Identifying Theoretical Predictors of Risky Alcohol Use Among Noncollege Emerging Adults

Abstract

Objective:

Studies show that emerging adults who do not obtain postsecondary education are at greater risk for developing alcohol use disorders later in life relative to their college-attending peers. Research examining constructs amenable to change within this population is necessary to inform intervention efforts. Thus, the current study aimed to identify psychosocial correlates of risky alcohol use for noncollege emerging adults. A secondary goal was to examine whether gender moderated the relationships between the psychosocial constructs and alcohol use.

Method:

Participants were a nationally representative sample of noncollege emerging adults (18–22 years old) who reported using alcohol in the past year, recruited through an established Internet panel (N = 209; 125 women). A path model was used to examine the relationship between theoretically derived constructs (expectancies, attitudes, normative beliefs) and risky (peak) drinking. A second model examined a multigroup solution to assess moderating effects of gender.

Results:

The full-sample model revealed significant associations between attitudes toward drinking and risky drinking. The model assessing gender differences revealed association between normative beliefs and drinking for women but not men, whereas attitudes were significantly associated with risky drinking for both men and women.

Conclusions:

Findings highlight the importance of attitudes and, for women, descriptive norms in the etiology of risky drinking among noncollege emerging adults, which emphasizes their potential utility in the development and adaptation of interventions for this at-risk population.

Significant effort has been put forth to reduce high-risk drinking and consequences among emerging adults (18- to 25-year-olds) (see Cronce and Larimer, 2011; Larimer and Cronce, 2002, 2007). The majority of studies examining this age group have focused on college students (typically ages 18–22 years), despite the actuality that less than half of high school seniors transition directly to 4-year institutions of higher education (Aud et al., 2012). Although there are similarities between profiles of drinking among college and noncollege emerging adults, Cleveland and associates (2012, 2013) identified a small but discernible group of noncollege emerging adults who were more likely to drink alcohol on a daily basis. Further, studies have shown that noncollege emerging adults are at increased risk for developing alcohol use disorders later in life (Bingham et al., 2005; Carter et al., 2010; Slutske, 2005). Given the different environmental and interpersonal contexts associated with attending or not attending college (e.g., White et al., 2008), it is possible that drinking among noncollege emerging adults is associated with different factors. Although studies have identified changes in social context and residence as contributing to differences in college and noncollege drinking (White et al., 2005, 2008), the examination of psychosocial factors has been limited to descriptive norms and expectancies (Quinn and Fromme, 2011; Slutske et al., 2002). Other interpersonal factors that have been reliably associated with college alcohol use, such as attitudes and injunctive norms (Baer, 2002; Turrisi et al., 2000; Wood et al., 1992), have been unexplored within the noncollege population. The current study aims to address this gap by identifying the association between theoretically derived psychosocial factors and high-risk drinking among noncollege emerging adults.

Predictors of risky drinking

The current study proposes a framework grounded in theories of decision-making (e.g., Ajzen and Fishbein, 1980; Gerrard et al., 2008; Janz and Becker, 1984), which focus heavily on psychosocial characteristics amenable to change in brief interventions (i.e., expectancies, attitude toward drinking, and both descriptive and injunctive normative beliefs). The constructs and their hypothesized relationships with drinking for noncollege emerging adults will now be discussed in turn.

Expectancies.

Expectancies refer to the perceived advantages and disadvantages of performing a given behavior. Within the college literature, holding more positive expectancies regarding potential outcomes of alcohol use (e.g., having positive social or sexual experiences, having fun) tends to be associated with higher quantity and frequency of alcohol consumption (Palfai and Wood, 2001). This relationship also has been established in work focused on female adolescents and emerging adults (Slutske et al., 2002). It is hypothesized that noncollege emerging adults will be more likely to engage in risky drinking if they perceive that doing so will result in more positive expected outcomes.

Attitude toward drinking.

This component refers to the affective or emotional aspects of behavior, consisting of a degree of arousal and the affective direction (positive or negative) of that arousal (Ekman and Davidson, 1994). In general, individuals with a strong positive attitude toward a behavior will be more inclined to engage in the behavior, whereas those with a strong negative attitude will be less likely to engage in the behavior (Ajzen and Fishbein, 1980). Previous work within the college population suggests that attitudes are consistently associated with risky drinking (Mallett et al., 2009). We hypothesized that noncollege emerging adults with more positive feelings toward risky drinking would be more likely to engage in risky drinking.

Normative beliefs.

Social norms can be divided into two separate influences: descriptive and injunctive (Cialdini, 2003). Descriptive norms typically inform perceptions of how many of one’s friends are engaging in a behavior, or in this case, to what extent (how much) they are engaging in that behavior. Individuals who endorse higher base rates of risky drinking among their peers will be more likely to engage in risky drinking themselves. It is hypothesized that noncollege emerging adults will also follow this pattern. Injunctive norms consist of perceptions regarding the extent to which important referents will approve or disapprove of the behavior. It is hypothesized that the more noncollege emerging adults perceive their peers to approve of risky drinking behaviors, the more likely they will be to endorse those behaviors.

