Does Alcohol Outlet Density Affect Youth Access to Alcohol?

Abstract

Purpose

To investigate how community alcohol outlet density may be associated with alcohol access among adolescents.

Methods

Data were collected through a three-wave panel study with youth aged 14-16 at baseline using computer-assisted telephone interviews. Study participants were recruited from 50 zip codes with varying alcohol outlet density and median household income in California. Data analyses were conducted using multilevel, linear growth models and data from 1,028 youth (52% male; 51% White).

Results

After taking into account individual-level factors and zip code median household income, zip code alcohol outlet density was significantly and positively related to the initial levels of the likelihood and frequency of getting alcohol through various sources including commercial outlets, shoulder tapping, home or family members, and underage acquaintances.

Conclusions

High levels of alcohol outlets in the community enable youth access to alcohol through commercial outlets, family, and social networks.

One widely used intervention strategy to reduce underage drinking in the United States has focused on restricting youth ability to purchase alcohol from commercial establishments. This approach is proved to be effective. For example, increasing minimum drinking age is associated with reduction in alcohol consumption among young people [1,2]. Community-based intervention activities such as responsible beverage service training and enforcement of underage sales laws also appear to reduce sales of alcohol to minors [3-5]. Yet, American youth still find alcohol readily available. The Monitoring the Future study reports that in 2006, 63% of the 8^th graders, 83% of the 10^th graders, and 93% of the 12^th graders considered alcohol “fairly easy” or “very easy” to get [6]. This should not be surprising given the various sources through which youth can acquire alcohol.

Research examining alcohol availability through retail outlets suggests that, despite the intervention efforts, such availability is quite widespread. For example, alcohol purchase studies indicate that underage-looking individuals could purchase alcohol from off-premise outlets 30-70% of the time [3,7-11]. In addition, the likelihood that purchase attempts succeed at one outlet is higher if similar outlets exist nearby [8]. In this respect, youth who wish to drink may have greater access to alcohol through commercial outlets in areas with higher alcohol outlet densities.

A related form of underage alcohol purchase occurs when a youth “shoulder taps” an adult stranger outside an off-premise establishment asking that person to buy alcohol on his or her behalf. A recent study reported that 8% of a general sample of approached adults agreed to buy alcohol for minors and adults in urban areas were more likely to do so than adults in suburban areas [12]. As urban areas have greater alcohol outlet densities, shoulder tap would be more prevalent in areas with higher alcohol outlet densities.

High alcohol outlet densities have been found to be associated with negative youth well-being, including underage drinking and driving [13], violent assaults [14], and injuries related to accidents and assaults [15]. That youth alcohol access and negative health outcomes are related to alcohol outlet densities seems surprising because commercial outlets are not the major venue through which youth acquire alcohol. Instead, youth access alcohol primarily through social contacts such as friends and relatives [16-19]. This finding is reinforced by research suggesting that getting alcohol through of-age peers happens more frequently among youth who perceive it as less risky to ask of-age peers to purchase alcohol than to buy it themselves [20].

Moreover, important variations are observed in patterns of youth access to alcohol. A recent longitudinal study that followed a group of 6^th graders through grade eight reported that parents are the primary source for alcohol during early teen ages, although the importance of this source decreases with age [17]. Adults other than parents and persons under age 21 are the next most important sources; their importance increases over time with adults other than parents being the most common source by grade eight. Although relatively few teens get alcohol from commercial sources, use of commercial sources increases with age. Females are more apt to rely upon social sources for alcohol [16], whereas males are more likely to use commercial sources [17]. Moreover, youth with higher discretionary income are more likely to access alcohol through commercial sources [19]. Lastly, racial/ethnic minorities may be more likely to use commercial sources [16,17]. Research suggests that this is likely because racial/ethnic minorities tend to live in poor neighborhoods with relatively high concentrations of alcohol outlets [21-23].

In summary, this review suggests that youth alcohol access behavior is complex and conditioned by individual and contextual factors including local alcohol outlet densities. To investigate how community alcohol outlet density may influence youth alcohol access, the present study employed a longitudinal design and a geographic sampling method to recruit participants throughout the state of California from areas with varying levels of alcohol outlet density and household income. Specifically, it was hypothesized that youth who reside in areas with higher alcohol outlet densities would be more likely to use both commercial and social sources for alcohol and use them more often than youth who reside in lower-density areas. The assumption about social sources is based on the notion that, in areas with higher alcohol outlet densities, youth would have greater access to alcohol through social networks because social contacts would have greater access to commercial outlets.

