OBJECTIVE. Studies of US adolescents have linked exposure to movie smoking with smoking behavior. It is unclear whether European adolescents are also responsive to movie tobacco imagery.
METHODS. A longitudinal study was conducted to assess exposure to movie smoking in 2711 German never-smokers (aged 10 to 16 years). Movie smoking exposure was estimated by asking adolescents if they had seen movies from a list of 50 movie titles, randomly selected for each adolescent from 398 internationally distributed movies released between 1994 and 2004 that became box-office hits in Germany. These films were reviewed for smoking content. Adolescents were resurveyed 12 to 13 months later to determine smoking status, and results were compared with a similarly designed survey of 2603 white US adolescents. We hypothesized replication of the main effect of the exposure on trying smoking, and an interaction, with a significantly larger response among adolescents whose parents did not smoke.
RESULTS. The 398 internationally distributed movies represented 80% of the German box-office hits within this time frame, with the majority (388) produced and/or distributed internationally by US companies. Smoking was present in 74% of the movies. Overall, 503 (19%) of the students tried smoking during the follow-up period. The incidence of trying smoking was associated with increased exposure to movie smoking. The form of the dose-response was similar to the US sample, with the strongest response to movies seen in the lower 2 quartiles of exposure. After controlling for baseline covariates, exposure to movie smoking remained a significant predictor of trying smoking in German adolescents, and the effect was significantly stronger in adolescents whose parents did not smoke.
CONCLUSIONS. Smoking in internationally distributed US movies predicts trying smoking among German adolescents, closely replicating findings from a longitudinal study of white US adolescents. Smoking in these movies could have important worldwide public health implications.
Smoking is responsible for ∼4 million deaths worldwide each year,1 and current global patterns of youth smoking suggest little abatement of cigarette use.2 The World Health Organization Framework Convention on Tobacco Control (WHO FCTC) is an evidence-based treaty based on the recognition that tobacco use is an epidemic spread by a number of complex factors, including “global marketing, trans-national tobacco advertising, promotion and sponsorship.”3 Global marketing affects smoking onset by positioning the behavior as something to which young people aspire.4 Advertising accomplishes this, in part, by showing “aspirational” models smoking.5 Among other elements, the treaty aims to reduce the impact of advertising on smoking prevalence through regulation.
One risk factor for adolescent smoking that is conspicuously absent from the WHO FCTC is smoking in movies.6,7 US media companies (Hollywood studios) produce and/or distribute 80% to 90% of all box-office hits that are viewed in many countries, ∼80% of which contain depictions of smoking.8 Tobacco industry documents indicate awareness of leadership and marketing executives regarding the importance of cinema in communicating strong, positive images for cigarettes and smoking9 to position the behavior as something to which people aspire. American movie stars participated in a contemporary Japanese tobacco advertising campaign,10 and tobacco companies paid for the placement of their products in movies as recently as 1990.9 Hollywood movies continue to include cigarette brands,11 and when these movies are distributed internationally, the brand appearances in them are seen by millions worldwide.12 Finally, the most compelling reason for including movie smoking in the WHO FCTC is the strong relationship that has been demonstrated between exposure to smoking in movies and the adoption of smoking among adolescents.
Smoking in movies has been linked with adolescent smoking in 3 longitudinal US studies,13–15 but to our knowledge, there are no published data on exposure of adolescents outside the US to internationally distributed movies. Our aim was to replicate the results of 1 longitudinal US study of white adolescents13 in a German sample. The US study found that exposure to smoking in movies was associated with trying smoking in a dose-related manner, independent of confounders that included other social influences, personality characteristics of the adolescent, and parenting style. The study found effect modification, such that adolescents with nonsmoking parents were more responsive to movie smoking. Finally, the form of the dose-response was curvilinear, such that adolescents with relatively low levels of exposure were more responsive to movie smoking.16
Replicating methods that were used in the US study,13 we invited 42 randomly selected secondary schools from Schleswig-Holstein, Germany, to participate in a school-based survey. We distributed a self-administrated written survey in 2005 to adolescents (aged 10–16 years) who were enrolled in the 27 schools that agreed to participate,17 and we repeated the survey 1 year later. Trained research staff administered both confidential surveys during class time with parental written permission and student assent. The study was approved by the Ministry of Cultural Affairs of the Bundesland Schleswig-Holstein. We were able to complete follow-up surveys in 2864 (86.4%) of 3313 baseline never-smokers. After exclusion of 153 surveys for missing data on ≥1 variables in the analysis, the final sample consisted of 2711 adolescents.
