Estimating tobacco consumption in remote Aboriginal communities using retail sales data: some challenges and opportunities
Dr David MacLaren, School of Public Health, Tropical Medicine and Rehabilitation Sciences, James Cook University, PO Box 6811, Cairns, Queensland 4870. Fax: (07) 4042 1658;
Objective: To describe and discuss challenges and opportunities encountered when estimating tobacco consumption in six remote Aboriginal communities using tobacco sales data from retail outlets.
Approach: We consider tobacco sales data collected from retail outlets selling tobacco to six Aboriginal communities in two similar but separate studies. Despite challenges – including: not all outlets provided data; data not uniform across outlets (sales and invoice data); change in format of data; personnel change or management restructures; and anomalies in data and changes in community populations – tobacco consumption was estimated and returned through project newsletters and community feedback sessions. Amounts of tobacco sold were returned using graphs in newsletters and pictures of items common to the community in community feedback sessions.
Conclusions: Despite inherent limitations of estimating tobacco consumption using tobacco sales data, returning the amount of tobacco sold to communities provided an opportunity to discuss tobacco consumption and provide a focal point for individual and community action. Using this method, however, may require large and sustained changes be observed over time to evaluate whether initiatives to reduce tobacco consumption have been effective.
Implications: Estimating tobacco consumption in remote Aboriginal communities using tobacco sales data from retail outlets requires careful consideration of many logistical, social, cultural and geographic challenges.
Addressing high levels of tobacco use among Indigenous Australians is increasingly seen as an important factor to closing the life-expectancy gap between Indigenous and non-Indigenous Australians.1 The challenges are considerable given that smoking rates of 51%2 among Indigenous Australian adults appear to have changed little over the past 20 years, while rates in the general population have declined steadily to less than 20%.3 The challenges are even greater in some remote Aboriginal communities in the ‘Top End’ of the Northern Territory (NT), where smoking rates from 68%4 to 83%5 in men and from 65%4 to 73%6 in women have been documented; with little change since the mid-1980s.7 Smoking rates in these communities are among the highest reported in Indigenous populations in Australia, New Zealand or North America.2,8–11
To determine whether program initiatives are having the desired effect, reliable monitoring of tobacco consumption over time is crucial. At the national level, tobacco consumption is estimated using a combination of customs and excise data, industry sales figures and Australia-wide and regional surveys.12 A small number of local and regional studies in the NT's ‘Top End’ have attempted to estimate tobacco consumption in remote Aboriginal communities using community store sales and/or data from wholesalers describing purchases-into-store.6,13–16 An indicator of tobacco consumption progressively measured in these discrete, isolated localities potentially offers a reliable way to assess responses to tobacco program initiatives without the considerable expense of other methods of monitoring trends in tobacco use such as repeated population surveys. Challenges documented in published studies included the inability to quantify tobacco purchased outside community stores, highly mobile populations and disparities between self-reported consumption and consumption estimated from store sales. A recent study14 suggests that, in spite of such limitations, “large and sustained changes in tobacco consumption should still be apparent and worthy of further investigation”.
This paper describes and discusses approaches to using community store sales and ordering data to estimate tobacco consumption in six Aboriginal communities and one mining town in two separate, but related, studies. We describe the challenges experienced by a university research team in compiling, managing and interpreting sales data, and report on how sales data and estimated tobacco consumption have been returned to community members. We also discuss how returning sales data has provided a focal point for individual and collective action.
Collecting, managing and interpreting tobacco sales data
Tobacco sales data were collected in a one-year pilot study (2006–2007) and current five-year study (2007–2011) of tobacco use in remote Aboriginal communities in the ‘Top End’ of the Northern Territory. The 12-month pilot study was conducted in three Aboriginal communities (total population 1,500) located near a mining town (population 1,000 and predominantly non-Indigenous) to assess feasibility of collecting tobacco sales data from all tobacco outlets in these communities. The current five-year study, the ‘Top End Tobacco Project’ is being conducted in three Aboriginal communities (total population 3,100) located in isolated corners of Arnhem Land. Here we comment on tobacco sales data collected for the period July 2007 – March 2009.
