Heavy Episodic Drinking and Alcohol Consumption in French Colleges: The Role of Perceived Social Norms
Abstract
Background: The effect of normative perceptions (social norms) on heavy episodic drinking (HED) behavior is well known in the U.S. college setting, but little work is available in other cultural contexts. The objective of this study is therefore to assess whether social norms of alcohol use are related to HED in France, taking account of other influential predictors.
Methods: A cross-sectional survey was carried out among 731 second-year university students in the Paris region to explore the role of 29 potential alcohol use risk factors. The probability of heavy episodic drinking and the frequency of HED among heavy episodic drinkers were modeled independently. Monthly alcohol consumption was also assessed.
Results: Of the students, 56% overestimate peer student prevalence of HED (37% for alcohol drinking prevalence). HED frequency rises with perceived peer student prevalence of HED. Other social norms associated with HED are perceived friends’ approval of HED (increasing both HED probability and HED frequency) and perceived friend prevalence of alcohol drinking (increasing HED probability only). Cannabis and tobacco use, academic discipline, gender, and the number of friends are also identified as being associated with HED.
Conclusions: Overestimation of peer student prevalence is not uncommon among French university students. Furthermore, perceived peer student prevalence of HED is linked to HED frequency, even after adjusting for other correlates. Interventions correcting misperceived prevalences of HED among peer students have therefore the potential to reduce the frequency of HED in this population.
Alcohol use is widespread among students in higher education, and heavy episodic drinking (HED; at least 5 alcoholic drinks on 1 occasion) is frequent (Wechsler et al., 2002). Alcohol dependence can lead to academic failure (Aertgeerts and Buntinx, 2002) and alcohol abuse has serious health consequences (Hingson et al., 2005). Heavy drinking in the university is a risk factor for adverse outcomes up to ten years later (Jennison, 2004). Heavy episodic drinking in young adults has been the object of multiple investigations, which have been recently reviewed (Courtney and Polich, 2009). Effects of HED can be difficult to differentiate from effects from alcohol dependence, as heavy episodic drinkers are much more likely to be classified with alcohol dependence (odds ratio, OR of 19) and with alcohol abuse (OR = 13) (Knight et al., 2002). HED can affect general health: frequent heavy episodic drinkers (at least 3 times in the past 30 days) are more likely to experience more sick days (≥ 14 days in the past month, especially for mental health) than nonheavy episodic drinkers (Okoro et al., 2004). Heavy episodic drinking has been associated with cognitive impairments such as frontal lobe and working memory deficits (Nichols and Martin, 1997; Townshend and Duka, 2005), It is hypothesized that binging increases the number of alcohol withdrawals, which produce long-term deficits (Glenn et al., 1988; Stephens et al., 2005).
Knowledge of the factors that influence alcohol consumption in the university is necessary to design prevention approaches. A recent review of the literature has been conducted in the United States for alcohol use (Borsari et al., 2007) and another was conducted for HED in Europe specifically, but for the general population (Kuntsche et al., 2004). These reviews list many identified risk factors for alcohol use and HED, including one particular class of risk factors, perceived norms, that has the potential to be useful for prevention purposes.
The social norms theory, applied to alcohol drinking, states that students believe other students’ attitudes towards alcohol (injunctive norms, such as perceived approval of alcohol use) are more permissive than they actually are, and that other students drink more than they actually do (descriptive norms, such as perceived prevalence of alcohol use). Their behavior will be influenced by these incorrect perceptions, inducing them to drink more. Misperception of the drinking norms can be caused by pluralistic ignorance (the majority of moderate drinkers falsely believe their peers drink more than they really do), false consensus (the minority of heavy drinkers falsely believe the others drink as they do) and by false uniqueness (alcohol abstainers will falsely believe abstention is more unique than it really is) (Berkowitz, 2004). The social norms approach will try to correct these misperceptions by providing more accurate norms for drinking and abstention, with the assumption that correcting the students’ misperceptions about alcohol drinking will decrease their consumption.
Research has shown that alcohol use misperceptions do exist (Borsari and Carey, 2001, 2003; Perkins et al., 1999) and that perceived social norms are among the strongest predictors of alcohol use (Kypri and Langley, 2003; Perkins et al., 2005). Furthermore, there is some evidence in favor of the effectiveness of social norm interventions (DeJong et al., 2006; Mattern and Neighbors, 2004; Perkins and Craig, 2006), although not all interventions have proven successful (Campo and Cameron, 2006; Clapp et al., 2003; DeJong et al., 2009; Thombs and Hamilton, 2002; Wechsler et al., 2003).
