Article Text
PDF
Abstract
Background Tobacco industry documents illustrate how packaging promotes smoking experimentation and reinforces existing smokers' behaviour. Plain packaging reduces the perceived attractiveness of smoking and creates an opportunity to introduce larger pictorial warnings that could promote cessation-linked behaviours. However, little is known about the effects such a combined policy measure would have on smokers' behaviour.
Methods A 3 (warning size) *4 (branding level) plus control (completely plain pack) best-worst experiment was conducted via face-to-face interviews with 292 young adult smokers from a New Zealand provincial city. The Juster Scale was also used to estimate cessation-linked behaviours among participants.
Results Of the 13 options tested, respondents were significantly less likely to choose those featuring fewer branding elements or larger health warnings. Options that featured more branding elements were still preferred even when they also featured a 50% health warning, but were significantly less likely to be chosen when they featured a 75% warning. Comparison of a control pack representing the status quo (branded with 30% front of pack warning) and a plain pack (with a 75% warning) revealed the latter would be significantly more likely to elicit cessation-related behaviours.
Conclusions Plain packs that feature large graphic health warnings are significantly more likely to promote cessation among young adult smokers than fully or partially branded packs. The findings support the introduction of plain packaging and suggest use of unbranded package space to feature larger health warnings would further promote cessation.
- Plain packaging
- graphic health warnings
- young adult smokers
- packaging and labelling
- young adults
Statistics from Altmetric.com
Introduction
The World Heath Organization Framework Convention on Tobacco Control (FCTC) recognises smoking as the leading cause of preventable death and disease internationally and aims to limit the harm caused by tobacco.1 Signatories have agreed to restrict tobacco marketing and several have banned tobacco advertising and sponsorship, and restricted retail marketing. Nevertheless, marketing continues through other media, particularly the branding that appears prominently on tobacco packaging.2
At its most fundamental level, branding uses distinguishing symbols to help consumers identify products or services.3 However, brands comprise more than a series of signs or colours; they also possess symbolic meanings that marketers and consumers create jointly, to the extent that brands become relationship partners.4–6 This anthropomorphic status enables brands to develop, communicate and reinforce ideas, images and personalities; these attributes have particular significance in relatively homogeneous markets such as the tobacco market, where each market partition contains brands with similar functional properties.7 8 Cigarette companies create, promote and sell images that they communicate via branding, specifically brand livery, which includes colours, fonts, logos and images.9–11 The distinctive imagery associated with tobacco brands means this imagery functions as a ‘badge product’, for which consumers' choice depends less on the brands' physical properties than on their alignment with smokers' (or potential smokers') aspirations.2 9 Branding theory suggests the bonds between consumers and the brands they use are often robust, entrenched and resistant to change.5
Analyses of industry documents also highlight the crucial importance of branding, and of cigarette packaging as a medium to communicate brand attributes.9 Cigarette packets are meticulously researched and designed, their livery reassures smokers about risk12 and their brand imagery reinforces smokers' self-image.11 Tobacco packaging thus ensures smokers and potential smokers continue to receive messages that promote smoking.
Increased regulation has led tobacco companies to rely on packaging to establish and reinforce brand imagery.2 13–15 Packaging is largely unregulated and, as cigarette packets are retained until empty, they are highly visible and frequently displayed to others. Moodie and Hastings concluded: ‘Tobacco packaging is no longer the ‘silent salesman’ it once was…it now has to shout loudly. This increased volume is used to defy advertising bans and drown out health warnings’.16
Tobacco control researchers have argued that packaging's role in promoting tobacco brands contravenes Articles 11 and 13 of the FCTC, and called for its regulation.17 Previous studies found the progressive removal of brand elements led adult smokers to view cigarette packs as increasingly unattractive. Furthermore, perceptions of the people who would smoke cigarettes from plain packages, the inferred experience of smoking these cigarettes and the risk of forming misleading beliefs were lower than for those smoking branded cigarettes.11 18
Young people are disproportionately at risk of taking up smoking and many appear resistant to interventions designed to deter smoking initiation among non-smokers and encourage smokers to make quit attempts.19 This group is also particularly susceptible to branding, so may be responsive to plain packaging and larger graphic warning labels,20 21 although their effects may be moderated by brand familiarity.22 Recent work examined how manipulation of warning size and branding elements affected young adults' perceptions of smoking.23 The findings suggest reducing branding imagery while introducing larger warnings decreased perceptions of the pack and smoking experience among non-susceptible and susceptible non-smokers, experimenters and established smokers.