Methods

Procedure

The data were from a panel study of youth aged 14-16 at baseline. Computer-assisted telephone interviews (CATI) in English or Spanish were used to collect data during November 2003 - March 2006. Three interviews were conducted at approximately 1-year intervals. Zip codes were the community units in this study. The sampling plan included all Californian zip codes that had at least 200 youth of ages 14-16 (N = 1,076) according to the 2000 Census [24]. The off-premise license data in 2000 were obtained from the California Department of Alcoholic Beverage Control. We focused on off-premise outlets because few youth of this age group get alcohol from bars, pubs, or restaurants [16,19]. Zip codes were stratified into “low,” “medium” and “high” outlet density groups based on count of active off-premise outlets per roadway mile. The literature indicates that drinking is positively related to neighborhood, family and personal income [25-28]. Zip codes were also stratified into “low,” “medium” and “high” income groups using the median household income data from the 2000 Census [29]. This resulted in a 3-by-3 cross-classification matrix, representing areas that varied orthogonally across alcohol outlet density and household income. The central and diagonal cells (i.e., high density-high income, high density-low income, medium density-medium income, low density-high income, and low density-low income) were then selected. Ten zip codes were randomly selected from each cell and each selected zip code was separated by at least two unselected zip codes using ArcGIS 8.1¹[30].

Listed household telephone numbers within the selected zip codes were purchased and used to recruit participants. CATI randomly selected one youth aged 14-16 per household to complete about 300 interviews in each zip code group (~30 interviews per zip code). When a residential household was contacted, a brief screening was conducted to determine eligibility (e.g., presence of a youth aged 14-16). When more than one qualified youth existed in a household, one was randomly selected. If this youth refused, the next eligible person was invited to participate. Parental consent and respondent assent were then obtained and the interview was conducted or scheduled. About 85% of the purchased telephone numbers were usable; of which 66% were ineligible (e.g., no eligible youth, language barrier, quota met) and 23% eligibility unknown. Of the eligible households (N = 3,567), 43% completed the interview (N = 1,541). The cooperation rate among contacted households was 56%. At Waves 2 and 3, 1,300 (84%) and 1,123 (73% of the baseline), respectively, completed the interviews.

Individual-level variables

At each wave of survey, respondents who reported drinking in the past 12 months were asked about how often they acquired alcohol through various sources. Commercial access consisted of “bought it yourself with a fake ID” and “bought it yourself without a fake ID.” Shoulder tap was measured by asking youth how often they had a stranger buy it for them. Social access included “got it from someone you know who is 21 or older,” “got it from someone you know who is under age 21,” “got it from home with parent's permission,” “got it from home without parent's permission,” “got it from your brother of sister,” and “got it from another relative.” Responses to these items were measured using a 5-point scale (never - more than 6 times) and were recoded into number of times (i.e., 0, 1, 3, 5, and 7).

Because youth may have greater access to alcohol if social contacts drink and drink more often and this social network for drinking may change over time, social network for drinking was considered a time-varying covariate of alcohol access. Thus, measures of parent and peer drinking were constructed for Waves 1-3. Respondents were asked how often in the past 12 months they thought that their parents (or guardians) had at least one whole drink of alcohol, on a 10-point scale (0-9: not at all - every day). Data were obtained separately for each parent and a mean score was taken over two parents. For those having only one parent, parent drinking was indicated by that parent's drinking. Those who reported no parents/guardians were given a score of 0. Respondents were also asked to name up to five friends that they met regularly. For each friend, we asked about age, frequency of meeting with this person in a typical month (0-6: not at all - everyday), and frequency of drinking by this person in the past 12 months (the same 10-point scale as parent drinking). Frequency of drinking by each friend was multiplied by frequency of meeting with this person and then a mean score was taken over multiple friends. Considering that association with of-age persons might have additional bearing on alcohol access, a dummy variable was generated to indicate if any friend they named was of age 21 or older. Personal income was another time-varying covariate considered, indicated by the amount of money that respondents received or earned in a typical week (7-point scale, $0 - more than $125).