Most of the studies focusing on alcohol in universities have been carried out in North America. In a literature review on HED in Europe, Kuntsche et al. question the transferability of existing U.S. data, particularly on the subject of peer influences, to other countries (Kuntsche et al., 2004). For instance, in France, a southern, wine-producing country, the legal drinking age was 16 years until March 2009, it is now 18 years (21 in the United States), accommodation in student residences concerns a minority (13%) of students (Gruel et al., 2007) and regular drinking is more culture-specific than in the United States (explaining a consumption of 11.4 L of pure alcohol per 15+ capita in France versus 8.4 L in the United States, according to the WHO’s Global Information System on Alcohol and Health).
Misperceptions of the norms are not systematic in all universities (Wechsler and Kuo, 2000) and the cultural contexts of alcohol consumption vary from country to country. To date, few data have been provided outside the United States. Misperceptions of alcohol drinking norms and an association with students’ patterns of use have been reported in universities from New Zealand (Kypri and Langley, 2003) and Scotland (McAlaney and McMahon, 2007). There is a need to assess the role of perceived peer norms in other, non-English-speaking, cultural contexts to decide if the social norm prevention approach is likely to help reduce alcohol use in these settings. Finally, other predictors could be more salient than those measuring social norms and the association between alcohol and social norms could disappear after adjustment for these (Cameron and Campo, 2006).
Factors other than perceived norms identified for alcohol use among higher education students include gender, alcohol expectancies and drinking motives, perceived harm, other substance use and academic discipline (Borsari et al., 2007; Cameron and Campo, 2006; Kuntsche et al., 2004; Sher and Rutledge, 2007; Webb et al., 1997).
This study reports the results of a survey carried out among French higher education students and examines the association between perceived peer norms and alcohol consumption. We study more specifically HED and do not differentiate it from other alcohol use patterns such as alcohol abuse and dependence. The specific questions addressed are as follows:
- (i)
In the French cultural setting, are perceived social norms associated with alcohol use?
- (ii)
Are perceived social norms among the best correlates of alcohol use?
Materials and Methods
Participants
The data originate from a survey on tobacco, alcohol and cannabis use among second-year students from 12 classes. The study was approved by the institutional review board of the French institute of health. Teaching establishments were selected at random among 4 specific academic disciplines in the Ile-de-France region. We included 4 nursing schools, 3 sociology faculties, 3 foreign language faculties and 2 medicine faculties. The 13-page questionnaires, anonymous and voluntary, were completed between October 2005 and February 2006. Approximately 10 students attending the lecture during which the questionnaire was handed out refused to complete the survey, leading to an estimated participation rate of about 98%.
A total of 731 students answered the survey. Table 1 summarizes the characteristics of the sample. Of the respondents, 79% were female, since in France, as in most European countries, females are more represented than males in higher education (Schnitzer and Middendorff, 2005). Nationally, the proportion of females was of 88% for nursing students in 2004 (Marquier, 2006) and of 75% for foreign language students, 70% for sociology students, and 66% for medicine students in 2007 (ministry of education data).
| Academic discipline | Age: mean (standard deviation) | Gender: percentage of females (%) |
|---|---|---|
| Sociology (36%) | 20.9 (3.8) | 70.3 |
| Nursing school (34%) | 24.6 (5.8) | 86.6 |
| Medicine (19%) | 19.9 (0.9) | 77.6 |
| English as a foreign language (12%) | 20.4 (1.5) | 81.6 |
| Overall (n = 731) | 21.9 (4.6) | 78.5 |
Dependent Variables
We assessed both the quantity of alcohol usually drunk in a month and the monthly occurrences of HED (defined as having 5 or more alcoholic drinks on 1 occasion).
Students were asked about (i) the units of alcohol that they usually drank in a week (1 unit = 10 g of pure ethanol, the usual quantity of alcohol contained in a glass of wine served in a bar in France—a reference scale for most common beverages was provided in the questionnaire) and about (ii) the number of times they had 5 or more drinks on 1 occasion in the previous month.
