Cultivation and Spiral of Silence Effects: The Case of Smoking

MASS COMMUNICATION & SOCIETY, 2004, 7(4), 413–428

Cultivation and Spiral of Silence Effects: The Case of Smoking James Shanahan Department of Communication Cornell University

Dietram Scheufele Department of Journalism and Mass Communication University of Wisconsin

Fang Yang and Sonia Hizi Department of Communication Cornell Univeristy

Public opinion about regulating smoking is characterized by considerable ambivalence. On the one hand, more than 80% of Americans recognize the health risks of smoking cigarettes; the majority of the public favors various antismoking policies, such as restricting smoking in public places. On the other hand, a majority also agrees that a person should have the right to smoke in public. Studies about communication and media effects on smoking are mostly concerned about the effectiveness of antismoking campaigns. How general media use, where fewer intentional antismoking messages are present, might affect attitudes about smoking is less studied. This study compares the relative impact of mass media on attitudes about smoking behavior and on various smoking-related policy stances. In particular, we are interested in the role that mass media play in influencing perceptions of the social climate surrounding cigarette smoking. Data come from a survey of 794 respondents about media use, smoking behaviors, and attitudes toward smoking. Results show that TV viewing (exposure) is indirectly related to perceptions of public support for smoking. TV exposure is positively related to prevalence estimates of smoking, which in turn are related to support for a smoking ban and to perceptions

Requests for reprints should be sent to James Shanahan, Department of Communication, Cornell University, 314 Kennedy Hall, Ithaca, NY 14853. E-mail: [email protected]

414

SHANAHAN, SCHEUFELE, YANG, HIZI

of public support for such a ban. Implications for communication theories such as spiral of silence and cultivation are discussed.

Mass media have long been regarded as a valuable resource for promotion of public health (Atkin & Wallack, 1990). How media might affect prohealth policy attitudes and prohealth advocacy efforts is a particularly important topic, especially insofar as public health professionals emphasize the media’s role in portraying images to improve public health. In this article, we explore media theories in the context of the issue of smoking, analyzing how people’s attitudes about smoking policy and their advocacy activity can be related to their general media use. Studies of media effects on smoking are mostly concerned with the effectiveness of antismoking campaigns. Most studies thus focus on messages in the context of purposive, short-term campaigns. Despite the obvious importance of these campaigns, people are arguably more exposed to protobacco commercial messages than antitobacco messages (Pierce et al., 1998). Even with a cigarette advertising ban on television and radio since 1971 in the United States, and a variety of consent agreements that have limited tobacco companies’ promotional activities, companies still spend billions of dollars on sales promotion each year through print media, billboards, sports sponsorships, and so on (Levine, 1988). How general media use, within which fewer intentional antismoking messages are present, might affect people’s perceptions about smoking, attitudes about antismoking policies, and motives to carry out antismoking advocacy action is less studied. We emphasize the media’s direct influence on people’s perceptions of the social climate surrounding cigarette smoking, which in turn affects individual attitudes about smoking and potential willingness to engage in advocacy efforts. Perceptions of what most others are doing and perceptions of others’ moral approval or disapproval can both be counted as forms of social influence (Cialdini, Reno, & Kallgren, 1990). People’s own smoking behavior and attitudes about smoking are both found to be subject to such forms of social influence. For instance, prevalence estimates of peer smoking are found to influence adolescents’ smoking behavior (Collins et al., 1987). Unfavorable public sentiment and opinion climate for cigarette smoking and smokers have also been found to be negatively related to state smoking rates and positively related to people’s willingness to quit smoking (Kim & Shanahan, 2003). Our approach was intended to explore first how media affect smoking prevalence estimates and perceptions of public attitudes toward antismoking policy, and then to look at how these perceptions affect individual policy attitudes and opinion expression. We explore media theories in the smoking context, particularly theories of cultivation (Shanahan & Morgan, 1999) and the spiral of silence (Noelle-Neumann, 1973). Not much work has been done in these areas that is directly related to smoking, although we will review some related studies that exist. Both theories posit a role for mass media in explaining how people develop

