examine the conditional associations for males and females, separately. Given our expectations that the effects of AE and valuations on alcohol use and negative consequences may differ for men and women, we explored simple slopes for all significant interactions, even when there was not a significant main effect. This was done because of the potential for cross-over interactions; for example, if there were a significant positive association for men and significant negative association for women, the main-effect for the total sample would be an amalgam of the two subgroups and could be non-significant. Analyses were done using R software: Package ‘lavaan’ for CFA (Rosseel, 2012), package ‘psych’ for calculating ICC values (Revelle, 2020), ‘glmmTMB’ for multilevel modeling (Brooks et al., 2017), package ‘emmeans’ for calculating simple slopes (Lenth, 2020).

Given the conceptual similarities between dimensions of AE and valuations, we also inspected both models for potential multicollinearity (i.e., when a modeled covariate is too strongly related to the other covariates in the model). Variance inflation factors (VIF) were calculated and examined whereby VIF values less than 5 indicate low correlation, values between 5 and 10 are moderately correlated but acceptable, and values higher than 10 are too highly correlated to be tolerable. In both models (i.e., weekly drinks and negative consequences), all VIF values were well below the threshold, indicating that multicollinearity was of minimal concern (see supplemental material for more details on checks for multicollinearity).

Results

In a preliminary step, we examined whether attrition was greater among participants who engaged in more alcohol use or reported more negative consequences at baseline, but attrition was not significantly associated with either (weekly drinks t=-1.04, p=.297; negative consequences t=-0.88, p=.381). Descriptive statistics and sex comparisons for all study variables are displayed in Table 1. Notably, men (on average) drank more than women; no sex differences were observed for negative consequences experienced. Relative to women, men also reported greater expectancies pertaining to tension reduction, liquid courage, and risk/aggression, and reported more favorable valuations of all seven dimensions of alcohol experiences, on average.

Table 1.

Descriptive statistics and estimates of within-person variability across time points for study variables (N = 433).

	Mean	SD	Possible Range	Gender Differences in Person-Mean Values Across All Four Waves2	Intraclass Correlation Coefficient	Within-Person Standard Deviation	Gender Differences in Within-Person Standard Deviations2
Weekly Drinks1.	10.72	9.63	0 – 45	t = 6.76***	.85	3.86	t = −1.53
Negative Consequences	4.57	5.00	0 – 23	t = 1.80	.62	2.28	t = 0.66
EXP- Sociability	3.19	0.54	1 – 4	t = −0.05	.50	0.32	t = −0.13
EXP- Tension Reduction	2.52	0.66	1 – 4	t = 4.32***	.52	0.41	t = −2.91**
EXP- Liquid Courage	2.60	0.65	1 – 4	t = 2.85**	.59	0.37	t = −2.67**
EXP- Sexuality	2.45	0.67	1 – 4	t = 1.43	.61	0.37	t = −0.61
EXP- Cog. Beh. Impairment	2.62	0.54	1 – 4	t = −1.66	.60	0.30	t = −1.78
EXP- Risk/Aggression	2.46	0.67	1 – 4	t = 2.53*	.66	0.35	t = −2.29*
EXP- Self-Perception	1.85	0.60	1 – 4	t = −0.70	.57	0.34	t = −0.20
VAL- Sociability	4.02	0.71	1 – 5	t = 2.67**	.54	0.39	t = −1.31
VAL- Tension Reduction	3.77	0.86	1 – 5	t = 2.10*	.53	0.50	t = −1.79
VAL- Liquid Courage	3.43	0.94	1 – 5	t = 3.08**	.65	0.44	t = −1.82
VAL- Sexuality	3.11	0.84	1 – 5	t = 3.62***	.61	0.48	t = −2.33*
VAL- Cog. Beh. Impairment	1.81	0.61	1 – 5	t = 6.47***	.61	0.30	t = 0.85
VAL- Risk/Aggression	2.27	0.75	1 – 5	t = 4.80***	.63	0.38	t = −1.32
VAL- Self-Perception	1.60	0.63	1 – 5	t = 4.97***	.53	0.34	t = 0.52

Note:

^1.Extreme values were recoded to three standard deviations above the mean (Tabachnick & Fidell, 2019).