The association between the dependent variables and hazardous alcohol consumption was explored using the AUDIT score (Saunders et al., 1993). A score of 8 and above is predictive of alcohol abuse or dependence. We do not differentiate between nondependent and dependent alcohol users in the main analyses. However, all analyses were also done excluding alcohol-dependent (according to the AUDIT score) students, as these students might have different motives for drinking.
Independent Variables
Potential predictors for HED or alcohol consumption (AC) or were included in the questionnaire. These variables were selected on the basis of those used in the currently existing questionnaires for the epidemiological surveillance of substance use in France (Andersson et al., 2007; Legleye et al., 2007). To explore normative perceptions, perceived friend and peer student, descriptive (prevalence of use) and injunctive (approval of use) norms were assessed: (a–b) perceived peer student an friend prevalence of alcohol drinking, (c) perceived peer student prevalence of HED, (d–f) perceived friends’ approval of HED, moderate and excess drinking. Table 2 presents the 29 variables considered.
| Covariate | Description |
|---|---|
| Student characteristics | |
| Gender | Male or female |
| Age | Age ranged from 18 to 65 years |
| Tobacco use | No use, occasional use, or daily use, in the previous month |
| Cannabis use | No use, use ≤1/week, use >1/week in the previous year |
| Partner status | Being alone or having a partner |
| Family structure | Parents living together or not |
| Baccalauréat grade | Secondary school final exam grade, as a measure of academic achievement |
| Number of friends | 0–4, 5–7, 8–10, or ≥ 11 |
| Number of friends in the class | 0–1, 2–3, 4–5, or ≥ 6 |
| Self-esteem | Rosenberg’s Self-Esteem Scale |
| BMI | Body mass index |
| University environment | |
| Academic discipline | Sociology, medicine, English as a foreign language, or nursing studies |
| Smoking prevalence in the class | Estimated from the students’ answers |
| Cannabis use prevalence | |
| Drinking prevalence | |
| Heavy episodic drinking prevalence | |
| Knowledge of university’s alcohol policy | “Does your campus have a policy (e.g., consumption ban, …) against alcohol”: no, yes, or don’t know |
| Social norms | |
| Previous exposure to substance use questionnaires | Yes or no |
| Perceived exposure to alcohol prevention campaigns | 0 times, 1–3 times, 4–5 times, 6–30 times in the previous month |
| Perceived prevalence of alcohol drinking among peer students | “In your teaching establishment, among 10 students, how many drink alcohol?” |
| Perceived prevalence of heavy episodic drinking among peer students | “In your teaching establishment, among 10 students, how many sometimes have 5 or more drinks in a row?” |
| Perceived approval of moderate alcohol drinking by friends | “What would your close friends think if you had 1 or 2 alcoholic drinks almost every day?”: wouldn’t disapprove, would disapprove, or strongly disapprove |
| Perceived approval of excess alcohol drinking by friends | “What would your close friends think if you had 4 or 5 alcoholic drinks almost every day?”: wouldn’t disapprove, would disapprove, or strongly disapprove |
| Perceived approval of heavy episodic drinking by friends | “What would your close friends think if you had 5 alcoholic drinks in a row?”: wouldn’t disapprove, would disapprove, or strongly disapprove |
| Perceived prevalence of alcohol drinking among friends | “Among your friends, how many drink alcohol?”: none, less than one third, about half, more than two thirds, or all |
| Measures related to alcohol | |
| Alcohol representation score | “Do you think the following qualifiers are relevant to alcohol?” (“harmful,”“a trap,”“a pleasure,”“healthy,”“a scourge,”“friendly and sociable”) 6-item Likert scale from 0 = “Not at all” to 4 = “Entirely” |
| Attitudes towards the alcohol “industry” | Derived from a tobacco industry scale. 6-item Likert scale. Example: “the state has to allow the alcohol industry to sell its products” |
| Beliefs about the alcohol “industry” | Derived from a tobacco industry scale. 7-item Likert scale. Example: “the alcohol industry lies” |
| Alcohol prevention campaigns perception score | 6-item Likert scale about alcohol prevention campaigns: “There are too many,”“They are convincing,”“I don't feel concerned,”“They do not give the right reasons to change behavior,”“They catch attention,”“They have more to do with political issues than with public health issues” |
The Baccalauréat grade, which corresponds to the national secondary school final exam grade, allows entry to the university and is therefore comparable for all students. It is used as an indicator for academic achievement. We used Rosemberg’s scale to measure self-esteem (Vallieres and Vallerand, 1990). The perception of the alcohol-related “industry” derives from a scale originally used for tobacco (Hersey et al., 2003). Students were asked to give an estimate of alcohol drinking and HED prevalence among students of the same teaching establishment, which in France comprises students from the same academic discipline (but not necessarily the same class) and sometimes from similar fields.
