Cigarette manufacturers in the USA spend the majority of their marketing and sales promotion dollars in retail stores. These venues have become the main communication channel used by cigarette companies to promote the use of their products.1 Their concentration of resources on the retail environment may be due to the fact that it is relatively free of marketing restrictions, while many traditional advertising methods are no longer available to cigarette companies. Cigarette advertising on television and radio was banned in 1971,2 and the Master Settlement Agreement (MSA) between attorneys general from 46 US states and the tobacco companies eliminated billboard and transit advertising, and restricted sponsored events by tobacco companies in 1998.3

Since the implementation of the MSA in 1999, cigarette companies have increased their overall marketing expenditures and have also allocated a larger share of those expenditures to the retail outlet.4 In 1997, the year prior to the MSA, cigarette companies spent about 66% of their $5.6 billion budget to market their products in retail stores.5 By 2003, total expenditures had more than doubled to $15.1 billion, with $12.9 billion (85.2%) spent on retail stores. In 2005, overall marketing expenditures by cigarette companies decreased to $13.1 billion, but the proportion of money spent in stores rose to 88% of the total. In 2002, the Federal Trade Commission required more detailed information about several expenditure categories, so we now know that most of the marketing expenditures are spent on strategies to reduce cigarette prices at the point of sale. In 2005, 75% was spent on price discounts (eg, off-invoice discounts, buydowns, and voluntary price reductions), and 4.5% was spent on multi-pack discounts (eg, buy two, get one free). In the present work, we refer to these price-discounting strategies as sales promotions.

Between 1999 and 2002, increases in the amount of cigarette advertising on storefronts was documented.6 Also, the proportion of stores with various types of advertising and sales promotions increased.7 8 Increases in the proportion of cigarettes sold with sales promotions were also observed in the same timeframe. However, no studies were found that document changes in the amount and type of advertising and sales promotions from more recent years.

Analyses of tobacco industry documents show that cigarette manufacturers have used advertising to target racial and ethnic minorities.9 10 Several studies conducted in single communities confirm that cigarette advertising in stores has been used to target members of racial and ethnic groups, and lower income populations.11 12 However, no broad scale or longitudinal studies of demographic characteristics of store neighbourhoods were found.

We also found no longitudinal studies of changes in the amount of cigarette advertising and sales promotions in stores, nor any on how the neighbourhood environments around the store might affect changes in these marketing practices. This study was designed to: (1) examine trends in the amount and type of cigarette advertising in California stores that sold cigarettes from 2002–2005, (2) examine how sales promotions have changed over time, and (3) assess neighbourhood influences such as demographics (population density, socioeconomic status, and race/ethnicity) on trends in cigarette marketing in stores.

METHODS

This study used data from the California Tobacco Assessment Study (CTAS), a longitudinal prospective cohort study of California stores that sell tobacco. Between 2002 and 2005, standardised observations of tobacco-related advertising in stores were made annually.

Sample

The baseline sample for the study was based on a previous randomly selected sample of stores that sold cigarettes throughout the state of California.13 The initial store sample was part of a 1997 Stanford University study and the sampling frame consisted of 40 186 stores identified as selling tobacco from a list provided by the California Board of Equalization. Annual sample sizes ranged from 554 for 2002 to 574 in 2005. Each year, stores identified by telephone verification or in the field as “no longer in business” were replaced with new stores randomly selected from the original sampling frame.

Measures

Stores were classified into one of seven types: chain convenience with gas (petrol), chain convenience without gas (petrol), large pharmacy (three or more cash registers), gas (petrol) only (kiosk without interior shopping section), liquor store, corner store or small market (less than three cash registers), and supermarket (at least three cash registers).

Data collectors were trained using a combination of classroom and field training. Information was recorded for all branded cigarette advertising including signs (eg, posters, shelf labels, Perspex-enclosed packs, small tags with member benefit sales), shelving units, portable displays that hold cigarette packs (“displays”) and branded functional items, (eg, trashcans and coin trays). Presence of a sales promotion for each sign or display was noted. A sales promotion was recorded if a price was posted, and noted as a “special price”, eg “special offer”, “cents-off”, “reduced price”, “save xxx”, or if there was an offer of free packs with the purchase of cigarettes, eg, “buy two packs get one free” (usually pre-packaged).

