• Open Access

The social gradient of alcohol availability in Victoria, Australia

Authors

  • Michael Livingston

    1. School of Population Health, University of Melbourne, Victoria and AER Centre for Alcohol Policy Research, Turning Point Alcohol and Drug Centre, Victoria
    Search for more papers by this author

Correspondence to: Michael Livingston, AER Centre for Alcohol Policy Research, Turning Point Alcohol and Drug Centre, 54 Gertrude Street, Fitzroy, VIC, 3065; e-mail: michaell@turningpoint.org.au

Abstract

Objectives: To explore the socioeconomic gradient of alcohol availability in Victoria, Australia.

Methods: Using data for the smallest geographic units available, per-capita and area-based measures of alcohol outlet density were calculated for four types of alcohol outlet (broadly: restaurants, hotels, packaged liquor outlets and licensed clubs). These densities were aggregated across deciles of socioeconomic disadvantage, to compare the average density of alcohol outlets across ten levels of socioeconomic status. In addition, negative binomial regression models were developed, assessing the relationship between density and socioeconomic status while controlling for relevant covariates.

Results: In urban areas, both takeaway liquor outlets and licensed clubs were significantly more likely to be located in areas of socioeconomic disadvantage, while hotel and restaurant licences were more prevalent in advantaged areas. In rural and regional Victoria, all types of outlet types were more prevalent in areas of socioeconomic disadvantage.

Conclusions: The findings suggest that disparities in exposure to alcohol outlets in Victoria may play a role in explaining socioeconomic disparities in health outcomes, with Victorians living in disadvantaged areas generally faced with higher levels of alcohol availability.

This study examines how alcohol outlets are distributed in Victoria, Australia. In particular, the study explores the socioeconomic spread of alcohol outlets to determine whether their distribution may help explain socioeconomic inequalities in alcohol-related harms.

There has been substantial international research highlighting the significant socioeconomic inequalities in health outcomes in developed countries.1 Within Australia these inequalities are considerable, with mortality rates significantly higher among disadvantaged communities across a range of age groups and causes of death.2 The causes of health inequalities are varied and complex,3 but it is clear that risky health behaviours contribute to them, with notable socioeconomic gradients found for health risk factors such as obesity and smoking internationally4,5 and within Australia.6,7 The situation for alcohol is less straightforward. The relationship between socioeconomic disadvantage and drinking pattern varies substantially between countries and sub-populations.8,9 In Australia, there is some evidence that males from lower socioeconomic areas are more likely to drink at risky levels,10 although this is not a consistent finding.11 Despite the varied findings on the relationship between socioeconomic disadvantage and drinking, there is clearer evidence that less advantaged people experience a greater burden of alcohol-related harm. For example, research in Finland found that for a given level of drinking, poorer Finns were around twice as likely as wealthier Finns to die or be hospitalised due to alcohol.12 Higher rates of alcohol-related mortality among disadvantaged sections of the population have been found in a number of countries,13–15 and a recent study in Australia suggested blue-collar workers had a cirrhosis mortality rate more than twice that of white-collar workers.16 Given the disparities in harm experienced, there is clear evidence that alcohol contributes to health inequalities in many parts of the world, including Australia.

Many researchers have looked to environmental factors to explain the disparities in health observed across socioeconomic groups.17 This is particularly true when the focus is on health risk behaviours.

For example, a range of studies have examined the association between socioeconomic deprivation and access to healthy food. In the US, Walker et al.18 reviewed 31 studies examining food access, finding that, in general, higher densities of fast-food outlets and less access to supermarkets occurred in disadvantaged neighbourhoods. Similar work in the US has demonstrated lower access to facilities related to physical activity in disadvantaged neighbourhoods.19 Findings outside of the US have been more mixed.20 For example, a recent study in New Zealand demonstrated higher densities of fast-food outlets in deprived neighbourhoods, but similarly high rates of access to healthy foods in these neighbourhoods.21 In Australia, a number of studies have found higher access to fast-food outlets in socioeconomic disadvantaged neighbourhoods, including three studies from Victoria.22–24

Similar analyses have been undertaken examining access to attempt to explain socioeconomic disparities in smoking rates. These studies have been limited to the US, but have found support for the argument that the local environment contributes to health inequalities. Multiple studies have found evidence linking rates of smoking to tobacco outlet densities25–27 and it has been shown repeatedly that tobacco outlets are more prevalent in poor or minority neighbourhoods.26,28,29

