Article Text

Download PDFPDF

Alcohol outlet density, perceived availability and adolescent alcohol use: a multilevel structural equation model
  1. E Kuntsche1,
  2. H Kuendig1,
  3. G Gmel1,2
  1. 1
    Swiss Institute for the Prevention of Alcohol and Drug Problems (SIPA), Research Department, Lausanne, Switzerland
  2. 2
    Alcohol Treatment Centre, Lausanne University Hospital, Lausanne, Switzerland
  1. Dr E Kuntsche, Swiss Institute for the Prevention of Alcohol and Drug Problems (SIPA), Research Department, PO Box 870, CH-1001 Lausanne, Switzerland; ekuntsche{at}sfa-ispa.ch

Abstract

Objectives: Perceived availability is commonly associated with adolescent alcohol use. Little is known about the factors which shape this perceived availability. The present study investigates (1) whether perceived alcohol availability is related to the characteristics of the adolescents’ social environment and the per capita outlet density in the community and (2) whether adolescent alcohol use is related to perceived availability, social environment characteristics and outlet density.

Methods: Multilevel structural equation models were estimated based on data from a national representative sample of 6183 adolescents in the 8th and 9th grades of school (mean age 14.8 years) in 254 communities in Switzerland.

Findings: Social environment characteristics, that is, having peers and siblings who drink, going out without parental knowledge of the adolescents’ whereabouts, drinking in public settings and the density of on-premises but not off-premises alcohol outlets, were related to perceived availability. Adolescent alcohol use increased with the permissiveness of social environment characteristics and with increasing perceived availability. Community-level perceived availability and the density of on-premises but not off-premises outlets were related to volume drinking but not to the frequency of risky drinking occasions.

Conclusions: Perceived availability and drinking volume appear to be shaped by the adolescents’ social and physical environments. Adolescents who have a variety of opportunities to obtain alcohol might develop the impression that underage drinking is common and socially endorsed. Consequently, preventive actions to curb adolescent alcohol consumption should take into account the social acceptance of drinking and the physical availability of alcohol in the community.

View Full Text

Statistics from Altmetric.com

It is widely accepted that among adolescents perceived availability of alcohol is positively associated with their alcohol consumption.13 However, research on adolescent populations is scarce and often fails to include or disentangle the effects of the various determinants of availability and adolescent alcohol consumption. It is still unclear which characteristics of the physical and social environment shape adolescents’ perception of alcohol availability. Using a multilevel approach with two different data sources, the present study investigates (1) individually perceived availability as a function of alcohol outlet density and the characteristics of the adolescents’ social environment and (2) adolescent alcohol use as a function of perceived availability and social and physical characteristics.

Empirical evidence has demonstrated that the density of alcohol outlets is associated with alcohol use and heavy drinking in the general population4 5 and among college students.6 7 Authors argue that a high number of alcohol outlets in a given environment makes alcohol more easily available. In turn, easy availability was associated with higher levels of alcohol use1 8 and alcohol-related consequences, such as drinking and driving,4 car crashes and fatalities,9 injuries10 and violent crime.11 Therefore, reducing the number of alcohol outlets has been put forward as a means of reducing overall drinking levels, risky single occasion drinking (RSOD; also called binge drinking or heavy episodic drinking), and alcohol-related problems.1 5 1214

However, most of the literature on alcohol outlet density so far has focused on adult populations.4 5 15 16 As a result, little is known about whether young people are affected to the same degree by outlet densities as adults.17 Some notable exceptions do exist, however. In multilevel analyses, perceived alcohol outlet density and illegal merchant sales in the community were directly related to adolescent alcohol use at the individual level.18 19 However, these studies did not include perceived availability or characteristics of the adolescents’ social environment. Solely focusing on structural measures such as the density of alcohol outlets neglects the fact that adolescents commonly obtain alcohol from parents, peers and other non-commercial sources.2022

The first aim of the present study was to test to what degree perceived availability among adolescents depends on their social environment (peers or siblings who drink, drinking in public places or without parental supervision) and on their physical environment (the per capita density of on-premises and off-premises alcohol outlets). All these factors may contribute to adolescents’ perception that alcohol is easily available by creating the impression that underage drinking is common and is socially endorsed in their physical or social environment.2 23 The second aim was to determine the number of drinks consumed in the last 12 months and the frequency of RSOD as a function of the characteristics of the social environment and the perceived availability at an individual level, and as a function of the density of on-premises and off-premises alcohol outlets in the municipality, that is, within the smallest political unit such as a village or a core city (but without any suburban areas that are municipalities themselves), and perceived availability at the community level.