Naming Conventions
We make use of several naming conventions in this article, which are summarized here. We use the term “alcohol consumption” to designate the number of alcohol units usually drunk in a month by students. “Heavy episodic drinking” (HED) refers to having 5 or more drinks in a row (we used the same threshold for males and females). We prefer this term to “binge drinking,” since the term binge is associated with alcohol intoxication, and many students drinking above the threshold of 5 drinks in a row have still moderate blood alcohol concentrations (Thombs et al., 2003). It has therefore been recommended to stop using the term “binge drinking” (Berkowitz, 2003; Inter-Association Task Force on Alcohol and Other Substance Abuse Issues, 2000). We use the term “alcohol use” to refer to both alcohol consumption and HED.
Finally, we use the term “peer students” to designate the students from the same teaching establishment.
Statistical Analysis
Our main concern was to model HED, although the same analyses have been conducted for alcohol consumption.
The methodology used for the analysis is the same as the one used to explore tobacco use in the same study (Riou França et al., 2009).
We used multiple imputation (Schafer, 1999) to deal with missing data, as 220 students (30%) did not provide information for at least one of the independent variables. Multiple imputation is a technique in which incomplete data are filled in with values drawn from a distribution (conditional on the observed data) several times (5 here). The resulting datasets are then analyzed in order to obtain combined estimates that take fully into account the increased variability due to the missing values. Since the individuals with missing values are not discarded from the analyses, multiple imputation is less likely to produce biased estimates than complete case analysis (Harrell, 2001a).
The probability of engaging in HED and the frequency of HED among heavy episodic drinkers were modeled separately in a 2-part model (Blough et al., 1999). Two-part-models predict separately the zeroes (i.e., not engaging in HED) and the positive values (i.e., HED intensity among heavy episodic drinkers). Since engaging in HED is a binary variable, a multiple logistic regression model, using the full sample of students, was used to model its probability. The logistic regression model’s coefficients are easily interpretable: their exponential yields an estimate of the odds ratio (OR) of being an heavy episodic drinker. In a second part, a linear regression, using the subsample of students engaging in HED, was used to model the logarithm of HED frequency among heavy episodic drinkers. We modeled log-frequencies instead of the raw frequency to take into account the skewness of the distribution of this variable. The exponential of the coefficients therefore yields, in this part of the model, the (multiplicative) amount by which HED is increased for a particular independent variable.
The variables explored in the study include social norm measures and other factors identified in the literature or commonly measured in France. All these variables are therefore expected to be associated with alcohol use. Given the moderate sample size and the risk of colinearity, we cannot include all variables in the model and we need to select only those with the stronger association patterns. Stepwise variable selection procedures are problematic since they often involve multiple testing of model coefficients (retaining only variables significant at a given level), thus inflating the risk of the type I error. In order to select which variables were to be retained in the regression models, we used a bootstrap procedure (Heymans et al., 2007). For each bootstrap sample, we start with the model with containing all independent variables and iteratively remove the variable that mostly improves model fit, as measured by Akaike’s information criterion (AIC), which penalizes the model likelihood by model complexity. Once no further improvement of model fit is possible by removing independent variables, the variables retained in the model are memorized and the procedure is repeated in the next bootstrap sample. We therefore simulated 1,000 bootstrap samples (200 per imputed dataset) of the data and conducted for each sample a backward, AIC-based, variable selection. The final model contains only the independent variables included in at least 75% of the bootstrap samples.
The model with the retained variables is estimated in Stata. It should be noted that, since we are using multiple regression, the effect of each variable in the model is estimated adjusting for the other included variables. The significance of the variables was tested using Wald tests (the ratio between the estimated parameter and its standard error follows a normal distribution in the logistic regression case and a Student’s t-distribution in the linear regression case) (Harrell, 2001b). Since students are clustered within classes, there is a need to adjust for this sampling design when estimating the standard errors of the estimates. For simplicity’s sake, we prefered a cluster robust variance estimator (Rogers, 1993; Williams, 2000) to a multilevel model (which would be more difficult to estimate, especially in the presence of multiply imputed values). The significance threshold used (type I error) was of 5%.
