Christine Brittle & Nithya Muthuswamy George Mason University

Transcription

1 SCIENTIFIC ELITES AND CONCERN FOR GLOBAL WARMING: THE IMPACT OF DISAGREEMENT, EVIDENCE STRENGTH, PARTISAN CUES, AND EXPOSURE TO NEWS CONTENT ON CONCERN FOR CLIMATE CHANGE George Mason University Abstract Scientific elites are a key source of environmental news. Yet, how these elites influence public concern for environmental topics via the news media is relatively unknown. This study experimentally examined the influence of such elites using the test issue of global climate change. We hypothesized that elite scientific agreement would increase central message processing, and that partisan cues would activate partisan beliefs. The results showed that none of the main experimental manipulations had any independent influence on concern for this issue: whether scientists were presented as agreeing or disagreeing, whether scientific evidence was strong or weak, and whether political cues were present or absent did not influence concern for global warming. Concern was primarily explained by environmental concern, and, to a lesser extent, by political ideology. Thus, in this study, attitudes about global warming do not appear to be influenced by strong scientific evidence even when scientists are presented as agreeing. However, such evidence has a reasonable claim to make as important information upon which the public should base attitudes about issues such as global warming 23

2 1. Introduction Global warming is an immensely important societal concern, as the consequences of human-induced climate change will affect all life on Earth (Drake 2000; Hitz and Smith 2004; Parry and Carter 1998). Despite a broad scientific consensus that global warming constitutes a legitimate environmental problem (Parry and Carter 1998; Drake 2000; Pollack 2003), several scientific uncertainties remain on this issue (Parry and Carter 1998; Swart et al. 2008). In addition to these scientific uncertainties, journalistic adherence to media norms such as attention to controversy, dramatization, and source balance may create the impression that additional uncertainties exist, even in the face of overwhelming scientific agreement (Boykoff and Boykoff 2007). Attention to scientific uncertainty positions scientific elites as a key source of information on this topic. This paper examines the role of scientists as elites in influencing public concern for the specific environmental issue of global warming. Zaller (1992) defines elites as persons who devote themselves full time to some aspect of politics or public affairs, including politicians, higher-level government officials, journalists, some activists, and many kinds of experts and policy specialists. Elites are generally recognized as influential in shaping and informing public opinion on a wide range of topics (Zaller 1992; Koch 1998; Page, Shapiro, and Dempsey 1987; Kinder and Sanders 1996). For example, elites influence how the public thinks about policy issues (Kinder and Sanders 1996; Nelson and Kinder 1996; Shapiro 1998), and public opinion tends to move in the direction of elite consensus and fractures when elites are divided (Paul and Brown 2001; Zaller 1992). While scientific elites are a key source of environmental news, how these elites influence public concern for environmental topics via the news media is relatively unknown. 2. Portrayal of climate change issues in the media Because significant scientific uncertainties remain prominent in discussions of climate change (Parry and Carter 1998; Boykoff and Boykoff 2007), these uncertainties present a problem for journalists seeking to cover this issue fairly. Journalists may have a difficult time evaluating scientific information, and may be uncomfortable making judgments as to when uncertainty is significant and when it is not (especially when such uncertainties are being highlighted in public debates). Given this, the journalistic norm when covering such issues is to present opposing viewpoints and leave readers to sort out the truth (Stocking 1999; Sandman et al. 1987; Dunwoody 1999; Bushman and Anderson 2001; Boykoff and Boykoff 2007). 1 Journalists refer to this as source balance. This balance often occurs despite the fact that one side of an issue may have a stronger evidentiary basis than the other (Bushman and Anderson 2001). Thus: 1 See Gelbspan (1997) for examples of efforts by specific societal groups (e.g. business or industry coalitions) to highlight global warming uncertainties in an effort to sway opinion. 24