THE CASE OF SMOKING

415

conceptions of social reality and of social climate. Cultivation argues that exposure to message systems, particularly television, subtly encourages heavy viewers to accept the worldviews embodied and encoded in media messages. Typically, cultivation studies examine relationships between frequency of exposure to TV messages and conceptions about the real world. Spiral of silence theory argues that perceptions of majority opinion are likely to affect personal opinion formation and an individual’s willingness to speak out about issues. Individuals who perceive that their opinions are less supported by the majority are theorized to be less willing to speak out about those opinions. In this article, we try to combine both approaches by including relevant media measures (such as television exposure, normally an independent variable in cultivation research), measures of perception of smoking prevalence (a potential dependent variable in cultivation research), perceptions of public support for stringent smoking regulation (an independent variable in spiral of silence research), and finally, actual willingness to speak out in public about smoking (a potential dependent variable in spiral of silence research). We link these measures in a model that combines aspects of both cultivation and spiral of silence theory. Thus, the unique aspect of this study is that it explores a much-discussed issue (the connection between media and smoking behavior) in a direction opposite to that usually taken. Though the literature on purposive campaigns has much to offer, our approach opens up the debate to ideas about social aspects of smoking, within a social climate, that need to be considered.

SMOKING AND THE MEDIA Content Analysis We begin with a review of some of the smoking content analyses that have been conducted. The tobacco control movement positions mass media as one of its main targets, assuming that mass media, particularly advertising, are a potent force for influencing people’s decisions to smoke. Content analytic studies are used to show to what extent smoking is prevalent in media content. Particularly after the decision was taken to ban smoking ads on U.S. television, the rate of appearance of smoking in mass media program content became an important issue. Breed and DeFoe (1984) found that media tobacco use declined and alcohol use increased over the three decades of their study. Very little smoking was portrayed in the 1981–1982 season. Examining trends, Cruz and Wallack (1986) found one smoking act per hour of programming, with more smoking in dramas than situation comedies; two thirds of smokers were lead characters, with 70% cast in strong, enduring roles.

416


Hazan and Glantz (1995) found that 24% of programs contain at least one tobacco event; 92% were protobacco events, whereas 8% were antismoking messages. In terms of character analysis, 55% of smokers were “good guys”; high-status characters were more likely than medium- or low-status characters to smoke. They argue that higher rates of smoking occur on television than in real life. Based on a more recent study on television during the 1998–1999 season, however, only [o]ne in five episodes (19%) portrayed tobacco use. No characters under the age of 18 were shown using tobacco. Eight percent of adult major characters used tobacco. Twenty-three percent of episodes that showed tobacco use expressed negative statements about smoking. Overall, teen viewers were exposed to relatively little tobacco use. These episodes avoided underage tobacco use and presented adult use to a limited extent. (Christenson, Henriksen, Roberts, with Kelly & Carbone, 2000, p. 4)

This study also concluded that smoking was portrayed with decreasing frequency on television. Though the studies are not easily comparable, the recent data tend to confirm the impression that we see smoking less on television than we used to. If the Mediascope study is correct, TV characters smoke at a rate lower than in the “real world.” Recent studies show actual smoking rates in the low 20% range, varying by state (Centers for Disease Control, 2000). If so, this phenomenon suggests a cultivation hypothesis: H1: Heavy television viewers will underestimate the extent to which people smoke in the real world. At the same time, we should note that heavy viewers are themselves more likely to actually smoke (Sidney et al., 1996). We test this as our second hypothesis: H2: Smoking will be positively correlated with television exposure. Other studies have looked more at direct media content influences on smoking. Difranza et al. (1991) found that “Joe Camel” advertisements were effective marketing tools to children. Bauman, Laprelle, Brown, Koch, and Padgett (1991) found that media campaigns could directly influence smoking decisions. Pierce, Anderson, Romano, Meissner, and Odenkirchen (1992) found that TV spots affected people’s use of a cancer information line. But Klein et al. (1993) found that adolescents who used media more were also more likely to engage in risky behaviors, including tobacco use. Popham et al. (1993) estimated that 33,000 individuals quit smoking as a result of a tested media campaign. Pollay et al. (1996) found that advertising rates are related to market share, suggesting a direct influence. Again,

THE CASE OF SMOKING

417

FIGURE 1 Support for restricting smoking in restaurants, percentages. From Gallup; retrieved from http://web.lexis-nexis.com/statuniv.

the studies are not completely consistent, but some researchers believe that targeted campaigns can influence viewers, both for and against smoking. Though our study does not include a measure of exposure to anticigarette campaigns, it should be remembered that overall exposure to television should be correlated to exposure to antismoking information.