Gender, age, and race/ethnicity were included as stable personal factors that may affect youth alcohol access. Gender was indicated by a dummy variable (male =1). Race/ethnicity was indicated by two dummy variables representing Latino and other races (non-Hispanic white were the reference group).

Community-level variables

Community alcohol outlet density was indicated by off-premise outlet counts per 1,000 roadway miles at the zip code level. Community income was indicated by zip code mean median household income. For data analysis purpose, income was rescaled by rounding at thousands.

Data Analysis

Cases were excluded from the analyses if they had any of the following conditions: (a) dropped out of study (N = 418); (b) moved out of California (N = 32); or (c) had missing data on the variables of interest (N = 63). All analyses were restricted to the 1,028 cases with complete data. This study sample consisted of 52% male and 32% Latino, 51% non-Hispanic white, and 17% other (5% African American, 5% Asian or Pacific Island American, and 7% multi- or other races). Mean age at baseline was 14.9 years (SD = .81). Logistic regression analysis indicated that those who were excluded tended to be Latino (p < .001), other races (p < .01), and of age 16 at baseline (p < .05). Differences in gender and drinker status at baseline were not significant.

The main data analyses used linear growth models and were implemented with 3-level models in HLM 6 [31]. Outcome variables were five categories of alcohol access: (a) commercial access, (b) shoulder tap, (c) home or family members, (d) of-age acquaintances, and (e) underage acquaintances. Level-1 models assessed the repeated measures of each category of access. Intercepts (initial levels) and linear growth rates of alcohol access and slopes for time-varying covariates were estimated. To appropriately estimate the level-1 association of the outcome variable with a time-varying covariate, each time-varying covariate was centered around the individual mean and the individual mean was then added in the level-2 models as a predictor for the intercept [32,33]. The level-2 models assessed the between-persons differences in intercepts and growth rates. Predictors included age, gender, and race/ethnicity and for intercept only, mean scores of time-varying covariates. Intercepts and growth rates estimated at level-2 were the outcome variables in the level-3 models and predicted by zip code variables². Two sets of analyses were conducted. The first set used a Bernoulli model to assess the likelihood of use. The second set assessed frequency of use, which was treated as count data with an over-dispersed Poisson distribution [34,35].

Results

Descriptive Statistics

One-third of the study sample (32%) reported past-12-month drinking at baseline. This rate went up to 46% and 56% at Waves 2 and 3, respectively. Only a small number of respondents reported commercial access (5-11%) or shoulder tap (7-13%) throughout the study period. Many more youth got alcohol through social sources; 32-56% had got alcohol from home or through someone they knew. Table 1 presents the descriptive statistics for alcohol access.

Table 1.

Descriptive statistics for alcohol access by sources

	Wave 1	Wave 2	Wave 3
Commercial Access
Bought it yourself with a fake ID, %	1.3	1.2	2.8
Bought it yourself without a fake ID, %	3.9	6.3	9.8
∎ Overall %	4.6	6.8	10.8
∎ Mean frequency (SD)	2.8 (3.6)	3.6 (3.1)	4.1 (3.3)
Shoulder Tap
Had a stranger buy it for you, %	6.7	10.2	12.6
∎ Mean frequency (SD)	2.3 (2.1)	2.6 (2.2)	3.0 (2.3)
Social Access
∎ Overall %	32.4	45.6	55.8
a. Of-age Acquaintances
Someone you know-21 or older, %	22.6	32.2	43.2
∎ Mean frequency (SD)	2.9 (2.3)	2.9 (2.4)	3.5 (2.5)
b. Underage Acquaintances
Someone you know-under age 21, %	20.7	29.4	34.7
∎ Mean frequency (SD)	2.9 (2.3)	3.2 (2.4)	3.4 (2.5)
c. Home or Family Members
Home-with parents' permission, %	7.3	9.8	12.2
Home-without parents' permission, %	12.3	14.8	15.7
Siblings, %	8.2	10.8	14.2
Other relatives, %	10.7	11.0	12.5
∎ Overall %	23.7	29.9	33.5
∎ Mean frequency (SD)	3.4 (3.6)	3.5 (3.5)	4.2 (4.5)

Note. Non-users were excluded from the calculation of mean frequency.

