Article Text

Who sells tobacco, who stops? A comparison across different tobacco retailing schemes
Free
1. Suzan Burton1,
2. Fiona Phillips2,
3. Christina Watts3,4,
4. Kelly Kennington2,
5. Michelle Scollo5,
6. Kylie Lindorff6,
7. Sam Egger7
1. 1 School of Business, Western Sydney University, Sydney, New South Wales, Australia
2. 2 Cancer Prevention and Research Division, Cancer Council Western Australia, Perth, Western Australia, Australia
3. 3 Cancer Prevention and Advocacy Division, Cancer Council New South Wales, Sydney, New South Wales, Australia
4. 4 Sydney School of Public Health, The University of Sydney, Sydney, New South Wales, Australia
5. 5 Cancer Council Victoria, Melbourne, Victoria, Australia
6. 6 Quit Victoria, Cancer Council Victoria, Melbourne, Victoria, Australia
7. 7 Cancer Research Division, Cancer Council New South Wales, Sydney, New South Wales, Australia
1. Correspondence to Professor Suzan Burton, Western Sydney University School of Business, Sydney, NSW 2747, Australia; S.Burton{at}westernsydney.edu.au

Abstract

Background Licensing of tobacco retailers has been proposed as a mechanism to encourage retailers to stop selling tobacco. However, previous studies of tobacco licensing and/or of retailers who have stopped selling have been restricted to one legislative environment. This study examines patterns of tobacco retailing across three legislative environments with three different licensing schemes (an annual fee-based licence, a zero-cost, one-off notification scheme and no notification/licensing scheme).

Method A telephone survey was conducted of 2928 potential tobacco retailers who could personally choose whether or not to sell tobacco (rather than the decision being made at a head office).

Results Unexpectedly, the annual licence fee to sell tobacco was not significantly associated with a lower rate of selling tobacco or a higher rate of stopping. After allowing for other factors, probability of selling, stopping selling and reported importance of tobacco sales varied across outlet types (p<0.001 for all three outcomes), and according to the remoteness of the retailer (p<0.001, p trend=0.041 and p=0.025 respectively).

Participants and methodology

Since alcohol-licensed premises were excluded in NSW, and previous research has shown those outlets are most likely to stop selling,12 alcohol-licensed premises in other states were excluded from the analysis. (Including alcohol-licensed premises in the models would make comparisons of any unadjusted statistics (eg, means and proportions) across NSW and the other states difficult.) However, in order to test whether the exclusion of alcohol-licensed premises selectively biased the results, additional analyses were run including alcohol-licensed outlets in WA and Victoria. As expected, the proportion of outlets that had stopped selling was then higher (because alcohol-licensed outlets were more likely to stop selling), but otherwise the results were not materially different. (Additional analysis included in online supplementary tables 1-3).

Statistical methods

Logistic regression was used to examine associations between the dichotomous outcomes: (1) selling tobacco (vs not currently selling) and (2) stopping selling tobacco (vs not stopping) and the following outlet characteristics: retailer type (convenience store/general store, grocery stores, fruit and vegetable stores, newsagent/post office), accommodation), socioeconomic status (SES) of location (quintiles of socioeconomic advantage/disadvantage score),17 remoteness of location18 (five categories) and state (NSW, Victoria and WA), with OR as effect measure. Linear regression was used to examine associations between reported importance of selling tobacco scores (the seller’s response on a scale from 1 to 7 of increasing importance) and the same outlet characteristics listed above, with differences in mean score as effect measure. Two additional variables asked only of current retailers—distance to the nearest alternative tobacco seller and number of staff employed at the outlet—were also included in the importance score analysis. Ordinal versions for those variables (where respondents had chosen an answer from a range) were coded with the left-hand cut-points (the minimum value from the selected range of values) and with ‘don’t know/can’t say’ responses coded as missing values. For each categorical independent variable (ie, the outlet characteristics), global tests of the null hypothesis of equal effects across categories were performed (with p values represented by the term ‘p value’ in tables) and we report ORs and 95% CIs for individual categories using dummy coding. In addition, tests for linear trends were performed by inclusion of continuous/ordinal versions of independent variables where appropriate (with p values represented by the term ‘p trend’ in tables). ‘Don’t know’/’can’t say’ responses were excluded as missing values in regression models but were included in the overall numbers and calculations of proportions. When testing for linear trends, interval scaled variables (number of staff, distance to nearest retailer) were coded ordinally and for variables that were not interval scaled (SES of location, remoteness of location), consecutive integers were used for coding.17 In post hoc supplementary analyses suggested by a reviewer, we refitted the regression models used to produce the main results (tables 1–3) using clustered robust standard errors with the cluster variable being the postcode of outlet location.18