METHODS

Data sources

Individual student data came from Swiss participation in the 2003 European School Survey Project on Alcohol and Drugs (ESPAD),24 which has been conducted every 4 years in Europe since 1995. These data were collected by means of a paper–pencil questionnaire which was sent to schools to be administered to each pupil in the selected classes. According to the Helsinki Declaration,25 the students could freely choose to participate and confidentiality was ensured at all stages of the study. Based on a list of all state-run school classes in Switzerland, 473 classes were randomly sampled, of which 408 participated (ie, an 86.3% response rate at the class level). Only 4.1% of the students in the participating classes did not take part in the survey because they were absent or refused (ie, a 95.9% response rate at an individual level). Taken together, the overall response rate was 83.1%. The sample of 6337 students can be considered as representative of all the 8th and 9th grades of the schools in the German-, French- and Italian-speaking regions in Switzerland. For five school classes (n = 81, 1.3%), community-level information was not available; these classes were therefore excluded from the analyses. Missing values at the individual level varied between 0.2% and 1.3%. Those students who did not answer two or more questions (n = 73; 1.2%) were excluded. When a student answered all but one question (n = 257; 4.1%), the missing values were replaced by Markov Chain Monte Carlo estimates.26 27 The advantage of such an imputation method is that it uses the maximum available information for an individual derived from other items. The LISREL 8.51 programme28 was used to impute missing values. The final sample consisted of 6183 12–17-year-old students clustered in 358 school classes in 254 communities (ie, an average of 1.4 classes per community).

Community-level data were taken from the 2001 Swiss national company census.2931 This census was based on information from every company registered in Switzerland on 28 September 2001. This information was cross-checked using other data sources such as the taxation register and employment statistics. In total, there were 480 107 companies, of which 98.2% participated in the national company census. However, some registered companies employed a particularly small workforce (eg, they had less than one full-time employee; n = 91 077, 18.9%) and thus had to be excluded by the Swiss Federal Statistical Office before the geospatial indicators were calculated. In the end, these indicators were based on information provided by 382 979 companies in 2840 Swiss municipalities, which had a total of 5433 off-premises and 23 408 on-premises alcohol outlets.

Measures

Individual level

Drinking volume (QF)

The total number of drinks consumed in the last 12 months was obtained by multiplying the number of drinks consumed at a typical occasion (Q) by the frequency of alcohol use (F). To determine frequency, the question dealt with the number of drinking occasions in the last 12 months with answer categories ranging from zero to 40 or more. To create a linear measure, midpoints of categories were used, with 45 occasions for the upper category (40 times plus half the range to the midpoint of the adjacent category). The usual quantity of alcohol use assessed the total number of standard drinks of any alcoholic beverage (eg, beer, wine, spirits and alcopops) consumed on a typical occasion. The answer categories ranged from “less than one drink” to “five or more drinks”. Midpoints of categories were used, with 0.5 drinks for the lowest category and 5.5 drinks for the highest.

Risky single occasion drinking

The question used to assess RSOD was “Think back once more over the last 30 days. How many times (if any) have you had five or more drinks in a row?”, with answer categories ranging from none to 10 or more times. Midpoints of categories were used, with 11.25 occasions for the upper category.