In the two studies a total of 20 outlets were reported as selling tobacco – 11 outlets in the pilot study and nine in the current study (Table 1). Management for each retail outlet were requested to provide data describing the type and quantity of tobacco products sold each month, or, where point-of-sale figures were not available, purchased-into-store from wholesalers. Data were sent via e-mail or fax to the study team every three months. The Aboriginal Community Councils in each community were consulted and endorsed these studies prior to commencement. Data were electronically imported or manually entered into Microsoft Excel spreadsheets.
Table 1. Types of retail outlets and factors influencing the provision of tobacco sales information from 20 tobacco retailers in seven localities in two studies of tobacco use in Arnhem Land, Northern Territory.
|1||Community store – community controlled retail group||1||Yes||Sales||Yes||Electronic||Yesa||Yesa|
|4||Community store – recent transfer to large retail group||2||Yes||Sales||Yes||Electronic||Changes||No|
|8||Community store – recent transfer to community controlled retail group||3||Yes||Sales||Yes||Electronic||Multiple changes||No|
|9||Community store – community controlled organisation||3||Yes||Invoices||No||Manual||Yesb||Yes|
|10||Community store – community controlled retail group||4||Yes||Sales||Yes||Electronic||Yesa||Yesa|
|11||Community store – local council||5||Yes||Invoices||Yes||Manual||Multiple changes||No|
|19||Private Enterprise||6||Yes||Invoices||Yes||Electronic||Multiple changes||No|
|20||Community store – local council||7||Yes||Invoices||No||Manual||Multiple changes||No|
Table one describes characteristics and data provided by each outlet in terms of type, regularity, format and stability of management structures through which tobacco sales data was sent. Three of the 20 retail outlets did not provide data. Two retailers provided repeated assurances that data would soon be forthcoming, however provided no data. The third, an individual reportedly selling tobacco from a residence in a pilot study community, could not be contacted for inclusion in the study.
Eight of the 17 retailers provided transcriptions (three handwritten or typed) or photocopies of invoices, four provided copies of invoice information generated by an accounting software package and five provided sales data recorded by barcode readers at retail outlet checkouts or, in one case, a vending machine.
Six of the 17 retailers altered the format in which data were provided at least once during the observation period: all six of these retail outlets underwent changes to personnel or management responsible for providing data. Two retailers, both of whom provided data in an efficient and timely manner, had a change in personnel managing the extraction of data from computer records. The format in which it was provided changed slightly in both cases. A further retail outlet decided to participate in the study after a staff member who had provided data to the study when employed by another local retailer joined its organisation – photocopies of invoices were provided retrospectively.
Changes to the format of data and personnel providing data required vigilance to ensure the continuity and quality of data provided for the studies. Each change required renegotiation with new personnel or corporate structures to provide ongoing data for the study. Delays of up to three months occurred because of such changes. Some retailers provided data without prompting each quarter, while others required regular reminders to send data. Limitations of human and technical resources and/or time were reported by some smaller retailers as delaying provision of quarterly sales data.
Retailers involved with the pilot study were offered no incentives or compensation to provide data to that study. However, nine retailers involved in the current five-year study were offered $500 per annum to cover costs for providing data. One retailer (a private enterprise) refused to accept the $500 per annum offer, stating it was a service to the community to provide data to the study. This retailer provided prompt quarterly sales data throughout the observation period. The retailers that accepted the $500 per annum provided data through the observation period with varying levels of promptness. The one retailer in the current study that did not provide sales data during the observation period (also a private enterprise) was aware of the $500 per annum offered by the study. Although the utility of the $500 per annum payment to facilitate the initial and ongoing provision of sales data was not systematically assessed, its effect as an incentive to participate and promptly provide data to the study was not uniform.