^2.Positive t-values indicate that men scored higher, on average, than women. Intraclass correlation coefficients (ICC) are calculated by dividing the between-person variance by the total variance (i.e., the sum of between- and within-person variance) and 1- ICC represents proportion of total variance that is attributed to within-person variance. Coefficient of variation is the ratio of the standard deviation to the mean and indicates the level of within-person fluctuation around the mean (i.e., variability).

^*p<.05

^**p<.01

^***p<.001.

Within-Person Variability

Regarding the first study aim, the ICCs listed in Table 1 signify the amount of variance in study variables that was explained by between-person differences. Approximately 85% of the variance in weekly drinks and 62% of variance in negative consequences was attributed to between person-differences as opposed to 15% and 38% being attributed to within-person differences, respectively. ICC estimates for alcohol expectancies ranged from .50 to .66, indicating that a substantial amount of the variability in these perceptions (between 50% and 34%, respectively) is attributed to within-person fluctuations across the one-year study period. Similarly, between 47% and 35% of the variability in valuations was attributed to within-person fluctuations. Taken together, the ICC estimates indicated that college students vary substantially across time on their reports of alcohol expectancies and valuations.

A secondary component of Aim 1 was to examine sex-differences in within-person fluctuations as estimated via participants’ within-person standard deviations on each study variable. Men and women did not differ in terms of variability in alcohol use and negative consequences, but several significant differences were found for expectancies and valuations. On average, women had more within-person variability on expectancies of tension reduction, liquid courage, and risk/aggression, as well as more within-person variability in valuations of sexuality.

Multilevel Models (Aims 2 and 3)

Number of Weekly Drinks.

The first multilevel model estimated between- and within-person associations between expectancies and valuations and number of weekly drinks (Table 2). At the between-person level, weekly number of drinks was positively associated with sexuality and risk/aggression expectancies, and inversely associated with cognitive/behavioral impairment expectancies. No valuation dimensions were significantly related to weekly drinking at the between-person level. Central to the aims of this study, within-person effects indicated significant positive associations between weekly number of drinks and sociability, sexuality, and risk/aggression expectancies, and a significant inverse association with self-perception expectancies. Pertaining to within-person effects of valuations, a significant positive effect was found for risk/aggression valuations such that participants, on average, engaged in more alcohol use at timepoints that they more favorably valued risk/aggression.

Table 2.

Mutlilevel regression models estimating between- and within-person associations between expectancies/valuations and weekly drinks, with interactions and simple slopes by sex.