We assessed lifetime smoking experience by asking, “How many cigarettes have you smoked in your life?” (never smoked, just a few puffs, 1–19 cigarettes, 20–100 cigarettes, >100 cigarettes). The cohort consisted of never-smokers at baseline, and any smoking by the follow-up survey, even just a few puffs, was considered initiation of smoking. This definition is consistent with the definition used for incident smoking in the comparison US study.13
To obtain a population-based estimate of exposure to smoking in international movies that were distributed in Germany, we randomly selected a unique list of 50 movie titles for each adolescent from a sample of 398 popular contemporary movies that were released between 1994 and 2004 in German cinemas. These movies were box-office hits in the United States and Germany as well (see Appendix 1, which lists the names of all 398 movies). German box-office success was determined by the German Federal Film Board (www.ffa.de), a German institution that was incorporated under public law and gives detailed information on the number of cinema visitors per movie per year. We selected the top 25 German box-office hits each year from 1994 to 2001 (172 were internationally distributed movies) and the top 100 German box-office hits from 2002 to 2004 (226 were internationally distributed movies). This list of movies represents 80% of the German box-office hits within this time frame, the majority (388) of which were produced and/or distributed internationally by American companies. The excluded movies were German or European in origin.
Trained coders counted the number of occurrences of smoking in each movie using methods previously described.18 Exposure to these movies was assessed at baseline by asking each student to indicate which films he or she had seen. We calculated exposure to movie smoking for each respondent by summing the number of smoking occurrences for each movie that the respondent reported seeing. We adjusted for variation in the amount of smoking contained in movie lists by expressing individual exposure to movie smoking as a proportion of the total number of possible smoking occurrences on the student movie list. We then multiplied the proportion by 2566 (the number of smoking occurrences in all 398 movies) to obtain an estimate of exposure to the entire sample.
Controlling for Confounding
We controlled for a number of covariates that could confound the relation between exposure to smoking in movies and trying smoking, including characteristics of the adolescent, specific social influences, and parenting style (Appendix 2). The assessment of the covariates mirrored that of the US study, with the exception of controls for socioeconomic status (SES) and tobacco marketing receptivity.13 In the US study, adolescent reports on parent education were used to control for parent SES; in Germany, family SES is uniform within schools, so school was used as a proxy for family SES (the Ministry of Cultural Affairs of Schleswig-Holstein would not allow us to assess parent SES directly from the survey). In the US study, ownership of tobacco branded merchandise was assessed because it was common in 1999, when that survey was conducted; in Germany, the distribution of such merchandise is not common.19 To assess receptivity to tobacco marketing in German adolescents, we asked them whether they had a favorite tobacco advertisement. For the analysis, responses to individual items that measured students' personality (sensation seeking/rebelliousness) and parenting style (maternal responsiveness and demandingness)20 were summed, such that higher scores signified more of each characteristic.
Generalized linear models21 were used to assess smoking initiation as a function of both movie exposure and baseline covariates. We used a log link, rather than a logistic regression, so that the relative risks could be estimated directly.13 We report results with school entered as a cluster variable but found similar results when it was added as a fixed covariate. For the regression analyses, exposure to movie smoking was classified into population quartiles as follows: 0 to 89 occurrences for the first quartile, 90 to 279 for the second quartile, 280 to 580 for the third quartile, and 581 to 2566 for the fourth quartile. For the interaction analysis, we fit linear interaction terms between quartile of exposure and age, gender, friend, sibling, and parent smoking. For the main and interaction analysis, we used fully adjusted models. All reported P values are 2-sided, and P < .05 was considered statistically significant. Movie smoking exposure was used as a continuous variable to construct lowess (locally weighted scatter plot) plots,22 comparing the form of the relationship between exposure to movie smoking and adolescent smoking initiation in the German and US longitudinal samples. The number of box-office hits was larger for the US sample (601 box-office hits in the North American market), so those individuals had higher estimated levels of exposure to movie smoking. To compare the dose-response curves, we standardized movie smoking exposure for each study so that the lowest value was 0 and the highest was 100. Approximately one third of each curve was created by ∼25 outliers who had extremely high exposure to movie smoking, so we trimmed the distributions, recoding movie exposure to the 95th percentile for adolescents above it. For both samples, there were no adolescents whose exposure to movie smoking was 0, so we first subtracted the lowest exposure level from each value so that the lowest level was 0. We then divided the distribution by the highest exposure number and multiplied by 100. For the German sample, the median and interquartile range was 23 (7–48); for the US sample, it was 32 (18–56). All analyses were performed using Stata 9.2 (Stata Corp, College Station, TX).