3 A scientific debate is often reported as a he-said, she-said encounter with little insight provided to the reader or viewer to help them understand the subtleties. Because journalists wish to avoid being branded as one-sided advocates, and because they recognize their own inadequacies to analyze a situation, they often feel compelled to give equal time to opposing points of view, irrespective of the strength of the scientific arguments supporting one side or the other. The result is often to give unwarranted attention and thereby bestow credence to frivolous pronouncements and marginal debate. (Pollack 2003) In the specific case of media attention to global warming, one consequence of balanced reporting has been that scientific disagreement and uncertainty are frequent components of media coverage, thereby underestimating the larger consensus on this issue. For example, approximately 20% of newspaper articles on global warming from 1998 to 2003 featured scientific disagreement in one study, and some element of scientific uncertainty appeared in three-quarters of this coverage (Brittle 2005). Other studies have also found that disagreement and uncertainty are primary themes in media coverage (Bell 1994; Corbett et al. 2002; Williams 2001; Zehr 2000; Boykoff and Boykoff 2004, 2007). Given this, some researchers argue that that global warming opponents have used routine scientific uncertainty (i.e. the uncertainty that accompanies all scientific research) to undermine the scientific consensus on this issue. Likewise, Boykoff and Boykoff (2004) argue that media balance in regards to global warming has provided skeptics with unmerited public attention. Similarly, Gelbspan (1997) argues that global warming skeptics have successfully lulled the public into a deep apathy on this issue. In short, these researchers claim that media portrayals of scientific disagreement and uncertainty reduce public concern for global warming. However, this claim has not been empirically tested, and little is known about the mechanisms of how this reduction might occur. This paper investigates this claim, and proposes a causal model integrating the literature on elite influence and the dual cognitive models of attitudinal change. 3. Theoretical framework and hypotheses Literature pertaining to elite influence on attitude formation in the context of scientific elites is particularly relevant to this paper. Attitudes tend to polarize when elite opinion is divided according to awareness of differences in value orientations (Zaller 1992). In other words, to the extent that people are paying attention to the news on a particular issue, and that expert opinion on that issue is polarized (e.g. according to political party or some other set of values), public opinion tends to diverge according to each individual s selfidentified values and affiliations. Thus, media portrayals of scientific disagreement should result in a polarization of concern to the extent that such debates contain cues to value orientations. Alternately, media portrayals of scientific consensus should lead to a mainstreaming of attitudes, consistent with elite consensus. That is, if media coverage indicates that elites agree about an issue, public opinion should move in the direction of this elite consensus (Zaller 1992). 25

4 Furthermore, dual-cognition models of attitude change, such as the Elaboration Likelihood Model (ELM), also offer crucial insights that shed light on these phenomena. According to the ELM, message processing occurs along a continuum from very high cognitive effort (called central processing in the ELM) to very low cognitive effort (called peripheral processing in the ELM) (Petty et al. 2004; Petty and Cacioppo 1986). The portrayal of scientific disagreement should lead to less elaboration (as people focus on the value orientations of each disagreeing side as a heuristic cue to which position they should support), while the portrayal of scientific agreement should lead to more elaboration (as, absent such heuristic cues, people are more likely to process the content). Therefore, scientific disagreement about global warming (i.e. any mention of scientific disagreement, or any caveat on scientific agreement) will alter the cognitive processing strategies used to understand scientific information consistent with the Elaboration Likelihood Model. Under conditions of high cognitive effort, the ELM specifies that argument strength is an important variable driving attitude change. In other words, under conditions of high cognitive effort stronger arguments are more persuasive than weaker arguments, since people are paying attention to argument quality. In contrast, under conditions of limited cognitive effort, simple message cues are sufficient to influence attitudes (e.g. I believe this message because the source of the message is attractive ), because people are not paying attention to argument strength but to other factors. Based on the integration of the literature on elite influence and the tenets of ELM, this paper posits that that scientific agreement serves as a moderating variable affecting how cognition occurs. Scientific elite arguments in the news media should be processed differently depending on whether the information is presented as reflecting scientific agreement or scientific disagreement. Specifically, under conditions of perceived scientific agreement, scientific messages should be (appropriately) processed centrally. If scientists are perceived to be making strong, consistent arguments, attitudes should reflect this scientific agreement. However, to the extent that scientists are perceived to be in disagreement, messages will be processed peripherally, and media consumers will rely on shortcuts such as their own predispositions, both issue specific and in general (Zaller 1992; Lord, Ross, and Lepper 1979; Kuhn 2000; Slovic 1999), to process messages, as opposed to scientific evidence strength. Therefore: H1(a): Under conditions of relatively high scientific agreement, messages will be processed centrally as indicated by a higher reliance on argument strength in judgment formation, and a lesser reliance on pre-existing beliefs such as environmentalism or ideology. H1(b): Under conditions of relatively low scientific agreement, messages will be processed peripherally as indicated by participants low reliance on argument strength in judgment formation, and a stronger reliance on preexisting beliefs such as environmentalism or ideology. 26