Perceptions of smoking. Turning to spiral of silence and perceptions of smoking, the research is newer and perhaps less clear (though see Dixon, Lowery, Levy, & Ferraro, 1991). Certainly very little spiral of silence research exists dealing directly with smoking. In this article the main attitude that concerns us is support for regulations restricting smoking. Public opinion data show that a majority of people do support some type of restriction on where people can smoke. New York City Mayor Michael Bloomberg led a well-publicized campaign in 2003 to ban indoor smoking in his city. As an example, Figure 1 shows an opinion trend for public support for either banning smoking in restaurants or supporting designated areas for smoking.1 The figure shows that about 60% favored setting aside special areas, whereas almost 40% favor completely banning smoking in restaurants. This growing support, of course, coincides with actual smoking bans in some communities, states, and cities but may also reflect a decreasing tolerance for smoking. Some surveys have even looked at the possibility of banning tobacco outright. Though this may seem to be something of a “modest proposal,” a 2001 Gallup survey found that 14% of respondents would favor making smoking “totally illegal.” A 1995 CBS–New York Times survey, using somewhat less ominous wording, 1All survey data in this paragraph are derived from poll results archived at Lexis-Nexis (http://web.lexis-nexis.com/statuniv).

418


found that 43% of respondents would favor “legislation banning tobacco altogether.” Thus, in some sense, subtle but broad changes seem to be under way with respect to public tolerance for smoking. As majority opinion about smoking drifts toward the negative direction, issues of social norms and tolerance for smoking become that much more interesting. Social norms have been considered important in recent work on health prevention communication, but much of this work is not aware of mass communication research dealing with media effects on social norms. For instance, recently two economists argued that Smokers do care about social acceptance. Inconsiderate smoking is assumed to trigger negative emotions on [the] non-smokers’ part, and these reactions are stronger the less used non-smokers are to such inconsiderate behavior. Smokers decide whether to be considerate or not by weighting the inconvenience costs of being considerate against the benefits of social acceptance. (Nyborg & Rege, 2001, p. 2)

They were unwittingly stating the axioms of spiral of silence theory, in which willingness to express opinions publicly is weighed against social costs that might come from majority pressure. To some degree, these days, smoking in public is a very sensitive form of opinion expression. Social norms have been built into recent popular campus campaigns to reduce drinking and now even smoking. They use estimates of drinking or smoking behavior in advertising to convince abusing students that they are in the minority (Haines & Spear, 1996). This practice raises questions about how media portrayals play into perceptions of social norms. If the media are portraying smoking less, does this amplify the trends toward cementing the taboos developing around smoking? How do perceptions of smoking relate to media use? Spiral of silence studies have typically dealt with such questions by asking respondents to imagine themselves in a hypothetical situation, such as a party or traveling in a train compartment, where they meet strangers discussing political issues (Scheufele & Moy, 2000). The respondent is asked to imagine that the strangers are discussing a political issue, expressing opinions that would be different from the respondent’s own. How willing would the participant be to express his or her own opinion knowing that it differs from the strangers’? This so-called “train-test” measure has been used in many different studies (see Glynn, Hayes, & Shanahan, 2001, for a review of all of them). In our study we use a variant of the question: “Now, imagine a situation in which you are in a restaurant eating with friends. At the table next to you, a man begins smoking a cigarette, even though the section is clearly marked ‘no smoking.’ How likely would you be to ask the man to stop smoking?” In addition, we also included a more general measure of how likely the respondent would be to speak up in a social setting where the group’s opinion differed from the respondent’s own.

