Portrayal of tobacco use in prime-time TV dramas: trends and associations with adult cigarette consumption—USA, 1955–2010
- Correspondence to Dr Patrick E Jamieson, Annenberg Public Policy Center, University of Pennsylvania, 202 S. 36th St., Philadelphia, PA 19104, USA;
- Received 14 November 2012
- Revised 23 September 2013
- Accepted 30 December 2013
- Published Online First 3 April 2014
Objective Although portrayal of television (TV) and movie tobacco use has been linked with initiation of cigarette smoking in adolescents, its association with smoking in adults has not been assessed. Therefore, we examined long-term and annual changes in tobacco portrayal in popular US TV dramas and their associations with comparable trends in national adult cigarette consumption.
Methods Tobacco use in 1838 h of popular US TV dramas was coded from 1955–2010. The long-term trend and annual deviations from trend were studied in relation to comparable trends in adult per capita cigarette consumption using correlational and time-series methods that controlled for other potential predictors.
Results TV tobacco portrayal has trended downward since 1955 in line with the historical trend in cigarette consumption. Controlling for changes in cigarette prices and other factors, annual changes of one tobacco instance per episode hour across 2 years of programming were associated with annual change of 38.5 cigarettes per US adult. The decline in TV tobacco portrayal was associated with nearly half the effect of increases in cigarette prices over the study period.
Conclusions The correlation between tobacco portrayal in TV dramas and adult cigarette consumption is consistent with well-established effects of exposure to tobacco cues that create craving for cigarettes in adult smokers. Although tobacco use in TV dramas along with movies has declined over time, portrayal of smoking on screen media should be a focus for future adult tobacco control research and policy.
Tobacco use is the leading cause of preventable mortality in the USA, resulting in about 443 000 deaths annually.1 Exposure to tobacco use in movies and on television (TV) has been identified as a potential causal contributor to adolescent smoking initiation.2 ,3 However, evidence regarding adults is largely absent. There is suggestive evidence that long-term trends in tobacco use in popular movies closely track per capita adult cigarette consumption from 1950–2006.1 ,4 Less clear is whether this pattern is a consequence of societal factors that influenced cigarette consumption and media portrayal or an effect of movie portrayal on smoking.
One way to study the potential long-term influence of tobacco portrayal on screen media is to evaluate its presence on television (TV). TV has been popular in the USA with 65% of households having at least one TV set in 1955 and 93% by 1965.5 The 4 h 51 min per day the average American household spent watching TV in 1955 grew steadily to 8 h 14 min by 2006,5 an escalation in potential exposure to tobacco content reaching more people than movies. For example, even when watching movies on TV is combined with movie theatre viewing, it is measured in hours per week, while TV viewing is captured in hours per day.6 Since exposure to TV, tobacco has been linked with adolescent initiation and amount of smoking,7 ,8 we evaluated its potential influence on adult smoking.
Observational learning of novel and potentially exciting adult behaviours can explain media influence on adolescent smoking initiation.9 But adults are more likely to have initiated smoking than adolescents.1 Hence, we expect portrayal would reinforce an existing adult smoking habit by eliciting craving for cigarettes.10 Exposure to smoking cues in scenes of people smoking has predicted increased craving11 and may hinder smokers from quitting.12 For example, tests using EEG (electroencephalography), a method of recording electrical activity in the brain, have found that cigarette-related stimuli capture the attention of smokers who are trying to quit, and may cause relapse.13 Suggestive evidence of this cue-reactivity effect comes from Sargent et al,14 who found that adult smokers who had watched a movie with smoking scenes reported increased postexposure craving for cigarettes. Hence, exposure to televised tobacco use may elicit craving in smokers that could reinforce their smoking or encourage relapse among those attempting to quit.