The research officers managing the sales data required vigilance in checking inconsistencies or anomalies across the wide variety of formats in which the data were supplied and entered into Microsoft Excel. Unexpected variations in tobacco sales, orders or type of tobacco products required further investigation. For example, one outlet showed a systematic peak in monthly sales every three months. On contacting the retailer it was reported that computer-generated ‘monthly’ sales routinely reported four-week ‘months’ with one five-week ‘month’ each quarter depending on the last business day of the calendar month. Although anomalies apparent to the study team were followed up with retailers, there was no other independent auditing or validation of data sent by retailers.
The research team faced numerous challenges when using tobacco sales data to estimate tobacco consumption in study communities. These included:
Retailers not exclusively servicing study communities: Five retailers in the pilot study and one retailer in the current study did not exclusively service study communities. A retailer in the current study is near a highway and also serves tourist traffic when the road is open in the dry season. The manager estimates 80–90% of tobacco is sold to residents from the nearby Aboriginal community in the ‘dry’ season, the remainder sold to passing travellers. In the ‘wet’ season, when there is no road traffic, 100% of tobacco is sold to community residents.
Tobacco sourced from locations other than community outlets: Two communities in the current project have road access to mining or service towns in the dry season and all communities in the pilot had access to outlets in the nearby mining town in both the wet and dry seasons.
Variability of retail outlet tobacco data: Invoice data from small outlets tended to vary month by month, while data from the point-of-sale tended to vary less. Smaller retailers (that provided wholesale invoices rather than sales data) reported taking advantage of special prices from wholesalers and increasing stocks. Subsequent months’ orders were negligible, reportedly because of excess tobacco in stock. One outlet also reported $30,000 of tobacco stolen during a burglary; subsequent invoices reflected a spike in orders when the tobacco was re-ordered.
Variability in community populations: Community populations varied between the ‘wet’ and ‘dry’ seasons, often in response to social or cultural events, such as ceremonies and funerals. Populations were generally more mobile in the dry season. The study did not quantify population variations but, consistent with similar studies in these communities, used Australian Bureau of Statistics (ABS) census data.
Changing policy environment: Major social and financial change occurred including the Federal Government's 2007 Northern Territory Emergency Response (NTER). Under the NTER, a proportion of government benefits are issued onto a ‘basics card’ to be spent on government-approved goods. The remaining is available for discretionary purchases, including tobacco. How these changes have impacted tobacco consumption in the study communities has yet to be assessed.
The ease of data collection and interpretation of the data varied across the outlets. There were no uniform features across stores that led to better or easier data collection, management or interpretation. It was our experience that information from outlets exclusively servicing local Aboriginal communities and stores providing point-of-sales data proved the most convenient to collect and manage. Data from communities with minimal access to tobacco from other locations proved most straightforward to interpret. Maintaining open and regular communication with retail outlet managers proved important to enable the reliable collection, management and interpretation of tobacco sales data.
Returning tobacco sales data to communities
Despite the challenges experienced in collecting, managing and analysing tobacco sales data, these data have been returned to each community involved in the current ‘Top End Tobacco Project’. This has been through quarterly newsletters to each community which includes an updated bar graph of tobacco sales each three months (from July 2007). Data have also been returned through a series of individual and public meetings, workshops and presentations at local councils, health centres, workplaces, sporting venues and with family groups during quarterly community visits between August 2008 – August 2009.7 Electronic Microsoft PowerPoint and hard-copy flip-chart presentations include the total number of cigarettes (or cigarette equivalents for loose tobacco) sold per day, week, month and year in that community. Presentations also included the financial value of tobacco for a day, week, month and year in that community.
