Article Text

Download PDFPDF

Impact of a point-of-sale tobacco display ban on smokers’ spontaneous purchases: comparisons from postpurchase interviews before and after the ban in Western Australia
  1. Owen B J Carter1,2,
  2. Tina Phan1,2,
  3. Brennen W Mills1,2
  1. 1Centre for Behavioural Research in Cancer Control, Curtin University, Shenton Park, Western Australia
  2. 2Now at the Office of the Pro-Vice-Chancellor (Health Advancement), Edith Cowan University, Perth, Western Australia
  1. Correspondence to Dr Owen Carter, Office of the Pro-Vice-Chancellor (Health Advancement), Edith Cowan University, Rm 21.544, Building 21. 270 Joondalup Drive, Edith Cowan University, Joondalup 6027, Western Australia; o.carter{at}


Objective To assess the impact of the Western Australian tobacco point-of-sale display ban on spontaneous purchase behaviours.

Methods Daily adult smokers (n=402) observed purchasing cigarettes were recruited via exit interviews either 2 months before or after the implementation of the display ban. Smokers were asked if they had intended to purchase cigarettes before entering the store to assess spontaneous purchase behaviours. Whether smokers had noticed the displays before their purchase and the extent to which this influenced their purchase decision was also assessed via non-prompting questions.

Results When compared with before the ban, fewer smokers after the ban noticed the displays (27.1% vs 1.1%, p<0.001), fewer reported making spontaneous purchases (28.2% vs 19.8%, p<0.05) and fewer claimed the displays influenced their purchase decisions (free recall 5.0% vs 1.1%, p<0.05; cued recall 22.1% vs 3.8%, p<0.001). Before the ban, spontaneous purchasers were more likely than planned purchasers to suggest the displays influenced their purchase decisions (free recall 9.7% vs 3.2%, p<0.05; cued recall 40.0% vs 17.9%, p<0.01). After the ban, spontaneous purchasers nominating the influence of displays fell substantially (free recall 9.7% vs 5.6%, p=NS; cued recall 40.0% vs 11.1%, p<0.01) as it did for planned purchasers (free recall 3.2% vs 0.0%, p<0.05; cued recall 17.9% vs 2.1%, p<0.01).

Conclusions We observed a 30% reduction in smokers making spontaneous tobacco purchases after implementation of the Western Australian tobacco display ban and between a fivefold and sixfold reduction in the proportion suggesting displays influenced their decision to purchase cigarettes. These data are consistent with previous research suggesting tobacco displays encourage spontaneous purchases and their removal corresponds to reductions in the same.

  • Advertising and Promotion
  • Packaging and Labelling
  • Tobacco industry

Statistics from

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.

Restrictions on tobacco advertising are becoming ever tighter around the world, leaving the tobacco industry with fewer and fewer options to promote its products. Facing widespread advertising bans in the mass media, tobacco advertising within stores remains aggressive in jurisdictions where it is still permitted.1 Even in jurisdictions where in-store tobacco advertising in not permitted, the tobacco industry considers tobacco point-of-sale (POS) displays as a key tool remaining in its promotional arsenal.2 Tobacco companies have been documented offering retailers free shelving and substantial incentives to give prominence to their brands in POS display areas.3 ,4 However, some jurisdictions have decreed that visible displays of tobacco products still constitute a form of advertising inconsistent with comprehensive tobacco advertising bans. As such, POS display bans have been implemented in Australia, Canada, Finland, Iceland, Ireland, New Zealand, Norway and Thailand and are currently being phased in throughout the UK.5 The tobacco industry has ferociously opposed tobacco display bans wherever they are introduced and to date has threatened and/or launched legal challenges to such legislation in Iceland, Saskatchewan, Norway, England, Scotland and the village of Haverstraw in the state of New York.6 Haverstraw was forced into rescinding its ban in the face of a financially crippling legal challenge brought about by an alliance of seven tobacco companies.7 However, on all other occasions, in other jurisdictions such legal challenges have been defeated or withdrawn by the tobacco industry where governments have demonstrated a preparedness to defend their newly introduced legislation in court.6