There is good reason to focus on the equity of how alcohol outlets are distributed, as there is a substantial research literature linking the distribution of outlets to the distribution of alcohol-related harm.30,31 A large number of cross-sectional studies have identified associations between alcohol outlet density and harms including violence,32–34 accidents,35 sexually transmitted disease,36 morbidity,37 youth drinking,38,39 child maltreatment,40,41 and neighbourhood amenity problems.42 This link has been further validated by longitudinal studies demonstrating that alcohol-related harms change along with the density of alcohol outlets.43–46 It is worth noting that the specific effects of alcohol outlets on harm types vary substantially by type of outlet, type of harm and the setting of the study.31 Studies on Victorian data at the postcode level have found particularly strong effects for hotels and takeaway liquor outlets. For example, in a longitudinal study of violence across greater Melbourne, hotel numbers were linked to rates of assault in inner-city areas, while takeaway outlets were associated with assaults in the suburbs.45 Further work based on a similar design found significant relationships between packaged liquor and domestic violence over time,47 while a study based on hospital data linked hotel licences to assault-related admissions and packaged outlets to alcohol-use disorders.48 Thus, data from Victoria demonstrate significant relationships between alcohol outlet distributions and harm rates at the local level. This suggests that research into how these outlets are distributed across socioeconomic areas may provide some useful insights into the role of alcohol availability in perpetuating health inequalities.

An increasing number of researchers have examined these questions outside Australia. The first study in the field was undertaken by Gorman and Speer using data from a single city in New Jersey.49 They found that alcohol outlets were generally concentrated in more disadvantaged neighbourhoods, although the area with the highest number of outlets had a low level of poverty. LaVeist and Wallace50 conducted a similar analysis using data from Baltimore focusing only on takeaway liquor stores. As with the New Jersey study, they found substantially higher densities of liquor stores in poor, black neighbourhoods. Two more recent studies examined these relationships across the US,51,52 confirming that in urban areas poor communities face much higher exposure to liquor stores than advantaged communities. Romley et al.52 focused on urban areas of the US, finding that liquor stores in particular clustered in poor, black neighbourhoods, and that bars were also more concentrated in low income areas. This was supported by more recent work by Berke et al.51 who used data from regional and rural areas as well as cities across the US. Their results suggested that the relationship between poverty and alcohol availability is strongest in urban areas and much less notable in suburban or regional parts of the US. Outside the US, two national studies in New Zealand identified similar patterns, with bars, clubs and takeaway liquor outlets all substantially more likely to be located in more deprived neighbourhoods,53–54 while a study in Glasgow produced less clear results.55 It is worth noting that these studies use a range of methodologies and measures. For example, Hay et al.53 measured the mean distance to the nearest outlet, while Romley et al.52 used density per capita and per 100 roadway miles. Few of these previous studies have incorporated socio-demographic controls, although Romley et al.52 did examine whether similar effects existed in predominantly white and predominantly non-white areas. None of these studies attempted to control for measures of non-resident neighbourhood usage (e.g. tourism). Thus far there have been no studies of the distribution of alcohol outlets in Australia.

Method

This study examines the distribution of alcohol outlets in Victoria in 2006, the most recent Census year in Australia. Victoria is the second most populous state in Australia, with a population in 2006 of 5.13 million.

Geographic units

All data were aggregated to the smallest possible geographical unit, the Census Collection District (CD). In general, CDs contain around 500 residents, varying substantially in geographic size depending on population density.

Geographic units were classified as either metropolitan or regional based on the remoteness classification contained within the Australian Standard Geographical Classification System (ASGC)56 from the Australian Bureau of Statistics. Within the ASGC, collection districts are grouped into five categories: major cities, inner regional, outer regional, remote and very remote. For ease of analysis, the 22 ‘remote’ CDs, 688 ‘outer regional’ CDs and 2,135 ‘inner regional’ CDs were combined into a single ‘regional/remote’ group, with the remaining 6,453 CDs in the ‘major cities’ group.

Data

Unit level licensing data for all licences active at the end of June 2006 were provided by Responsible Alcohol Victoria, including full address details. Only four types of licences were examined: general licences (pubs, hotel, taverns), on-premise licences (cafes, restaurants, bars), club licences (sporting clubs, RSLs, etc) and packaged liquor licences (bottle shops, licensed supermarkets). This excluded a range of less relevant licence types, including wholesalers, producers and limited licences. Individual licences were geocoded by a commercial organisation, Mapdata Sciences Australia. Overall, 99.5% of licences were geocoded to at least the street level, comparable to the results of previous studies.42,53,57 Outlet densities were calculated for Census Collection Districts across each of the four licence categories in two ways: per capita and per square kilometre. The use of both methods allows some examination of whether any associations found between outlet density and socioeconomic disadvantage were due to the way in which outlet density was calculated.