In addition to store observation data, census tract data were obtained by a group specialising in Geographical Information Systems (http://www.census.gov/geo/www/tiger/tiger2k/tgr2000.html). Census data were based on the census tract in which a store was located and included quartile of population density, quartile of proportion of population under 18 years of age, and a composite measure of socioeconomic status (SES; http://www.census.gov/main/www/cen2000.html). The SES variable was created by using principal components analysis of the following standardised variables: median family income, median household income, median housing value, proportion of crowded housing units (1.5 or more persons per room), proportion of population over age 25 without a high school diploma, proportion of population living below the poverty level, and proportion of population unemployed. Additionally, three dichotomous variables were created to indicate whether the proportion of a specific racial/ethnic group in a census tract was above the state-wide proportion of population for the specific group. These variables were: African–American, Asian/Pacific Islander, and Hispanic. For example, the measure high proportion of African–American indicates whether the proportion of African–Americans in the census tract was greater than the state-wide proportion of the population of African–Americans (7%). The population proportions in California were as follows: Asian/Pacific Islander, 13%; Hispanic, 35%. The state race/ethnicity proportions were based 2005 values (http://quickfacts.census.gov/qfd/states/06000.html).

Analysis

Using longitudinal store observation data from 2002 to 2005, trends in the amount and type of tobacco advertising over time were examined. Descriptive statistics were generated for amount and type of advertising over the study period. To assess changes in stores over time, the data were conceptualised as forming a hierarchy where data at each time point was nested within individual stores. Using multilevel modelling techniques, change over time and correlates of change were assessed.

The first set of multilevel models did not adjust for neighbourhood characteristics, thus the sole level one predictor was time point, and there were no level two predictors (data not shown). These models assessed whether there were statistically significant changes over time for the following measures: number of cigarette displays, percentage of interior signs and displays with a sales promotion, and whether or not the store had at least one sales promotion. Because existence of a sales promotion was a dichotomous response variable, a hierarchical generalised linear model was estimated.

To assess growth rates adjusted for neighbourhood characteristics (census variables), the models consisted of two levels: response variable and time point were level one variables, and store and neighbourhood characteristics level two variables. Thus all level two variables were predictors and included store type and neighbourhood characteristics. Adjusted models were estimated for two response variables: the total number of advertisements, and the proportion of all advertisements with a sales promotion. Proportion of advertisements with a sales promotion was used to illustrate changes in the type of advertising and to evaluate whether increases in cigarette company expenditures on price discounting strategies in this time period were uniformly observed in stores. There was minimal clustering of stores within census tracts; 93% of tracts had only one store. In tracts with two or more stores, the store with data for the most time points retained for analysis. If multiple stores had data for the same number of time points, one store was randomly selected. After removing stores clustered within tracts, the analysis sample consisted of 642 stores. Data were available for all four time points for 64% of the sample; 15% had valid store observation data for three time points, leaving approximately 21% of stores with data for two or one time points.

Due to protocol refinements, store type classification of six stores changed over time, so the classification made in the last year of available data was used in the models. Prior to entry into the models, the store type was dummy coded and the most prevalent store type, corner store, was the reference category.

In the adjusted models, level two predictors were added to identify correlates of change or variation in the slope of the level one predictor, time. We did not attempt to model variation in a level one intercept, because the goal of these analyses was to model growth rates, not variation in initial values (mean values at 2002 for total cigarette advertising or mean proportion of materials with a sales promotion). Model parsimony was compromised in part by store type being dummy coded, yielding six control variables at level two. To improve model parsimony, only predictors of growth rate with bivariate p values less than 0.05 were included in the final models. For the store type variable, if one of the six dummy variables was statistically significant, all store type variables were retained. Similarly, if a significant relationship was found for one race/ethnicity, all race/ethnic group variables were retained in the final model.

Growth was modelled as a linear function due to a limited number of data points across time. The combination of data from a maximum of four time points and an unbalanced data set restricted the ability to examine curvilinear growth rates.

Results

Changes in advertising and sales promotions from 2002 to 2005 are presented in table 1.

Overall, the average number of cigarette advertisements increased from 2002 to 2005 (22.7 to 24.9); in an unadjusted model this increase was not statistically significant (p = 0.398). However, there was significant variation in the growth estimate for total number of cigarette advertisements across neighbourhoods, warranting further study of changes over time (variance component estimate for growth rate = 41.5, p<0.001). There has been a steady decrease in the amount of cigarette displays over the study period – from a mean of 1.4 displays in 2002 to a mean of 0.5 in 2005 (p<0.001). The trend data also show an overall increase from 2002 to 2005 in the percentage of stores with at least one advert for a sales promotion (from 68.4% to 79.6% of stores, p<0.001). The percentage of interior signs or displays with a sales promotion per store increased annually from 2002 to 2004 (from 27.5% to 38.8%, p = 0.004), however this value decreased from 2004 to 2005 (31.6%, p<0.001).

