METHODS

We tested the following questions:

1. Has externally visible cigarette advertising increased since the signing of this agreement, especially in establishments frequented by youth?

2. Have brands with extensive billboard advertising pre-MSA redirected resources to retailer storefronts including windows, buildings and sidewalks?

3. Are retailers with more externally visible cigarette advertising more likely to sell to underage youth?

We assessed: (1) the changes in the mean number of cigarette ads for each brand/category by placement (building and detached area, sidewalk, windows and doors) visible from outside of five different types of tobacco retail establishments—that is, convenience, grocery, liquor, pharmacy and gas station/gas mini-marts (referred to here as gasoline retailers); (2) change in the mean totals for cigarette brands having high billboard expenditures pre-MSA; and (3) the relation of total cigarette advertising to illegal sales of cigarettes to youth.

Data complied by Competitive Media Reporting¹⁹ were used to determine cigarette brands with extensive billboard expenditures pre-MSA (that is, Marlboro, Camel, and Kool). Four teams of one or two youths and one adult were trained according to the same protocol as in 1998 (time 1 protocol),²⁰ and openly collected identical information for the premium brands originally counted—that is, the top five ranked ordered by Massachusetts youth who smoke Marlboro, Newport, Camel, Winston, and Kool,⁶ as well as two additional brands, Parliament and Lucky Strike, with changed ad campaigns during this time period. Cigarette advertisements, which included any incidence of brand or generic advertising, were counted for each of the three distinct areas of exterior advertising: (1) buildings and detached area (parking lots, fences, gas pumps, etc); (2) sidewalks; (3) windows and doors (which included advertisements affixed to the interior glass to be viewed from outside). Each of the seven specified brands of cigarettes was counted independently and any other cigarette brand was recorded in an “all other” category. Parliament and Lucky Strike were later recoded into the “all other” category to conform to the time 1 protocol. Two team members verified all counts and in all but one instance, pictures were taken. Printed information describing the advertising study was available for the merchant upon request, but was not offered in advance. Data entry, checking, and analyses were conducted or supervised by the authors.

We developed the paired sample from the original time 1 sample in a three step process (figs 1 and 2). Firstly, OTCTR data excluding department stores (n =4 2) collected pre-MSA (n = 2798) were cross matched against an archival database of results of underage purchase attempts from independent, representative samples of Massachusetts tobacco vendors collected annually during the same period (October 1997 through September 2000). The sample methodology and protocol for conducting these checks conform to federal monitoring requirements, is consistent over time, and is described in depth elsewhere.²⁰ Briefly, a randomised cluster design with stratification on geographic region and tobacco control funding status was utilised with census of all tobacco vendors located in the (randomly) selected clusters. OTCTRs that had not been subject to a compliance check in the previous three years (n = 1404) were eliminated. For vendors with checks in multiple years, the most recent outcome (sale or no sale) was used. This allowed us to assess the overall relation between retailers' illegal cigarette sales and the average amount of externally visible cigarette advertising.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Decision tree for derivation of time 2 sample.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Decision tree for paired sample time 2.

Secondly, we stratified on whether or not a picture was taken at time 1 and oversampled those without (64% v 50%). Since picture taking was optional in the original study, we were concerned that the communities not taking pictures were perhaps less organised in their tobacco control effort. This produced a sample of 758 establishments to be surveyed at time 2. Thirdly, to minimise the likelihood that changes detected pre- and post-MSA were the result of a change in ownership, franchise or type of store (which might affect the display of advertising and the propensity to sell to underage youth), all identifier information (establishment name, address, type, and franchise) collected at time 2 had to match that recorded at time 1. The resulting convenience sample of 556 tobacco retailers (73% response rate) accounts for approximately 9% of all Massachusetts OTCTRs. Table 1 displays the number and percentages of establishments in Operation Storefront, the derived sample, and the paired sample for each of the five establishment types (convenience, grocery, liquor, pharmacy, and gasoline retailers) as well as the illegal sales rate for each type of establishment from archival data.

Planned comparisons, tested with a series of paired sample t tests conducted in SAS version 8.1, assessed changes in mean number of cigarette advertisements pre- and post-MSA overall, by selected brands, exterior placement, and type of establishment. Spearman's ρ tested the strength of association for dichotomous variables, and χ² tests assessed the comparability of the samples (Operation Storefront, derived sample, matched sample).

RESULTS

The proportions of the five types of establishments as tested by χ² analyses are comparable across all samples.

Changes in cigarette advertising

Table 2 presents the mean number of cigarette advertisements for the matched sample of 556 over-the-counter retail tobacco vendors by brand (Marlboro, Newport, Camel, Winston, Kool) and “all others”, exterior placement (building/detached, sidewalk, windows/doors), and type of establishment. Subtotals are provided for each brand by placement, type by placement (far right hand column), and overall totals for brand, type, and placement at time 1 (pre-MSA) and time 2 (post-MSA). The time 1 to time 2 increase in externally visible cigarette advertising is 9%. Most cigarette advertisements were found on windows (80%) pre- and (79%) post- MSA, but the greatest percentage increase pre- to post-MSA was found for buildings (15%). The increase on buildings of gasoline retailers was over 200% and this type of retailer had the greatest percentage increases overall (56%).

