Article Text

PDF

The effects of variant descriptors on the potential effectiveness of plain packaging
  1. Ron Borland,
  2. Steven Savvas
  1. The Cancer Council Victoria, Melbourne, Victoria, Australia
  1. Correspondence to Dr Ron Borland, The Cancer Council Victoria, 1 Rathdowne St, Carlton, Melbourne, VIC 3053, Australia; Ron.Borland{at}cancervic.org.au

Abstract

Objectives To examine the effects that variant descriptor labels on cigarette packs have on smokers’ perceptions of those packs and the cigarettes contained within.

Method As part of two larger web-based studies (each involved 160 young adult ever-smokers 18–29 years old), respondents were shown a computer image of a plain cigarette pack and sets of related variant descriptors. The sets included terms that varied in terms of descriptors of colours as names, flavour strength, degrees of filter venting, filter types, quality, type of cigarette and numbers. For each set, respondents rated the highest and lowest of two or three of the following four characteristics: quality, strongest or weakest in taste, delivers most or least tar/nicotine, and most or least level of harm.

Results There were significant differences on all four ratings. Quality ratings were the least differentiated. Except for colour descriptors, where ‘Gold’ rated high in quality but medium in other ratings, ratings of quality, harm, strength and delivery were all positively associated when rated on the same descriptors.

Conclusions Descriptor labels on cigarette packs, can affect smokers’ perceptions of the characteristics of the cigarettes contained within. Therefore, they are a potential means by which product differentiation can occur. In particular, having variants differing in perceived strength while not differing in deliveries of harmful ingredients is particularly problematic. Any packaging policy should take into account the possibility that variant descriptors can mislead smokers into making inappropriate product attributions.

  • Packaging and Labelling
  • Advertising and Promotion
  • Harm Reduction

Statistics from Altmetric.com

Introduction

Brands are used to distinguish products, but they also play a role in communicating to the consumer a set of values, beliefs, aspirations, images and ideas.1 This is the case for cigarettes as with many other consumer products. Variant descriptors are also used on packs to inform, support and differentiate cigarettes within a brand.

There are an abundance of cigarette products on the market. In Australia, brands tend to be targeted at a particular price segment of the market (eg, premium, mainstream and budget), and the popular brands have a family of variants. As elsewhere, the variants, with the exception of menthol-flavoured ones, are primarily marketed around levels of strength, although there are a few that vary in terms of length and stick diameter.2 Variants that differ in experienced strength (essentially filter-vented ones3) used to be marketed using terms like ‘light’ and ‘mild’, but with the prohibition of the misleading use of these terms, are now described in other ways. A selection of variant descriptors found on cigarette packs in the Australian market in 20104 includes label descriptors from Benson & Hedges (eg, Classic, Smooth, Fine, Rich and Ultimate); Dunhill (eg, Premier, Distinct, Refined and Infinite); Holiday (eg, Bright blue, Sea green, Dawn grey and Sun gold); Horizon (eg, Red, Purple, Blue and Orange).

Variant descriptors such as ‘mild’; ‘light’, and ‘low-tar’ have been shown to mislead smokers.5 This is because cigarettes so labelled do not systematically deliver less nicotine or tar to smokers, and are, thus, not even plausibly less harmful, even though they are labelled as such, on the basis that they deliver lower yields of tar to a smoking machine.6 The main mechanism for this deception is the use of filter venting, which acts to dilute the smoke, especially when puffed gently, and allows for compensatory smoking, such that the smoker (but not the smoking machine) smokes the cigarette differently, and ends up taking in about the same amount of tar as they would from a ‘full strength’ variant.3 ,6 ,7 However, they experience the cigarette as somewhat milder or lighter in taste, leading to the incorrect, but understandable inference that they are less harmful,8 something the tobacco industry failed to correct.2 As a consequence, the misleading use of terms like light and mild to describe cigarettes is now banned in many countries, including Australia. Doing this is an obligation under the Framework Convention on Tobacco Control.9

Research on the impact of prohibiting the misleading use of terms such as light, mild and low-tar have had only a small, if any, effect on smokers perceptions,10 ,11 possibly because of the transfer of the inferred attributes to the alternate variant terms that replaced the banned terms. Experimental studies have also shown reduced harm and associated ratings for terms like ‘smooth’, ‘silver’ or ‘light’, and for lower numbers (eg, 6 vs 10).12–14