Multi-level analyses

Preliminary analyses showed that initial levels and linear growth rates for the likelihood and frequency of the five categories of alcohol access were all significantly different from zero. Fixed effects and estimates of variance components from the final models for commercial access and shoulder tap are presented in Table 2. Table 3 presents the analyses for social access. The likelihood and frequency models are discussed together.

Table 2.

Results from the final models for commercial access and shoulder tap

	Commercial Access				Shoulder Tap
	Likelihood of Use		Frequency of Use		Likelihood of Use		Frequency of Use
Effects	Fixed	Random	Fixed	Random	Fixed	Random	Fixed	Random
	Coeff.	Variance	Coeff.	Variance	Coeff.	Variance	Coeff.	Variance
Level-1, within-person change		--------a		0.1247		--------a		0.1183
Perceived parent drinking	0.0079		-0.1138		0.0928		-0.0447
Perceived peer drinking	0.0618‡		0.0780*		0.0550‡		0.0505*
Associations with of-age peers	-0.0491		-0.6603		-0.3323		-0.5950
Personal income	0.0050		-0.1691		0.0010		0.0802
Level-2, between-person variation
Initial level		0.8355		9.9362‡		1.4259		7.6568‡
Age	0.5377*		0.5937*		0.5469†		0.7022‡
Male	0.3752		0.7144		0.2413		0.4858
Mean perceived parent drinking	-0.0177		-0.0063		0.0049		-0.0005
Mean perceived peer drinking	0.1565‡		0.2006‡		0.1776‡		0.2334‡
Mean associations with of-age peers	0.5270		0.6746		0.6129		0.6441
Mean personal income	0.1276*		0.2058†		0.0371		0.0811
Growth rate		--------b		3.4466‡		--------b		2.3065‡
Age	-0.1271		-0.1202		-0.3334†		-0.4608‡
Male	0.2058		0.1122		0.1078		0.0539
Level-3, between-zip code variation
Initial level		0.1418		0.1326		0.0382		0.1297
Alcohol outlet density	0.0013‡		0.0024‡		0.0010†		0.0011*
Median household income	-0.0056		-0.0006		-0.0023		-0.0093
Growth rate		0.0318		--------b		0.0051		0.0154
Alcohol outlet density	-0.0000		-0.0005*		-0.0004		-0.0004
Median household income	0.0073*		0.0057		-0.0034		-0.0008

^aNote. The level-1 variance in the Bernoulli model is heteroscedastic and equals 1/p(1-p) where p is the predicted probability according to the level-1 model.

^bFor the model to converge, the variance of this outcome variable was set to 0.

^*p < .05,

^†p < .01,

^‡p < .001.

Table 3.

Results from the final models for alcohol access through social sources

	Home or Family Members				Qf-Age Acquaintances				Underage Acquaintances
	Likelihood of Use		Frequency of Use		Likelihood of Use		Frequency of Use		Likelihood of Use		Frequency of Use
Effects	Fixed	Random	Fixed	Random	Fixed	Random	Fixed	Random	Fixed	Random	Fixed	Random
	Coeff.	Variance	Coeff.	Variance	Coeff.	Variance	Coeff.	Variance	Coeff.	Variance	Coeff.	Variance
Level-1, within-person change		--------a		0.5789		--------a		0.4013		--------a		0.5069
Perceived parent drinking	0.1201‡		0.1201*		0.0901*		0.0482		0.0468		0.0602
Perceived peer drinking	0.0568‡		0.0168*		0.0708‡		0.0349‡		0.0527‡		0.0386‡
Associations with of-age peers	0.2576		0.2361		0.5031*		0.1431		-0.0680		-0.2200
Personal income	0.0331		0.0516		0.0851†		0.0248		-0.0568		-0.0437
Level-2, between-person variation
Initial level		1.2781‡		2.7526‡		1.2736		2.7611‡		0.7613‡		2.2224‡
Age	0.2499†		0.2132†		0.4044‡		0.4032‡		0.3792†		0.2640*
Male	-0.0515		-0.1750		0.0096		-0.0349		-0.6108‡		-0.4932‡
Mean perceived parent drinking	0.1027†		0.1159‡		0.0521		0.0540*		0.1395‡		0.1261‡
Mean perceived peer drinking	0.1622‡		0.1505‡		0.1912‡		0.1621‡		0.1773‡		0.1695‡
Mean associations with of-age peers	0.2798		0.5854*		0.7870*		0.5653*		-0.8594†		-0.5965*
Mean personal income	0.0991*		0.0917*		0.2016‡		0.1864‡		0.0392		0.0627
Growth rate		--------b		0.7326‡		--------b		0.6456‡		--------b		0.7002‡*
Age	-0.2000‡		-0.1540‡		-0.1325*		-0.1349†		-0.3005‡		-0.1766†
Male	-0.0459		0.1239		-0.0387		0.0772		0.2321		0.2293*
Level-3, between-zip code variation
Initial level		0.0240		0.0011		0.0248		0.0118		0.0349		0.0088
Alcohol outlet density	0.0006*		0.0004*		-0.0004		-0.0002		0.0006*		0.0007‡
Median household income	-0.0041		-0.0087†		-0.0107*		-0.0147‡		0.0009		0.0012
Growth rate		0.0181		--------b		0.0142		0.0048		0.0195		0.0005
Alcohol outlet density	-0.0004*		-0.0002		0.0002		0.0000		-0.0001		-0.0002
Median household income	-0.0016		0.0017		0.0023		0.0058†		0.0069*		0.0054*