Perceived availability

Three questions assessed the perceived availability of alcoholic beverages. These were prefaced by the sentence “How difficult do you think it would be for you to get each of the following, if you wanted?”. Adolescents could indicate whether they found beer, wine and spirits “impossible” (coded as 1), “very difficult” (coded as 2), “fairly difficult” (coded as 3), “fairly easy” (coded as 4) or “very easy” (coded as 5) to obtain. The internal consistency of the three perceived availability items was good (Cronbach’s alpha = 0.86; note that values above 0.7 are considered to be satisfactory).32 33

Drinking peers

The question was “How many of your friends would you estimate drink alcoholic beverages?”, with answer categories ranging from “none” (coded as 1) to “all” (coded as 5).

Drinking siblings

Adolescents who indicated that one or more older siblings drink alcoholic beverages were coded as 1. Adolescents whose older siblings did not drink or those who did not have older siblings were coded as zero.

Drinking in public settings

To distinguish between adolescents who usually drink in commercial public drinking places and those who usually drink in private settings — for example, at home, at a friend’s home or at private parties — respondents who indicated that the last time they drank alcohol was “at a bar or a pub”, “in a disco” or “in a restaurant” were coded as 1; all others were coded as zero.

Poor parental knowledge

The question was “Do your parents know where you spend Saturday nights?”, with answer categories ranging from “always know” (coded as 1) to “usually don’t know” (coded as 4).

Community level

Alcohol outlet density

Two indicators of the physical availability of alcohol were used: density of on-premises (eg, restaurants, bars) and that of off-premises (eg, shops, supermarkets) alcohol outlets per 1000 habitants. These per capita measures have the advantage of compensating for effects due to differences in population density in each community.34 35 Both density measures were highly correlated with the road mean distance of on-premises (r = −0.20, p<0.001) and off-premises (r = −0.29, p<0.001) outlets in the communities.

Statistical analyses

Both research questions mentioned above can be answered by estimating a comprehensive two-level structural equation model (fig 1). This was done for QF and RSOD separately by using the statistical software Mplus 3.13.36 The datasets of students and communities were matched by means of a community identification number. To evaluate the overall model fit, the comparative fit index (CFI), the Tucker–Lewis index (TLI), the root mean square error of approximation (RMSEA) and the standardised root mean square residual (SRMR) were used. The CFI and the TLI are related to the total variance accounted for in the model; values greater than 0.95 are desired. The RMSEA and the SRMR are related to the residual variance; values less than 0.05 are desired. To account for the skewness of alcohol use variables, a logarithmic transformation was applied. Since the log of zero is not defined, one drink was added before calculating the logarithms (eg, QFln = ln(QF+1)). Adding one drink puts the smallest possible value (ie, 1) of the logarithm back to zero.37 38 The intraclass correlation coefficient was 0.02 for QFln and RSODln. Since alcohol use in adolescence varies according to sex and age,39 the model was adjusted accordingly at the individual level. The perceived availability of beer, wine and spirits was used to create a continuous latent factor which assessed differences in perceived availability at the individual and the community levels.

To answer the first research question, perceived availability (PA) was regressed on drinking peers (DP), drinking siblings (DS), public drinking (PD) and parental knowledge (PK) at the individual level:

PAij = βPA0jPA1j DP+βPA2j DS+βPA3j PD+βPA4j PK+βPA5j sex+βPA6j age+rPAij withβPA0j = γPA00PA01 OnP+γPA02 OffP+uPA0j

where PA indicates the regression of perceived availability, j the community and i the student within a community, β and γ the regression weights at individual level and community level respectively, and r and u the error terms at individual and community level. At the community level, perceived availability (βPA0j; ie, the variation in perceived availability across communities when all individual-level variables were taken into account) was regressed on the per capita density of on-premises (OnP) and off-premises (OffP) alcohol outlets.