In support of the contention that tobacco displays are a form of advertising—and therefore incompatible with comprehensive tobacco advertising bans—there is evidence of displays impacting on both tobacco-related awareness and behaviours. For instance, implementation studies of display bans in Ireland, Canada and Australia demonstrate that the bans lead to substantial reductions in the proportion of smokers recalling any tobacco promotional activities.8 ,9 This is especially significant for youth uptake of smoking; a systematic review of 12 studies concludes there is a ‘moderately strong’ association between youth uptake and exposure to POS tobacco promotion due to influences on brand awareness and assumptions about social norms and availability.10 Consistent with this conclusion, a recent virtual reality experiment demonstrated that youths are significantly more likely to attempt a tobacco purchase in virtual stores where tobacco displays are present than when they are not.11

The promotional effect of tobacco displays on spontaneous purchases by adult smokers has been investigated in at least four previous studies. In 2006, Wakefield et al conducted a telephone survey of Australian adults—prior to the implementation of display bans—indicating that 25% of smokers reported spontaneous purchases of cigarettes ‘at least sometimes’ as a result of being exposed to tobacco displays.12 However, smokers were asked to recall all previous purchase experiences within the past 6 months and the limits of autobiographical memory are likely to have reduced accuracy and introduced biases.13 In 2008, Carter et al attempted to counter recall bias by using exit interviews with smokers immediately after they had been observed making tobacco purchases.14 Smokers were first asked whether they had just made a spontaneous tobacco purchase and then asked separately ‘Did the cigarette pack display encourage you in any way to purchase cigarettes in this instance?’ Spontaneous tobacco purchasers were nearly four times more likely than planned purchasers to suggest they had been influenced by tobacco displays (47% vs 12%, p<0.05).14 These two studies were criticised during a legal challenge by Japan Tobacco International to the Heath Act 2009 (including the Tobacco Advertising and Promotion (Display) (England) Regulations 2010. A tobacco industry consultant argued that the wording of the Wakefield et al and Carter et al survey questions was ‘leading’ and ‘suggestive’ to the point where smokers would have found it ‘very hard to say no’ (p. 77).15 (It should be noted that a large majority of smokers to both surveys actually did say ‘no’—75% and 78%, respectively12 ,14.) More recently, Clattenburg et al conducted postpurchase interviews in Vermont (USA) using indirect questions to assess what influenced smokers’ purchasing decisions. Spontaneous purchasers were asked ‘What, if anything in this store, prompted you to buy cigarettes?’ and were presented with five response options, including ‘lower price’, ‘recommendation from a friend’, ‘special promotion/coupon’, ‘tobacco advertising or display at the store’ and ‘other’.16 Of the 301 smokers interviewed, 8.6% reported making spontaneous purchases and suggested the tobacco display influenced their decision—a result consistent with the 10% reported by Carter et al.14 In addition, Clattenburg et al reported that younger smokers aged 18–24 years were twice as likely as older smokers to report spontaneous purchases—going someway to validate data that approached but did not achieved statistical significance reported by Wakefield et al (p=0.068) and Carter et al (p=0.072).12 ,14

Japan Tobacco International also criticised the Wakefield et al and Carter et al methodologies for not having a control group to demonstrate a difference in spontaneous purchases in the absence of tobacco displays (p. 78).15 This criticism would apply equally to the Clattenburg et al study. However, such data were provided by Li et al from the ITC four-country survey in 2010 that compared spontaneous purchases in countries where display bans did and did not exist at that time.8 They reported that in the US and UK, where no display bans existed, 6.13% of smokers recalled making spontaneous purchases influenced by tobacco displays compared with only 2.52% of smokers in Australia and Canada, where display bans existed.8 These data go some way to demonstrate the contribution of tobacco displays to spontaneous tobacco purchases but unfortunately share similar limitations to Wakefield et al as smokers were asked to respond retrospectively and to a leading question.

The Li et al study was also compromised by the fact that the method could not account separately for tobacco displays and the confounder of in-store advertising—still existing in the USA and partially in the UK at the time of the survey. Thus, their observed difference could potentially have been explained entirely by in-store tobacco advertising; a similar problem faced by the Clattenburg et al study that combined ‘tobacco advertising or display at the store’. As such, all four studies have a combination of limitations that give the tobacco industry continuing ability to suggest any evidence linking tobacco displays to spontaneous purchases remains ‘flawed and unreliable’.15

In 2010, a strict display ban was introduced in Western Australia that required all tobacco products be removed from view. Since 2006, state regulations had already banned tobacco discounting and in-store advertising of tobacco products. The ability of retailers to create ‘power walls’ of tobacco products3 had also been curtailed by restricting tobacco displays in shops to no more than one square metre, being allowed to display only one example of each brand and feature a maximum of 120 tobacco products.14 Four years later, the 2010 introduction of a full display ban required all tobacco products to be removed from line of sight. Typically, retailers complied by installing enclosed tobacco cabinets behind their counters. Figure 1 illustrates typical tobacco retail display arrangements before versus after the ban in Western Australia.