The TV tobacco portrayal trend has not been studied over a sufficiently long period to determine its true direction. Breed and DeFoe found that the trend in TV tobacco declined in the decades leading to 1982.15 Hazan et al's analysis suggested that from 1981 to 1993, TV smoking portrayal may have risen.16 The Coding of Health and Media Project (CHAMP) helps to fill this void by analysing tobacco portrayal on popular prime-time TV shows from 1955 to 2010 (see http://www.youthmedirisk.org/). This database provides a sensitive measure of long-term trends in TV tobacco portrayal comparable with the period covered in our previous analysis of movies.4
The CHAMP analysis for TV can also provide a more sensitive measure of year-to-year fluctuations in mass media tobacco portrayal than its analysis of movies, which only sampled 15 of the top 30 grossing films per year. Furthermore, the greater exposure to TV in US households makes it more likely that such annual changes will be related to aggregate measures of adult cigarette consumption. Thus, if exposure to smoking cues on TV influenced cigarette use, annual changes in TV tobacco portrayal should correlate with annual changes in consumption independent of changes in cigarette prices, an important historical predictor.17 This hypothesis was tested using annual tobacco portrayal on popular prime-time TV shows since 1955.
TV sample selection
The study's TV sample was selected from Nielsen's annual top 30 most popular prime-time drama network television shows (see http://www.nielsen.com/us/en.html). The number of hours of TV that were selected for coding varied by year depending on content availability. In each season, odd-numbered episodes (eg, 1st, 3rd, 5th) were sampled. If fewer than six episodes were available for purchase for a season, every available episode was coded. Because TV seasons span two calendar years, we combined shows that ran in the spring and fall of one calendar year. The year 1955 was chosen for the beginning of the study because national cigarette prices were not available prior to that year.18 IRB approval was not sought for this study because it did not involve human subjects.
We selected the drama genre because of its popularity, continuity from 1955 to 2010, and availability. Dramas spanned several overlapping subgenres, including police, detective, crime, legal and medical content, with medical and legal shows more frequent from the 1990s to the 2000s. From 1955 until the 1970s, the sample also included westerns (see http://youthmediarisk.org).
TV sample viewership
Examination of Nielsen household viewership shares, which estimate the percentage of TV households tuned to a programme,19 indicates that the dramas in our sample reflected a popular part of the prime-time TV landscape. Because household TV penetration is available for every fifth year since 1955,5 we calculated shares for those years from 1955 to 2005 as follows: (sum of Nielsen shares for TV shows in year)×(% of households with TV in the same year). These scores estimated the total shares of households that were exposed to shows in our sample. The shows had values of 22.6 in 1955, 19.7 in 1965, 43.0 in 1975, 63.5 in 1985, 58.4 in 1995 and 106.0 in 2005. Shares increased over time because more shows were available for coding in later years, and the proportion of households with TV increased and indicate that sizeable proportions of households were exposed to the shows each week of the fall and spring seasons (without reruns).
Twenty undergraduate students trained by mastering a codebook. About 21 h of film content were used for training (see http://youthmediarisk.org). Coders were required to exhibit a high level of reliability using Krippendorff's α formula (Kα > 0.80).20 This formula controls for chance agreement across all coders.
The typical coding unit was a 4 min segment based on dividing half-hour episodes minus advertising (about 20 mins) into five equal segments. In order to make hour-long episode coding segments comparable to the more dominant half-hour shows, we divided hour-long episodes minus advertising (about 45 min) into 10 equal segments, yielding a 4.5 min coding segment. We converted all tobacco instance rates to a time-based, rather than segment-based metric (ie, instances per hour). The coded sample included 23 350 segments from 2521 episodes in 76 shows for an average of 32.8 advertising-free hours of programming per year (STD=22.0). In terms of total airtime, the sample represented 73.9 advertising-free hours of shows per year (STD=45.7).
Tobacco incidents were recorded for each character's tobacco use per segment (0 vs 1). Characters were considered to exhibit tobacco use when they purchased, handled, or smoked a tobacco product including cigarettes, cigars, pipes, or chewing tobacco. Tobacco use was counted even if the act of inhaling cigarettes or using smokeless tobacco was not shown. The total number of characters using tobacco in a segment constituted the instance measure, a score that ranged from 0 to 4. TV characters were identified using either the DVD packaging or the TV episode guide in http://www.imdb.com. Characters that were not part of the plot, had no speaking role, were not named, and were not the focus of the scene were not coded. Characters could include non-human and animated actors.