THE CASE OF SMOKING

419

This question differs slightly from the typical train test question in a few ways. It is similar in that it deals with opinion expression in a public setting. However, the assumed opinion of the interlocutor (the smoker, in this case) is known. That is, the smoker is smoking and therefore assumed to support the right to smoke in the restaurant. If the respondent also supports such a right, then there is less need for the respondent to ask to stop the behavior. Social pressure is less important in this case. Thus, we are more interested in the case in which the respondent would prefer that the smoker stop. In this case, what are the effects of media exposure and social pressure on the respondent’s willingness to speak up in this admittedly sensitive situation? Spiral of silence research normally predicts willingness to speak out from perceptions of public support for one’s own opinion. Thus, we included two variables, a measure of one’s own support for legislation banning tobacco and a measure of perception of public support for such legislation, which asked respondents to estimate the percentage of people that would support such legislation. Using these measures, we generate two additional hypotheses: H3: Personal support for legislation banning tobacco will be positively related to willingness to speak out in public. H4: Perceptions of public support will influence willingness to speak out, positively.

Media use and perceptions. Finally, we need to examine how media conceptions influence perceptions, and how all of our variables relate to willingness to speak out. Spiral of silence theory typically proposes that the “media tenor” is an important factor in how people determine where majority opinion is centered. We measure respondent’s exposure to television and newspapers, as well as attention to both comedy and drama on television, to supply general measures of media exposure. We hypothesize that H5: Perceptions of public support for antismoking legislation are related positively to media use. This hypothesis is based on the assumption that smoking frequency is declining in the media. As viewers see fewer people smoking in the media, spiral of silence theory predicts that they will internalize these perceptions into their own beliefs about how prevalent smoking actually is.

Policy relevance. The discussion thus far brings us to wonder how the phenomena we have examined might affect policy decisions. One possibility is increased attention to social norms campaigns, which have been widely used in alcohol awareness work. Also, greater understanding of the role of media in these

420


broad social trends would complement well the work on individual media campaigns. Reception of individual media messages will be placed in the context of broader message systems. Those designing campaigns can expect greater effectiveness to be derived from such awareness. Individual programs and campaigns may show short-term effectiveness, but long-term message patterns are more likely to encourage subtle social shifts in behavior.

METHOD We conducted a telephone survey in Tompkins County, New York, in October 2001. The fieldwork was conducted by the Survey Research Institute at Cornell University. The survey used a random digit dialing method, eventually yielding a final sample size of 794. In the sample, the mean age was 45, 61% female, and highly educated (75% had graduated college). These demographics closely parallel the demographics of the larger county.2 The smoking rate was 15%; of the remainder 38.3% had smoked at one time. Response rate for the survey (following AAPOR3 definitions) was 44%. We used unweighted data in all analyses. Thus, the data used for this study came from a local survey based on a carefully constructed probability sample. Regardless of the quality of the sample, however, the question arises as to whether the local nature of the sample limits its generalizability. Despite the geographic limitations of the sample, the relatively elaborate and specific measures of opinion perception, opinion expression, and media use are not available in large national data sets, such as the National Election Study or the General Social Survey. In this work, then, there is a tradeoff between the generalizability offered by large-scale samples and the quality of measurement we could obtain in this sample. Given the fact that this study focused on relatively subtle distinctions between different types of climate perceptions and opinion expression, we feel that quality of measurement contributes heavily to the overall validity of the dataset and its interest. We included five main indicators of smoking and perceptions about smoking:

• An estimate of the percentage of population that smokes (ESTIMATE). • An estimate of the percent of the population that would support a total ban on tobacco (BAN ESTIMATE).

• A measure of personal support for such a ban (SUPPORT BAN). • Personal smoking status; has respondent ever smoked (SMOKE). • Willingness to ask person to stop smoking in a restaurant (ASK STOP). 2Our mean age differs from the reported county mean of 28.6 in the 2000 Census, due to the fact that we intentionally excluded student housing units from our sample. 3American Association of Public Opinion Research.

THE CASE OF SMOKING

421

The full text of these items is presented in Appendix A. Most of the aforementioned measures are straightforward. Though it may seem unusual that anyone would actually support a total ban on tobacco, we were looking for an issue imbued with a relatively high degree of moral controversy. This method is consistent with Noelle-Neumann’s original conceptualization of the “morally loaded” issue (1993). As well, we have seen earlier that at least one opinion survey has used such a question. Recognizing that the question does not capture the current state of thinking about tobacco control, we felt justified in using the measure based on these conceptual considerations. We also included a measure of how likely the respondent would be to speak up in a social setting where the group’s opinion differed from the respondent’s own. This is a typical measure used in spiral of silence research to indicate willingness to express opinion in controversial contexts. Though we did not use the opinion “congruency” measures that are often used in spiral of silence research, our variables allow tests parallel to those actually used by Noelle-Neumann on the smoking issue (1993, p. 47), one of the few extant tests of smoking as a spiral of silence issue. Also, these operationalizations allowed us to combine tests of cultivation and spiral of silence in a single analytic model. We also included measures of television viewing (hours–minutes of exposure per day) and newspaper readership (days per week). We also had measures of attention to both drama and comedy on television. Though other measures of media attention and exposure could have been useful, and possibly could have provided more sensitive indexes, we followed the normal cultivation practice of using exposure measures. These were accompanied by basic demographic indicators including sex, age, and education level.