The final sample of 2711 adolescents had a mean age at baseline of 12.5 years (SD: 1.1). Participants who were lost to follow-up were significantly older, were male, were higher on sensation seeking/rebelliousness, had higher exposure to movie smoking, were more often from lower SES schools, reported lower school performance, and had more friends who smoke.
Table 1 shows covariates and their relationship to movie smoking at baseline. Exposure to movie smoking increased with older age; male gender; lower SES school; smoking by peers, siblings, or parents; poorer school performance, having a favorite tobacco ad; and having higher sensation seeking/rebelliousness. Exposure was lower when parents were perceived as more responsive and demanding. To account for these differences, we adjusted for these variables in the multivariate analysis.
Movie smoking was present in 74% of the 398 movies. Adolescents had seen a mean of 11 (SD: 8) movies, for an estimated mean movie smoking exposure (contained in the 398 movies) of 400 (SD: 407) smoking occurrences. A total of 503 (19%) of participants initiated smoking during the follow-up period, compared with 10% on during 1- to 2-year follow-up for the US sample.13 Adolescent smoking was also significantly associated with school type, sibling and friend smoking, having a favorite tobacco ad, and sensation seeking/rebelliousness. The incidence of trying smoking was associated with increased exposure to movie smoking: 10% in quartile 1, 15% in quartile 2, 21% in quartile 3, and 28% in quartile 4. Exposure to movie smoking remained a significant predictor of adolescent smoking after controlling for all covariates in Table 1, with the adjusted relative risk ranging from 1.36 to 1.96 comparing quartiles 2 and 4, respectively, with quartile 1 (Table 2). Other predictors of smoking onset included sibling and friend smoking, having a favorite tobacco ad, and sensation seeking/rebelliousness. After controlling for covariates, there was no effect of age, parent smoking, school performance, or parenting style.
In the model that tested for an interaction between exposure to movie smoking and parent smoking, adolescents with nonsmoking parents had a significantly greater response to movie smoking (P < .05), as illustrated in Fig 1, which shows adjusted relative risks for movie smoking exposure categories for adolescents with and without parent smoking. Also in parallel with the findings in the US sample, there were no other significant interactions with any of the other variables that we tested, with the exception of a negative interaction between movie smoking exposure and sensation seeking/rebelliousness that was close to significance (P = .06).
Figure 2 shows crude dose-response curves for incidence of trying smoking as a function of baseline exposure to movie smoking for the German and the US samples. For both samples, there was a positive curvilinear association with the response attenuating beyond the 60th percentile for exposure. The curve for the US sample runs below the curve for the German sample, reflecting different overall rate of trying smoking between samples.
In this prospective study of German adolescents, we observed a two-fold increased risk for trying smoking with higher exposure to smoking in internationally distributed American movies, independent of other risk factors for smoking, and replicating the main effect results for another cohort study of US adolescents. Other similarities to the American findings are the form of the dose-response curve, with larger effects for adolescents in the low end of the exposure distribution, and effect modification, such that adolescents with nonsmoking parents were more responsive to movie smoking compared with those whose parents smoke. The prospective design minimizes chances of recall and selection bias, and the high follow-up rate reduces concern that results might be affected by differential follow-up. The large number of controls reduces the likelihood that there is residual confounding; we may have overcontrolled, because we included as covariates some variables that may actually be intermediate variables. For example, in another study, Wills et al23 showed that some of the effects of movie smoking on behavior were mediated through changes in peer smoking status.
The cross-cultural replication of a movie smoking effect on smoking among German children is notable not only for its similarity to the US findings but also because these 2 cultures treat smoking differently, with smoking being more socially acceptable to Germans and tobacco companies enjoying fewer restrictions in Germany. For example, in the United States, indoor smoking laws now cover half the population; in Germany, few towns or jurisdictions ban indoor or workplace smoking. Moreover, in the United States, many advertising venues are limited by the Master Settlement Agreement; Germany allows outdoor, point-of-sale, and cinema advertising and has no limits on indirect advertisements or sponsorship. Consequently, adult and peer smoking prevalences are much higher in Germany. Thus, German adolescents sustain higher exposure to many social influences to smoke, all potential confounders of the movie smoking–adolescent smoking relationship. Higher incidence of tried smoking and higher exposure to these confounders may explain why the adjusted movie effect is ∼50% lower in German adolescents compared with their US counterparts (the reduction in effect size was 0.35, 0.67, and 0.56, respectively, for quartiles 2, 3, and 4 relative risks).13
Despite its strengths, this study is subject to limitations that are inherent in any observational study. The sample of adolescents retained in the study were different from the baseline sample, being generally lower risk for trying smoking; the findings therefore may not generalize to adolescents who have multiple risk factors for smoking and who may be less reactive to movie smoking cues. The most important caution is with respect to a causal interpretation that it is the movie smoking exposure that leads the adolescents to smoke. Risky behaviors cluster in movies, and adolescents with high exposure to movie smoking are also highly exposed to movie alcohol use, movie violence, and depictions of other risky behaviors. Social influence theory suggests that the most simple explanation is that children's attitudes toward smoking become more favorable when they see movie stars smoke; indeed, studies have shown more favorable attitudes toward smoking among never-smoker adolescents with high exposure to movie smoking.24 Social influence theory could also explain why adolescents whose parents do not smoke are more responsive to movie smoking. These adolescents are naive to addicted adult smoking that adolescents whose parents smoke observe every day; we suspect that the naive adolescent would be more attitudinally susceptible to glamorized movie smoking cues. Despite the appeal of a direct social influence explanation, it is also possible that the movie exposure effect is mediated through the adoption of other risk behaviors, such as alcohol use.