Coders met a high level of reliability in identifying tobacco use incidents (Kα=0.85). The sample included 2522 tobacco use instances; over 94% of segments with tobacco use portrayed cigarettes or cigars. The study's measure of average annual TV tobacco use incidents was calculated by taking the ratio of the sum of incidents per episode divided by the number of TV hours of programming in the episode without advertising and averaged over all episodes per year from 1955 to 2010. This produced a mean rate of 1.5 incidents per episode hour across the 56 years (median=1.2, and STD=1.3). A split-half analysis of programming per year found tobacco content was highly consistent within each year of programming from 1955 to 2010, (r=0.79, p<0.001).
The US annual adult (ages 18+) per capita cigarette consumption from 1955 to 2010 was taken from Centers for Disease Control and Prevention (CDC), and from the tobacco industry and industry advisors.21 ,22 Consumption started at 3597 in 1955, rose to 4345 in 1963, and dropped to 1278 in 2010 (mean=3210.6, STD=967.1).
US annual weighted average cigarette prices were taken from the Tax Burden on Tobacco from 1955 to 201018 and adjusted for inflation using 1982–1984 prices (1982–1984=100 cents) using the historical annual Consumer Price Index, Unadjusted (CPI-U). These prices ranged from 86.6 cents in 1955 to 273.1 cents in 2010 (mean=125.7, STD=52.8).
TV and radio cigarette ad ban (1971)
To assess the potential impact of the removal of TV cigarette advertising in 1971, we contrasted the preban (1955–1970=0) versus postban (1971–2010=1) period. This controlled for the drop in advertising expenditures since 1971.23
SPSS 20.0 was used for descriptive statistics and curve fitting. The STATA 12.0 ARIMA program was used to analyse relations between time series for tobacco portrayal and consumption from 1955 to 2010 using autoregressive models.24 These models used full information maximum likelihood with robust SEs in order to protect against heteroscedasticity. Two-tailed significance tests were conducted with 95% CIs.
Before analysing annual changes in TV tobacco incidents and price as predictors of tobacco consumption, we estimated their best-fitting linear to quintic trend components. The best-fitting trend was identified by the significance test (p<0.05) for each increasing component and by visual inspection of the detrended series. Significant trends in TV tobacco, price and consumption were removed resulting in a stationary series for each variable.24
We anticipated that TV tobacco and price would be contemporaneous predictors of consumption. However, we also evaluated models with one and two lags in predictors. A 1-year lag in TV instances might be predictive because the fall season of the prior year would be shown as reruns in the off-season of the subsequent year, and there could also be a lag in consumption due to warehouse withdrawals reflecting prior retail demand for cigarettes. We could not locate annual personal disposable income prior to 1964. However, after being detrended and evaluated it was omitted from the final model because it did not predict post-1963 (b=0.031, 95% CI=−0.008, 0.070, p=0.121).
The raw trends from 1955 to 2010 in TV tobacco instances and cigarette consumption were highly correlated, r=0.658, p<0.001 (figure 1). Tobacco instances ranged from 2.89 every episode hour in 1955–1964 to 0.31 instances in 2001–2010. However, the peak in tobacco instances occurred in 1961 at 4.96 per hour. The best-fitting trend in cigarette consumption included a cubic component, which reflected a slight rise at the end of the series (adj. r2=0.990). Cigarette consumption rose from 1955 to 1963 and then declined. The best-fitting trend in TV tobacco instances also declined from the early 60s with a cubic component reflecting a rise at the beginning and end of the series (adj. r2=0.375). After removing these trends, TV tobacco residuals remained correlated with cigarette consumption residuals (r=0.397, p=0.002). As seen in figure 2, detrended TV tobacco tracked detrended raw consumption evidence of dissociation after 1995.
The results of the autoregressive model (see table 1) show that detrended TV tobacco instances were related to consumption contemporaneously (b=16.91, 95% CI=0.64, 33.18) and with 1-year lag (b=21.54, 95% CI=4.17, 38.92), while controlling for inflation-adjusted price of cigarettes (b=–1.95, 95% CI=−3.47, −0.44), a one-year lag in consumption (b=0.73, 95% CI=0.57, 0.89), a 1-year lag in price, and the 1971 TV ad ban. Because the total TV hours of programming represented by our sample fluctuated over the course of the study, the model also controlled for this factor. However, annual TV hours coded did not correlate with changes in TV tobacco instances (r=0.054, p=0.692) or consumption (r=0.130, p=0.340), and thus did not affect our model's results. There were no significant 2-year lags for any predictor and so none are included in the final model.