5 An important consideration in examining how scientific information influences people s perceptions towards global warming is to assess if scientific information is non-partisan. Scientific information in newspaper articles on global warming contains a great deal of political information; political cues are present in over half of all articles that contain scientific evidence, and likewise are present in over half of all articles that contain scientific disagreement (Brittle 2005). Such cues tend to be polarized, with Democrats favoring tough actions on global warming, while Republicans oppose such actions (Brittle 2005). Thus, the elite debate has effectively been polarized in much of the preexisting discourse on global warming. This same polarization is reflected in partisan attitudes about global warming, with a widening gap between the viewpoints of Democrats and Republicans on global warming (Dunlap and McCright 2008). This raises the question of the effects of the association of political cues with scientific content on cognitive message processing. It is likely that such cues will lead to polarization along ideological divides, thus making ideology a stronger predictor of concern, since ideological information will serve as a cue to peripheral processing of messages. Further, the literature on elite influence suggests that this effect should be especially pronounced when scientists disagree, as two-sided arguments are much more likely to lead to peripheral message processing. Therefore: H2(a): When partisan cues are present, these cues will increase the predictive power of ideology as an explanation of concern for global warming. H2(b): Partisan cues will make ideology an especially strong predictor of concern for global warming when scientists disagree. Further, this study also assessed if a mere exposure to a climate change topic regardless of the type of the stimulus (presence or absence of scientific disagreement or partisan cues) increased perceived importance of climate change among participants. Consistent with the theory of agenda setting that specifies that any attention to an issue increases the perceived importance of that issue (Ader 1995; Iyengar and Kinder 1987; Atwater, Salwen, and Anderson 1985; Erbring, Goldenberg, and Miller 1980; Trumbo 1995; McCombs and Shaw 1972), it was predicted that simply reading any article on global warming would impact concern for this issue positively: H3: Reading any newspaper article on global warming should produce higher levels of concern for global warming. Lastly, this study also included an assessment of the impact of overall concern for the environment, scientific knowledge, and need for cognition as possible alternate influences on concern for global warming. Overall, concern for the environment tends to be broadly predictive of specific attitudes about unique environmental issues (Dunlap 2002). Level of scientific knowledge (Miller 1998) generally influences how people understand and react to scientific information. And, need for cognition (Cacioppo et al. 1996) influences the tendency of any one individual to process information centrally (as opposed to 27

6 peripherally). These constructs were included in this study because they all may have a direct impact on concern for global warming or because they may interact with evidence strength to affect how messages on global warming are processed. In other words, it is likely that those high in need for cognition or in scientific knowledge may be more sensitive to differences in evidence strength. 4. Method 4.1 Participants Four hundred and eighty seven adults were recruited from multiple public settings, including parks, shopping plazas, and cafeterias, in Maryland, Virginia, Washington, D.C., and North Carolina. 2 Overall, participants were relatively diverse. About half of the subjects characterized themselves as being from a large city, with an additional third from medium size cities, and the remainder from small cities or farms. Participants represented twenty-eight states, reflecting a geographically diverse sample, were more or less evenly distributed between men (51%) and women (49%), and averaged 40 years of age (SD = 14.6 years). Two-thirds of the sample was both White and non-hispanic, while the remaining third were non-white, Hispanic, or both. A third of the sample self-identified as conservative, just under half (43%) self-identified as liberal, and the remaining fifth categorized themselves as politically moderate. Participants were well educated, with 69% having a college degree or more. 4.2 Design A 2 x 2 x 2 factorial with two offset controls was designed to assess the influence of scientific agreement (agreement/disagreement), scientific evidence strength (strong/weak), and the presence of political cues (present/absent) in newspaper portrayals of climate change issues. The two control conditions comprised a no-information condition and another condition in which participants were exposed to information on global warming devoid of any scientific or political content. The two offset control condition allowed for testing Hypothesis 3 related to the theory of agenda setting. Participants were randomly assigned to one of the 10 conditions. The planned cell size of 50 subjects per condition was established based on a power analysis, using data from the 2002 Detroit Area Study (DAS). In the DAS dataset, for each change in level of perceived scientific agreement regarding global warming, there was a change in each variable measuring concern. Thus, for each one-unit increase in level of perceived scientific agreement, concern for global warming increased about half a unit. For the purposes of 2 This work was supported by the National Science Foundation under Grant No. SES Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. 28