^aNote. The level-1 variance in the Bernoulli model is heteroscedastic and equals l/p(l-p) where p is the predicted probability according to the level-1 model.

^bFor the model to converge, the variance of this outcome variable was set to 0.

^*p < .05,

^†p < .01,

^‡p < .001.

A. Commercial access

The level-1 models showed that individuals' likelihood and frequency of commercial access increased with perceived peer drinking (Table 2). The level-2 models indicated that initially, rates of commercial access were higher among those who were older, reported greater peer drinking, and had more personal income. The level-3 models indicated that initially, rates of commercial access were higher in zip codes with higher alcohol outlet densities. Growth rates of the likelihood were greater in zip codes with higher median household incomes, whereas growth rates of frequency were greater in zip codes with lower alcohol outlet densities.

B. Shoulder tap

The level-1 models showed that individuals' likelihood and frequency of shoulder tap increased with perceived peer drinking (Table 2). The level-2 models indicated that initially, rates of shoulder tap were higher among those who were older and reported greater peer drinking; growth rates were negatively related to age. The level-3 models indicated that initially, rates of shoulder tap were higher in zip codes with higher alcohol outlet densities.

C. Home or family members

The level-1 models showed that individuals' likelihood and frequency of home or family access increased with perceived parent and peer drinking (Table 3). The level-2 models indicated that initially, rates of this access were higher among those who were older, reported greater parent and peer drinking, and had greater personal income; frequency rates were also higher among those who had greater association with of-age peers. Growth rates were negatively related to age. The level-3 models indicated that initially, likelihoods of this access were higher in zip codes with higher alcohol outlet densities, but growth rates were higher in zip codes with lower densities. Initial levels of frequency were higher in zip codes with higher alcohol outlet densities and lower median household incomes.

D. Of-age acquaintances

The level-1 models showed that individuals' likelihood of access through of-age friends increased with perceived parent and peer drinking, association with of-age peers, and personal income, whereas frequency increased with perceived peer drinking (Table 3). The level-2 models indicated that initially, rates of this access were higher among those who were older, reported greater peer drinking, had greater association with of-age peers and had greater personal income; frequency rates were also higher among those who perceived greater parent drinking. Growth rates were negatively related to age. The level-3 models showed that initially, rates of this access were higher in zip codes with lower median household incomes. Growth rates of frequency were greater in zip codes with higher median household incomes.

E. Underage acquaintances

The level-1 models showed that individuals' likelihood and frequency of access through underage friends increased with perceived peer drinking (Table 3). The level-2 models indicated that initially, rates of this access were higher among those who were older, female, reported greater parent and peer drinking, and had less association with of-age peers. Growth rates were negatively related to age. Growth rates of frequency were also greater among males. The level-3 models showed that initially, rates of this access were higher in zip codes with higher alcohol outlet densities; growth rates were greater in zip codes with higher median household incomes.