The finding that exposure to internationally marketed movies predicts smoking across cultures could have important public health implications. The majority of movies in the German movie market are produced and distributed by US companies; therefore, most movie smoking exposure comes from these movies. If one accepts a causal interpretation, then this suggests that reductions in the prevalence of smoking in US-produced movies could affect youth smoking outside the United States. Given that youth smoking is prevalent almost everywhere,2 this finding also suggests that movie smoking is a risk factor that is worthy of consideration in treaties such as the WHO FCTC. A number of policy considerations could, in theory, affect adolescent exposure to smoking in movies. Specifically, movie rating boards could rate smoking just as they rate violence for public health reasons.
This study showed that exposure to smoking in internationally distributed American movies predicted trying smoking in German adolescents, closely replicating a finding in a sample of white US adolescents and suggesting that exposure to internationally distributed movies is an important risk factor for adolescents who have access to these movies just as they are in the United States.
APPENDIX 1 Listing of the 398 Movies Used in This Study
APPENDIX 2 Covariates and Their Assessment
This study was funded by the Ministry of Health of the Federal Republic of Germany. The content analysis of American movies was funded by CA-77026, CA-94273, and the American Legacy Foundation. The sponsor of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report and in the decision to submit the article for publication.
We thank Dan Nassau and Balvinder Rakhra for coding the movies; Elana Bergamini for supervising the content analysis; Catharina Banneck, Gabriele Banneck, Lars Grabbe, Patrick Kruse, Asja Maass, Christa Panzlaff, Bilge Sayim, Björn Sülter, and Gudrun Wiborg for assistance in conducting the surveys and data preparation; and Susanne Tanski for helpful comments on the manuscript.
- Accepted June 1, 2007.
- Address correspondence to Reiner Hanewinkel, PhD, Institute for Therapy and Health Research, Düsternbrooker Weg 2, 24105 Kiel, Germany. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Shibuya K, Ciecierski C, Guindon E, Bettcher DW, Evans DB, Murray CJ. WHO Framework Convention on Tobacco Control: development of an evidence based global public health treaty. BMJ.2003;327 :154– 157
- ↵Lynch BS, Bonnie RJ. Growing Up Tobacco Free: Preventing Nicotine Addiction in Children and Youths—A Report of the Institute of Medicine. Washington, DC: National Academy Press; 1994
- ↵US Department of Health and Human Services. Preventing Tobacco Use Among Young People: A Report of the Surgeon General. Atlanta, GA: Office on Smoking and Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, US Department of Health and Human Services; 1994
- ↵Charlesworth A, Glantz SA. Smoking in the movies increases adolescent smoking: a review. Pediatrics.2005;116 :1516– 1528
- ↵Worth K, Tanski S, Sargent JD. Trends in Top Box Office Movie Tobacco Use 1996–2004. Washington, DC: American Legacy Foundation; 2006
- ↵Mekemson C, Glantz SA. How the tobacco industry built its relationship with Hollywood. Tob Control.2002;11(suppl 1) :I81– I91
- ↵Lambert A, Sargent JD, Glantz SA, Ling PM. How Philip Morris unlocked the Japanese cigarette market: lessons for global tobacco control. Tob Control.2004;13 :379– 387
- ↵Jackson C, Henriksen L, Foshee VA. The Authoritative Parenting Index: predicting health risk behaviors among children and adolescents. Health Educ Behav.1998;25 :319– 337
- ↵McCullagh P, Nelder JA. Generalized Linear Models. 2nd ed. London, England: Chapman and Hall; 1989
- Copyright © 2008 by the American Academy of Pediatrics