7 this experiment, the lowest threshold for a meaningful effect size was set at the equivalent of a one-category shift in perceptions of scientific agreement Therefore, using Cohen s (1992) formula for calculating effect size, the cell size of approximately 50 participants per condition was chosen to detect such an effect. The dependent measures in the study were concern for global warming. The strength of other correlated factors, such as concern for the environment and ideology, was also examined. Additionally, need for cognition (a measure of the tendency to engage in cognitive activity) (Cacioppo et al. 1996) and scientific knowledge (a measure of how well people understand science and scientific issues) (Miller 1998; Laugksch 2000; National Science Board 2000) were also measured in this study. Finally, demographic variables such as gender, age, race, and level of education were also statistically controlled. 43 Pretest of manipulations on scientific agreement and evidence strength A pretest was conducted to test and select the experimental manipulations on scientific agreement and evidence strength. Seventy-one undergraduate students in a large Midwestern University rated forty-four sample newspaper statements on either the strength of scientific evidence (24 statements) or the level of scientific agreement (20 statements) using a 7-point response format. Based on pretest results, three final manipulations were selected for this study (two for strength of scientific evidence and one for level of scientific agreement). Pretest participants who were exposed to the first selected example of stronger evidence (M=5.1, SD=.94) were significantly more likely to rate the message as stronger on evidence when compared to those exposed to the weak version (M=4.5, SD=1.31), t=2.31, df=69, p=.01). Similarly, participants who read the second example of stronger evidence rated the evidence as somewhat more important (M=4.4, SD=1.50) than those in the weaker version (M=3.9, SD=1.16), t=1.61, df =69, p=.06. The selected manipulation on scientific agreement was also successful. Participants also felt that the statement that portrayed scientific agreement represented stronger agreement (M = 5.4, SD=1.02) than those who read the article that presented scientific disagreement (M=3.5, SD =1.12), t =7.48, df= 69, p< Experimental Materials The experimental materials used in the study consisted of a series of three short newspaper articles. The middle article comprised the experimental manipulation. The story concerned a new interactive computer model on climate change being developed by scientists at Yale University. The scientific agreement manipulation was created by inserting the words but skeptics are questioning the accuracy of the data into the disagreement version. The evidence strength was manipulated by saying that surface temperatures have increased either 1.1 or 8.1. degrees, and that ocean temperatures have increased by either.5 or 5 degrees as a result of climate change. Finally, the political cue was that the announcement was made either by Yale s President or the Democratic Senator from Connecticut. (This political cue intentionally made reference to a political 29

8 party and not an actual political figure so that general political beliefs are activated and not beliefs about specific political figures.) The agenda setting control condition was exposure to an article on global warming that contained no scientific or political information. The second control condition was exposure to an article on video games and contained no scientific or political information. 4.5 Instrumentation The questionnaire in this study consisted of six main sections; a series of newspaper articles constituting the main experimental treatment, three additional independent variable scales designed to explore the influences on the dependent variable, a dependent variable scale, and a series of demographic questions. Of the three independent variables included in the study, the concern for the environment scale measured overall concern for environmental protection and preservation, self-association with the environmental movement, perceived importance of environmental quality, and importance of taking actions to protect the environment. The need for cognition scale developed by Cacioppo and Petty (1996) was condensed for the purpose of this study, and the scientific knowledge scale contained questions from Miller (1998) and Bann and Schwerin (2004). All scales were measured on a 5-point response format. The construct concern for environment scale (M = 3.79, SD = 0.71) yielded an alpha of.80, and the need for cognition scale (M = 3.68, SD =.59) yielded a reliability of.83. Lastly, the scientific knowledge scale (M = = 3.76, SD =.62) resulted in a reliability score of.70. After reading these newspaper articles and answering a short series of questions about each one, subjects completed a series of items on concern for global warming that comprised the main dependent variable for the study. Seven items asked subjects questions such as how personally concerned they are about global warming and how serious they rate global warming as an environmental problem. Each item was measured on a four-point scale. The items scaled well, with a Cronbach s alpha of.93. A higher score on this scale indicated more concern for global warming (M = 2.64, SD =.78). 4.6 Assignment to Condition Prior to testing the central hypotheses in the study, the 10 conditions were analyzed for equivalence. Participants varied somewhat in their characteristics across these ten conditions (e.g. gender, age, ethnicity, and concern for the environment), and the means for each of these variables in each condition are shown in Table 1. Regression equations for each of the variables reflected in Table 1 were computed using the experimental conditions as predictors. None of the regression equations were significant (regressions not shown), suggesting that random assignment to condition was more or less successful. 30

9 Table 1: Participant Characteristics Across Condition: Concern for Environment, Scientific Knowledge, Need for Cognition, Gender, Age, Educational Level, Race, and Ideology Table shows means for each variable in each condition; refer to experimental design in methods section for full explanation of condition; n=487 adults, approximately 50 adults per condition. 4.7 Hypothesis Testing The hypotheses were tested with a series of regression equations using experimental condition, concern for the environment, scientific knowledge, need for cognition, ideology, gender, age, education, and race as independent variables, and concern for global warming as the dependent measure. To test hypotheses 1: A series of regression equations were performed, including an interaction term for the relation between scientific agreement and strong evidence. To test hypotheses 2: A regression equation predicting concern for global warming for the eight main experimental conditions in this study was calculated. Dummy variables for scientific agreement, strong evidence, and political cues in the experimental stimulus article (including the interactions between these terms), as well as variables measuring concern for the environment, scientific knowledge, need for cognition, ideology, gender, age, education, and race were included in this equation. Additionally, five two way interaction terms were included to determine whether the presence of political cues or the presence of scientific disagreement in a newspaper article increased the importance of concern for the environment or if ideology was a predictor of concern for global warming. 31