The current study investigated the effect of a range of cigarette descriptors on smoker perception towards cigarettes in a context where there was minimal other information available, as is now the case as plain or standardised packaging was introduced in Australia in December 2012.15 At the time of the study, descriptors such as colour (eg, Winfield Blue, Holiday Gold, Pall Mall Red etc), flavour (Dunhill Infinite, Escort Ultimate, Peter Jackson Fine, etc) and other labels, such as inferred quality (Dunhill Refined, Alpine Rich, etc) were all used by cigarette manufacturers to brand their product. Written descriptors are often used with complementary colours or designs on the pack, something the Australian legislation now prohibits. It should be noted that this study was completed before the Australian government announced its specific proposals as to what the plain packs would look like.15 We were interested in exploring inferences drawn from sets of descriptors, including those used currently and some hypothetical ones. We also explore inferences about names that describe physical characteristics of the cigarettes, and ones that refer more to qualitative differences.

Method

Participants

Participants were 320 adults, aged 18–29 years old, who were current or ex-smokers (recent, as they were recruited from a panel where they had identified themselves as smokers), were recruited into two parallel surveys. An equal number of males and females were enrolled (quota sampling). The demographic breakdown is shown in table 1.

Table 1

Demographic and smoking characteristics of participants

Design

As part of two studies by our group investigating the effect of the cigarette pack16 and stick design on smokers’ perceptions,17 additional questions were incorporated following the focal study to investigate the effect of variant descriptors on smokers’ perceptions.

Participants were exposed to a prototype plain cigarette pack with generic markings and a circled area on the front of the pack labelled VARIANT DESCRIPTOR (see figure 1). The prototype plain package cigarette pack was designed to simulate the (then) expected design features that would be permitted under the Australian plain packaging legislation. The pack was designed to contain 20 standard-sized cigarettes, a standard beige colour (subsequent regulation specifies the colour of Australian plain packs as dark olive brown15), with standard font for the brand, the variant descriptor, and the number of cigarettes. The graphic health warning on the face of the pack was SMOKING CAUSES PERIPHERAL VASCULAR DISEASE, with the appropriate picture. Note, that our prototype used 10-point font for brand and variant, whereas the new packs have no greater than 14-point font for the brand and 10-point font for the variant name, both in the typeface Lucinda Sans.15

Figure 1

Screen shot of prototype pack as shown with variant descriptor circled. Access the article online to view this figure in colour.

Participants were asked to make judgements about sets of descriptors using two or three of four dimensions: quality (highest to lowest) strength (strongest to weakest in taste); tar and nicotine levels (highest to lowest); and harmfulness (most to least).

The sets of variant descriptor terms shown in the first survey and ratings made were Colour (Red, Blue, Gold, Silver, White)—taste and tar/nicotine levels; Flavour (Infinite, Ultimate, Full Flavoured, Smooth)—taste and tar/nicotine; Venting (Highly vented, Medium vented, Low venting, Unvented)—taste, tar/nicotine and harmfulness; Filter (Advanced filter technology, Charcoal filter, Filter tipped, Dual filter)—tar/nicotine and harmfulness. The sets were presented in this order.

The second survey only asked about taste and quality for four sets of variants. It used variants differing in Colour (Red, Blue, Gold, Silver, White); Quality (Rich, Distinct, Premium, Premier, Refined); Style (Standard, Virginia, American blend, Oriental blend); Blend (Blend 001, Blend 004, Blend 012, Blend 333, Blend 879) presented in that order.

The venting and blend descriptor sets were shown in either the order as listed, or in reverse order. All other sets were randomised for presentation.

Ratings were generally made by identifying the highest and lowest (or best and worst) in each set. For the colour set in survey one, subjects, instead, ranked each colour from lowest to highest quality (1–5). Thus, colour was rated in quality in different ways in the two samples.

Procedures

A registered market research company (the Social Research Centre) was commissioned to implement these web-based questionnaires, using a national panel of previously identified smokers they had generated. As compensation, participants were rewarded with credits as part of a company incentive scheme. Ethical approval was obtained from the Cancer Council Victoria Human Research Ethics Committee.