	Number of Weekly Drinks (DDQ)
	Main Effects Model (Step 1)	Interaction Effects by Sex (Step 2)	Simple Slopes for Male (n = 174)	Simple Slopes for Female (n = 259)
	RR [95% CI]	RR [95% CI]	RR [95% CI]	RR [95% CI]
Control Covariates
Intercept	2.10 [0.41, 10.70]
Cohort (RCT-1=0, RCT-2=1)	1.14 [0.98, 1.32]
Campus (Public=0, Private=1)	1.29 [1.09, 1.52]**
Birth Sex (Male=0, Female=1)	0.75 [0.63, 0.88]***
Age	0.95 [0.90, 1.01]
Greek Status (No=0, Yes=1)	1.18 [1.07, 1.30]**
Between-Person Effects (Level-2)
EXP- Sociability	1.00 [0.75, 1.32]	1.31 [0.71, 2.39]
EXP- Tension Reduction	1.10 [0.92, 1.32]	0.86 [0.59, 1.27]
EXP- Liquid Courage	0.76 [0.55, 1.06]	0.46 [0.23, 0.93]*	1.33 [0.77, 2.30]	0.60 [0.41, 0.90] *
EXP- Sexuality	1.28 [1.04, 1.58]*	0.78 [0.50, 1.20]
EXP- Cog. Beh. Impairment	0.78 [0.62, 0.97]*	2.01 [1.27, 3.17]**	0.52 [0.36, 0.74]***	1.00 [0.76, 1.32]
EXP- Risk/Aggression	1.58 [1.19, 2.11]**	0.77 [0.42, 1.38]
EXP- Self-Perception	0.92 [0.75, 1.14]	1.05 [0.70, 1.56]
VAL- Sociability	1.27 [0.97, 1.65]	0.63 [0.36, 1.10]
VAL- Tension Reduction	0.99 [0.82, 1.20]	1.24 [0.84, 1.82]
VAL- Liquid Courage	0.98 [0.77, 1.26]	1.80 [1.09, 2.97]*	0.69 [0.48, 1.01]	1.31 [0.95, 1.80]
VAL- Sexuality	0.91 [0.77, 1.08]	1.31 [0.91, 1.88]
VAL- Cog. Beh. Impairment	1.16 [0.92, 1.46]	0.99 [0.63, 1.56]
VAL- Risk/Aggression	1.06 [0.82, 1.38]	0.74 [0.44, 1.23]
VAL- Self-Perception	1.23 [0.97, 1.56]	1.41 [0.89, 2.24]
Within-Person Effects (Level-1)
EXP- Sociability	1.09 [1.01, 1.17]*	1.13 [0.97, 1.32]
EXP- Tension Reduction	1.01 [0.96, 1.06]	1.09 [0.99, 1.21]
EXP- Liquid Courage	1.00 [0.93, 1.07]	0.91 [0.79, 1.05]
EXP- Sexuality	1.08 [1.02, 1.15]**	1.02 [0.91, 1.15]
EXP- Cog. Beh. Impairment	0.98 [0.91, 1.06]	1.11 [0.95, 1.29]
EXP- Risk/Aggression	1.10 [1.02, 1.18]*	1.03 [0.89, 1.19]
EXP- Self-Perception	0.92 [0.86, 0.97]**	0.86 [0.77, 0.97]*	0.99 [0.91, 1.07]	0.85 [0.78, 0.92]***
VAL- Sociability	0.98 [0.92, 1.04]	0.92 [0.81, 1.05]
VAL- Tension Reduction	0.96 [0.92, 1.01]	0.94 [0.86, 1.03]
VAL- Liquid Courage	1.00 [0.95, 1.05]	1.01 [0.90, 1.12]
VAL- Sexuality	0.97 [0.93, 1.02]	1.04 [0.95, 1.14]
VAL- Cog. Beh. Impairment	1.05 [0.98, 1.12]	1.22 [1.07, 1.40]**	0.96 [0.89, 1.05]	1.18 [1.06, 1.31]**
VAL- Risk/Aggression	1.12 [1.06, 1.18]***	1.00 [0.89, 1.11]
VAL- Self-Perception	1.00 [0.94, 1.05]	0.94 [0.84, 1.06]

Note: EXP = Expectancy item; VAL = Valuation item. Cog. Beh.= Cognitive Behavioral. Because weekly drinks is a count variable, Poisson count regression was used to estimate Rate Ratios that represent proportional change for each unit increase in the predictor (e.g., a Rate Ratio of 1.10 = 10% increase for each one-unit change in the predictor). The estimates shown in the column for Step 2 are the coefficients for the interaction effects for each expectancy/valuation by birth sex. When significant interactions by sex were detected, simple-slopes estimates are reported for men and women (significant simple-slopes estimates are highlighted in bold).

^*p<.05

^**p<.01

^***p<.001.

In the second step to examine Aim 3 — the extent that associations between alcohol use and expectancies and valuations were moderated by sex — several significant interactions were detected. At the between-person level, women who typically expected greater liquid courage effects engaged in significantly less alcohol use compared to those who typically expected less liquid courage effects, but this was not the case for men. Men who typically expected more cognitive/behavioral impairment drank significantly less than men who expected less impairment, but this association was not found for women. At the within-person level, the inverse effect of self-perception AE on weekly alcohol use was significantly moderated by sex; simple-slopes indicated that this effect only held for women, but not for men. Finally, there was a significant within-person association between cognitive/behavioral impairment valuations and weekly drinks for women, but no significant association was detected for men.