10 A three-way interaction term between scientific disagreement, political cues, and ideology was also included to see if ideology was especially influential in predicting concern when scientists disagreed and political cues were present (as predicted by Hypothesis 2b). To test hypothesis 3: Regression equations were conducted to compare concern for global warming in baseline condition 1 (where the experimental article did not mention global warming), to concern for global warming in baseline condition 2 (which mentioned global warming but contained no scientific or political information). Further, regression equations were also conducted to examine the influence of any information on global warming (in either the second baseline condition or in one of the eight main experimental conditions) vs. the condition with no information on global warming. Several exploratory regression analyses were also conducted to investigate further some of the results. 5. Results The first set of hypotheses predicted that scientific agreement will lead to central message processing when scientists agree (indicated by reliance on evidence strength in judgment formation), and peripheral processing when scientists disagree (indicated by a disregard for evidence strength). Our expectation was that strong evidence should lead to more concern for global warming when scientists agree, but only when scientists agree. We first tested this relationship for the four experimental conditions where political cues were not present. However, the interaction term for the relationship between scientific agreement and strong evidence was not significant, suggesting that central message processing did not occur. (Alternately, it is possible that central processing did occur, but that even under conditions of high cognitive engagement subjects were not able to effectively incorporate scientific information into their attitudes about this issue. Since cognition was not directly measured, this possibility cannot be ruled out.) As predicted, stronger evidence also did not inform attitudes in the conditions where scientists were presented as disagreeing. In other words, strong evidence did not lead subjects to be more concerned about global warming, either directly or indirectly. Two constructs were significant in predicting concern for global warming. First, concern for the environment was related to concern for global warming, such that a one unit increase in concern for the environment increased concern for global warming.56 units. Additionally, a one-unit increase in ideology moving from conservative to liberal resulted in a.05 unit increase in concern for global warming. Thus, it appears that environmentalism, and, to a lesser extent, ideology, account for almost all of the variance in concern for global warming. Interestingly, a very different picture emerged when the four conditions with political cues were examined (Table 2). Here, a large and significant three-way interaction exists between political cues, scientific agreement, and strong evidence. For the conditions with 32

11 political cues, as seen before, evidence strength does not affect attitudes when scientists disagree, and both concern for the environment and ideology continued to predict overall concern. One additional factor worth noting about the regression with political cues is that while ideology is an important predictor of concern for global warming in both the political cues and no political cues conditions, the regression coefficient on this variable is more than twice as strong when political cues are present as opposed to when such cues are absent. Table 2: Regression Model Predicting Concern for Global Warming and the Interaction Between Scientific Agreement and Evidence Strength 33

12 Hypothesis 2 concerned the role of partisan cues on message processing. This hypothesis predicted that partisan cues would increase the importance of ideology in predicting concern, and that this effect will be magnified when scientists disagree. As shown in Table 3, concern for the environment remains the primary predictor of concern for global warming, with each one unit increase in concern for the environment leading to a halfunit increase in concern for global warming. Likewise, ideology remains a predictor of concern for global warming, with each one-unit increase in ideology (moving from conservative to liberal) leading to a.05 unit increase in concern for global warming. Table 3: Regression Model Predicting Concern for Global Warming and the Interaction Between Political Cues and Ideology 34

13 As predicted, the interaction term between the presence of political cue in a newspaper article on global warming and ideology was positive and significant, supporting Hypothesis 2(a). In other words, the role of ideological beliefs in predicting concern appears to increase when political cues are present. However, contrary to Hypothesis 2(b), this effect was not magnified when scientists disagreed. Notably, the presence of political cues does not change the role of concern for the environment in predicting concern (in other words, political cues neither strengthen nor decrease the influence of concern for the environment). Likewise, the presence of scientific disagreement does not change the role of concern for the environment or of ideology in predicting concern (in other words, scientific disagreement does not cause subjects to rely more on environmental beliefs or ideology in informing beliefs). Thus, it appears that most processing of scientific information on global warming across all conditions was peripheral in nature: concern for the environment and political ideology are the strongest predictors of concern about global warming in all conditions, and other factors such as scientific agreement, evidence strength, and the presence of political cues do not directly impact concern. There are two main exceptions to this general finding: first, there is evidence that central route processing occurred in the specific condition where scientists agreed and political cues were present. In this condition, strong evidence led to greater concern than weak evidence. Second, political cues in a newspaper article seem to increase the importance of ideology in predicting concern; in other words, political cues strengthen the effect of ideology on concern. The third hypothesis pertained to the theory of agenda setting. Data suggested that simply reading a newspaper article on global warming had no effect on concern for this topic: concern was just as high in the condition where no communication on this topic was present as it was in the other baseline condition and in the nine experimental conditions that mentioned global warming (see Table 4). To further explore the large and significant three-way interaction between political cues, scientific agreement, and strong evidence, regressions were re-run for five specific subgroups within the sample: Republicans, Democrats, Independents, and those high and low on environmental concern (using a median split on the environmental scale) (results not shown). These exploratory analyses were intended to determine if the three-way interaction was attributable to the effects of this combination (strong evidence, scientific agreement, political cues) on any one group of subjects, e.g. Republicans or environmentalists. The only group for which the interaction term on agreement and evidence strength was significant and positive was for those high in environmental concern (t=1.98, p=.05). Examining this effect in more detail, another regression equation was run for the four conditions with political cues (results not shown). A significant, positive, three-way interaction between scientific agreement, evidence strength, and concern for the environment was found (t=2.45, p=.02). 35