Statistical analysis

An overall mean for each descriptor in each rating was calculated by weighting as 5 points each ranking of most attractive, highest quality, strongest taste, or most harm; weighting as 1 point each ranking of the least attractive, lowest quality, weakest taste or least harm; and scoring all other cases 3 points (including cases where all were rated equivalent).

Data analysis was conducted using SPSS V.18.0. Repeated measures analysis of variance were used to test for mean differences between variant descriptor terms. Posthoc tests used Least Square Differences for multiple comparisons. We used a significance level of 0.05 throughout, but note that within-subject power to find effects was greater than for the between-subjects effects. As the data is not strictly interval, we also checked using the Kolmogorov–Smirnov test which, for the sample size of 160, gives any difference in proportions greater than 0.108 significant at p<0.05, and any greater than 0.129 significant at p<0.01. However, as we ranked both highest and lowest, the combination of this is lost from the non-parametric test, so the parametric one gives a better overall picture of differences.

Results

Sample characteristics

Table 1 shows the demographic breakdown of the two independent samples. Both samples were similar, with only small changes in smoking behaviour for ex-smokers. Age ranged from 18–29 years, with the median at 25 years. Approximately 80% of all respondents were current smokers. As differences between the samples were minor, this paper will not differentiate between the results from either sample.

Colour descriptors

Table 2 shows the perceived taste, quality and the tar/nicotine delivery for cigarettes in packets with varying colour descriptors. There were significant differences by descriptor in all: perceived taste (p<0.001), quality (p<0.001) and tar/nicotine delivered (p<0.001). It is notable that the results from the full ranking on quality mirrored those from the highest/lowest ratings, confirming the validity of this truncated means of elucidating preferences. There was a moderate association between colour descriptors rated strongest in taste, and delivers most tar/nicotine (ρ=0.58, p<0.01). Red was ranked strongest in taste (p<0.001), and delivering the most tar/nicotine (p<0.001), while White was lowest in both. By contrast, Gold was rated as highest in quality, and beyond that, the darker colours rated higher than the lighter ones. Quality and taste were not significantly correlated (p>0.05). The Quality and tar/nicotine delivery correlation could not be computed, as the measures were across different samples.

Table 2

Colour descriptors—taste, quality and tar/nicotine (nic) delivery: mean rankings (SE) and extreme rankings

Flavour descriptors

Table 3 shows the flavour descriptors affected rankings on perceived taste (p<0.001) and tar/nicotine delivery (p<0.001), with the two ratings moderately correlated (ρ=0.57, p<0.01). Full flavoured was ranked strongest in taste (p=0.013) and delivering the most tar/nicotine (p=0.014), though a sizeable number of respondents also rated the Ultimate as the strongest tasting (29%) and most tar/nicotine delivering (24%). By comparison, Fine was ranked as having the weakest taste and the least tar/nicotine. Posthoc tests revealed all between descriptor ratings, except between Infinite and Smooth, were significant for both rankings.

Table 3

Flavour descriptors—taste and tar/nicotine (nic) delivery: mean ranking (SE) and extreme rankings

Inferred quality descriptors

For the inferred quality descriptors, both taste (p<0.001), and tar/nicotine delivery (p<0.001) rankings varied, with the two moderately correlated (ρ=0.37, p<0.01). Table 4A shows that Rich was rated as the strongest taste (p=0.05) and Premium as the highest quality (p=0.002). Conversely, Refined was rated as the weakest in taste (p<0.001) and equal lowest in quality with Distinct (p>0.05). Distinct and Premier were not differentiated on taste (p>0.05). Rich and Premier were not differentiated on quality (p>0.05).

Style descriptors

Main effects in the style descriptor were found for taste (p<0.001) and quality (p=0.041), with the two moderately correlated (ρ=0.49, p<0.01). Table 4B shows that respondents rated Virginia and American blends strongest in taste, but could not differentiate between the two (p>0.05), but were rated as stronger in taste (ps<0.001) than both Standard and Oriental blend which were not differentiable (p>0.05). A similar pattern occurred for quality, however, only Oriental blend was significantly lower in quality than the Virginia (p=0.023) and American blends (p=0.033).

Blend descriptors

For the blend descriptors (table 4C), there were main effects for taste (p<0.001) and quality (p=0.048), with the two moderately correlated (ρ=0.49, p<0.01). Blends were rated in ascending order of strength based on number on both ratings, but a few reversed this order, and half thought them to be all equivalent.