Negative Consequences.

The second multilevel model estimated associations between expectancies and valuations and consequences of alcohol use (Table 3). Between-person estimates indicated relatively strong associations between negative consequences and sexuality, risk/aggression, and self-perception AE; those who typically expect alcohol to have these effects tended to report experiencing more negative consequences, on average. None of the valuation dimensions were significantly associated with negative consequences at the between-person level. At the within-person level, students reported experiencing relatively more negative consequences at timepoints they reported greater AE regarding sexuality and self-perception. Regarding valuations, students reported relatively more negative consequences at timepoints they reported more favorable valuations of altered self-perception; no valuations were associated with experiencing fewer negative consequences.

Table 3.

Mutlilevel regression models estimating between- and within-person associations between expectancies/valuations and negative consequences of alcohol use, with interactions and simple slopes by sex.

	Negative Consequences of Alcohol Use
	Main Effects Model (Step 1)	Interaction Effects by Sex (Step 2)	Simple Slopes for Male (n = 174)	Simple Slopes for Female (n = 259)
	RR [95% CI]	RR [95% CI]	RR [95% CI]	RR [95% CI]
Control Covariates
Intercept	0.15 [0.02, 1.07]
Cohort (RCT-1=0, RCT-2=1)	1.08 [0.90, 1.29]
Campus (Public=0, Private=1)	1.22 [1.00, 1.48]*
Birth Sex (Male=0, Female=1)	1.05 [0.86, 1.29]
Age	0.96 [0.90, 1.03]
Greek Status (No=0, Yes=1)	1.10 [0.95, 1.26]
Between-Person Effects (Level-2)
EXP- Sociability	0.93 [0.66, 1.30]	0.85 [0.40, 1.77]
EXP- Tension Reduction	0.96 [0.78, 1.20]	1.07 [0.67, 1.72]
EXP- Liquid Courage	0.82 [0.55, 1.21]	0.43 [0.19, 0.99]*	1.54 [0.76, 3.12]	0.66 [0.42, 1.05]
EXP- Sexuality	1.54 [1.20, 1.98]***	0.94 [0.55, 1.59]
EXP- Cog. Beh. Impairment	0.94 [0.72, 1.24]	1.23 [0.71, 2.15]
EXP- Risk/Aggression	1.82 [1.29, 2.56]***	0.82 [0.40, 1.67]
EXP- Self-Perception	1.35 [1.06, 1.72]*	0.92 [0.57, 1.49]
VAL- Sociability	1.29 [0.94, 1.78]	0.62 [0.31, 1.22]
VAL- Tension Reduction	1.06 [0.84, 1.33]	0.84 [0.52, 1.34]
VAL- Liquid Courage	0.95 [0.71, 1.27]	2.04 [1.12, 3.72]*	0.66 [0.41, 1.07]	1.35 [0.94, 1.94]
VAL- Sexuality	0.82 [0.67, 1.01]	1.57 [1.01, 2.42]*	0.60 [0.41, 0.89]**	0.97 [0.76, 1.23]
VAL- Cog. Beh. Impairment	1.11 [0.84, 1.46]	1.02 [0.59, 1.76]
VAL- Risk/Aggression	1.25 [0.92, 1.71]	0.81 [0.44, 1.50]
VAL- Self-Perception	1.02 [0.77, 1.35]	0.83 [0.47, 1.46]
Within-Person Effects (Level-1)
EXP- Sociability	1.03 [0.92, 1.15]	1.29 [1.01, 1.65]*	0.88 [0.72, 1.09]	1.13 [0.99, 1.29]
EXP- Tension Reduction	1.01 [0.94, 1.09]	1.06 [0.91, 1.23]
EXP- Liquid Courage	1.06 [0.95, 1.18]	0.81 [0.64, 1.02]
EXP- Sexuality	1.16 [1.07, 1.27]***	1.02 [0.85, 1.21]
EXP- Cog. Beh. Impairment	1.02 [0.91, 1.14]	1.11 [0.88, 1.42]
EXP- Risk/Aggression	1.06 [0.95, 1.18]	0.87 [0.68, 1.10]
EXP- Self-Perception	1.14 [1.04, 1.24]**	1.04 [0.87, 1.25]
VAL- Sociability	0.94 [0.85, 1.04]	1.17 [0.95, 1.43]
VAL- Tension Reduction	1.05 [0.98, 1.12]	1.09 [0.95, 1.25]
VAL- Liquid Courage	0.99 [0.91, 1.08]	0.97 [0.82, 1.15]
VAL- Sexuality	1.03 [0.96, 1.10]	1.07 [0.92, 1.23]
VAL- Cog. Beh. Impairment	0.98 [0.89, 1.08]	1.30 [1.06, 1.59]*	0.88 [0.76, 1.01]	1.14 [0.98, 1.33]
VAL- Risk/Aggression	1.06 [0.98, 1.15]	0.86 [0.73, 1.02]
VAL- Self-Perception	1.12 [1.03, 1.22]*	0.80 [0.68, 0.96]*	1.23 [1.09, 1.39]***	0.99 [0.87, 1.12]

