King III C, Siegel M, Celebucki C, Connolly GN. Adolescent Exposure to Cigarette Advertising in MagazinesAn Evaluation of Brand-Specific Advertising in Relation to Youth Readership. JAMA. 1998;279(7):516-520. doi:10.1001/jama.279.7.516
From the Graduate School of Business Administration, Harvard University, Boston, Mass (Dr King); Social and Behavioral Sciences Department, Boston University School of Public Health (Dr Siegel); and the Massachusetts Tobacco Control Program, Massachusetts Department of Public Health, Boston (Drs Celebucki and Connolly).
Context.— Understanding the relationship between cigarette advertising and youth
smoking is essential to develop effective interventions. Magazine advertising
accounts for nearly half of all cigarette advertising expenditures.
Objective.— To investigate whether cigarette brands popular among adolescent smokers
are more likely than adult brands to advertise in magazines with high adolescent
Design.— Cross-sectional analysis of 1994 data on (1) the presence of advertising
by 12 cigarette brands in a sample of 39 popular US magazines; and (2) the
youth (ages 12-17 years), young adult (ages 18-24 years), and total readership
for each magazine.
Main Outcome Measures.— The presence or absence of advertising in each of the 39 magazines in
1994 for each of the 12 cigarette brands.
Results.— After controlling for total magazine readership, the percentage of young
adult readers, advertising costs and expenditures, and magazine demographics,
youth cigarette brands (those smoked by more than 2.5% of 10- to 15-year-old
smokers in 1993) were more likely than adult brands to advertise in magazines
with a higher percentage of youth readers. Holding all other variables constant
at their sample means, the estimated probability of an adult brand advertising
in a magazine decreased over the observed range of youth readership from 0.73
(95% confidence interval [CI], 0.50-0.96) for magazines with 4% youth readers
to 0.18 (95% CI, 0.00-0.47) for magazines with 34% youth readers. In contrast,
the estimated probability of a youth brand advertising in a magazine increased
from 0.32 (95% CI, 0.00-0.65) at 4% youth readership to 0.92 (95% CI, 0.67-1.00)
at 34% youth readership.
Conclusion.— Cigarette brands popular among young adolescents are more likely than
adult brands to advertise in magazines with high youth readerships.
AT THE HEART of the public health debate about interventions to reduce
teenage smoking lies the question of whether cigarette advertising influences
youth.1 Of all the media by which cigarettes
are advertised—newspapers, magazines, billboards, and mass transit—magazines
receive the largest share of tobacco company expenditures.2
In 1994, the tobacco industry spent $252 million, or 46% of its total cigarette
advertising expenditures, on magazine advertising.2
Although the Food and Drug Administration's tobacco regulations3 and the proposed global tobacco settlement4 address magazine advertising, the specific impact of
cigarette advertising in magazines on youth smoking behavior has not been
well studied. Previous studies have provided indirect evidence that cigarette
advertising in magazines targets youth readers. This evidence derives primarily
from studies of differences in the number or proportion of cigarette advertisements
in youth-oriented compared with adult-oriented magazines,5- 12
changes in the number of advertisements in youth-oriented magazines over time,5,7- 9,13
demonstrations of the appeal of magazines' cigarette advertisement themes
and images to youth,14- 16
and differences in the themes and images used in cigarette advertisements
in youth-oriented and adult-oriented magazines.7,8,12,17
Two methodological problems limit the ability of the existing studies
to draw definitive conclusions. First, since most youth-oriented magazines
have many more adult than youth readers, these studies cannot exclude the
possibility that cigarette advertisements in these magazines target adult,
rather than youth, readers. Cigarette advertisements in these magazines may
be targeting young adult readers (18- to 24-year-olds), rather than those
younger than 18 years. The tobacco industry itself has defended its advertising
practices on the grounds that its advertising is reaching young adults, rather
Second, these studies only examined aggregate cigarette advertising.
Youth tend to smoke only a few cigarette brands.3
By aggregating all brands, even those smoked almost exclusively by adults,
previous analyses may have reduced their power to detect a significant association
between cigarette advertising and youth readership.
In this study, we examine whether brands smoked by a significant number
of adolescents are more likely to advertise in magazines with higher youth
readerships than cigarette brands smoked almost exclusively by adults. The
analysis addresses the major limitations of previous research by (1) using
data on adult and youth magazine readership as continuous variables instead
of classifying magazines as either adult or youth; (2) controlling for adult
readership and young adult (ages 18-24 years) readership in the analysis;
(3) using brand-specific, rather than aggregate, cigarette advertising data;
and (4) comparing brands smoked by a substantial proportion of youths with
those smoked almost exclusively by adults.
Although a finding that youth cigarette brands are more likely to advertise
in magazines with more youth readers does not demonstrate an intent on the
part of cigarette advertisers to expose adolescents to their advertising messages,
such a finding would demonstrate that adolescents are more likely to be exposed
to advertising by cigarette brands that are popular among youth smokers.