RESULTS Our sample, on average, thought that about 40% of the population smokes, which is an overestimate. Indeed, we found that many of our open-ended estimates of smoking behaviors and attitudes tended toward the high side. Similarly, the sample thought that about 40% of the public would support a smoking ban. Recall that only about 15% of Americans actually do support such a ban, although our question wording might have prompted some respondents to think about the question more in terms of public-place bans. The actual percentage of people supporting a ban in our sample was about 41%, definitely higher than the national statistic from Gallup. This may also suggest that question wording was prompting respondents to think about smoking in public places. Additionally, the political profile of Tompkins County, New York, is such that more residents probably would be in favor of antismoking legislation than what one would expect to see nationally. Finally, almost 70% said they would be willing to ask someone to stop smoking in a place that was labeled nonsmoking.

422


The relative frequencies confirm some of our foregoing discussion: People think that there is broad support for regulation of smoking; there is in our sample relatively broad support; and people would be willing to speak up in public about smoking. Thus, in some sense, our own data confirm what the national data show, which is that smoking is becoming less socially acceptable. Next we turn to examination of our hypotheses. First, we hypothesized that heavy television viewers would underestimate the extent to which people smoke. We test this hypothesis with a simple correlation. In fact, we found a positive correlation between television exposure and estimates of smoking in the real world (r = .22, p < .001). Thus, our first hypothesis did not receive support. Television viewing was also correlated to smoking itself (r = .14, p < .001), corroborating our second hypothesis, so it may be that heavy viewers are simply applying a projective bias to their estimates. In any case, the effect of actual television content, with its reduced smoking frequency, is unclear given these data. The cultivation hypothesis that reduced smoking rates on television would be associated with lowered estimates of smoking prevalence is not supported. Future content analysis needs to examine this issue; we take up some problems related to this finding in our discussion section. Our third and fourth hypotheses dealt with support for a smoking ban, perceptions of public support for such a ban, and their effect on willingness to speak out about smoking. Hypothesis 3 was supported, with a correlation showing that personal support for a smoking ban is positively related to willingness to speak out (r = .12, p < .01). Thus, as expected, personal opinion about public smoking plays a role. However, Hypothesis 4 was not supported; a correlation showed no significant direct relationship between perception of how many people support a smoking ban and willingness to speak out. This runs counter to the spiral of silence explanation of willingness to speak out. However, we should note that there is an interesting interaction between personal support, perception of public support, and willingness to speak out. Table 1 shows this interaction. Willingness to ask someone to stop smoking is coded on a 4-point scale ranging from 1 (very likely to do so) to 4 (not at all likely). The means show that one group in particular is least willing to speak out. Clearly the group that thinks a majority supports a ban but that does not themselves support the ban is least willing to speak up, whereas the other groups all show relatively high willingness. This “discrepant” group probably feels somewhat defensive about smokers’rights. Because this group probably would be the least likely to speak up in any case, their motives are not necessarily rooted in trying to conform to the majority, but perceptions of majority opinion do mark them off from the others who also do not support a ban. All in all, then, apart from a defensive minority of individuals, social opinion and social norms are strongly in favor of smoking regulation. In examining Hypothesis 5, we wanted to tie all of our variables together. To do this, we used regression path analysis to test a causal model of media effects on smoking perceptions, attitudes, and willingness to speak out. The model begins with basic endogenous demographic variables, then includes media variables

THE CASE OF SMOKING

423

TABLE 1 Willingness to Ask Someone to Stop Smoking in a Restaurant by Personal Support for Smoking Ban and Perceptions of Such Support Estimate of Support