Note: EXP = Expectancy item; VAL = Valuation item. Cog. Beh.= Cognitive Behavioral. Because negative consequences is a count variable, Poisson count regression was used to estimate Rate Ratios that represent proportional change for each unit increase in the predictor (e.g., a Rate Ratio of 1.10 = 10% increase for each one-unit change in the predictor). The estimates shown in the column for Step 2 are the coefficients for the interaction effects for each expectancy/valuation by birth sex. When significant interactions by sex were detected, simple-slopes estimates are reported for men and women (significant simple-slopes estimates are highlighted in bold).

^*p<.05

^**p<.01

^***p<.001.

Adding the interaction effects in a second step revealed that several associations were moderated by sex (Aim 3). At the between-person level, sex moderated the associations between liquid courage and negative consequences, such that a positive association was found for men and an inverse association for women, but simple slopes were not significant. In contrast, a positive association between liquid courage valuations and negative consequences was found for women and an inverse association for men, although simple slopes were also not significant. Finally, a significant inverse between-person association between sexuality valuations and negative consequences was found for men, but not for women, indicating that men who typically have more favorable valuations for sexuality effects tend to report fewer negative consequences. At the within-person level, sex moderated the associations between sociability AE and negative consequences, such that a positive association was found for women and an inverse association was found for men, though neither simple slope was significant in its own right. Regarding valuations, sex moderated the association between cognitive and behavioral impairment and negative consequences, whereby an inverse association was found for men and a positive association was found for women, but neither simple slopes were significant. Valuations pertaining to self-perception effects of alcohol revealed a significant positive association for men, but not women, indicating that men experience relatively more negative consequence at timepoints they have relatively more favorable valuations of altered self-perception effects.

Discussion

The current study sought to evaluate the dynamic relationship of AE, valuations, and alcohol-related outcomes among a sample of college student drinkers. While positive AE and valuations are associated with increased alcohol use and negative consequences, the role of negative AE and valuations has been less delineated. Thus, we sought to utilize a more nuanced approach to assessing the role of negative AE and valuations, which was accomplished by examining longitudinal data from four timepoints spanning a 12-month period to assess dynamic relations between AE, valuations and alcohol-related outcomes. Further, the current study is the first to examine the moderating role of sex in AE and valuation variability. Utilizing these strategies, several novel associations between AE, valuations, sex, and alcohol-related outcomes emerged, which have important implications for assessment, prevention, and intervention.