Using a probit model, we analyze whether—controlling for the other
factors that might affect a cigarette brand's magazine advertising—youth
brands are more likely than adult brands to advertise in magazines with a
high percentage of youth readers. We also compare the effect of different
demographic readership characteristics on the likelihood that a cigarette
brand advertises in a magazine.
Since only the outcome of the advertising decision is observed, the
empirical specification uses a binary choice model of advertising behavior.
The dependent variable is the presence or absence of advertising for a specific
brand in a given magazine in 1994. The predictor variables in our model include
the following: (1) the demographic characteristics of a magazine's readership,
including the total number of readers (ages 12 years and older) and the percentage
of readers in various demographic subgroups (youth [ages 12-17 years], young
adults [ages 18-24 years], females, blacks, smokers, heavy smokers, and menthol
smokers); (2) the cost per reader of placing an advertisement in a given magazine
(the cost of a full-page, 4-color advertisement divided by the total number
of magazine readers); (3) the number of annual magazine issues; (4) the total
magazine advertising expenditures of a cigarette brand; (5) the popularity
of the magazine (the percentage of readers who consider that magazine to be
their favorite); and (6) the median income of magazine readers.
For each cigarette brand, we determine whether that brand advertised
in each of the 39 magazines in our sample. We create a record for each of
these brand-magazine pairs. Since there are 12 brands and 39 magazines, the
data set comprises 468 records. For each record, the dependent variable is
1 if the cigarette brand advertised in that magazine and 0 if the brand did
not advertise in that magazine.
To assess possible differences in the advertising behavior of adult
and youth cigarette brands, we constructed an indicator variable, δ,
that is 0 for adult brands and 1 for youth brands and created an additional
series of regressors by multiplying each explanatory variable by δ.
These interaction variables allowed us to estimate separate regression coefficients
for youth and adult brands. For example, the interaction variable for youth
readership is defined as δ × (% youth readers). This interaction variable enabled us to measure differences in the
advertising patterns for adult and youth brands with respect to the level
of youth readership in magazines in which they advertise.
In our complete probit model, the probability, P,
that a given brand advertises in a particular magazine is P = Φ (y*), where Φ is the cumulative distribution function
for the standard normal and
y* = A + Aiδ + (B + Biδ) ×
(% Youth Readers) + (C + Ciδ) × (Total Number of Readers)
+ (D + Diδ) × (Advertising Cost per Reader) + (E +
Eiδ) × (Total Advertising Expenditures for Brand Among
All 39 Magazines) + (F + Fiδ) × (Number of Annual Issues
of Magazine) + (G + Giδ) × (% Young Readers) + (H +
Hi δ) × (% Female Readers) + (I + Ii δ)
× (% Black Readers) + (J + Ji δ) × (% Hispanic
Readers) + (K + Ki δ) × (% Smokers) + (L + Li δ) × (% Heavy Smokers) +(M + Mi δ) ×
(% Menthol Smokers) +(N + Ni δ) × (% Favorite Magazine)
+(O + Oi δ) × (Income) + Error.
Here δ = 1 for youth brands and δ = 0 for adult brands.
By including both a variable and its interaction term in the regression
specification, we can determine whether differences in the probability that
adult and youth brands advertise in a magazine are statistically significant
for each independent variable in the model. For example, the coefficient B reflects the change in likelihood of advertising as youth
readership increases for adult cigarette brands, and the coefficient (B + Bi)reflects the change in likelihood of
advertising as youth readership increases for youth brands. Under the null
hypothesis—that the probability of a cigarette brand advertising in
a magazine is unrelated to the magazine's youth readership—both the
coefficients B and Bi
would be 0. If adult brands, but not youth brands, were more likely to advertise
in magazines with higher youth readership, then the coefficient B would be positive and the coefficient (B
+ Bi) would be 0 (Bi would be negative and equal in magnitude to B ). If youth brands, but not adult brands, were more likely to advertise
in magazines with higher youth readership, then the coefficient B would be 0, but the coefficient (B + Bi) would be positive (Bi would be positive).
The statistical significance of the coefficient Bi allows us to assess the significance of any difference between
adult and youth brands in the likelihood of advertising in magazines at varying
levels of youth readership.
To select a sample of magazines, we identified the 60 national magazines
with the highest overall readership for 1994 using data from Simmons Market
Research Bureau, Inc.19 Of these 60 magazines,
we included in the sample only those for which 1994 information on adult and
youth readership and brand-specific cigarette advertising was available. Ten
magazines were excluded because these data were unavailable. An additional
10 magazines were excluded because, as a policy, they did not accept tobacco
advertising in 1994. One magazine was excluded because it contained cigarette
advertisements for only 1 brand in 1994, and the other magazines in which
that brand advertised were not among the 60 in our study. The final sample
consisted of 39 magazines (Table 1).