An initial prototype presentation, trialled in community one in August 2008, used bar graphs to represent tobacco sold in each community. Community leaders suggested that sales data together with cumulative financial value may be of more use to community members if compared with items significant to the community. Subsequent electronic and flip-chart presentations used pictures of items such as domestic power meter tokens (‘power cards’), refrigerators and freezers, outboard motors, four-wheel-drive vehicles and houses to represent the value of the tobacco sold in that community. For example, the cumulative amount of money spent on tobacco in one study community in a day was equivalent to the value of a 40-horsepower outboard motor; in a week equivalent to a four-wheel-drive vehicle and in a year equivalent to three new houses. This information has been translated into local languages and is now being used by health staff, school staff and community-based organisations to raise awareness about the amount of tobacco consumed in the community and its financial impact in each community.
It has been our experience, similar to other studies in the region, that when research findings are provided to local community-level decision makers in an appropriate format, it can have a locally empowering effect to inform community-based action.17 Our experience was that the most animated responses by community-based audiences to the return of results from the ‘Top End Tobacco Project’ did not come when returning the smoking prevalence results from the community-based survey (community one 70%; community two 76%; community three 82%), rather when returning information about the amount of tobacco sold. This was of particular significance when the amount of money spent on tobacco in each community was compared to items of value to community members. This has stimulated both individual and collective action to address tobacco consumption in these communities. A 2010 follow-up survey will assess the impact and outcomes of these and other community-based initiatives to inform community members about tobacco use and its role to assist communities to reduce tobacco smoking.
Further research potential
In yet-to-be-published data, collected during the 2008 ‘Top End Tobacco Project’ tobacco use survey, the average self-reported number of cigarettes consumed by survey participants appears to closely reflect the average number of cigarettes sold in each community. Further investigation is required to assess whether such preliminary associations observed at a single point in time, will remain through time and/or occur in similar communities. Further investigation is also required into how returning tobacco sales data to community members may act as an impetus for individual or collective action to address tobacco in the community.
Although monitoring is essential, and measures for effective tobacco control policy and program initiatives are required in remote communities, estimating tobacco consumption using tobacco sales data from remote community retail outlets requires careful consideration of inherent challenges. As with national data used to estimate national tobacco consumption, it is our experience that such data from remote Indigenous communities in the NT's ‘Top End’ have also proved to be inconsistent in format and reporting periods, not uniformly subject to external audit and often difficult to manage and interpret.12 This has been compounded by many other geographic and social variables. A number of studies in NT Indigenous communities have used monthly sales and ordering data to monitor tobacco consumption and calculate changes over time.13,15,16 Although such estimates may suggest a high degree of accuracy and precision, and allow an apparent mechanism for the efficient monitoring of tobacco consumption in remote communities, some authors have emphasised that the limitations of using tobacco sales data to estimate tobacco consumption may mean that only quantum changes in consumption over time may become apparent using this method.14 Given the considerable challenges experienced in this and other studies, including the variability in estimated tobacco consumption that can occur based on store turnover,16 we agree that quantum changes in tobacco sales would need to be observed over an extended period to evaluate whether initiatives to reduce tobacco use have been effective.
Numerous challenges have been encountered when collecting, managing and interpreting tobacco sales data from retail outlets in these two separate but similar studies in remote Aboriginal communities. Despite these challenges, sales data have been used in a practical way at the community level; both presenting the data formally in newsletters and directly to community decision makers and community members in a less-formal visual format. Further research is required to assess the utility of tobacco sales data from remote Aboriginal communities as both a tool to monitor tobacco consumption and an impetus to stimulate individual and collective action to reduce tobacco use in these communities.
Ethical approval for the study was provided by the Human Research Ethics Committee of the NT Department of Health and Families and the Menzies School of Health Research.
The research was supported by the Commonwealth Department of Health and Ageing and the National Health and Medical Research Council (NHMRC) project grant 436012. The authors thank the co-investigators on NHMRC project grant 436012: Professors Rob Sanson-Fisher and Robert Gibberd (University of Newcastle), Professor Sandra Eades (Baker Institute), Associate Professor Rowena Ivers (University of New South Wales) and Associate Professor Kate Conigrave (University of Sydney). We would also like to thank community members who are participating in the study, community-based research assistants, retail outlet management and staff and former project officer Anil Raichur.