Results
Substance Use and Perceived Norms
Alcohol Use Patterns. About 89% of the students had already drunk alcohol at least once in their lifetime and 69% were current alcohol drinkers (average alcohol units consumed in a month of 10.3, ranging from 1 to 160, SD = 17.8). About one third (37%) of the students overestimate the prevalence of alcohol use among peer students (19% correctly perceive it and 44% underestimate it). The mean perceived prevalence is of 67% (SD = 21%).
About 29% of the students reported at least 1 episode of HED in the month (average of 2.7 episodes per month, ranging from 1 to 15, SD = 2.2). A majority (57%) of students overestimates the prevalence of HED among peer students (20% correctly perceive it and 23% underestimate it). The mean perceived prevalence is of 36% (SD = 20%).
Among alcohol users, 78% only drank at week-ends and 15% wanted to reduce their consumption.
Among alcohol users, the mean AUDIT score was of 4.9. Of the students, 21% having drunk alcohol in the previous month (and 44% of those classified as heavy episodic drinkers) had an AUDIT score ≥ 8 and could be considered as alcohol dependent. There is a significant linear correlation between the monthly quantity of alcohol consumed and the AUDIT score (ρ = 0.70, p < 0.001). This correlation is even stronger with the monthly frequency of HED episodes (ρ = 0.75, p < 0.001).
Other Substance Use. About 20% of the students were classified as occasional and 15% as daily tobacco smokers. Of the students, 22% were occasional and 7% regular (use > once per week) cannabis users.
Heavy Episodic Drinking Correlates
Probability of Engaging in Heavy Episodic Drinking. This part of the model was estimated on 729 students for whom the monthly frequency of HED was known (29% of the students in this sample engaged in HED at least once in the past 30 days).
8 dependent variables are above the 75% selection threshold for inclusion in the model (see column A of Table 3 for the inclusion probabilities in the model).
| Variable | A: HED probability | B: HED frequency | ||
|---|---|---|---|---|
| % retained | Rank | % retained | Rank | |
| Tobacco use | 100.0 | 1 | 58.2 | 15 |
| Perceived friends’ approval of heavy episodic drinking | 100.0 | 1 | 82.4 | 4 |
| Perceived prevalence of alcohol drinking among friends | 99.8 | 3 | 65.9 | 12 |
| Alcohol representation score | 97.9 | 4 | 30.5 | 26 |
| Gender | 97.6 | 5 | 77.8 | 6 |
| Heavy episodic drinking prevalence in class | 95.1 | 6 | 36.5 | 23 |
| Cannabis use | 93.7 | 7 | 81.5 | 5 |
| Number of friends in class | 85.5 | 8 | 62.2 | 14 |
| Perceived exposure to alcohol prevention campaigns | 74.8 | 9 | 31.3 | 25 |
| Cannabis use prevalence in class | 70.3 | 10 | 40.6 | 18 |
| Smoking prevalence in class | 62.7 | 11 | 75.1 | 8 |
| Perceived friends’ approval of moderate drinking | 62.2 | 12 | 73.0 | 9 |
| Self-esteem | 53.4 | 13 | 62.5 | 13 |
| Alcohol drinking prevalence in class | 50.9 | 14 | 36.8 | 22 |
| Perceived HED prevalence among peer students | 46.5 | 15 | 93.2 | 1 |
| Academic discipline | 43.0 | 16 | 89.4 | 2 |
| Knowledge of university alcohol policy | 36.9 | 17 | 48.8 | 17 |
| Number of friends | 36.8 | 18 | 83.6 | 3 |
| Alcohol prevention campaigns perception score | 35.7 | 19 | 36.1 | 24 |
| Age | 35.4 | 20 | 24.1 | 29 |
| Previous exposure to substance use questionnaires | 35.0 | 21 | 40.4 | 19 |
| Perceived drinking prevalence among peer students | 32.6 | 22 | 75.5 | 7 |
| Attitudes towards alcohol “industry” score | 28.9 | 23 | 68.4 | 10 |
| Perceived friends’ approval of excess alcohol drinking | 28.1 | 24 | 39.7 | 20 |
| Parents together | 26.0 | 25 | 39.0 | 21 |
| Beliefs about alcohol “industry” score | 25.7 | 26 | 29.1 | 28 |
| BMI | 24.1 | 27 | 29.6 | 27 |
| Partner status | 20.2 | 28 | 51.0 | 16 |
| Baccalauréat grade | 17.9 | 29 | 66.1 | 11 |
The model estimates are given in column A of Table 4: perceived approval of HED by friends, a high perceived prevalence of alcohol use among friends, a positive representation of alcohol, being in a class with a high prevalence of HED, occasional cannabis use, tobacco smoking and reporting having more than 5 friends in class all increase the probability of reporting HED episodes.