14 Table 4: Regression Model Predicting Concern for Global Warming and the Interaction Between Political Cues and Ideology 36

15 Table 5: Regression Model Predicting Concern for Global Warming and Agenda Setting Effects This study also included measurements of overall scientific knowledge and need for cognition as possible alternate influences on concern for global warming. However, neither of these variables was significant as a main effect in any of the regression equations predicting concern for global warming. Likewise, scientific knowledge and need for cognition did not interact with other variables in the model. Lastly, demographic variables such as gender, education, and race impact concern for climate change to a lesser degree when compared to concern for environment, although these demographic variables are not uniformly consistent in predicting the dependent outcome. 4. Discussion The focus of this study was threefold: First, the moderating effects of scientific agreement on cognitive processing of news related to climate change were assessed. Second, the 37

16 agenda setting effects of global warming news, and the relationship between partisan cues, political ideology, and concern for global warming were explored. Third, the impact of concern for the environment, scientific knowledge, need for cognition on concern for climate change was assessed. Toward this end, participants (n = 487) were randomly assigned to one of eight conditions that were experimentally manipulated to form eight main conditions, varying the level of scientific agreement, the strength of scientific evidence, and the presence of political cues, and two control conditions. The results showed that none of the main experimental manipulations had any independent influence on concern for this issue: in other words, whether scientists were presented as agreeing or disagreeing, whether scientific evidence was strong or weak, and whether political cues were present or absent did not influence concern for global warming. In all cases (including for the two baseline conditions), concern for global warming was primarily explained by environmental concern, and, to a lesser extent, by political ideology. Two experimental effects did appear, however. First, among those high in environmental concern, evidence strength influenced concern (that is, stronger evidence led to more concern) when scientists agreed and political cues were present. Second, when political cues appeared in a newspaper article on global warming, ideology became a stronger predictor of concern than when such cues were not present. Several implications emerge from these findings. This study found that elite scientific cues did not have a substantial impact on subjects concern levels for global warming: in other words, even when scientists were portrayed as agreeing, public concern for global warming did not increase. This finding is inconsistent with the literature on elite influence, which predicts that when elites agree public attitudes should move in the direction of elite consensus. This finding suggests that scientists ability to influence public concern may differ from that of other elites, perhaps because scientific information is more difficult for members of the public to use in informing their opinions. Likewise, this research did not support the hypothesis that elite agreement should have a moderating effect on cognition. This study hypothesized that strong scientific evidence should lead to more concern than weak evidence (indicating central route processing has occurred), but that this effect should not occur when scientists disagree (indicating that peripheral route processing may have occurred). This effect was not observed in all but one case. When scientists agreed and political cues were present, strong evidence led to greater concern than weak evidence among those high in environmental concern. This finding is contrary to the hypothesized role of political cues, and to research on elite influence which suggests that political cues serve as heuristic shortcuts and not as cues to attention. The most likely explanation for this finding is offered by the theory of affective intelligence (Marcus, Neuman, and MacKuen 2000). Affective intelligence specifies that cognitive processing is low effort unless there is some reason to expend additional cognitive effort. That is, the theory predicts that participants do not pay attention to scientific evidence unless they have some reason to do so: one reason for such attention is if the message content makes participants anxious. Thus, if political cues make those high in environmental concern anxious (e.g. because such cues remind environmentalists that 38