Table 3 summarises the significant results. Externally visible cigarette advertisements on retail establishments have increased pre- to post-MSA for the brand category “all others” and decreased for Winston. These changes are driven primarily by changes on windows. Increases in total cigarette advertising was marginally significant (p = 0.07, data not shown).

Increases from time 1 to time 2 in the overall amount of externally visible cigarette advertisements were revealed for establishments where youth regularly purchase cigarettes—for example, gasoline retailers—and were significant overall, on buildings and windows.

Brands categories with increased exterior advertising on gasoline retailers included “all others”, Camel, Kool, and Newport. Marlboro advertisements significantly increased on the buildings of gasoline retailers and results were suggestive for sidewalks (pavements) as well (p = 0.072, data not shown). Close to threefold increases are found on buildings of gasoline retailers when the three brands with strong billboard presence pre-MSA (Marlboro, Camel, and Kool) are combined; findings were also suggestive that Kool increased advertising pre- to post-MSA (p = 0.068, data not shown). Increases in advertising on windows for gasoline retailers were found for “all others”, Camel, and Newport. All other brands also showed significant increases on convenience store windows.

While we found no significant reductions in total cigarette advertising visible from outside of any type of establishment, significantly fewer Camel ads were visible from outside of liquor stores post- versus pre-MSA, with declines in Kool (p = 0.050) and Winston (p = 0.057) marginally significant (data not shown). The significant declines revealed on the windows of liquor stores were attributable to changes in Winston, Camel, and Kool. Similarly, convenience stores exhibited declines in Winston and Newport advertising with reductions somewhat suggestive for Camel (p = 0.07, data not shown). Significant declines in Winston advertising occurred both on buildings and windows of this establishment type.

Relation of advertising to illegal sales

Overall around 15% of minors who attempted to purchase cigarettes from these establishments were successful, although the rate varied with type of establishment as shown in table 1.

Although the effect is modest, it is evident that average amounts of cigarette advertising and illegal sales rates recorded across the period of this study are correlated (ρ (555) = 0.097, p = 0.022). While none reached significance, within type correlations (with the exception of drug stores) were positive and consistent with the overall results (ρ(61) = −0.137, p = 0.29; ρ(261) = 0.02, p = 0.76; ρ(38) = 0.097, p = 0.56; ρ(83) = 0.101, p = 0.36; ρ(108) = 0.138, p = 0.15.)

DISCUSSION

Results of this study cannot be generalised to all of Massachusetts, as the convenience sample, which formed the basis of the longitudinal study, is not representative and may favour areas with more active tobacco control. This could underestimate any increase in exposure as the more active communities may be more likely to seek voluntary reduction in storefront advertising or enforce community zoning or safety concerns that would affect cigarette advertising as part of all exterior advertising. No interior advertising (other than that affixed to windows and doors and meant to be viewed from outside the store), promotional items, catalogues, displays or functional items were counted in this study and we may therefore be underestimating the amount of youth exposure in Massachusetts. Wakefield and colleagues¹⁸ documented increases in interior as well as exterior exposure pre- and post-MSA.

Longitudinal studies are subject to selection bias. Although no sample differences were detected, the matching criteria we employed may further advantage the more established stores or the more stable neighbourhoods. We believe that the disadvantages in our conservative approach are more than offset by our ability to infer that if exterior tobacco advertising has not decreased among these vendors, it is unlikely that it has decreased in any group of vendors; and that changes over time in exterior retail cigarette advertising are less likely caused by individual differences in establishments than by environmental pressures, either pro- or anti-tobacco control.

While it is true that many of the 27 significant results of the 120 planned paired comparisons would not survive a conservative correction for inflated α error, and chance alone could be expected to produce six significant results, random fluctuations should be distributed across all five types of storefronts and all three external areas. Furthermore, they should be equally likely to produce decreases as increases. This was not the case. We found over four times as many significant results as could be expected by chance alone, and increases (16) outnumbered decreases (11) by a margin of close 3 to 2; this ratio was slightly greater when considering only those with a probability level of p < 0.01 (8 to 5). Although the overall change in total cigarette advertising (9%) is comparable to that found by Wakefield and somewhat suggestive of increases (p = 0.078), and the increase in “all others” reached significance, this masks differences by brands, establishments, and placement which can differentially impact youth exposure. With the exception of Winston, there is no evidence that youth are less exposed to any externally visible cigarette brand advertisement after the MSA than before. Planned comparisons provide evidence that two years after the MSA, youth are still at risk for continued involuntary exposure from external retail cigarette advertising, particularly from gasoline retailers which have high rates of illegal sales and are often frequented by youth.¹⁰ Retailers where youth prefer to purchase cigarettes (gasoline retailers and convenience stores) have increased their cigarette advertising exposure from time 1 to time 2, and the increase is significant and substantial (150%) for gasoline retailers. Changes in brand/category exposure for convenience stores is mixed with increases for “all other”, decreases in Winston and Newport, and suggested increases for Kool.