Table 2 presents descriptive statistics for total cigarette advertisements by store type from 2002 to 2005 for those stores included in the multilevel analysis. The data suggest that changes in the amount of advertising varied by store type. The greatest increases were observed in supermarkets and convenience stores. However, in petrol only stores and large pharmacies, the amount of advertisements appeared stable. To test variation in changes over time multilevel models were estimated.

Summary of multilevel model for total number of cigarette advertisements

The results of the initial multilevel model, with number of cigarette advertisements as the response variable and time as the sole level one predictor, indicate a non-significant overall growth rate (b = 0.36, p = 0.389). This suggests a non-significant change in the number of advertisements across all stores types. However, the model estimated statistically significant non-zero variance of individual growth rates (variance component estimate 23.9, p<0.001). Thus, based on model estimates, the growth rate varied significantly across stores.

Level two predictors of the growth rate were added to subsequent models, and the final model included the following predictors of growth: store type, population density quartile 4 versus all other quartiles, SES, and a high proportion of African–Americans, Asian/Pacific Islanders and Hispanics (table 3). In this model, the level two intercept for growth rate represented small markets in neighbourhoods of average SES, low to moderate population density, and census tracts without a high proportion of African–Americans, Asian/Pacific Islanders, or Hispanics. No statistically significant growth rate was observed for the reference group, meaning the intercept (representing the growth rate) did not significantly differ from 0. However, the model did estimate significant annual increases in the number of advertisements for three store types: supermarkets, convenience stores without petrol, and liquor stores. Supermarkets had the greatest estimated growth rate, 5.3 items/year (p<0.001). For convenience stores without petrol, the estimated annual increase was 2.8 items, and for liquor stores it was 2.3 items (p<0.001 for both store types). A significant annual decrease of 1.6 advertisements was estimated for the store type petrol only (p = 0.005). The number of advertisements decreased 1.4 items annually for stores in neighbourhoods with the highest population density (p = 0.001). Finally, the number of advertisements in stores in neighbourhoods with an above average proportion of African–Americans increased by an estimated 2.5 items/year (p<0.001).

Multilevel models also were estimated for the proportion of advertisements with a sales promotion (table 4). Significant annual increases in the proportion of advertisements with a sales promotion were observed for all store types with the exception of small markets (p<0.05). The greatest increase was in supermarkets and stores that sold petrol-only, both with an estimated annual increase of four percentage points, followed by convenience with petrol and large pharmacies with 3.5 percentage points and 3.3 percentage points increases respectively. Compared to stores in neighbourhoods with the low to moderate population density, stores in the most densely populated neighbourhoods were estimated to have an annual 1.3 percentage point decrease in the proportion of advertisements with a sales promotion (p<0.039). Additionally, stores in neighbourhoods with a high proportion of African–Americans had an estimated annual 1.0 percentage point increase in the proportion of promotional items (p<0.031). Significant relationships were not identified in neighbourhoods with high proportions of other ethnic/racial groups or between SES and the proportion of items with a sales promotion.

DISCUSSION

This study examines trends in the amount and type of cigarette advertising and sales promotions in California stores selling cigarettes from 2002–2005, and assesses the effects of neighbourhood characteristics on cigarette marketing in stores. The longitudinal data confirm that the pattern of changes in the amount of cigarette advertising roughly matches the direction of changes in cigarette company marketing expenditures in stores over the same time period with increases from 2002 (22.7 advertisements) to 2004 (26.1 advertisements) and a decrease in 2005 (24.9 advertisements). Variation by store type was noted, with significant increases in supermarkets and convenience stores without petrol. Supermarket increases may be attributed to the amount of small tags advertising member benefit sales; these are commonly found in these types of stores but an analysis of small tags was not conducted.

The percentage of stores with at least one advert for a sales promotion increased from 68.4% in 2002 to 79.4% in 2005. Increases in the proportion of interior signs and displays with a sales promotion also were observed over the study period. In California, self-service cigarette displays were legislatively prohibited in 2001 (Business & Professions Code no. 17200) but displays in other locations of the store, eg, behind the counter, continue to be allowed. As also observed in an earlier study of community bans on self-service displays,14 this legislation may have stimulated the decreases in displays observed during the course of the study.