Current study

Using the proposed framework, the goals of the current study were twofold. First, we identified correlates of risky alcohol use for a non-college-attending emerging adult population using a nationally representative sample. Based on previous research linking each of the proposed constructs to risky alcohol use within the college population (Bonar et al., 2011; Borsari and Carey, 2001; Rosenberg et al., 2011), we hypothesized that each (expectancies, attitudes, and norms) would contribute uniquely to risky alcohol use among noncollege emerging adults.

A secondary goal of the current study was to examine whether the influence of identified correlates within the framework remains invariant by gender. Gender differences in high-risk drinking (Wood et al., 2004) have been noted among studies of college students, and it is thought that these differences are even more pronounced among same-age noncollege peers (Borsari et al., 2007). However, previous research using the same sample as the current study found that male and female participants reported similar rates of drinking behaviors (Cleveland et al., 2013).

Method

Participants

Participants for the current study were noncollege emerging adults (N = 209) between ages 18 and 22 years (M = 20.78, SD = 1.21). The majority of participants completed high school and received a diploma (59%) or a General Educational Development (GED) credential (19%); 22% completed less than a high school education. Most participants identified as female (59.8%; n = 125) and White (71.8%; n = 148), although a significant portion of the sample also identified as Hispanic/Latino/a (18.8%; n = 39) or African American (11.2%; n = 23). Only participants who identified as drinkers were used in the current study, and drinking behaviors were widely varied, with participants reporting an average of 6.41 (SD = 13.71) drinks per week.

Recruitment and procedures

Participants were recruited from a predetermined Internet panel (KnowledgePanel; Knowledge Networks, Palo Alto, CA) designed to be representative of the general population of the United States and maintained by the consumer information company Knowledge Networks. KnowledgePanel members are randomly recruited through probability-based sampling of addresses from the U.S. Postal Service’s Delivery Sequence File, which covers approximately 97% of U.S. households. The selected households are provided access to the Internet and hardware, if needed. KnowledgePanel recruitment uses dual sampling frames that include both listed and unlisted telephone numbers, telephone and nontelephone households, cell-phone-only households, and households with and without Internet access. Only persons sampled through these probability-based techniques are eligible to participle (see Cleveland et al., 2013, and Chang and Krosnick, 2009, for more information).

At entry, members of KnowledgePanel complete a profile survey of demographic information, including gender, age, race/ethnicity, income, and education. Based on information provided in the profile survey, it was determined that 1,631 KnowledgePanel members met initial eligibility criteria for the current study (age 18–22 years, not enrolled in high school, and no more than a high school education). In fall 2011, these eligible panelists received an email invitation from Knowledge Networks that described the study and included a link to an online screening survey with criteria to select panelists who had received no postsecondary training of any type (e.g., technical or trade school). Of those contacted, 666 consented to participate and completed the screening items, yielding a study completion rate of 41%—a number consistent with web-based recruitment (Larimer et al., 2007). Of those who completed the screening survey, 264 (40%) were confirmed as nonstudents and completed the anonymous, web-based survey. Approximately 80% of those who completed the survey reported using alcohol in the past year, yielding a final sample size of 209. All 264 panelists who completed the survey were compensated with $5 via a “point” system administered by Knowledge Networks. All study procedures were reviewed and approved by the appropriate institutional review boards.

Measures

Risky drinking.

The Quantity/Frequency/Peak Questionnaire (Dimeff et al., 1999; Marlatt et al., 1998) was used to assess peak alcohol use as an indicator of risky drinking. Participants were asked to think of the occasion when they drank the most in the previous month and then report the maximum number of drinks consumed. A standard drink definition was included (12 oz. of beer or wine cooler, 8.5 oz. of malt liquor, 5 oz. of wine, 3.5 oz. of fortified wine, or 1.5 oz. of distilled spirits).

Positive alcohol expectancies.

Five items from the Alcohol Expectancies Questionnaire (Brown et al., 1987) were used to assess positive expectancies related to alcohol use. Response options ranged, on a 5-point Likert scale, from strongly disagree (-2) to strongly agree (2) (α = .85). Sample items included, “Drinking makes me feel good” and “I often feel sexier after I’ve had a few drinks.”

Attitude toward drinking.

Three items assessed attitudes toward drinking (Turrisi et al., 2009). Participants responded to the statement, “I would feel favorable toward going to a party or a bar on a weekend night and …” (a) not drinking, (b) having a few drinks [e.g., 1–3], and (c) getting drunk.” Response options ranged from strongly disagree (1) to strongly agree (5). To maintain measurement compatibility between attitude toward drinking and risky drinking, we found it most theoretically relevant to use the single item “getting drunk,” which most closely matched with the outcome of risky drinking (Fishbein and Ajzen, 2010).

Normative beliefs.