However, although both adult and young adult smokers' perceptions are less favourable when exposed to plain packaging and, in the latter case, larger warnings, we have not identified any studies that examined more behaviourally oriented outcomes. Such studies would be particularly valuable as the tobacco industry has trenchantly opposed proposals to introduce plain packaging and questioned the effects this would have on smoking prevalence.24 We extended earlier work by measuring young adult smokers' choice behaviours and projected responses to plain packaging and larger graphic health warnings. More specifically, we hypothesised that reducing branding elements and increasing the health warning size would progressively reduce the attractiveness of cigarette packaging. Given findings from the ITC study that suggest graphic health warnings have promoted quit attempts,25 we also hypothesised that the propensity to engage in cessation-related behaviours would be greater following exposure to a plain pack with a 75% graphic health warning relative to a branded pack with a 30% graphic health warning.
Methods
Design
The study estimated responses to graphic health warnings in two ways. First, a ‘best-worst’ experiment using a 3 (warning size) * 4 (level of branding) plus control (pack with neither branding nor a warning) design exposed participants to a balanced series of combined brand-warning images. Table 1 illustrates the design employed. Holiday was the test brand as this is the most popular young adult brand in New Zealand.26 A graphic warning label similar to one used in the European Union and Asian countries was used to eliminate the likelihood respondents would have conditioned reactions to the graphic health warning.
Experimental design
Second, the Juster Scale (an 11 point probability scale designed to estimate conditional behaviours) was used to estimate responses to one of two test packages (see supplementary file 1 for an illustration of the Juster Scale).27 The research design was reviewed by the departmental ethics administrator and an iwi (indigenous peoples) consultative committee.
Sample
A face-to-face survey of 292 young adult smokers aged between 18 and 30 was conducted. This demographic has the highest smoking prevalence rates and is thus of particular interest.19 The sample was selected purposively from young people observed smoking in the University of Otago and Otago Polytechnic campuses, throughout the central city, and in a low socioeconomic area of Dunedin, New Zealand (South Dunedin). Although this approach means sample representativeness cannot be assumed, the research tested the relative effects of different options rather than estimating population parameters. The fact that the sample was non-random was therefore not regarded as problematic. The overall response rate was 79% (valid interviews/(total contacts − ineligible individuals)).
Procedure and questionnaire
Respondents first participated in a ‘best-worst’ (BWS) experiment using the approach that Finn and Louviere developed as a multiple comparison extension to paired comparison tests.28 BWS is an indirect measurement approach that infers participants' levels of preference for the test variables (in this case, varying levels of branding and different-sized graphic warning labels) from their choice patterns.29 We used BWS to measure participants' choices of specific cigarette packs; these choices should more closely approximate actual choices than attitude or intention measures that require explicit, direct questions about the test stimuli.
BWS produces an ordinal ranking of items for each person, and interval or ratio scale estimates of each items' attractiveness based on aggregating responses at a sample or segment-level.29–31 BWS estimates the preference strength of each choice option (whether a brand or attribute); the resulting measures can be directly compared across subjects, avoiding the indeterminancy of rating.28 The associated choice tasks engage respondents and are easy to do, particularly compared with the cognitive effort required to provide ranking information for large sets of items.
To implement a BWS task, one must construct multiple comparison sets. A straightforward way to do this with desirable statistical properties is by using a balanced incomplete block design (BIBD). In this study, we used a BIBD to create 13 sets that each presented four cigarette packs for comparison; each pack appeared four times and co-appeared once with each other pack. Thus, respondents viewed 13 show cards; each featuring four cigarette pack images that varied by the number of brand elements shown and by the warning label size and format. Respondents were told that each pack contained the same type of cigarettes and cost the same price, and that the options they saw were the only ones available to them when they went to purchase cigarettes. Respondents were then asked for each show card “Which pack would you be most likely to choose” and “Which pack would you be least likely to choose”. Two versions of the show cards were used to average any order effects. Figure 1 contains an example of the showcards respondents saw.
Showcard stimulus.
All respondents then used the Juster Scale to estimate their responses to one of two pack options: a control pack that was fully branded pack and featured a 30% graphic health warning, and text pack that featured the brand name in a standardised font and had a 75% graphic health warning. The Juster Scale addresses reliability problems caused when respondents interpret points on intentions scales differently; it also eliminates the difficulty of determining which intentions categories correspond to likely behaviour.