Frequency of HED Episodes Among Heavy Episodic Drinkers. This part of the model was estimated on the 208 students reporting at least 1 HED episode in the previous month.
Again, 8 dependent variables are above the 75% selection threshold for inclusion in the model (see column B of Table 3). The model estimates are presented in column B of Table 4: gender (p = 0.126, Wald test) and perceived alcohol consumption prevalence in class (p = 0.102, Wald test) fail to reach statistical significance. The overall number of friends reaches significance (p = 0.041, global Wald test) but the pattern of association is unclear. Perceived approval by friends of HED, greater perceived prevalence of HED in class, cannabis use, being a sociology student, and being in a class with low smoking prevalence all increase the frequency of HED among students reporting at least 1 episode.
Alcohol Drinking Correlates
The analysis performed for HED was also carried out to model alcohol consumption in general. Alcohol use probability was estimated on 725 students for whom the usual monthly alcohol intake was known. Alcohol use intensity was estimated on the 499 students usually drinking at least 1 unit of alcohol per month. Table 5 gives the model estimates.
Probability of Being an Alcohol User. Positive representations of alcohol and of the alcohol industry, tobacco and cannabis use, and being in a class with a higher prevalence of alcohol use all increase the probability of alcohol use. Occasional use of tobacco or cannabis is associated with higher probability of alcohol use as compared with daily or more regular use.
Monthly Quantity of Alcohol Drank by Users. Cannabis use fails to reach significance (p = 0.0656, Wald test). Social norms play an important role as correlates of the quantities drunk by alcohol users. Alcohol users reporting that their friends approve of moderate alcohol intake and HED report alcohol consumption 80% higher compared with students reporting that their friends strongly disapprove of these behaviors. Taking the quantities drank by alcohol users reporting that none of their friends drink alcohol as a reference, students reporting having more than two thirds two thirds iends who are drinkers consume twice as much. Female alcohol users drink 61% of the amount drunk by males. Sociology students appear to be the group with the highest consumption among alcohol users and medicine students the one with the lowest (74% of the intake by sociology students) ones. Alcohol users who smoked tobacco reported alcohol consumption that was 31% higher for daily smokers and 43% higher for occasional smokers.
Alcohol Use Correlates Among Nondependent Students
The same analyses were repeated excluding the 106 students who could be considered as alcohol dependent according to their AUDIT score. The model estimates are presented between parenthesis in Tables 4 and 5. The results are comparable although, due to a reduced sample size (especially in part B of the models), some variables are now nonsignificant.
Discussion
Using data from second-year higher education students in the Paris area in France, this study has tested whether the associations observed in the United States between misperceptions of alcohol use and drinking behavior hold in another cultural context (independently of the actual extent of misperceptions). We focussed on HED, and modeled separately the probability of engaging in HED and the frequency of HED episodes among students reporting at least 1 episode in the month. Table 6 summarizes the factors associated with HED and alcohol consumption.
| Variable | Heavy Episodic Drinking | Alcohol consumption | ||
|---|---|---|---|---|
| Use probability | Frequency among users | Use probability | Frequency among users | |
| Descriptive norms: perceived prevalence… | ||||
| of AD among peer students | ||||
| of AD among friends | X | X | ||
| of HED among peer students | X | |||
| Injunctive norms; perceived friends’ approval… | ||||
| of HED (5 drinks in a row) | X | X | X | |
| of excess drinking (4–5 drinks/day) | ||||
| of moderate drinking (1–2/day) | X | |||
| Other variables | Tobacco use, alc. representations, gender, HED prevalence in class, cannabis use, number of friends in class | Academic discipline, number of friends, cannabis use, smoking prevalence in class | Alc. representations, cannabis use, alc. industry attitudes, tobacco use, Alc. use prevalence in class | Alc. representations, tobacco use, gender, academic discipline |
- AD: alcohol drinking; HED: heavy episodic drinking; X: variable included in the model, positive association.