Table 1

Predictors of selling tobacco

Table 2

Predictors of stopping selling tobacco

Table 3

Mean importance and predictors of importance (on a scale from 1 (‘not at all important’) to 7 (‘very important’))

Results

Response rate

After exclusion of the businesses described above that are not typical tobacco retailers, 7271 potential tobacco retailers remained on the calling list, of whom 2744 (37.8%) were not contactable. The remaining 4527 potential tobacco retailers were contacted and 3279 (72.4%) agreed to participate in the survey. Tobacco sell status could not be determined for 24 retailers, which were therefore excluded. A further 327 were found to be a business type that is not a typical tobacco retailer (eg, wholesalers) and were also excluded. After also excluding alcohol-licensed outlets, the sample included 2140 retailers, including 646 current tobacco sellers (30.2% of outlets), 249 former-sellers (11.6%) and 1245 never-sellers (58.2%). The sampling process is summarised in figure 1.

Figure 1

Flow diagram showing outlet sampling process. NSW, New South Wale; WA, Western Australia.

Analysis

Results for the logistic regression examining associations between outlet characteristics and (1) selling tobacco (vs not currently selling) are shown in table 1, and for (2) stopping selling tobacco (vs not stopping) in table 3, both with OR as effect measure. Results for the linear regression examining associations between reported importance of selling tobacco scores and the same outlet characteristics, with the addition of the two additional variables asked only of current retailers—distance to the nearest alternative tobacco seller and number of staff employed at the outlet—are shown in table 3 (with differences in mean score as effect measure). Results using clustered robust standard errors were practically the same as the original results obtained using ordinary standard errors (online supplementary tables 4-6).

What characterises outlets that sell tobacco, compared to those that do not?

Table 1 shows the results of a logistic regression model examining differences in currently selling tobacco by outlets in the three states depending on outlet type, the SES of its postcode and its remoteness. There were significant differences in the likelihood of current selling depending on outlet type (p<0.001), with newsagents/post offices and accommodation outlets significantly less likely to sell than convenience stores (the reference category). The likelihood of selling tended to be higher in more remote geographical areas (both p value and p trend <0.001). There was no difference in the probability of selling tobacco depending on the SES of the outlet’s postcode (p=0.402). Outlets in Victoria were significantly less likely to sell than in WA (OR=0.63 95% CI (0.42 to 0.94)). On raw percentages, outlets in NSW were more likely to sell than those in WA (39.1% in NSW vs 27.0% in WA). However, after allowing for other factors in the model, the probability of selling by NSW retailers was not significantly different from those in WA (adjusted OR=0.86 95% CI (0.59 to 1.25)).

What characterises outlets that stop selling tobacco?

Across the three states, 27.8% of outlets had stopped selling (see table 2), some within the previous year, and others more than 10 years previously. Convenience and grocery-type stores were less likely to stop selling than other outlets (p<0.001), and the likelihood of outlets stopping selling tended to decrease with increasing remoteness (p trend=0.041). NSW outlets were significantly less likely to stop selling than Victorian outlets (p=0.003 for test of equality of NSW and Victoria OR estimates). NSW outlets were also less likely to stop than WA, but the difference was not significant (Adjusted OR=0.74 95% CI (0.44 to 1.27)).

Which outlet types rate tobacco sales as most important?

The perceived importance of tobacco sales was not able to be factored into the models predicting which outlets sell and which have stopped selling, because outlets not currently selling were not asked about the importance of tobacco sales. A separate model was therefore run to examine potential associations between self-reported importance of tobacco sales scores (ranging from 1 to 7, with higher scores indicating higher importance) for current tobacco retailers (see table 3). Consistent with the results for the outlets most likely to sell, and least likely to stop selling, convenience and grocery type stores rated tobacco sales as more important than newsagents/post offices (adjusted mean difference=−1.75 95% CI(−2.14 to to 1.36)) and accommodation outlets (adjusted mean difference=−2.25 95% CI(−3.16 to to 1.35)). Importance scores tended to be lower with increasing remoteness of the retailer (p trend=0.028). No other retailer characteristics were associated with importance scores. Notable, however, is that while the mean importance score for all outlets was 4.8 (95%CI (4.6 to 4.9), some outlet types (Newsagents/post offices and accommodation outlets) rated tobacco sales below the midpoint of 4 on the 7-point importance scale.