Juster argued that verbal intentions were actually disguised probability statements and proposed collecting these directly. The resulting 11-point Juster Scale combines verbal and numeric descriptions of the probabilities respondents will perform a future behaviour and recognises that all estimates of future behaviour are conditional (ie, some respondents with high probabilities will subsequently fail to carry out the behaviour concerned, while some respondents with very low probabilities will perform it). By producing a weighted average mean probability, the scale focuses on aggregate not individual behaviour; furthermore, the scale elicits metric level data (rather than ordinal data) with desirable analytical properties.27 Several studies have demonstrated the superior predictive ability of the Juster Scale when compared to traditional intentions scales.32 33
To test how the two brand-warning combinations affected cessation-related behaviours respondents were asked to imagine that the cigarettes they usually smoked were now only available in packs like the image shown. They were then asked: “What are the chances you would: seek cessation support (such as phoning the quitline), reduce the number of cigarettes you smoke, and make a quit attempt?” Because these estimates may be affected by social desirability error they were not used to predict population parameters, such as the proportion of smokers who would give up smoking, but to compare the relative effects of the two formats.34 Each option was tested across different day parts, in different areas and by different interviewers to randomise any location, timing or interviewer effects. The remaining questions examined respondents' smoking behaviour and demographic traits.
Statistical analysis
The BWS data were analysed with multinomial logit regression. To ensure the sub-samples administered the Juster Scale questions did not differ, current smoking behaviour, daily smoking behaviour and age at which regular smoking commenced were compared and found to be equivalent (see table 2). Independent samples t tests were used to compare the mean probability that each subsample would engage in cessation-related behaviours.
Respondents' characteristics
Results
Sample characteristics
Overall, 292 young adult smokers participated in the research. Table 2 contains details of respondents’ characteristics and compares these across the subsamples exposed to different stimuli in the between-subjects phase of the study.
Effect of pack manipulations on choice
In the multinomial regression model fitted to the best-worst data, the completely plain pack (no branding and no health warning) served as the benchmark against which other options were compared. The model estimates are interpreted as the log-odds of choosing a particular option relative to the plain pack. Thus, positive estimates indicate that an option is more attractive than the plain pack and negative coefficients indicate the opposite. Table 3 contains these results.
Best-worst estimates
Table 3 clearly shows that as branding elements are reduced, choice of those options as ‘best’ declines and of those chosen as ‘worst’ increases. The same pattern is evident for increases in warning label size. Only two options (fully and partially branded images with 30% graphic health warnings) were significantly more attractive than the control. Increasing the graphic health warning to 50% of the pack front significantly reduced the attractiveness of each option, but this was moderated by the number of brand elements present. Notably, 75% graphic health warnings were very unattractive, regardless of the level of branding present. We also estimated linear probability models (LPMs) from each respondent's choices to obtain a statistical description of the best and worst choices. We plot the mean estimates for combinations of brand elements and warning size in figure 2 to illustrate the resulting relation.
Response to brand elements and warning size.
Differences between each brand level were largest when a 30% graphic health warning was used and declined progressively as larger graphic health warnings were introduced. Predictably, as the pack surface available for branding decreased and the warning size increased, the effect of the brand elements diminished. As expected, the full brand level was chosen more often than the reduced brand levels; and choice patterns were similar across all brand levels. The two least branded options—the Holiday text and Cigarette text—did not differ in attractiveness, but were both significantly less attractive than the branded options.
Predicted effect of pack manipulations on cessation behaviours
Following exposure to either a control (branded 30% graphic health warning) pack or test (standardised brand name, 75% graphic health warning) pack, respondents used the Juster Scale to estimate the likelihood they would engage in cessation linked behaviours. The mean Juster Scale values for each of the three cessation related behaviours examined were computed and independent samples t tests were performed to examine whether the estimates differed across pack formats. Table 4 contains these results.
Estimated cessation-related behaviours
In each case, respondents exposed to the test pack (minimally branded with the larger warning label) were significantly more likely to indicate they would engage in cessation-related behaviours than those who saw the control (fully branded) pack.
Discussion
We hypothesised that young adult smokers would be less likely to select options that had reduced brand imagery and larger graphic health warnings. The findings from the best-worst experiment support this hypothesis and suggest that removing brand elements and increasing the graphic warning size would significantly decrease the attractiveness of tobacco packaging.
The results suggest branding partially offsets the effect of larger warnings and imply that larger warning labels would be most effective if placed on unbranded (or minimally branded) packages. The findings are consistent with earlier studies, which also concluded that familiar branding reassures smokers and overrides the negative and disconcerting effects graphic health warnings create.9 However, where graphic health warnings covered at least 50% of the package, or where minimal branding elements were present, respondents preferred plain packages; thus branding's ability to reassure smokers and maintain preference has limits and decreases when graphic health warnings match brand elements in size. These results provide important guidance to policymakers, who need to consider not only how people perceive possible interventions, but also how they are likely to respond to these.