Table 4

Taste and quality mean rankings (SE) and extreme scores on three sets of variant descriptors

Venting descriptors

Table 5 shows the rankings of the venting descriptors varied by taste (p<0.001), delivers the most tar/nicotine (p<0.001) and harm (p<0.001). There was a strong association (ρ=0.69, p<0.01) between taste and tar/nicotine delivery, a medium association (ρ=0.48, p<0.01) between taste and rated harm, and a strong association (ρ=0.66, p<0.01) between rated harm and tar/nicotine delivery. Unvented was ranked the strongest taste (all p<0.001), delivering the most tar/nicotine (p<0.001), and most harmful (p<0.001), with highly vented ranked lowest on all.

Table 5

Venting descriptors—taste, quality and tar/nicotine delivery ranking means (SE) and extreme scores

Filter descriptors

Table 6 shows the filter descriptors affected rankings of both delivers most on tar/nicotine (p<0.001) and most harm (p<0.001), with the two highly correlated (ρ=0.68, p<0.01). Charcoal filter and filter tipped were not differentiable in terms of tar delivery (p>0.05) but were ranked as equal greatest in delivering the most tar/nicotine (p<0.001) and most harm, while advanced filter was seen as delivering the least tar/nicotine and being least harmful. Rankings for filter descriptors of most harmful followed the same pattern as tar/nicotine delivery.

Table 6

Filter descriptors—tar/nicotine (nic) delivery and harm ranking means (SE) and extreme scores

Consensus within variant descriptors

Ratings of quality and taste were positively correlated in two of three comparisons (not for colour). Taste and tar/nicotine delivery were significantly correlated on all four comparisons, as were the two between taste and harm, and the only one tested between tar delivery and harm. In all cases, except for colour, the various ratings made by the same respondents on a set of descriptors were correlated. While a majority of respondents saw consistent differences, a minority (usually) saw none, and this varied by descriptor set (see tables 36).

Discussion

The results from this study show that all sets of pack-label variant descriptors systematically affected respondents’ judgements of the cigarettes associated with them, and this is so even when presented on plain, standardised packs. This is not surprising; however, it is of concern when it leads to erroneous inferences about relative harm.

For the most part, rankings were intercorrelated with stronger taste, higher quality, higher tar and higher harm all related. The exception, where it was studied, was for the colour variants, where Gold was seen by many as the highest in quality, but Red the strongest in taste. This undoubtedly related to broader conceptual meanings, where gold is highly valued as a commodity, and by inference, things coloured gold are good. Furthermore, in Australia, two of the premium brands, Dunhill and Benson & Hedges both use gold in their packaging. Clearly, strong cultural preconceptions can shape these judgements, undoubtedly in interaction with use by the companies of these cultural preconceptions to position their products in the market.

It is notable that stronger cigarettes are rated as being higher in quality even though they are seen to be higher in harm, an attribute not normally associated with high quality. Unfortunately, we did not have any direct comparisons of quality and harm, so cannot be certain that harm and quality are positively associated, but tar level and quality are, as are tar level and harm.

While responding is consistent with what we know about human behaviour, it is useful to demonstrate that this is happening or can happen with cigarette brand variants.

We suspect the implicit positive association between strength, which is seen as an indicator of quality, and harmfulness is unique to products where there is a risk associated with use that is linked to the experiences sought from use. Thus, whisky is a stronger alcoholic beverage than beer, and we guess it would generally be seen as potentially more harmful. That smokers are acknowledging some harm is hardly surprising, Australian smokers have been educated for decades about the harms, and the current sample would have been exposed to such information all their lives. However, this should not be taken to show that these smokers are adequately informed. Smokers underestimate the magnitude of the harms,18 ,19 thus, while they are accepting some risk, they are underestimating the magnitude of this risk.

The most compelling conclusion, although hardly a surprising one from this study, is that given any set of attributes, people try to find order among them, and that some of this order is socially shared (ie, they come up with similar judgements). For variant sets that are currently used, some of the commonality is likely due to some combination of industry communications and taste differences between the variants. For those that are novel, it is clear that more universal heuristics are being used: for instance, a higher number means more of something than a low one. However, the nature of the comparison can affect this, with highly vented being seen as weaker, and less harmful than unvented. Whether this is due to understanding of how filter venting works, or relies on a more universal understanding that ventilation is something that ‘lets in air’, is uncertain.