Figure 1

Examples of tobacco point-of-sale displays in Western Australia before (left) and after (right) implementation of the ban of September 2010.

The implementation of a strict display ban in a jurisdiction without POS tobacco advertising created a new opportunity to isolate the unique influence of tobacco displays. We used a quasi-experimental design to examine the policy's impact on smokers’ spontaneous purchases. Specifically, we compared postpurchase interviews of smokers before and after the ban using different types of questions to address tobacco industry criticisms about recall and response bias in previous research.


Ethics approval for the study was obtained from the Curtin University Human Research Ethics Committee (PH-58-2010).


Data were collected 2 months before and 2 months after the implementation of the Western Australian tobacco display ban on 23 September 2010. A 2-month time period was chosen in order to avoid confounding from early introductions of cabinet doors leading up to the ban. Supermarkets were chosen as the sites for recruitment as they account for more tobacco sales within Australia than all other tobacco retailers combined (51%).17 Two supermarket chains representing a majority share (71%) of the Australian grocery market were then selected.18 In the state's capital city of Perth (population 1.9 million), both supermarket chains were located in four suburbs: two suburbs in the highest quartile of socioeconomic status, one in the third quartile and one in the second quartile, as suggested by the Australian Bureau of Statistics.19


After being observed purchasing cigarettes from one of the eight supermarkets, customers were immediately approached upon exiting the store and asked to participate in a 5 min survey on ‘health issues’ by student volunteers. No incentives were offered for participation. Participants were screened to ensure they were aged 18 years or older, smoked at least one cigarette per day and had not participated in our survey previously.


Two types of questions (free and cued recall) were used to assess what influenced smokers’ purchasing decisions. In the first instance, to assess the proportion of smokers who had made spontaneous tobacco purchases, participants were asked ‘Did you intend to purchase cigarettes before entering this store today?’ Participants responding ‘no’ were considered as ‘spontaneous purchasers’ and those responding ‘yes’ as ‘intentional purchasers’. Next, to avoid ordering effects within the survey, free recall of displays was assessed first by asking participants ‘What did you see, if anything, in the store that prompted you to purchase cigarettes in this instance?’ Responses were recorded verbatim and later coded into a dichotomous variable to differentiate those who mentioned the display from those who did not. Next, cued recall of tobacco displays was assessed by asking participants ‘Please state which of the following influenced your decision to purchase cigarettes in this instance?’ Smokers then selected from a list of five responses: the cigarette cabinet, the shop counter, the cigarette price list, the Quit-line advertisement and ‘seeing someone else purchase cigarettes’. The starting order of the list was systematically rotated for successive respondents to minimise ordering effects. The term ‘cigarette cabinet’ was chosen in order to be equally relevant both before and after the implementation of the ban. Finally, to measure the extent to which the ban had successfully removed cigarettes from line of sight, smokers were asked ‘Did you notice cigarette packets on display before purchasing your cigarettes today?’


Fisher's exact tests were used to compare bivariate associations between free, cued and prompted recall of the displays before versus after the ban, by spontaneous versus intentional purchasers and by younger (18–24 years) versus older adults (25+ years). It was specifically hypothesised that the salience of tobacco displays would be lower after the ban, so one-tailed tests were deemed appropriate, consistent with the recommendations of Ferguson.20


In total, 529 people observed purchasing cigarettes were approached to participate in the survey, of whom 68% consented, resulting in 402 daily smokers being recruited for the study: 220 recruited 2 months prior to implementation of the display ban and 182 recruited 2 months afterwards. The sample characteristics of the two samples were similar by sex, age and tobacco consumption (see table 1).

Table 1

Demographics of predisplay versus post-display ban participants

As can be seen in figure 2, the proportion of smokers making spontaneous tobacco purchases was significantly greater before the ban than afterwards (28.2% vs 19.8%, p<0.05). Likewise, the proportion of smokers whose decision to purchase cigarettes was influenced by the displays was significantly greater before the ban than afterwards, for both free recall (5.0% vs 1.1%, p<0.05) and cued recall measures. The proportion of smokers reporting they had noticed the display at all before making their tobacco purchase was significantly greater before the ban than afterwards.