17 partisan politics have reduced U.S. attention to global warming) this anxiety may increase attention to message content and thus make them more concerned about global warming when scientific evidence is strong and scientists are portrayed as agreeing. Another finding from this study, consistent with our predictions, is that when political cues were present, participants relied more heavily on ideology in influencing concern. That is, ideology became a stronger predictor of concern when political cues were present in a newspaper article. Whereas it may be evident that political cues may activate ideological beliefs, one surprising finding is that this effect was not more pronounced in the face of scientific disagreement. That is, political cues were as influential when scientists agree as when they disagree. This is surprising because research on elite influence predicts that under elite disagreement heuristic cues (e.g. political cues) are more likely to inform opinions, and this effect was not observed here. Of note, however, is that a slightly different finding appears if partisanship (i.e. Republican (1), Independent (0), Democratic (-1)) is used instead of ideology in predicting these effects (regressions not shown). When partisanship is used instead of ideology, a strong three-way interaction appears between disagreement, political cues, and party identification, such that Republicans are much less concerned about global warming, and Democrats are much more concerned, when political cues are present and scientists disagree. 3 This finding is consistent with the literature on elite influence, which suggests that polarization occurs under conditions of elite disagreement. Other possible explanations for concern for global warming examined in this study, including scientific knowledge levels, need for cognition, and age, did not appear to have any effect on concern for global warming. However, there was some evidence that being White and non-hispanic, and, possibly, having a higher level of education or being male, reduced concern. It is difficult to draw very strong conclusions from these demographic findings, however, because this sample is not representative of any particular population. Demographic measures were included in the study to serve as controls only. Thus, it is inappropriate to generalize from these findings to any larger population than that of the study. However, these findings are broadly consistent with other literature on risk perceptions, which has shown that well-educated White males, in particular, appear to be less concerned about risks than other population groups (Slovic 1999). At least three possible explanations exist for why no moderating effect of agreement was observed except among those high in environmental concern when political cues were present. Ultimately however, we believe that the impact of these issues on this study is likely to be negligible. 3 Ideology and party identification were highly correlated in this study (r =.62), and the data on ideology were more complete than those on party identification. Thus, ideology in this study is used as a primary measure of political beliefs. However, to verify that findings would not be substantively different if party identification was used instead, regressions were re-run (Tables 2 and 3) using this alternate measure of political beliefs. This is the one case where the two measures yield substantively different results. 39

18 First, it is likely that the results observed are a function of the topic chosen for the study, i.e., global warming. This topic was chosen because previous research suggested that the public was still making up its mind about this issue, leaving the media well positioned to influence concern (Rogers 1999; Stamm, Clark, and Eblacas 2000). However, these findings suggest that perhaps this is no longer the case. For example, almost one in ten participants in the study (8%) selected all their answers from either the highest concern choice or the lowest concern choice, indicating that they have strong opinions on this issue. (This problem is likely magnified by the fact that this sample was over-educated, thus making it more likely that subjects had previously heard of this issue and reached some conclusions on it.) Additionally, there was not a great deal of variation in concern levels for global warming that is not accounted for by environmentalism or political ideology. These two variables alone explained 32% of concern on this topic, with demographic variables explaining an additional 2-4% of variation. The fact that even a basic agenda setting effect did not appear in this study suggests that this is an issue about which many people have already made up their minds. If this is the case, it is not surprising that the manipulations did not lead to significant impacts on overall concern. The extreme politicization of global warming that has taken place over the last decade and a half may also provide one reason why opinions on this issue are relatively fixed (Gelbspan 1997; Dunlap and McCright 2008). It is possible that elite influence on concern for global warming has already taken place via self-association with the environmental movement and via political party alignment. If this is the case, elite scientists may now have only a limited ability to impact public concern. Such an interpretation is consistent with other studies that have found that concern for global warming is primarily linked to general environmental beliefs, as opposed to correct scientific understanding of this issue (Bord, O'Connor, and Fisher 2000). Thus, it is possible that the hypotheses tested here would have been better supported by examining an environmental issue less familiar to the public or one where conflict has been less prominent. However, in the context of pre-test results, if opinions on this topic are indeed fixed, they remain fixed in the presence of what participants recognize as scientific agreement and strong evidence. Another possible explanation for why these hypotheses were not supported is that the experiment itself took place in conditions of relatively high distraction. Most participants completed their questionnaires in the relatively distracting outdoors, as opposed to the quiet of a lab setting. While this adds to the external validity of the findings that did appear (as newspaper reading usual occurs under conditions of distraction), it may have induced peripheral message processing. To the extent that participants were not attentive to the content of what they were reading, it is not surprising that message characteristics did not influence concern. Again, however, this explanation must be viewed alongside the finding that political cues did appear to have an impact on concern, both by increasing the importance of ideology in shaping attitudes, and by stimulating those high in environmental concern to weigh evidence strength under conditions of scientific agreement. Thus, the fact that message content did affect some attitudes suggests that such content was at least partially processed by the participants. 40