Although increases in total cigarette advertising were noted on buildings, sidewalks, and windows, results were significant only for the windows and buildings of gasoline retailers which, because of the size of their lots and placement of buildings (often stand alone with three exposed sides), may have more available space than convenience stores for replacement advertising redirected from the billboard campaign. Significant increases overall and on the buildings and windows of gasoline retailers are troubling as this may signal a move towards more expansive and far reaching advertisements with increased involuntary exposure for younger impressionable children²¹ in vehicles, especially when those brands with strong pre-MSA billboard campaigns are combined.

While the design of this study did not allow us to assess changes in size of cigarette advertisements, Wakefield and colleagues noted an 8% increase in vendors with any tobacco advertising post-MSA, but a 22% increase in those with “high” levels of tobacco advertising (five or more ads or at least one ad with a dimension of at least 1 foot (30 cm)); the greatest increase in “high” levels (32%) was noted in parking lots (exclusively found in gasoline retailers either with or without convenience stores).

Furthermore, a national study of youth access to tobacco products found middle school smokers as likely as high school smokers to purchase their cigarettes in gasoline retail establishments, but less likely to purchase in convenience stores.¹⁰ Significant decreases in total cigarette advertising on windows occurred on liquor stores only. Liquor stores have reduced levels of cigarette advertising¹ and lower rates of illegal sales to minors relative to other tobacco vendors.²⁰ There is anecdotal evidence that liquor stores, experienced in age verification for alcohol sales and sometimes staffed by undercover police personnel, may have incorporated age checking for tobacco sales earlier than other establishments. Also, age verification is highly predictive of sales outcomes.²¹ Furthermore, liquor stores in Massachusetts are also less likely to be within 1000 feet (300 m) of schools¹ and presumably less likely then to contribute to involuntary youth exposure.

It is probable that the restrictions on cigarette advertising near schools promulgated by Massachusetts (and the FDA) still influenced merchants during this time period. When one team was kept from taking pictures and left per instructions, another adult upon revisit found that all storefront advertisements had been removed. Upon inquiry, he was told “that the store was now in full compliance with all regulations”. This shows not only the need for unobtrusive surveillance when conducting community based research, but the power of formal regulations and monitoring by a government entity. Unfortunately, any constraint exercised by individual vendors or the tobacco industry while awaiting the final decision of the Supreme Court is now unnecessary.*

Finally, while the effect size is modest, underage youth are more successful in obtaining cigarettes in establishments with greater amounts of cigarette advertisements. The association between amount of advertising and sales to youth is especially intriguing as buyer's age (older), lack of age verification, and self service displays are highly predictive of illegal sales.²⁰ It is possible that dependence on revenue from tobacco sales affects both the amount of advertising visible from the outside of stores, and the effort given to age verification. In related research, Cowling²² found the type (tobacco industry, or tobacco control programme) of age notification sign displayed by tobacco merchants to be associated with differential rates of illegal sales to youth, with those vendors having only the industry sign most likely to sell to underage youth. It is probable that communities in California and Massachusetts with active tobacco control programmes that enforce age related sales laws also encourage the voluntary removal of cigarette advertisements as well as the display of non-industry tobacco control signs; vendors who voluntarily reduce advertisements may also be more careful about selling to youth. Youth who smoke may view stores that display only industry sponsored age requirement signs, or have greater amount of exterior cigarette advertising, as being more tobacco friendly and may then frequent these stores more often. More research is needed to understand this relationship better.

What is evident from our study is that advertising over this 30 month period in a well established sample of tobacco retailers is not static; that brands and establishments and even placement on establishments change. While we applaud any decrease in total cigarette advertising on the windows of retail establishments, the reduction on liquor stores that depend upon and cater to an adult clientele is more than offset by the increases in gasoline retailers that do not. The industry may not be specifically targeting youth, but exposure to storefront cigarette advertising has increased in spite of the MSA, and any residual effects during this collection period of the regulations promulgated by the FDA and the Commonwealth will likely diminish. We are also unsure how the recent withdrawal of Philip Morris advertising from magazines with substantial youth readership will impact upon the amount of externally visible Marlboro advertising, a brand which accounted for almost one third of the observed advertising and is overwhelmingly preferred by youth who smoke.⁶ That cigarette advertising has not decreased post-MSA but has increased so dramatically in gasoline retailers calls into question the effectiveness of the MSA in reducing involuntary youth exposure, and begs that government regulation be strengthened and supported to protect youth from intended or unintended exposure.