Figure 2

Proportions (with 95% CIs) of smokers before versus after the tobacco display ban making spontaneous purchases, mentioning the display via free recall and cued recall and reporting noticing the displays at all.

Prior to the ban, spontaneous purchasers (n=62) were significantly more likely than planned purchasers (n=158) to mention the display either via free recall (9.7% vs 3.2%, p<0.05) or cued recall (40.0% vs 17.9%, p<0.05). After the ban, the difference between spontaneous (n=36) and planned (n=146) purchasing groups persisted for both free recall (5.6% vs 0.0%, p<0.05) and cued recall measures (11.1% vs 2.1%, p<0.05). As can be seen in figure 3, prior to the ban nearly twice as many spontaneous purchasers mentioned the display via free recall compared with after ban, but this difference did not achieve statistical significance (9.7% vs 5.6%, p=0.380).i However, significantly more spontaneous purchasers said the displays influenced them using cued recall before versus after the ban (40.0% vs 11.1%, p<0.01). Similarly, significantly more planned purchasers mentioned the displays before versus after the ban both via free recall (3.2% vs 0.0%, p<0.05) and cued recall (17.9% vs 2.1%, p<0.001).

Figure 3

Proportions (with 95% CIs) of spontaneous versus planned purchases mentioning the display via free recall and cued recall before versus after the ban.

Before the ban, smokers aged 18–24 years (n=30) were significantly more likely to mention the displays free recall than smokers aged 25+ years (n=190) (20.0% vs 2.6%, p<0.001). A non-significant difference was also observed between the two age groups via cued recall (36.8% vs 21.3%, p=0.129). After the ban, very few smokers aged 18–24 years (n=26) or 25+ years (n=156) mentioned the tobacco displays either via free recall (0.0% vs 1.3%, p=0.734) or cued recall (3.8% vs 3.8%, p=0.667).


The present study holds many advantages over previous studies investigating the influence of tobacco displays on spontaneous purchase behaviours: for the first time comparison data are available within a single jurisdiction both in the presence and absence of tobacco displays; the effect of the display on spontaneous purchases was not confounded by the presence of advertising; the exit interview method minimises recall bias; and the use of free recall and cued recall items minimises leading and suggestive questions. However, the study also has limitations. The reliance upon a convenience sample, even in light of our good response rate, may limit the ability to generalise these findings to the wider smoking population. In addition, the choice of supermarkets, while representing the majority of tobacco sales in Australia, may not be generalisable to other tobacco retailers.

With these caveats in mind, the results suggest that the ban effectively reduced smokers’ awareness of tobacco displays: one in four smokers reported noticing the displays before the ban, and one-in-a-hundred noticed them afterwards. This is consistent with observed high levels of retailer compliance when display bans were introduced in Canada, Ireland and Victoria, Australia.9 ,21 ,22 It also supports previously published data suggesting sharp declines in smoker awareness of tobacco promotion after such bans were implemented in the same countries.8 ,9 However, these studies used autobiographical retrospective telephone surveys prone to recall bias, whereas the contribution of this study was to demonstrate the dramatic effect of such a ban on tobacco purchasers’ immediate recall of tobacco displays.

Almost one-third fewer smokers reporting a spontaneous tobacco purchase after the display ban was implemented. It is important to emphasise that our method did not recruit all smokers who left the store—only those who made a tobacco purchase. As such, our data imply that the proportion of smokers who entered the store and left without making a spontaneous tobacco purchase increased as a result of the display ban. However, it is beyond the scope of the present study to directly attribute our observations to the display ban itself—the convenience sampling and non-random allocation of our preban versus postban comparison groups means we cannot rule out the difference being explained by other factors, such as sampling biases. Notwithstanding, our data are consistent with the hypothesis that the absence of tobacco displays is associated with fewer spontaneous tobacco purchases.

Previous research also suggested that spontaneous tobacco purchasers are significantly more likely than planned purchasers to claim tobacco displays influence their purchase decisions.14 ,16 Our data replicated these findings and went further to demonstrate that far fewer smokers mention the influence of tobacco displays on purchases after a ban is implemented. Our preban data also replicate previous research suggesting that younger smokers, aged 18– 24 years, are particularly susceptible to displays motivating spontaneous purchases.16 This is consistent with age differences reported in other studies.12 ,14 Considered together, these findings suggest that young adult smokers are more influenced than older smokers by the presence of displays. Our results also add that this age effect disappears with the implementation of display bans.