The results confirm findings from earlier qualitative studies that reported young people dislike plain packages and regard them as less attractive than branded tobacco packages,35 36 and support the conjoint findings reported in a Canadian expert review on packaging.37 The results also extend recent studies examining perceptual responses to plain packaging and larger graphic health warnings.9 21 This work found that respondents rated cigarettes from increasingly plain packages, the people who smoked these and the likely experience of smoking them, consistently less positively than they rated the same branded cigarettes. Our data show young adult smokers' choice behaviours reflect the patterns in perceptual data, with progressively plainer packs featuring larger warnings chosen less frequently as the ‘best’ option and more frequently as the ‘worst’.
Projected behavioural responses to plainer packages with larger warnings are also consistent with the choice patterns identified. These findings extend earlier studies that examined preferences or intentions rather than conditional behaviour 9 16 21 and so address criticisms raised by tobacco companies, which dispute the likely effects of plain packaging.24
The study had some limitations. First, BWS experiments involve choice behaviour, not actual behaviour; however, there is no reason to believe the patterns observed would not be evident in actual behaviour. Measuring such behaviour would require the introduction and evaluation of the proposed interventions, an inevitable limitation of any experimental study examining a policy initiative such as dissuasive packaging.
Second, respondents' use of the Juster Scale to estimate likely cessation behaviours was higher for the control pack than suggested by their current behaviour, a discrepancy likely to have resulted from the novel graphic health warning used. Respondents had seen this in the BWS experiment and informal feedback suggested this was more impactful than many of the graphic health warnings in current rotation. However, as the research was designed to examine the relative effects of the two different pack designs, our interest was in whether the designs tested elicited different responses not in deriving population estimates. Nevertheless, additional research is required to assess the predictive validity of the Juster Scale when used to estimate population health behaviours, and future longitudinal research could examine its accuracy in predicting smoking cessation.
Future research could also explore other demographics to examine whether the response patterns we have identified can be generalised beyond young adults; it could also employ cohort designs, so questions of habituation and wear-out may be tested. In addition, research with susceptible non-smokers could extend earlier studies by testing whether reduced branding and larger graphic health warnings deter smoking initiation as well as promote cessation.22 Given the tobacco industry's resourcefulness, research also needs to explore policy options that could prevent the proliferation of covers designed to obscure larger warnings and recreate the appeal of branded packages.
Conclusions
Researchers recognise that cigarette packaging promotes brand imagery, maintains brand salience and influences smoking uptake and continuation, and have called for plain (generic) packaging and larger graphic health warnings to be introduced. Our results show that removing brand imagery and increasing graphic health warning size has a marked and significant effect on the attractiveness of cigarette packaging to young adult smokers and increases the likelihood that they will engage in cessation-related behaviours. These findings are the first to estimate the interaction between brand elements and graphic warning size. Although our study now requires replication to test whether the findings hold in different jurisdictions and with different population groups, it has quantified the importance of tobacco branding and supports calls to remove this and introduce larger graphic health warnings.
What this paper adds
Plain packaging has assumed greater importance following news of Australia's intention to introduce legislation that implements this policy by June 2012. Although recent research demonstrates that removing brand imagery and increasing graphic health warning size will weaken brand perceptions, little is known about how these measures will affect behaviour.
Findings from a best-worst experiment reveal that the attractiveness of cigarette packaging declines sharply and consistently as brand elements are removed and graphic health warnings increased in size. Estimates from a behavioural probability scale suggest a plain package with a 75% graphic health warning would be significantly more likely to stimulate cessation-linked behaviours than the status quo (a fully branded pack with a 30% graphic health warning).
The results address questions about the likely behavioural effects of plain packaging, and suggest that increasing the size of graphic health warnings and reducing branding to a brand name in standard font would increase cessation-related behaviours among young adult smokers.
References
Supplementary materials
Web Only Data tc.2010.037861
Files in this Data Supplement:
Footnotes
Funding Funding for the research was provided by the Heart Foundation of New Zealand (Grant 1386). This project was a pilot study that informed a proposal subsequently funded by the Health Research Council of New Zealand (Grant 09/195R).
Ethics approval Ethics approval was obtained from the departmental ethics administrator. The full HRC study has been approved by the University of Otago Human Ethics Committee. Both studies have involved consultation with the Ngai Tahu Consultative Committee.
Provenance and peer review Not commissioned; externally peer reviewed.