Normative beliefs consisted of two components: descriptive norms and injunctive norms. Descriptive norms were assessed using the Drinking Norms Rating Form (Baer et al., 1991). Participants were asked to estimate the typical number of drinks their best friend consumed on each day of the week, averaged over the last 3 months. Open-ended responses were summed to create a perceived total number of drinks consumed by one’s best friend during a typical week (α = .83). Injunctive norms assessed participants’ perceptions of their friends’ approval or disapproval of five specific drinking behaviors (Turrisi et al., 2009). An example is: “My closest friends would approve of me playing drinking games.” Response options ranged from strongly disagree (1) to strongly agree (5) (α = .91).

Data analysis plan

Following sample weighting, preliminary analyses were conducted to test for missing data. Missing responses were minimal (<5% on any variable) and addressed using full information maximum likelihood, the default missing data method applied by Mplus (Version 6.1; Muthén and Muthén, 2010).

To identify unique contributions of each factor, a fully saturated path model was conducted in which peak drinking was regressed on all hypothesized correlates (expectancies, attitudes, and norms) simultaneously. To assess gender differences, the same path model was estimated using a multiple group solution, with gender defined as the grouping variable. Participant age was added as a covariate in both models to control for effects attributable to age differences or maturation. A sample-specific poststratification sample weight was used in all analyses to ensure generalizability of findings to a population larger than the study sample (for more information on creation of the sample weight, see Cleveland et al., 2013).

Results

In the overall model, expectancies, attitude toward drinking, and normative beliefs accounted for 19% of the variance in peak drinking. However, attitude toward drinking was the only factor with a significant path coefficient, indicating a significant positive relationship between attitude toward drinking and the amount consumed on peak drinking occasions. Parameter estimates for the full path model are presented in Table 1.

Table 1.

Standardized path coefficients (with standard errors in parentheses) for the effects of psychosocial factors on risky alcohol use, overall and by gender, controlling for age

	Association with risky (peak) drinking
Psychosocial factor	Overall (N = 209)	Men (n = 84)	Women (n = 125)
Positive expectancies	.04 (.11)	.05 (.16)	-.22 (.16)
Attitude toward drinking	.39 (.07)*	.37 (.14)*	.62 (.09)*
Descriptive norms	.10 (.09)	.08 (.14)	.18 (.09)*
Injunctive norms	-.004 (.09)	-.11 (.11)	.13 (.14)
Age	-.001 (.08)	-.01 (.10)	.06 (.12)
R2	.19	.14	.36

^*p < .05.

Adding gender as a grouping variable revealed that the examined correlates account for more variance in peak drinking among women (R² = .36) relative to men (R² = .14). Individual path differences also emerged when the unique contribution of each factor was examined. For women, attitude toward drinking and descriptive norms were significantly associated with risky drinking. However, for men, a significant association was found only for attitude toward drinking. Path coefficients and standard errors are presented by gender in Table 1.

Discussion

The current study examined psychosocial factors associated with risky alcohol use in a nationally representative sample of 18- to 22-year-old noncollege emerging adults. Overall, the tested predictors accounted for 19% of the variance in risky drinking. Contrary to our hypotheses, only attitude toward drinking was significantly and positively related to risky drinking. When the model accounted for gender, interesting differences between men and women emerged. Descriptive norms were significantly associated with risky drinking only for women, suggesting that women’s risky drinking behaviors are more strongly associated with their perceptions of their closest friends’ drinking behavior. Meanwhile, attitudes were associated with drinking behaviors for both men and women.

Although findings only partially supported our hypotheses, we found that attitudes were strongly associated with risky drinking for both men and women. This is consistent with the college literature (Mallett et al., 2009) and indicates that attitudes may be an appropriate target to consider for intervention efforts. Our findings also suggest that descriptive peer norms were significantly associated with women’s risky drinking behaviors in this noncollege sample. One issue to consider for future research, or before adapting normativebased interventions to a noncollege population, is the normative referent group that would be used. Recent studies indicate that key members of a person’s social network, described as “drinking buddies,” strongly influence alcohol use (Leonard et al., 2000; Reifman et al., 2006). Although same-age peers tend to be the primary normative influence among college students, very little research has attempted to identify the persons who serve this role among non-college-attending youth (Lau-Barraco and Collins, 2011). Compared with their college-attending peers, non-college-attending youth may have peer groups that vary in age, and they may socialize in a variety of settings. For instance, some individuals might identify their peer group as high school friends who are now in college, whereas others may identify with coworkers, friends from church, or family members. More work is needed to better understand normative influences in this population.

Limitations and future directions

Although steps were taken to control for weaknesses in the design of the study, it is not without limitations. First, despite the nationally representative sample, without longitudinal data we are unable to make inferences regarding the predictive validity of the psychosocial predictors on drinking outcomes. More information is also needed regarding the role of attitudes, as they consistently accounted for more variance in drinking than the other variables examined. One possible explanation for the strength of this association is the specificity of the attitudes measure, as it consisted of a single item pertaining to risky drinking. Future work that includes more comprehensive measures may further elucidate these mixed findings. Last, this study did not include the whole age range of emerging adults (18–25 years) or such environmental factors as living arrangements (e.g., living with one’s childhood family), relationship status, and employment status, that are known correlates of emerging adults’ alcohol use (Cleveland et al., 2013; Fleming et al., 2010; White et al., 2006). Future work should take these factors into consideration.