Consistent with our hypothesis, results suggest that both AE and valuations significantly fluctuated over-time, with a range of 34% to 50% of variance being explained by within-person fluctuations. Further, men and women differed significantly in terms of within-person variability of AE and valuations over time. The finding is consistent with studies examining daily associations between AE, valuations, and alcohol-related outcomes (Lee et al., 2015; Ramirez et al., 2020). However, the current study extends these findings in two ways. First, we examined data longitudinally over the span of 12-months, which adds further support to the notion that AE and valuations are dynamic, with significant fluctuations over time. Second, we examined sex differences in within-person variability, which produced novel findings. Specifically, within-person data showed that overall, women’s AE and valuations tended to fluctuate more so than men. These effects occurred for AE of tension reduction, liquid courage, and risk/aggression, as well as the valuation of sexuality.

Also consistent with our hypotheses, within-person variability in AE and valuations predicted both weekly drinking and negative alcohol consequences, with several interesting findings emerging. Generally, at both levels, few positive AE were predictive of greater weekly drinking and negative consequences, whereas there were no main effects for positive valuations. However, similar to previous research (Jones et al., 2001; Neighbors et al., 2007; Valdivia and Stewart, 2005; Young et al., 2006), our data suggest a mixed effect of negative AE and valuations on weekly drinking and negative consequences. By examining specific AE dimensions over time, a more nuanced assessment of their role emerged. For instance, cognitive and behavioral impairment was associated with fewer weekly drinks at the between-person level, but this relationship did not occur at the within-person level. In contrast, the opposite occurred for self-perception; no relationship to weekly drinking was observed at the between-person level, but greater expectations of impaired self-perception were associated with fewer weekly drinks at the within-person level. Similarly, the risk/aggression AE was associated with greater negative consequences at the between-person level, but not at the within-person level. From these findings, two important conclusions emerge. First, AE were predictive of weekly drinking and negative consequences at both the within- and between-person levels. While early studies conceptualized AE as trait characteristics (Fromme et al., 1993), findings from the current study are consistent with daily studies suggesting that AE are variable (Patrick et al., 2016; Ramirez et al., 2020) that are likely influenced by contextual factors (Monk and Heim, 2013). In contrast to AE which were associated with outcomes at both levels, within-person variability in some valuations was predictive of both outcomes, but between-person variability was not predictive of either. These findings suggest that valuations may be highly dependent on contextual factors, such that certain outcomes may be more valued depending on the situation and should be considered independently of AE (Lee et al., 2020). Second, these data suggest that not all negative AE and valuations are protective, highlighting the need to assess specific dimensions of negative AE and valuations (Mallett et al., 2008; Patrick and Maggs, 2011).

To our knowledge, this is the first study to examine the moderating effects of sex on AE, valuations, alcohol use, and alcohol-related consequences at both the between and within-person levels. Consistent with our hypothesis, we found evidence that AE and valuations at both within- and between-person levels were moderated by sex, with different associations emerging for men and women across predictors. In contrast to previous findings, men and women did not differ on weekly drinking or experiencing negative consequences as a result of greater expectancies or valuations of enhanced sociability (Read et al., 2004). Rather, several unique interactions emerged, including the finding that men were less likely to drink with greater expectation of cognitive and behavioral impairment (between-level), but women were less likely to drink with greater expectations of liquid courage (between-level) and altered self-perception (within-person). These findings may highlight the different social dynamics men and women engage in while consuming alcohol. Interestingly, sex moderated associations between negative consequences and liquid courage AE (between-level) and sociability AE (within-level), though neither case found significant simple slopes for men or women. When examining the moderating effect of sex on the relation between valuations and negative consequences, several significant interactions emerged at both between- and within-person levels. Notably, simple-slopes indicated these associations were only significant for men. Specifically, at the between-person level, greater valuation of sexuality was associated with fewer negative consequences. In contrast, greater valuation of altered self-perception was associated with experiencing negative consequences. These findings highlight the importance of assessing negative valuations dynamically and independently given their unique contributions for men and women’s self-reported negative consequences.