The model parameters indicate that an annual average of about two additional packs per adult (38.5 cigarettes) were consumed for each additional tobacco instance per hour based on the two lags in tobacco portrayal. Each additional cent increase in cigarette prices produced a decline of about two cigarettes per capita. We calculated price elasticity at the mean of price and consumption in the series (−0.00076). Although this value is below most elasticity estimates in the literature (−0.14 to −1.12),17 it is close to within one SD (0.43) of the reported mean elasticity (−0.48).25 The low estimate is not due to competition between detrended TV tobacco and price: the variables were uncorrelated at both lags over the study period: contemporaneous TV tobacco and price (r=–0.164, p=0.227), and lagged TV and price (r=−0.193, p=0.158), indicating that they were essentially independent correlates of cigarette consumption. Nevertheless, because price increased over the study period while consumption declined, elasticity increased over time.
Because the metric for TV tobacco cannot be directly compared to price, a comparison of elasticities would be misleading. Instead, we calculated the potential effects of each factor on change in consumption over the study period using the coefficients in the time-series model. Change in TV tobacco ranged from a high of 4.96 instances per hour in 1961 to 0.29 in 2010, while prices rose from 86.6 in 1955 to 273.1 in 2010. These changes resulted in a decline of 363.5 cigarettes per capita attributable to rising price and a decline of 179.3 cigarettes attributable to declining TV portrayal. Thus, these estimates suggest a potential effect of TV tobacco on per capita consumption that was nearly half (49.3%) as large as price. The estimates suggest that these two factors together accounted for a sizeable proportion of the decline in cigarette consumption (17.7%) since its peak in 1963.
While findings regarding tobacco-related TV trends have been inconsistent,15 ,16 our long-term study found that prime-time TV drama tobacco use has significantly declined since 1955, a trend that closely paralleled the historical decline in US tobacco consumption. This pattern is consistent with findings regarding tobacco content in top-grossing movies since 1950.4 Thus, portrayal of tobacco use in movies and TV has declined in parallel with US cigarette consumption since the 60s.
Although tobacco portrayal in movies and TV has been linked with adolescent smoking initiation, less is known about the effects of screen-media tobacco portrayal on adults. This study's time-series-analysis of annual changes found TV tobacco portrayal was associated with cigarette consumption contemporaneously and with a 1-year lag. Controls included changes in cigarette prices and the ban on cigarette advertising on TV since 1971. Indeed, changes in price and TV portrayal were independently predicted consumption since 1955.
While prices have increased and prime-time TV portrayal has declined over time, both may have influenced the decline in US cigarette consumption. Our estimates of the potential effects of each factor suggest that price had twice as large an effect on consumption over the study period as prime-time TV tobacco portrayal. Nevertheless, the effect of TV portrayal is sizeable and suggests that this potential influence should not be ignored in future tobacco control efforts. Because these measures of annual TV tobacco content were more sensitive than available data on annual movie content, we were better able to study the relation between TV tobacco content and adult tobacco consumption. Nevertheless, we hypothesise that the declining portrayal of tobacco use in top-grossing movies since the 1950s4 may have also played a role in slowing the US tobacco epidemic. In both cases, the effects of screen-media portrayal of smoking are consistent with the cue sensitisation literature, which indicates that exposure to cigarette cues captures smokers’ attention, reduces processing of competing stimuli, and activates craving for cigarettes.10–14 Watching TV at home would be an ideal environment to instigate such effects in adults who have already initiated smoking. Evidence of a cue-sensitisation effect when watching tobacco use portrayed in movie theatres14 suggests a potentially similar effect for TV.
Our finding that TV tobacco portrayal was related to cigarette consumption contemporaneously and with a 1-year lag was anticipated, because programming from the prior year would be shown as reruns in the off season of the next year. Additionally, there could be a lag between exposure to TV programming and measures of cigarette consumption. Nevertheless, the finding that the 1-year lag was as strongly related to consumption as the contemporaneous predictor was unexpected. This might reflect an additional carryover of prior year exposure to smoking in the subsequent year by slowing cessation among smokers. If increased exposure to TV smoking prevented smokers from quitting, it could maintain consumption into the following year.