Another contribution of the present study is to confirm that participants still nominate the tobacco display with minimal prompting. The tobacco industry claims that smokers’ suggestions of displays influencing their tobacco purchases can be fully attributed to leading questions15; our results repudiate this claim. However, it is important to note that around four times fewer participants mentioned the displays via free recall than cued recall. This is a reflection of the two questions assessing different things: the free recall data indicated the extent to which displays were the first thing smokers thought of when asked what influenced them, whereas the cued recall data indicated whether the displays influenced them to any extent at all—alone or in combination with other factors. As such, the 5.0% of our preban participants who nominated the display free recall serves as a valid but conservative estimate of the extent to which such influence tobacco purchases and suggests previous studies may have overestimated its influence. However, it is interesting that our cued recall item suggested 24.1% of the total preban sample nominated tobacco displays, which is remarkably consistent with the 22–25% previously reported via the prompted/free recall items of Wakefield et al and Carter et al.12 ,14 Likewise, the 40.0% of our preban spontaneous purchasers who nominated tobacco displays via cued recall is consistent with the 47% reported by Carter et al also via a prompted/free recall item.14 The congruence of these results, despite differing levels of prompting, goes some way to alleviate the concerns raised by the tobacco industry that smokers feel unable to say ‘no’ to leading questions.15

The main inconsistency between our results and previous studies is the proportion of our preban spontaneous purchasers who mentioned tobacco displays via cued recall (40.0%) compared with those of Clattenburg et al (76.5%) who also used a semiprompted/cued recall item.16 This difference can satisfactorily be explained by the fact that Clattenburg et al collected their data in Vermont, New England, where in-store tobacco advertising is still permitted, and their semiprompted/cued recall response item combined tobacco advertising and displays within the one response option.

Finally, it should be noted that participants, although fewer, were still reporting spontaneous tobacco purchases after the display ban was implemented. Visible tobacco products are therefore not the only factor influencing smokers’ spontaneous purchases. Even in the absence of visible tobacco products, it is likely that some customers in our sample would have recognised retail counters from which they had previously purchased cigarettes and the sight of these alone may have prompted them to make spontaneous purchases. This is consistent with the finding that some of our participants still mentioned the sight of the tobacco cabinet as an influence on spontaneous purchases even after the ban. However, the fact remains that after the ban significantly fewer participants made mentions of the display via either free or cued recall, coupled with significantly fewer reports of spontaneous purchases. In a context where tobacco advertising is not present, our results suggest that tobacco displays are an important influence on adult smokers’ spontaneous purchases and their removal from line-of-sight results in fewer such purchases.

What this paper adds

  • Four previous studies have suggested that tobacco displays encourage spontaneous cigarette purchasing behaviours among adult smokers and therefore constitute a form of advertising. However, two of these studies relied upon smokers’ memory of ‘usual’ purchases, thereby introducing recall bias; three used prompted questions likely to inflate respondent attributions; and two were confounded by coincidental tobacco advertising. Such limitations allow the tobacco industry to argue these studies are ‘flawed’ and ‘unreliable’. This study addresses these criticisms by using immediate, postpurchase intercept interviews, specifically avoiding the use of prompted questions and providing comparison data before and after the implementation of a display ban in the absence of other tobacco advertising. The results are entirely consistent with previous ‘flawed’ studies confirming that tobacco displays influence spontaneous purchases and their removal is associated with a reduction of such purchases.


Many thanks go to Ross Goulding and Kazuki Hayashi for assistance with data collection.



  • Correction notice This article has been corrected since it was published Online First. Affiliation 1 has been amended to ‘1Centre for Behavioural Research in Cancer Control, Curtin University, Shenton Park, Western Australia.’

  • Contributors OBJC was responsible for developing the research question, participated in the development of the methodological design and was primarily responsible for statistical analysis and preparation of the manuscript for publication. TP participated in refining the methodological design, was primarily responsible for data collection and aided in the preparation of the manuscript for publication. BWM aided in the development of the research question, in refining the methodological design and in the preparation of the manuscript for publication.

  • Funding This research was not funded. During the data collection period, OBJC's salary was partially funded by a Western Australian Health Promotion Foundation grant (15054).

  • Competing interests None.

  • Ethics approval Curtin University Human Research Ethics Committee.

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

  • i Low raw numbers for these subsamples resulted in an underpowered statistical comparison; a posthoc calculation using G*Power (v.3.0.1) suggests this sample size provided statistical power of 38% to detect a statistically significant difference.