There were two sets of ratings where responses were inconsistent with the science. The charcoal filter was seen as one of the filter descriptors delivering the most tar/nicotine, and being the most harmful. Charcoal is, in fact, a very good filter, but it is a dirty product, and thus, a heuristic of similarity would lead to the assumption that it was a more harmful product.

For the style descriptors, Oriental blend was rated as equally weakest in taste even though it is generally considered to be a very strong-tasting blend. This suggests that in both this case and with the filters, few respondents have had experience with these products and were relying on broader heuristics to make their judgements.

Currently, filter venting is the main engineering feature that affects the performance of cigarettes and, thus, the experiences of use.3 Highly vented filters are, erroneously,8 seen as lower in harm.5 As we have shown with the impacts of bans on misleading descriptors like ‘light’, ‘mild’ and ‘low tar’,10 ,11 just removing the descriptors has little impact when the engineering features that create the experiences of difference remain. The only viable solution is to prohibit engineering features that contribute to increased product attractiveness particularly of rated strength, especially when they do not actually reduce actual exposures, and thus, do not affect product harmfulness. While it might be possible to sustain some product differentiation solely on the basis of differential variant names, where physical differences exist, it is likely to be far easier to create strong product preferences for specific variants. This analysis suggests that merely considering names in isolation of product differentiation will be a less effective strategy than regulating both together.

There are a number of limitations to this study. Respondents were drawn from an internet panel, and though the sample is not representative, it was unselected on any aspect related to the study aims. Therefore, responses are unlikely to be much different than a more representative sample. That said, some caution should be taken in drawing conclusions about the absolute magnitude of effects, but this is of minor importance and does not threaten the conclusion that variant descriptors influence perceptions. It should also be stressed that these judgements were made in an experimental context, and we did not assess smokers’ prior experiences with different variants within each set, so cannot know what effects experience may have had for those descriptors that have been used. However, that we obtained similar effects for descriptor sets that are currently used, as well as novel ones, suggests that experience is not a critical feature of these judgements, although it may affect the magnitude of effects.

There are also a number of strengths with this study. First, we targeted younger smokers and ex-smokers (aged 18–29 years old), a group at an age more brand conscious.20 As such, they would be more affected (in their choices) by the characteristics we explored than older and more dependent smokers. Research has shown that loyalty to a cigarette brand, once fostered, is hard to break.21 ,22 We do not know to what extent these judgements would be shared by older smokers, but can see no reason why the same overall pattern would not occur. Second, the use of a within-subject design to explore key comparisons means that the study has considerable power, even with the relatively small sample size.

In summary, these results show that smokers impute differentiable characteristics to cigarettes based on variant labels. As such, variant descriptors on plain packaging can be used as a powerful branding tool. This role of variant descriptors will be even more important than it has been, as most other aspects of cigarette differentiation were removed from Australian packs from December 2012. While variant descriptors are allowed, it seems to us inevitable that smokers will try to make value judgements between them. If those differences are real and the judgements appropriate, this is a good thing. However, if these judgements lead to systematic misunderstandings, then they are a problem. It is particularly a problem when the incorrect attributions are around risk and harm, as is the case for many variants sets. Indeed we suspect any variant set that induces inferences about relative strength is likely to also result in inappropriate inferences about harmfulness. We believe that variants should only be allowed where the differences are meaningful, and can be shown not to unduly create misleading perceptions of risk. This is doubly the case for products like cigarettes which are apparently so benign, but which prematurely kill around half of all long-term users.

Key messages

  • Variant descriptors have the potential to be used to influence smokers’ attributions about cigarettes, including, in potentially misleading ways.

  • Attributions tend to be systematic with strength related to estimated tar/nicotine deliveries and, presumably, through this to perceived harmfulness.

  • These effects are likely to occur in the context of plain packaging as instituted in Australia.

Acknowledgments

Karen Moore assisted in the design of the study.

References

View Abstract

Footnotes

  • Contributors RB initiated and designed the whole trial, monitored data collection, and reviewed the paper. SS cleaned, analysed and interpreted the data and drafted the paper.

  • Funding From Quit Victoria and Cancer Council Australia.

  • Competing interests None.

  • Ethics approval Cancer Council Victoria Human Research Ethics Committee.

  • 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.