The amount of cigarette advertising and the proportion of cigarette advertising with a sales promotion are rising more rapidly in stores located in neighbourhoods with a higher proportion of African–Americans than the state-wide average. These findings are particularly noteworthy because we controlled for income and store type. Of note, faster increases in the amount of advertising and sales promotions were not found in neighbourhoods with relatively high proportions of Hispanics or Asian/Pacific Islanders. Our findings are consistent with previous studies and add to the concerns about targeting. That advertisements and sales promotions are increasing more rapidly in stores located in neighbourhoods with a higher proportion of African–Americans suggests that the companies may be segmenting the market and promoting lower prices to this subgroup. This finding is similar to studies of billboard advertising prior to the MSA, which contained a provision that eliminated outdoor advertising of tobacco.3 These studies found a disproportionate number of tobacco advertisements on billboards in predominantly African–American neighbourhoods in three US cities.15^–17

Sales promotions such as those studied here have been found to dampen the effect of tax and other price increases on smoking prevalence.18 19 The observed increases in sales promotions as well as the increasing proportion of stores offering them indicate that they are increasingly important to the companies as strategic tools to control prices and advertising in the retail environment. As noted in some of our previous work, cigarette companies tie retail incentives such as sales promotions to the strategic placement of both advertisements and products within the retail environment.20 21

This study has several limitations. First, it was confined to an examination of stores in California so the results may not be generalisable to other locales. For example, our findings on the proportion of stores with a sales promotion are higher than those found in an earlier national study.8 This may be attributed to increases in cigarette company marketing expenditures on price reductions, a difference in protocols, or a difference in California stores. Second, we did not gather advertising data on other types of products so we are unable to determine the relative dominance of tobacco advertising in stores. Third, neighbourhood characteristic data was based on the 2000 US Census. Census tract characteristics were treated as constant across the study period due to availability of data. However, some changes in census tract characteristics might have occurred between the census data collection and the store observation data collection.

Tobacco industry documents reveal that cigarette companies aggressively use retail cigarette advertising and product promotions to reach current and future customers.22^–24 These studies also indicate that cigarette companies are acutely aware of the importance of protecting the use of this venue in increasingly restrictive policy environments. The results of this study confirm an intense concentration of cigarette marketing resources in retail stores.

A case can be made for policies to restrict marketing strategies in the stores, particularly because a relationship between exposure to retail advertising and adolescent smoking has been found. Adolescents reporting frequent exposure to retail tobacco marketing had more positive attitudes and beliefs about smoking,25^–27 and were more likely to have experimented with smoking.28^–30 Teen smokers also preferred the brand most heavily advertised in the convenience store closest to their school.26 Importantly, higher levels of advertising, lower cigarette prices, and greater availability of cigarette promotions predicted smoking uptake among youth.31

Comprehensive restrictions on advertising and sales promotion can significantly reduce cigarette consumption, but partial bans are ineffective.32 To that end, Article 13 of the Framework Convention on Tobacco Control called for comprehensive bans on tobacco advertising and promotions.33 Although not explicitly stated, this would also include point of sale. Several countries including Iceland, Ireland, Thailand, and several Canadian provinces have banned tobacco advertising and product displays at the point of sale.23 34 Additionally, the advertising and promotion ban implemented in the UK in 2003 has significantly reduced exposure to pro-tobacco marketing influences.35 Where not pre-empted by existing law, other countries should follow suit.

In the US, there are serious legal constraints against restriction of tobacco marketing practices in stores. In 2001, the US Supreme Court struck down a Massachusetts regulation to restrict tobacco advertising within 1000 ft of schools because it found that the state regulations were pre-empted by the Federal Cigarette Labeling and Advertising Act (FCLAA; Lorillard Tobacco Co. v. Reilly, 533 US 525 (2001)) FCLAA contains a pre-emption clause that prohibits state and local laws and regulations from imposing any requirements or prohibitions on the advertising or promotion of cigarettes based on smoking and health (15 USC section 1334(b)).2 The removal of this pre-emption should be encouraged so that state and local jurisdictions can control the amount and type of tobacco advertising in their communities (within the parameters of the US First Amendment right to free speech). In addition, consideration should be given to strategies to grant federal control over tobacco industry marketing practices in stores.36 Either of these options could be used to limit the marketing of tobacco products as well as the promotional strategies used to reduce prices.

Given the overall importance of this venue to promote cigarettes, there is a need for periodic monitoring of the amounts and types of cigarette advertising in a nationally representative sample of stores. In addition, further research is warranted to assess the impact of company marketing practices on future and current smokers including effects of advertising exposure and price reductions on those attempting to quit smoking.