Discussion

The results show clear differences between outlet types in the likelihood of selling, stopping selling and the reported importance of tobacco sales for the outlets. Compared with other outlet types, convenience and grocery type stores were more likely to sell tobacco, less likely to stop selling and rated tobacco sales as more important. Precise comparisons are difficult because of the different descriptions used for different types of retailers, but those results are consistent with the higher market share of convenience stores, which captured 11.7% of Australian tobacco sales in 2017, and small grocers, with a market share of 5.9%, compared with newsagents/tobacconist kiosks (with 0.9%), and ‘others’, with 1.2% of sales.19 Those numbers need to be interpreted in terms of the numbers of each types of outlets, with one NSW study finding that convenience/grocery stores comprised 22.6% of tobacco retailers in the state, and newsagents only 9.8%.12 So the results show a logical pattern of tobacco retailing—outlets from a category of retailer with lower market share are likely to rate tobacco sales as less important, be less likely to sell, and more likely to stop selling if they sell. However, the probability of selling, of stopping selling and reported importance of tobacco sales were also associated with the remoteness of the outlet. That may be explained by fewer supermarkets and tobacconists—which dominate tobacco market share in Australia, probably because they tend to have lower prices20—being located in remote areas. That absence of key competitors would then mean that other tobacco retailers in those areas have less competition, make more sales, and so even though they rate tobacco sales as less important on average, they are less likely to stop selling.

In WA, where tobacco retailers need to pay an annual licence fee to sell tobacco, a lower proportion of retailers sold tobacco compared with NSW (with no licence fee), and WA retailers were more likely to stop selling than those in NSW (though neither difference was significant). However, analysis of the adjusted odds of selling suggests that the WA licence fee does not explain the difference: Victoria, with no licensing system, had the lowest proportion of tobacco retailers across the three states, and the highest proportion of ex-sellers. In NSW, after allowing for other factors, retailers were (like Victoria) actually less likely to sell than in WA (adjusted OR=0.86). Among those who had ever sold, NSW retailers were less likely to stop selling tobacco than retailers in WA (consistent with some effect due to a licence fee), but the difference was not significant. There is certainly other evidence that an annual licence fee encourages some retailers to stop selling,8 9 so these unexpected results may be due to different levels of population density across the states: at 2 529 875 km2, WA is more than 11 times the land area size of Victoria (at 227 416 km2),21 despite having only 39.8% of its population (2.62 million in WA vs 6.57 in Victoria).22 Consistent with that lower average population density, Victoria has no retailers in areas classified as very remote, and only seven retailers in areas classified as remote. NSW falls between the other two states in size (at 3.52 times Victoria’s land area size,21 and the largest population, with 1.2 times Victoria’s population size).22 Although the analysis allowed for remoteness of the location of the retailer, it is likely that the remoteness measure did not adequately capture those geographical differences across the states. For example, with Victoria having only seven retailers in very remote areas and none in remote areas, and the second highest population spread across the smallest area of the three states, the state is likely to have a much higher percentage of the population living closer to an urban centre. Urban centres will almost certainly contain a supermarket and/or tobacconist which, on average, both sell tobacco at lower prices that other outlets.20 Proximity to lower-priced competitors will result in other retailers making less money from tobacco sales, so a higher percentage of those other outlets are likely to stop selling. In contrast in NSW, despite having a smaller land area than WA, like WA, there are remote and very remote regions where there are less likely to be supermarkets, and so in the absence of an annual licence fee, retailers are more likely to persist in selling tobacco even, in the face of low sales. So the results are consistent with a competitive disadvantage for small retailers in less remote areas where they are more likely to be exposed to lower-cost competitors, and thus more likely to stop selling. Conversely, in more remote areas, small retailers are less likely to face low-cost competition, and thus less likely to stop selling. That explanation is supported by market share figures which show an increase in the tobacco market share of supermarkets and tobacconists, and a decrease in all other store-based retailers.19 So when they can, Australian smokers appear to be increasingly buying from lower cost retailers, possibly driven by steep annual tax increases in the price of tobacco driving smokers to the (average) lower-priced outlets of supermarkets and tobacconists.23 That behaviour would then contribute to the higher exit from the market of small retailers in Victoria due to their greater exposure to those lower-cost retailers.

Data availability statement

Data are available on reasonable request. Deidentified data are available on reasonable request by contacting the first author.

Ethics statements

Ethics approval

Ethics approval was obtained from the Ethics Committee of Western Sydney University.

Footnotes

• Contributors All authors contributed to the idea and to interpretation of the results. SB and SE led the analysis and wrote the first draft. All authors reviewed and approved the final version.

• Funding This study was funded by Cancer Council Western Australia.

• Competing interests No, there are no competing interests.

• Provenance and peer review Not commissioned; externally peer reviewed.

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.