To answer the second research question, alcohol use (AU; ie, QF and RSOD) was regressed on drinking peers, drinking siblings, public drinking, parental knowledge, and perceived availability at the individual level. At the community level, they were regressed on the density of on-premises and off-premises outlets:

AUij = βAU0jAU1j DP+βAU2j DS+βAU3j PD+βAU4j PA+βAU5j PK+βAU6j sex+βAU7j age+rAUij with

βAU0j = γAU00AU01 βPA0jAU02 OnP+γAU03 OffP+uAU0j

RESULTS

On average, participants indicated that either “some” or “most” of their peers drank alcoholic beverages (table 1). Nearly one-third had older siblings who drank alcohol. More than one-quarter had their last drink in a public setting (ie, in a bar, pub, restaurant or disco). On average, participants found it “fairly easy” or “very easy” to obtain beer, wine or spirits. Less than one out of five adolescents had not drunk alcohol in the last 12 months. Participants had consumed an average of 33 standard drinks in the last 12 months. More than two-thirds had at least one risky drinking occasion in the last 30 days; the mean was one occasion. On average, communities had approximately three on-premises alcohol outlets and one off-premises alcohol outlet per 1000 inhabitants.

Table 1 Sample and variable descriptions

Results from the multilevel structural equation model revealed that, at the individual level, drinking peers and siblings, drinking in public settings and poor parental knowledge were significantly related to the perceived availability of beer, wine and spirits (table 2). At the community level, a significant link was found between on-premises but not off-premises alcohol outlet density and perceived alcohol availability. The fits of the QF and the RSOD model were excellent (note that CFI and TLI values greater than 0.95 and RMSEA and SRMR values less than 0.05 indicate an acceptable model fit).40 41

Table 2 Individual-level and community-level predictors of perceived availability (standardised beta values; t values in brackets) and explained variance

At the individual level, having peers and siblings who drink, drinking in public settings, little parental knowledge and perceived availability were significantly related to QF and RSOD. At the community level, perceived availability and on-premises but not off-premises alcohol outlet density were positively related to QF but not to RSOD.

DISCUSSION

The present study tested to what degree perceived availability of alcohol among adolescents depended on the characteristics of their social and physical environments. The results demonstrate that various social characteristics, such as a high proportion of drinkers in the peer group, having siblings who drink, drinking in bars, restaurants or discos, or going out without their parents’ knowledge, were related to the perception that alcohol is easily available. At the community level, the density of on-premises but not off-premises alcohol outlets was positively related to perceived availability.

To actually obtain alcohol, the characteristics of the social environment and the density of on-premises alcohol outlets seem to be more important than the density of off-premises outlets. Previous research has shown that adolescents obtain alcoholic beverages frequently through social sources,2022 but no difference was reported for on-premises and off-premises outlets.21 Therefore, it is more likely that the perception that alcohol is easily available may be influenced by the visibility of drinking in the adolescents’ social and physical environments. This may explain why no relation was found with off-premises outlets, as alcohol consumption is directly visible only in on-premises outlets. Having peers and siblings who drink, or perceiving drinking in a range of public settings, may create normative expectations about the appropriateness of drinking and may shape the opinion of adolescents that drinking is common and socially endorsed.2 15 16 23

What is already known on this subject

Previous studies demonstrated that the greater the availability of alcohol—either subjectively perceived or obtained through commercial or social sources—the higher the consumption. However, research on adolescent populations is still scarce and often fails to include or disentangle the effects of different determinants of availability and adolescent alcohol consumption.

Adolescent alcohol use and heavy drinking were related to characteristics of the adolescents’ social environment and to perceived availability. The mean level of perceived availability in the community and the number of on-premises but not off-premises alcohol outlets per 1000 inhabitants was related to the number of standard drinks consumed in the last 12 months. This was, however, not the case for the frequency of risky drinking occasions.

What this study adds

The results of the present study demonstrate that the characteristics of the adolescents’ social environment, such as a high proportion of drinkers in the peer group, having siblings who drink, drinking in bars, restaurants or discos or going out without their parents’ knowledge, were related to the perception that alcohol is easily available. Adolescent drinking volume and risky drinking were directly related to characteristics of the adolescents’ social environment and indirectly to increased perceived availability. At the community level, the mean level of perceived availability and the number of on-premises but not off-premises alcohol outlets were related to the number of standard drinks consumed in the last 12 months but not to the frequency of risky drinking occasions.