Four main demographic variables were examined in relation to alcohol availability. Firstly, the ABS produced Socioeconomic Index for Areas (SEIFA) index of relative socioeconomic disadvantage was used as a composite measure of local area disadvantage.58 This index was used rather than individual markers of disadvantage (e.g. unemployment, income, education levels, etc.) due to the high correlations between these single measures. Collection Districts were assigned to deciles of socioeconomic disadvantage based on cut-offs produced by the ABS using the total Australian population. SEIFA disadvantage indices have been produced at each of the last four Censuses and, despite different underlying methodologies, comparisons of deciles across time is appropriate.58

In addition to socioeconomic disadvantage, median age, proportion of residents who were male and the proportion of people counted within the CD on Census night who were visitors from elsewhere were included in more detailed analyses to assess whether alcohol availability was related to the age and sex structure of the population or to tourism in the area. These data were sourced from the Census community profiles59 and via a specific data request to the Australian Bureau of Statistics where necessary.

Analysis

Following Hay et al.,53 all analyses were conducted separately for urban and regional areas. For the initial analyses, Census Collection Districts were grouped into ten groups based on the decile of socioeconomic disadvantage. Thus the metropolitan areas in Victoria were collapsed into 10 areas based on disadvantage, from decile 1 (most disadvantaged) to decile 10 (least disadvantaged). Populations, licence numbers and other descriptive statistics were aggregated for these 10 groups of CDs to provide overall average outlet densities for people living in each decile. Simple linear regression was used on these 10 aggregate units (with outlet density as the dependent variable and socioeconomic disadvantage as the independent variable) to determine whether socioeconomic gradients were statistically significant.

To delve more deeply into the cross-sectional associations between socio-demographics and alcohol availability in 2006, regression models were developed using the 9,095 CDs as the study units, with licence counts as the dependent variables and the population size, area, SEIFA decile, median age, proportion of male residents and proportion of visitors as independent variables. Due to the nature of the outcome variables (counts of outlets), negative binomial regression models were utilised. Poisson models (which are also used for count data) were not utilised as the data was not distributed appropriately.

Results

Descriptive results

Socio-demographic descriptive data for the SEIFA decile aggregations are presented in Table 1. Unsurprisingly, CD-level unemployment rate declines consistently from the most disadvantaged group of CDs (decile 1) to the least disadvantaged CDs (decile 10). Similarly, median income varies in the expected direction, with higher income in less disadvantaged CDs. Notably, in metropolitan areas, population density was substantially higher in disadvantaged CDs compared with those in the less disadvantaged deciles, while in regional and remote areas population densities were highest at both ends of the socioeconomic spectrum.

Table 1 –.  Descriptive statistics for SEIFA deciles, Victoria, 2006, by remoteness.
SEIFA decileNMedian ageUnemployment rateMedian incomePopulationArea (sq km)Population per sq. km
Metropolitan areas
1 (most disadvantaged)62236.812.75293.2352,5682021,749
249837.58.44363.2284,7801721,655
349636.26.92408.9286,8052081,377
455635.96.14439.4340,0562481,369
557636.45.33472.6332,383359926
657736.74.85491.1338,524540627
764736.44.42524.3379,403442859
871236.53.95564.2411,968508811
977636.93.66601.5454,519517880
10 (least disadvantaged)84238.13.01689.0483,803661732
Regional and remote areas
1 (most disadvantaged)12939.511.55321.4136,2045,47224.9
214742.07.55356.8187,80611,22116.7
312240.86.12383.2187,29916,57811.3
428040.65.11411.3160,65529,2335.5
518040.14.57427.8146,51731,8354.6
614839.13.80457.9135,18244,0923.1
736439.73.43469.0109,68939,1582.8
834338.63.17495.580,76021,1283.8
937237.32.92544.464,25710,3766.2
10 (least disadvantaged)70838.02.47590.332,8781,98716.5

Bivariate results

Data on alcohol outlet density by SEIFA quintile are presented in Table 2. Outlet density is measured in two ways: per 1,000 residents and per 10 km2. The relationship between socioeconomic disadvantage and alcohol availability was tested using simple linear regression with decile as the independent variable and the measures of outlet density as dependent variables.