From the Leading National Advertisers Brand Detail
Report for 1994, we determined whether each cigarette brand advertised
in each of the 39 magazines in 1994 and estimated each brand's total expenditures
for advertising in the 39 magazines in 1994.20
These estimates of advertising expenditures are based on the number of pages
of advertising and the price per advertising page for the magazine, not on
actual dollars negotiated with a publisher.
We used the SRDS Consumer Magazine Advertising Source to obtain the cost for a single, full-page, 4-color advertisement
in each magazine in 1994 and the annual number of issues for each magazine.21
Data on the adult (ages 18 years and older) readership for each magazine
were obtained from the 1994 Study of Media and Markets,19,22,23 produced
by the Simmons Market Research Bureau, Inc. From the Simmons data, we also
collected the following demographic information about adult readers for each
magazine: median individual income; percentage of female, black, Hispanic,
and young adult (ages 18-24 years) readers; percentage of readers who are
smokers, heavy smokers (≥30 cigarettes per day), and smokers of menthol
brands; and percentage of readers who reported a magazine to be their favorite.
Data on the number of youth (ages 12-17 years) readers for each magazine
were obtained from the 1994 Simmons Teen Age Research Study
(STARS),24 produced by the Simmons Market
Research Bureau, Inc, and the Mediamark Research Inc (MRI)
Twelve Plus report,25 produced by MRI.
The data were extracted from the above publications and entered into
an Excel spreadsheet. Data entry was 100% verified by comparing printouts
of the spreadsheet with the data in each publication. After verification,
a SAS data set was created by converting the Excel spreadsheet using DBMS/COPY.26 We used SAS27 and Stata28 to conduct all analyses.
In classifying adult and youth cigarette brands, we used data from the
national Teenage Attitudes and Practices Survey-II (TAPS-II).18,29
Using data obtained from 70 smokers in a cross-sectional, probability sample
of 4992 youths between the ages of 10 and 15 years, we divided cigarette brands
into 2 groups: those smoked almost exclusively by adults ("adult" brands)
and those smoked by a substantial proportion of adolescent smokers ("youth"
brands). Although there were only 70 smokers in our sample, the classification
of adult and youth cigarette brands was identical to that obtained using the
full sample of 438 smokers aged 12 through 17 years.
Because the TAPS-II survey did not record the name of every cigarette
brand, the usual brand smoked for 2.5% of 10- to 15-year-old smokers was reported
as "other." We therefore defined youth brands as those smoked by at least
2.5% of smokers aged 10 to 15 years in TAPS-II and adult brands as the usual
brand smoked by less than 2.5% of 10- to 15-year-old smokers in TAPS-II. Based
on these criteria, we classified 7 brands as adult brands (Salem [smoked by
0.6% of youth smokers], Virginia Slims [<2.5%], Benson & Hedges [<2.5%],
Parliament [<2.5%], Merit [<2.5%], Capri [<2.5%], and Kent [<2.5%])
and 5 brands as youth brands (Marlboro [42.9%], Newport [24.6%], Camel [13.2%],
Kool [4.1%], and Winston [2.8%]). Two generic brands (Basic and Doral) were
classified as unknown and were excluded from analyses that compared adult
and youth brands because we could not determine whether they were smoked by
2.5% or more of 10- to 15-year-old smokers (TAPS-II did not record the specific
names of generic cigarette brands).
Since 1994 youth market share data were not available, the brand market
share data were obtained from a 1993 survey. It is unlikely that changes in
brand use among youth smokers from 1993 to 1994 would have been large enough
to change the classification of brands as adult or youth brands in this study.
Moreover, using 1993 youth market shares and then examining brand advertising
behavior in 1994 alleviates the potential problem of advertising simultaneously
affecting youth market share.
The 39 magazines in this study accepted $4.1 billion in total advertising
in 1994, of which tobacco advertisements accounted for $232.0 million (5.7%)
(Table 1) and cigarette advertisements
accounted for $207.1 million (5.1%). There were 51579 pages of total advertising
in these magazines, of which 2737 (5.3%) were cigarette advertisements. These
cigarette advertisements represented 2085 separate insertions.
Youth readership ranged from 674000 (Entertainment
Weekly) to 6.7 million (TV Guide), and the
proportion of total readership made up of youths ranged from 4.2% (Family Circle) to 33.8% (Sport) (Table 1).