Analyses examining racial/ethnic differences

Variables for race/ethnicity were added to the level-2 models that examined the between-person differences in initial levels. The results showed that being Latino was positively associated with all forms of access, but only the associations with home or family access were significant: coeff. = .4486 (p < .001) in likelihood and coeff. = .4169 (p < .001) in frequency. Being “other” (i.e., neither Caucasian nor Latino) was mostly negatively associated with alcohol access; none of these associations were significant. Alcohol outlet density was still significantly related to the likelihood of access through home or family members (coeff. = .0006, p < .05), but marginally related to frequency of this access (coeff. = .0004, p = .055). The negative effect of household income in the frequency model also became less significant (coeff. = -0.0051, p = .110). These findings, together, indicated that Latinos tended to reside in lower-income zip codes and that Latino youth generally were more likely to access alcohol through home or family members. After controlling for race/ethnicity, youth in zip codes with higher alcohol outlet densities were still significantly more likely to access alcohol through home or family members.

Discussion

This paper investigated the association between alcohol outlet density and youth alcohol access. Results from the multilevel analyses showed that after controlling for individual-level factors and zip code household income, initial levels of alcohol access through direct purchase, shoulder tapping, home/family members, and underage acquaintances were all higher in zip codes with greater alcohol outlet densities. These findings support the argument that high alcohol outlet densities enable youth access to alcohol through commercial outlets and social networks (including family members).

The positive effects of alcohol outlet density found in the commercial access models are consistent with a previous study showing that off-premise establishments are more likely to sell alcohol to minors if they have similar outlets nearby [8]. Also consistent with the literature, only a small portion of study sample accessed alcohol through commercial outlets. What is at issue here, however, is that underage alcohol purchases succeed and that use of this mode of access increases with age. More importantly, a small number of risk-takers could supply alcohol to many underage drinkers. As youth predominantly get alcohol from home or someone they know, intervention practices that focus mainly on preventing alcohol sales to minors or shoulder taps will have a limited impact on overall youth alcohol access.

The association of alcohol access with social network for drinking (i.e., perceived parent and peer drinking) found in this study suggests that parents and peers may influence youth alcohol access in two ways: promotion of drinking and availability of alcohol. Youth who are exposed to parent drinking or socialize with drinkers are more likely to drink and have greater demand for alcohol. They may acquire alcohol through social or commercial sources, depending on convenience; youth with drinking parents find alcohol readily available at home and drinking peers may supply alcohol.

Personal income and neighborhood income appeared to be differentially associated with youth alcohol access. Personal income had a limited impact on within-person changes in alcohol access; but on average, greater personal income was associated with greater alcohol access. These findings suggest that in general, youth with high personal income may lead a lifestyle that involves drinking and demands greater alcohol access. In contrast, youth in lower-income zip codes were more likely to get alcohol from home or of-age friends whereas youth in higher-income zip codes were more likely to get alcohol through underage friends, suggesting that social network for drinking may differ for youth living in areas with different levels of family income.

Racial/ethnic differences in youth alcohol access were limited, except that Latinos were more likely to get alcohol from home or family members than Caucasian youth and youth in other racial groups. Although racial/ethnic minorities in this study tended to live in zip codes with lower household incomes and greater alcohol outlet densities (analysis not shown), they were not more likely to access alcohol through commercial outlets than Caucasian youth. Instead, youth residing in zip codes with higher alcohol outlet densities were generally more likely to use commercial access or shoulder tapping for alcohol when individual factors and household income were considered.

Findings from this study should be considered in light of several limitations. First, the study sample may not be representative of youth in California, as it under-represented African and Asian Americans. Neither was it representative of American youth. Rates of drinking among participants of this study were, however, comparable with the National Survey on Drug Use and Health (NSDUH) data. At baseline, 10%, 20% and 28% of our 14-to-16 year olds reported past-30-day drinking compared to 13%, 21%, and 29% in the 2003 NSDUH [36]. Two years later, 24%, 40% and 49% of our 16-to-18 year olds reported such behavior compared to 27%, 32% and 46% in the 2006 NSDUH [37]. Nonetheless, whether a study sample is representative or not does not affect the structural relationships among variables of interest. The second limitation relates to telephone survey's low response rate. Thirdly, using listed telephone numbers excludes unlisted numbers and households without telephone service, possibly leading to bias. We are unable to address biases associated with non-responders, i.e., households which might have met our criteria but were never reached. Despite these limitations, our study provides insights into how community alcohol outlet density may influence alcohol access behavior among adolescents.