Table 2 –.  Alcohol outlet density by SEIFA decile of socioeconomic disadvantage, Victoria, 2006, by remoteness.
DecileLicences per 1,000 residentsLicences per 10 square kilometres
 GeneralPackagedOn-premiseClubGeneralPackagedOn-premiseClub
Metropolitan areas
10.140.400.600.122.437.0410.512.03
20.180.450.890.142.967.5014.712.27
30.280.361.060.103.794.9514.551.44
40.300.341.170.124.154.6716.031.61
50.290.341.130.122.703.1710.471.11
60.360.331.450.192.282.079.071.18
70.270.281.030.102.312.458.810.86
80.280.271.010.092.282.228.230.77
90.310.241.170.092.732.1310.260.83
100.270.210.970.092.011.577.130.68
Linear regression coefficient0.012–0.0230.030–0.004–0.114–0.656–0.704–0.168
p-value0.07<0.010.230.260.14<0.010.03<0.01
Regional and remote areas
10.730.570.810.290.180.140.200.07
21.250.771.260.400.210.130.210.07
31.070.641.580.300.120.070.180.03
40.760.370.970.340.040.020.050.02
50.730.410.980.220.030.020.040.01
60.440.280.960.240.010.010.030.01
70.450.320.880.280.010.010.020.01
80.280.220.610.190.010.010.020.01
90.330.220.790.250.020.010.050.02
100.180.090.670.180.030.020.110.03
Linear regression coefficient–0.100–0.064–0.059–0.064–0.020–0.014–.017–0.006
p-value<0.01<0.010.06<0.010.04<0.010.030.03

In metropolitan areas, only packaged liquor availability had a significant socioeconomic gradient across both measures of alcohol outlet density, with substantially higher densities of packaged outlets in more disadvantaged areas regardless of the measure used. Based on per-capita outlet density, people living in the most disadvantaged areas were exposed to almost twice as many packaged outlets as those in the least disadvantaged. Using an area-based measure, this gradient was even steeper, with 4.5 times as many outlets per square kilometre in the poorest areas as in the richest areas. Using an area-based measure, clubs and on-premise outlets were also significantly more common in more disadvantaged areas, although the magnitude of these differences were smaller than those seen for packaged liquor outlets. In regional and remote areas outlets of all types were significantly more prevalent in less advantaged areas.

Multivariate regression models

To overcome the problem of using two different denominators in the rates presented above, negative binomial models were developed using the absolute number of alcohol outlets as the outcome variable and controlling for both population and CD area. These analyses used the disaggregated data from individual CDs (n=6,302 in metropolitan Victoria and n=2,793 in regional and remote Victoria). In addition, three further control variables (median age, a measure of tourism and the proportion of the population that were male) were incorporated to ensure that the relationship between socioeconomic disadvantage and alcohol availability was not being confounded by other factors. Thus, Table 3 provides more robust estimates of the relationship between socioeconomic disadvantage and alcohol outlet density.

Table 3 –.  Negative binomial models of the association between alcohol outlet density and socioeconomic disadvantage, Victoria, 2006, by remoteness.
 GeneralPackagedOn-premiseClub
 CoefP-valueCoefP-valueCoefP-valueCoefP-value
Metropolitan areas
Decile0.0340.045–0.070<0.0010.0440.001–0.0370.039
Population (1,000s)–1.891<0.001–0.1930.24–1.412<0.0010.2440.329
Area (km2)0.0320.0420.0030.7790.0270.030.053<0.001
% visitors0.093<0.0010.034<0.0010.108<0.0010.044<0.001
% male–0.0020.89–0.0170.137–0.0270.0170.0050.760
Median age–0.458<0.0010.0130.025–0.010.0760.0280.001
Regional and remote areas
Decile–0.173<0.001–0.158<0.001–0.0150.489–0.0460.063
Population (1,000s)0.1800.3960.7220.0060.5180.0441.038<0.001
Area (km2)<0.0010.994<0.0010.219–0.001<0.001<0.0010.738
% visitors0.050<0.0010.050<0.0010.133<0.0010.027<0.001
% male0.0110.386–0.0020.900–0.0080.626–0.0120.525
Median age0.030<0.0010.042<0.0010.044<0.0010.051<0.001

In metropolitan areas, in controlling for other factors (including both the size and the population of the Collection Districts), the results of the models show varying relationships between socioeconomic status and alcohol availability, with general and on-premise licences more common in advantaged areas, while club and packaged licences were more common in disadvantaged areas. The model coefficients represent the percentage increase in the number of outlets likely in an area for a unit increase in the relevant independent variable. For example, an increase in the decile of disadvantage of 1 unit (i.e. to a more advantaged decile) is associated with an increase of 3.4% in general outlets and 4.4% in on-premise outlets and with a decline of 7.0% in packaged outlets and 3.7% in club outlets.