Again, there are different explanations. A high number of on-premises outlets in the community might provide adolescents with greater opportunities to obtain alcohol, so that even underage drinkers (note that 89.2% of the sample were younger than the legal purchase age of 16) are more likely to find a bar, pub, disco or restaurant that will sell them alcohol. A high number of possibilities for alcohol consumption in public settings might signify that adolescents often go out to drink in public places, which is in turn related to a high number of drinks consumed in the last 12 months. However, the higher degree of social control in public drinking places compared with private settings might account for the fact that no significant link was found between the density of on-premises outlets and the frequency of risky drinking occasions. In addition, as mentioned above, perceived availability and social and physical characteristics may create normative expectations concerning the appropriateness of drinking in general,2 15 16 23 but not necessarily the appropriateness of heavy drinking.

POLICY IMPLICATIONS

Reducing underage sales of alcohol in bars, pubs, discos and restaurants and changing the acceptance of underage drinking in the community and the adolescents’ proximal social surrounding appear to be promising approaches to curbing adolescent alcohol use.

The major strength of the present study was the integration of different characteristics of the adolescents’ social and physical environments and the perceived availability at individual and community levels in one comprehensive multilevel structural equation model. This model was based on different data sources, namely a large national representative sample of adolescents and the geospatial data provided by the Swiss Federal Statistical Office. However, some limitations of the study should also be mentioned. For example, no direct measure of the frequency with which participants actually obtained alcoholic beverages from social and commercial sources was available. We were also restricted in choosing the ecological unit of analysis, as only municipality-level information could be obtained from the Swiss Federal Statistical Office. Although a rather good match was found between municipalities and the schools’ sampling, smaller ecological units such as neighbourhoods or street blocks might provide a more accurate picture of the links between individual and environmental characteristics.11 16 35 Another limitation concerns the cross-sectional nature of the data from which no causal conclusions can be drawn. For example, it was not possible to investigate how an increase in perceived availability leads to an increase in alcohol use over time and to what degree changes in alcohol use affect perceived availability. Such an examination could be carried out only as part of a longitudinal study.

Conclusions

Results from the present study indicate that the social environment, the perceived availability at individual and community levels, and the per capita number of bars, pubs, discos and restaurants are important determinants of adolescent alcohol use. These factors might provide adolescents with numerous opportunities to obtain alcohol through different sources. Moreover, by living in a social and physical environment in which alcohol use is omnipresent — peers and siblings who drink and a high density of bars, pubs, discos and restaurants — adolescents might develop the impression that underage drinking is common and socially endorsed.

Implications for policy and prevention

Similar to the general population,1 5 14 it appears that reducing the number of bars, pubs, discos and restaurants and underage sales of alcohol in these establishments through law enforcement measures42 or server training programmes43 could prove promising in terms of lowering adolescent alcohol use. However, for effective prevention of adolescent access to alcohol, structural measures should be extended to cover the wider community, or even the family. Komro and Toomey,44 for example, suggest efficient strategies such as laws and law enforcement that address commercial purchasing and also interventions in private settings where alcohol is distributed to minors. They also claim, however, that there is still too little research on how to reduce social access to alcohol.

In addition, it appears promising to change the acceptance of underage drinking in the community and the adolescents’ proximal social surroundings including peers, teachers, parents and siblings.19 45 Based on the results from a study of adolescents in Switzerland, Kuntsche and Delgrande Jordan46 call for preventive programmes that include parents, teachers, school administrators and the community to create an overarching environment of disapproval.

Acknowledgments

The authors would like to thank Esther Grichting, Matthias Wicki and Jürgen Rehm for their participation in the Swiss ESPAD project, the Swiss Federal Statistical Office for providing data on the companies’ census, and Elaine Sheerin for her valuable comments on the editing of the manuscript.

REFERENCES

View Abstract

Footnotes

  • Funding: The Swiss ESPAD study was mainly funded by the Swiss Federal Office of Public Health (grant no. 02.000537). Additional funds for the analysis of environmental factors in adolescent alcohol use were obtained from the Swiss National Science Foundation (grant no. 105311-111849/1).

  • Competing interests: None declared.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • In this issue
    Mauricio L Barreto