Four variables—the total advertising expenditures of a brand,
the annual number of magazine issues, the percentage of readers who consider
a magazine their favorite, and the percentage of youth readers—were
found to affect significantly the probability that a cigarette brand would
advertise in a given magazine. Of all the demographic magazine readership
variables examined, only the percentage of youth readers was a significant
predictor of whether or not cigarette brands were advertised in a given magazine
The coefficient for the youth readership interaction variable was statistically
significant, indicating that the relationship between advertising and youth
readership differed for youth and adult brands (Table 2). The probability of advertising in a magazine decreased
with the percentage of youth readers for adult brands but increased significantly
with the percentage of youth readers for youth brands. In other words, adult
brands were increasingly less likely to advertise in magazines as the percentage
of youth readers increased, and youth brands were increasingly more likely
to advertise in magazines as the percentage of youth readers increased.
Holding all other variables constant at their sample means, the probability
of an adult brand advertising in a magazine decreased from 0.73 (95% confidence
interval [CI], 0.50-0.96) at a youth readership level of 4% (the lowest level
in the sample magazines) to 0.58 (95% CI, 0.48-0.68) at a youth readership
of 12% (the mean level for all magazines) to 0.18 (95% CI, 0.00-0.47) at a
youth readership level of 34% (the highest level in the sample magazines)
(Figure 1). In contrast, the probability
of a youth brand advertising in a magazine increased from 0.32 (95% CI, 0.00-0.65)
at a youth readership level of 4% to 0.51 (95% CI, 0.38-0.63) at a youth readership
level of 12% to 0.92 (95% CI, 0.67-1.00) at a youth readership level of 34%.
The ratio of the probability of advertising for a youth brand compared
to an adult brand increased with increasing youth readership (Figure 2). At a youth readership level of 14% with all other variables
evaluated at their mean values, the ratio of advertising probabilities was
1.04 (95% CI, 1.03-1.04), indicating that youth and adult brands were about
equally likely to advertise in these magazines. At a youth readership level
of 4%, the ratio of probabilities was 0.43 (95% CI, 0.29-0.58), indicating
that youth brands were about half as likely to advertise in these magazines.
At a youth readership level of 34%, the ratio was 5.21 (95% CI, 4.87-5.54),
indicating that youth brands were about 5 times more likely to advertise in
To the best of our knowledge, this article is the first to examine systematically
the relationship between cigarette brand–specific advertising and youth
readership among a large, nearly complete, sample of the most highly read
magazines over a full year. This is also the first study, to our knowledge,
of cigarette advertising in magazines that compares advertising patterns for
brands smoked by young adolescents with those smoked almost exclusively by
adults. We found that youth brands were more likely than adult brands to advertise
in magazines with a higher percentage of youth (ages 12-17 years) readers.
Although young adult readership is a potential confounder of the observed
relationship between advertising and youth readership in previous studies,
our analysis controlled for the effects of young adult readership on the likelihood
of cigarette brand advertising in magazines. Both adult and youth brands were
more likely (although not significantly) to advertise in magazines as young
adult readership increased, but even after controlling for this effect, youth
brands were still significantly more likely than adult brands to advertise
in magazines with higher youth readership. The magnitude of the effect of
youth readership was also greater than that observed for young adult readership.
The percentage of youth magazine readers was the only demographic variable
that was significantly related to cigarette advertising in magazines.
There are several important limitations to this study. First, although
it is unlikely that our findings are explained entirely by the hypothesis
that cigarette brand advertising is related to young adult, rather than adolescent,
readership, it is probable that some of the observed effect of youth readership
may arise from an overlapping effect of young adult readership. The presence
of minor multicolinearity is suggested by the simple, pair-wise correlation
of youth readers and young adult readers (r = +0.68).
A potential consequence of such multicolinearity would be to diminish the
precision with which the coefficients are estimated.
Nevertheless, in the presence of this multicolinearity, our finding
that cigarette advertising was significantly related to youth, but not young
adult, readership strengthens our conclusion. If the tobacco industry, as
it claims, were only attempting to reach 18- to 24-year-olds, one would expect
the relationship between advertising and young adult readership to be stronger
than that for advertising and youth readership.
Second, the study does not allow us to make inferences regarding the
potential role of cigarette advertising in magazines on smoking behavior,
including smoking initiation, among adolescents. Third, it is impossible to
demonstrate an intent to target youth from an analytic study such as this
one, even though the data indicate a brand-specific relationship between advertising
and youth readership.
Despite these limitations, our findings provide new evidence that cigarette
advertising in magazines is correlated with youth readership, and that this
relationship is different for youth and adult cigarette brands. Youths are
more heavily exposed to magazine cigarette advertisements for brands that
are popular among youth smokers than for brands smoked almost exclusively
This finding has important public health policy implications. By adding
to the evidence that cigarette advertising in magazines is related to youth
readership, the results of this study strengthen the justification for regulating
cigarette advertising in magazines. Based on the documentation in this and
other studies5- 17
of widespread and heavy exposure of youths to cigarette advertising in magazines,
public health considerations argue that cigarette advertising in all magazines
should be eliminated.