The proportion of the counted population that were visitors from elsewhere (i.e. our measure of tourism) was positively associated with all four types of alcohol outlet, as was the geographical size of the CD. Other measures produced mixed results, with on-premise and general licences more common in CDs with smaller populations, while packaged licences and clubs were more common in high population CDs. Gender had little relationship on alcohol availability, while CDs with younger populations had more general licences and fewer packaged or club licences.

In regional and remote areas of Victoria, both general outlets and packaged outlets were substantially more likely to be located in disadvantaged CDs. There were non-significant social gradients for on-premise and club licences. Effect sizes were substantially larger than in metropolitan areas with 17.3% and 15.8% decreases in general and packaged outlets with each decile of disadvantage.

As with the metropolitan models, tourism was significantly associated with alcohol availability. Population was generally positively associated with outlet numbers (except for general outlets), while CDs with higher median ages had higher numbers of all outlet types.

Discussion

The distribution of alcohol outlets observed in the metropolitan 2006 data in some ways makes simple economic sense. Outlets where alcohol is typically more expensive (general and on-premise outlets) are more likely to be located in areas of socioeconomic advantage, while outlets where alcohol is sold most cheaply (packaged outlets and clubs) are more prevalent in disadvantaged areas. Thus, the results found here are likely to be, at least in part, the result of careful planning by the operators of outlets, targeting their premises to the most appropriate market. Therefore it is possible that the distribution of outlets found in Victoria is simply a reflection of the demand for alcohol at particular price points. However, it is also plausible that the outcomes found here relate to inequalities in community-level power. For example, disadvantaged communities may find it harder to influence planning and zoning decisions and thus be unable to limit the proliferation of outlets in their neighbourhoods. This is partly supported by the repeated defeat of objections to packaged liquor licences in disadvantaged communities.60,61

The higher rates of exposure to packaged alcohol outlets in both urban and rural Victoria is of particular concern with respect to health inequalities in the State. As was discussed earlier, rates of alcohol related harm in Victoria are higher among people who are more socioeconomically disadvantaged.16,62 This is particularly the case for rates of chronic harm, such as liver cirrhosis.16 There is growing evidence that the density of packaged liquor outlets in a community is related to the experience of alcohol-related harm in that community.63–66 Recent studies using data from metropolitan Melbourne provide further evidence that the number of packaged outlets in a neighbourhood is associated with violence, domestic violence and chronic disease.45,65 Thus, the findings of this study provide an indication that some of the socioeconomic disparities in alcohol-related harm found in Victoria may be related to an inequitable distribution of packaged alcohol outlets by level of disadvantage. Contrastingly, the positive association between economic advantage and the density of general outlets (hotels, etc) may act to reduce socioeconomic disparities in rates of alcohol-related harm, as densities of these outlets have also been linked to alcohol-related harm (particularly violence).45 However, general outlets are more likely to serve patrons from outside their direct neighbourhood, meaning that these effects may be less likely to be experienced only by people living in the more advantaged neighbourhoods with high densities of general outlets.

The findings of this study should be considered in light of a number of significant limitations. The findings are based on aggregated data, which likely masks significant individual level variation. Similarly, the study doesn't account for the actual use of alcohol outlets. It is likely that outlets of all types are visited by large numbers of customers from outside the CD they are located in. Thus, a bottle shop in a poor area may serve many customers from nearby wealthier areas. This effect will be even more pronounced for general, club and on-premise outlets, which may serve as destinations in themselves, drawing custom from areas some distance away. Despite these issues, it is worth noting again that respondents living in areas with higher densities of outlets have been demonstrated to experience higher rates of harm, so the distribution of outlets remains important, even while their use may span a larger catchment than analysed here. Finally, a limited number of control measures were available for use in this study. The density of alcohol outlets is likely to be driven by a range of factors not considered here, including, for example, the distribution of other commercial outlets. Similarly, the measure used in this study as a proxy for tourism, captures only resident tourists at the Census time and doesn't handle the substantial variation likely to take place across the year (e.g. snow areas and beachside areas will have very different seasonal tourism patterns).

In spite of these limitations, the findings of this study are broadly consistent with a wider literature on health inequalities and alcohol availability,51 with Victorians living in more disadvantaged communities exposed to substantially higher rates of alcohol outlets (particularly packaged liquor). Thus, alcohol policies that are aimed at reducing health inequalities in Victoria should focus on reducing packaged liquor outlet numbers in disadvantaged neighbourhoods.

Acknowledgements

The geocoding of Victorian liquor licensing data used in this study was funded by a research grant from VicHealth. The author is funded by a scholarship from the Sidney Myer Fund and through research funds from the Alcohol Education and Rehabilitation Foundation.

Ancillary