19 Finally, it is possible that the manipulation on evidence strength presented in this study simply was not strong enough. This particular manipulation was selected because it was fairly subtle and straightforward. However, the loudest manipulations on scientific evidence strength in the pretest had to do with the volume of scientific evidence for global warming; in other words, more arguments were rated as stronger arguments. While less subtle, a volume manipulation may have been more obvious to subjects than the one tested here and thus produced different effects. Nevertheless, the manipulation on evidence strength in this study was found to be significant in the pretest, and, therefore, some caution should be used before concluding that this manipulation was not strong enough to produce effects. Thus, none of these explanations can completely discount the main finding from this study: even in the face of scientific agreement, attitudes about global warming do not appear to be influenced by strong scientific evidence under most circumstances and for most people. This may be because newspaper readers do not centrally process the scientific information they encounter in the news media, because they do not have the scientific knowledge to effectively incorporate this information into their attitudes, because such information is not viewed as important, or because attitudes are fixed. All of these potential explanations have negative consequences for society and for scientific decisionmaking. Scientific evidence has a reasonable claim to make as important information upon which the public should base attitudes about issues such as global warming. For an issue with as many social, political, and economic consequences as global warming, a situation where evidence does not inform attitudes is certainly less than ideal. Additionally, while it is less clear for other topical areas that public attitudes should follow those of elites, scientific elites have a potentially more substantial claim to make as legitimate leaders of public attitudes on scientific issues. At least in this study, however, scientific elites do not appear to be influential. Public attitudes on global warming were remarkably insensitive to scientific information in the form of both cues to scientific agreement and in objectively stronger scientific evidence. This suggests a significant gap between the way that scientific elites use scientific information and the way that the public uses this same information, and creates a potential mismatch between scientific opinion on topics such as global warming and attitudes held by the public. This mismatch may be especially pronounced to the extent that public attitudes about science influence policy decisions. 41

20 References Ader, Christine R A Longitudinal Study of Agenda Setting for the Issue of Environmental Pollution. J&MC Quarterly 72 (2): Atwater, Tony, Michael B. Salwen, and Ronald Anderson Media Agenda-Setting with Environmental Issues. Journalism Quarterly 62 (2): Bann, Carla M., and Michael J. Schwerin Public Knowledge and Attitudes Scale Construction: Development of Short Forms. Research Triangle Park, NC: RTI International. Bell, Allan Climate of Opinion: Public and Media Discourse on Global Environment. Discourse and Society 5 (1): Bord, Richard J., Robert E. O'Connor, and Ann Fisher In What Sense Does the Public Need to Understand Global Climate Change? Public Understanding of Science 9 (July): Boykoff, Maxwell T., and Jules M. Boykoff Balance as bias: global warming and the U.S. prestige press. Global Environmental Change 14: Climate change and journalistic norms: A case-study of US mass-media coverage. Geoforum 38 (6): Brittle, Christine Global Warnings: The Impact of Scientific Elite Disagreement on Public Opinion, Communication Studies, University of Michigan, Ann Arbor. Bushman, Brad J., and Craig A. Anderson Media Violence and the American Public. American Psychologist 56 (6/7): Cacioppo, John T., Richard E. Petty, Jeffrey A. Feinstein, and W. Blair G. Jarvis Dispositional Differences in Cognitive Motivation: The Life and Times of Individuals Varying in Need for Cognition. Psychological Bulletin 119 (2): Cohen, Jacob A Power Primer. Psychological Bulletin 112 (1): Corbett, Julia B., Jessica L. Durfee, Roger D. Gunn, K. Maja Krakowjak, and Jeffrey T. Nellermore Testing Public (Un)Certainty of Science: Media Representations of Global Warming. Paper read at International Conference on Public Communication of Science and Technology, at Capetown, South Africa. Drake, Frances Global Warming: The Science of Climate Change. London: Arnold. Dunlap, Riley E An Enduring Concern. Public Perspective (September/October): Dunlap, Riley E., and Aaron M. McCright A Widening Gap: Republican and Democratic Views on Climate Change. Environment: Science and Policy for Sustainable Development (September/October):12. Dunwoody, Sharon Scientists, Journalists, and the Meaning of Uncertainty. In Communicating Uncertainty: Media Coverage of New and Controversial Science, edited by S. M. Friedman, S. Dunwoody and C. L. Rogers. Mahwah, NJ: Erlbaum Associates. Erbring, Lutz, Edie N. Goldenberg, and Arthur H. Miller Front-Page News and Real-World Cues: A New Look at Agenda-Setting by the Media. American Journal of Political Science 24 (1): Gelbspan, Ross The Heat is On: The High Stakes Battle over Earth's Threatened Climate. New York: Addison-Wesley. Hitz, Samuel, and Joel Smith Estimating Global Impacts from Climate Change. Global Environmental Change 14: Iyengar, Shanto, and Donald R. Kinder News That Matters. Chicago: University of Chicago Press. 42