Clinical Implications

These findings have important clinical implications as they suggest that assessing the full range of AE and valuations could provide additional contextual information that might improve intervention effects to reduce risk for a developmental trajectory toward negative alcohol consequences. Most of the young adult drinking prevention literature has attempted to decrease positive alcohol expectancies as a mechanism for to reducing alcohol consumption and subsequent negative consequences. The Alcohol Expectancies Challenge paradigm (Darkes and Goldman, 1993, 1998; Dunn et al., 2000) has been extensively researched and has been offered in a variety of formats ranging from in-person group experiences to personalized feedback sessions (Labbe and Maisto, 2011). The paradigm is designed to demonstrate the incongruent relationship between the pharmacological and behavioral effects of alcohol by focusing on positive expectancies such as social, sexual, tension, and liquid courage expectancies to demonstrate that expectancies are in fact mostly placebo effects or can be achieved a low levels of alcohol consumption. Previous research has called on The Alcohol Expectancies Challenge paradigm to expand in the targeting negative expectancies by focusing on negative consequences associated with use (Labbe and Maisto, 2011). Likewise, the use of alcohol valuations could provide an additional component to enhance the efficacy of the intervention. The current study found that greater within-person variability for alcohol valuations over time, especially negative valuations which were associated with greater negative consequences, thus suggesting that negative alcohol valuations maybe more malleable intervention target and could be added the Alcohol Expectancies Challenge intervention. Additionally, interventions should seek to personalize feedback for males and females given the sex differences noted in the current study,

Limitations and Future Directions

There are several limitations to the current study that warrant mention. First, the current study utilized self-report data collected with several months between timepoints, limiting the ability to assess how AE and valuations, as well as other contextual variables, may contribute to specific drinking events. However, examining between- and within-person associations using longitudinal data mitigates limitations associated with findings from cross-sectional studies, and complements findings from daily or EMA studies by demonstrating that variability in AE and valuations occurs over both short (e.g., daily or weekly) and long (e.g., monthly) periods of time (Lee et al., 2015; Ramirez et al., 2020). Second, the CEOA measures positive and negative AE and valuations, and our outcome measure (RAPI) only assessed negative consequences. Previous research has shown that AE predict positive and negative consequences (Lee et al., 2020; Park and Grant, 2005). Given that the relationship between AE and alcohol-related outcomes is assumed to be directional, assessing both types of outcomes is important. Third, our study is limited by the inability to test interactions between AE and valuations. Specifically, social learning theories suggest that decisions to drink are not only affected by the expected outcome, but the value of that outcome. As such, it would be expected that AE and valuations interact such that an AE only results in alcohol consumption when that outcome is positively valued. While some authors have argued that expectancy-value interactions can be calculated (Nicolai et al., 2018), Fromme et al. (1993) suggests that such an approach is problematic because high likelihood, low valuation scores would be equivalent to low likelihood, high valuation scores. However, these interactions have very different implications for motivations to drink. Accordingly, it has been suggested that AE and valuations are scored and interpreted independently, which enable meaningful clinical implications to be drawn from the current findings (Fromme et al., 1993; Valdivia and Stewart, 2005). Ultimately, the unresolved discourse regarding the measurement of AE and valuations, and the suboptimal model fit indices for the CFA we conducted on the valuations scale highlight the need for ongoing improvements to the measurement of AE and valuations. In particular, there is a need for longitudinal measurement validation of these scales examining the extent that the factor structures for AE and valuations are stable (or change) over time.

Conclusion

Several important findings emerged from the current study which have important implications for assessment, prevention, and intervention, as well as future research. First, examining relations between AE, valuations, and alcohol-related outcomes without consideration of changes over time and individual factors (i.e., sex) limits the utility of the findings. This is evidenced by our findings suggesting that different patterns emerged at the between- and within-person levels, as well as the moderating effects of sex. Second, both negative and positive AE and alcohol valuations are important correlates of use and consequences. These findings have important implications for the application of AE theory, as well as important clinical implications for interventions that aim to shift AE with the goal of reducing alcohol-related risks.