The Influence of Social Norms
Social norms variables appear to be associated with alcohol use, but not in all instances. We find the perceived prevalence of HED among university peers to be associated with HED frequency among students engaging in this behavior, and the perceived proportion of alcohol use among friends to be associated with the quantity of alcohol consumed by drinkers and of engaging in HED. These results challenge the idea that normative social perceptions are not predictive of alcohol consumption (Cameron and Campo, 2006).
Other studies have reported that proximal peers, such as friends, appear to be stronger predictors of drinking behavior than more distal peers, such as peer students (Yanovitzky et al., 2006). Indeed, in our study, while perceived approval of HED by friends is in association with both the probability of engaging in HED and HED frequency, perceived prevalence of HED among peer students is only associated with HED frequency. This pattern is even stronger when modeling the quantities of alcohol drunk by alcohol users: while perceived prevalence of alcohol drinking by peer students is not retained in the model, perceived approval of HED and of moderate daily alcohol drinking, as well as perceived prevalence of alcohol use among friends, are.
The association between perceived norms of alcohol use among friends and alcohol use is in accordance with the social norms theory. However, this association could also be explained by peer selection. Students with a higher propensity to drink could select their friends on the basis of their own involvement with alcohol. Social selection could indeed have a stronger effect than social influence (Bullers et al., 2001). Social selection could explain both the link between perceived prevalence of alcohol use among friends and individual use and the ling between perceived approval of HED by the friends and individual HED.
Perceived prevalence of alcohol drinking among peer students was not retained in the model estimating the probability of alcohol consumption. This contradicts the theory of false uniqueness (Suls and Wan, 1987), according to which alcohol abstainers underestimate the prevalence of abstention among their peers. Alcohol consumption is very prevalent among university students (69% of the sample), which might explain why an alcohol drinking prevalence estimate is not associated with this behavior.
While perceived norms related to both friends and peer students are in association with HED frequency, only those related to friends are linked to the monthly quantity of alcoholic drinks. This could be explained by the culture of alcohol drinking in France (wine is regularly consumed, but in moderate quantities), explaining the accuracy of perception of the norms (only 37% of the students overestimate peer student drinking prevalence). In contrast, HED is seen in France as an emergent behavior, imported from non-Latin countries (Kuntsche et al., 2004). In the 2007 ESPAD survey, the prevalence of HED was of 43% among 15- to 16-year-old French students, as compared with 54% in the U.K. (Hibell et al., 2009). Its use is therefore more variable and more difficult to estimate, leading to more misperceptions (56% of the students overestimate peer students HED prevalence). Students might also consider HED, more than alcohol drinking in general, as a means of integration with other students from the campus, leading to a higher sensibility to the perception of the norms of HED use.
Knowledge of the university’s alcohol policy can also be viewed as 1 type of social norm perception. This variable was not retained in any of the models explaining alcohol use. University norms are related to more distal referents than are peer students and friends and proximal peers are more likely to affect behavior (Yanovitzky et al., 2006). Furthermore, while French universities all enforce anti-smoking policies, alcohol policies mostly do not exist.
Other Correlates of Alcohol Use
Social norms are not the only correlates of alcohol use. Females use less alcohol than males, at least in part because of the differing physiological impact of alcohol on the sexes. This finding is consistent with previous research (Borsari and Carey, 2001; Kuntsche et al., 2004). What our study adds is the knowledge that this association holds even when adjustment is made for other important predictors, such as other substance use, which can be considered as a proxy for the propensity to adopt risky behaviors.
Multiple substance use has also been found to be associated with HED in college (Clapp and Shillington, 2001; Digrande et al., 2000). In our analyses, occasional tobacco and cannabis use appears to be more associated than regular use with the probability of alcohol use. This might indicate a festive pattern of association.
A positive representation of alcohol is also found to be a salient correlate of alcohol use. This score includes alcohol expectancies and alcohol drinking motives, both identified as robust predictors of alcohol use (Borsari and Carey, 2001).
Student sociability (number of friends in general or in class) is associated with HED. Having 5 or more close student friends has also been identified as a risk factor for HED among U.S. students (Wechsler et al., 1995). Highly sociable students could be more receptive to social norms.