It is also possible that other events influenced annual changes in TV tobacco portrayal and consumption, such as news reports about tobacco, most of which have focused on the harms of the product. Pierce et al26 reported a series of annual tobacco news stories from 1950 to 1983 and found a visual pattern that suggested a relation with adult consumption. We used a computer-assisted method to recover the values in the Peirce et al series and correlated the detrended values for their shorter time period with our detrended data. That series was correlated with consumption (r=−0.32, p=0.09). However, it was uncorrelated with TV tobacco (r=−0.12, p=0.55) and did not affect the results when it was added to our model. Thus, it appears unlikely that news stories affected TV portrayal and consumption.
Strengths and limitations
Unlike other studies that covered fewer years with smaller samples,15 ,16 this study tracked 56 years from 1955 to 2010 and coded 1837.9 commercial-free hours of TV. Our top 30 prime-time TV sample consistently reached a considerable percentage of US households. Our time-series analysis allowed us to study changes in tobacco portrayal and price controlling for long-term influences that affect each series.24 Still, this study has limitations. Our coding was not sensitive to whether tobacco portrayal was approving or not, or whether it was more tied to heroes or villains in plots. Although TV dramas were the most available genre for building a historical sample, it did not include non-drama genres or cable shows, which grew in audience size after 1980. The omission of cable shows may explain why our trend in cigarette consumption tracked TV portrayal less closely after 1995, a time when cable penetration and programming grew dramatically. It is also possible that earlier popular shows with tobacco content were increasingly seen in reruns. The sample also had an unavoidable historical availability bias because it was built from TV shows that were available for coding, in some cases decades after airing. However, the fact that sampled shows were popular enough to have been preserved and sold many years later suggests their importance to audiences.
Although our findings are consistent with the cue reactivity literature, the study cannot establish this causal direction. Variables not controlled in the study could be responsible for the relation between TV portrayal and consumption; however, we examined several and found no such evidence. Additionally, our findings are based on aggregated data for the US population, and may not apply to all segments of the adult smoker population. In particular, changes in consumption are based on changes in increased and reduced smoking by different age groups over time, neither of which can be separately detected by our measure. Indeed, econometric studies have better modelled cigarette consumption by analysing effects on the prevalence of smoking and cigarette demand,27 but data for such an analysis since 1955 are unavailable and is beyond this paper's scope.
Our finding of a relation between TV tobacco portrayal and adult tobacco consumption suggests that tobacco control efforts should continue to examine this link across screen-based media platforms. US TV shows are popular internationally, and such tobacco portrayal could influence cigarette consumption beyond the USA.
What this paper adds
Portrayal of tobacco use on entertainment TV should be studied because of potential links with adolescent and adult smoking.
The literature is unclear regarding the long-term direction of tobacco portrayal on US TV programming. This study covered more years than earlier studies, coded a larger sample, and found that portrayal in popular TV dramas has trended down since 1955 in line with aggregate cigarette consumption.
The study also found that annual changes in smoking incidents in popular TV dramas were correlated with US adult cigarette consumption from 1955 to 2010, suggesting that TV portrayals of smoking may have influenced national cigarette consumption.
A comparison of the effects of price versus TV portrayal over the period of the study indicated that TV tobacco and its lag may have accounted for a drop in consumption that was nearly half the amount attributable to price.
Tobacco control efforts should continue to evaluate the role of tobacco portrayal on entertainment screen media as an influence on adult smoking.
Acknowledgements We thank The Robert Wood Johnson Foundation for its financial support of CHAMP and Ilana Weitz, Susan S Lee, Eian More and the CHAMP coders (see http://www.youthmediarisk.org/).
Contributors PEJ conceived of the study. PEJ directed the coding. PEJ and DR conducted the analysis. PEJ wrote the first draft. PEJ and DR revised the first and second drafts, and contributed important intellectual content. PEJ had full access to all the study data and takes responsibility for the data and the analysis.
Funding The Annenberg Public Policy Center and the Robert Wood Johnson Foundation grant #052539.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.