Actual prevalence of alcohol use in class is also associated with the probability of alcohol use. Furthermore, academic discipline is associated with alcohol use intensity. This may indicate that in some classes, alcohol is a strong component of student life, whereas in others, abstinence would tend to be the norm. Academic differences in drinking patterns have been documented in other settings (Martha et al., 2009; Webb et al., 1997). As the effectiveness of interventions targeting social norms could be linked to the levels of alcohol use (Campo and Cameron, 2006), this factor should be taken into consideration.
Limitations
When interpreting the results from this study, one should keep in mind that the data come from a cross-sectional sample of students. This design only allows for measures of association, and we do not know whether the independent variables occur before or after the dependent variable in the model.
We used the same threshold (5 drinks, in the past 30 days) for males and females to define HED. This insures a better comparability with results from other European surveys who use the same definition, such as the ESPAD study (Hibell et al., 2009). As females have a different metabolism of alcohol, some studies prefer to use a different threshold, of 4 drinks, for females. The definition of HED varies from one study to the other, the latest proposed definition being “A pattern of drinking alcohol that brings blood alcohol concentrations to 0.08 gram percent or above (≥ 5/4 for men/women in 2 hours) on more than 1 occasion within the past 6 months” (Courtney and Polich, 2009). While our definition of HED does not take gender into account, it has been shown that the 5/4 drinks threshold can lead to intoxication levels that are below those used to define drunkenness (Thombs et al., 2003). By adopting a more severe definition for women, our study discriminates more those more at risk of adverse health effects.
Furthermore, we only measured the student’s alcohol use patterns during the academic year (October to February). Alcohol consumptions during the summer period were not measured, and drinking behaviors may change during this period. If students are more apt to HED during vacations, as seems to be the case in the United States (Courtney and Polich, 2009), our study might underestimate the proportion of heavy episodic drinkers. The correlates of HED might be different in summer and during academic time; still, as this study explores the association between peer student social norms and alcohol use, and it has an interest for university-based prevention campaigns, focusing on consumption during university studies (as opposed to vacation time) is sensible. Moreover, HED behaviors during and outside vacation are likely to be highly associated.
Associations identified between perceived friend norms and individual use could be explained by peer selection: students more prone to drinking can preferentially elect their friends among drinkers. On the opposite, the association identified between perceived peer HED prevalence and individual HED frequency among heavy episodic drinkers can be attributed to a social norms effect, since it less plausible for a student to select his university on the basis of his alcohol use patterns.
Implications for Prevention
This study is in agreement with the hypothesis of an association between perceived alcohol use norms and alcohol use among French higher education students, and replicates North American findings in the French cultural context. Social norms are associated with the quantities drunk by students, with the probability of engaging in HED and with the frequencies of HED, even after adjusting for multiple substance use, gender, alcohol representations, and academic discipline (i.e., including these variables in the regression models). Still, since injunctive norms such as the friends’ approval of alcohol use are also linked to alcohol consumption, and as proximal peers such as friends appear to be more associated to these behaviors than university peers, social norm interventions focussing solely on the correction of misperceived norms among peer students might not be optimal.
Furthermore, a minority of students actually underestimates the norms of alcohol use (23% underestimate peer student HED prevalence but 44% underestimate peer student alcohol consumption prevalence). By providing more accurate norms of use to the students, there is a possibility of perverse effects among underestimators. Universities should decide on whether or not to engage in a social norm prevention campaign on the basis of local data, perhaps originating from a pilot study. Further research is needed to evaluate the effectiveness of social norm interventions in the French cultural setting and rule out the hypothesis of peer selection to explain the findings of this study.
Acknowledgments
The research was funded by a grant from ACTIF (Alliance Contre le Tabac en Île-de-France—Ile-de-France Alliance Against Tobacco), a nonprofit association, with the support of the INPES (National Institute of Prevention and Health Education) and the French Cancer Plan.
Bertrand Dautzenberg is President of ACTIF (Alliance Contre le Tabac en Île-de-France), a nonprofit association which promotes smoking prevention and provided funding for this study. His involvement with the ACTIF is on a voluntary basis and he does not receive any financial compensation for his work. He was responsible for decisions to partially fund this study.
The authors are indebted to Angela Verdier, who checked the manuscript for its English on behalf of the INSERM U669 unit.
