全球变化知识资源中心

The scientific community now recognizes that global climate change is primarily caused by human activities and is already having significant negative impacts⁸. Despite this link, less than half of Americans believe anthropogenic climate change is occurring and it continually ranks at the bottom of national priorities⁹. In light of this discrepancy, scholars have examined, in part, how attention to these news stories may influence relevant attitudes and beliefs¹. However, individuals often selectively view and interpret information in ways that reinforce previously held beliefs^{10, 11}. Thus, more information about climate change in the public sphere has the potential to amplify, rather than attenuate, public polarization and to fail to motivate public action on the issue¹².

The selective interpretation of factual information has raised challenges to a model of science communication often termed the scientific literacy model. Within the literacy model, increasing scientific literacy by disseminating of factually accurate scientific information through formal (for example, schools) and informal (for example, mass media) channels will move public opinion towards consensus scientific perspectives, and help promote public support in line with scientific views of societal issues such as global climate change^{3, 4, 5}.

In contrast, the motivated reasoning model posits that individuals ‘work backwards’ and process information in a biased manner to reach conclusions consistent with previously held beliefs⁶. People do not approach evidence and arguments about controversial issues in a purely rational, even-handed manner⁷. Instead, an individual’s beliefs and political ideology strongly biases how they respond to information through selective attention, comprehension and/or recall^{6, 7, 13}. For example, those with greater issue involvement or with strongly held opinions are less likely to modify their beliefs when confronted with new information and so will frequently ignore, misinterpret, counter-argue or derogate ideologically incongruent evidence^{14, 15, 16}.

Recent work has found that attention to environmental news can increase knowledge about climate change, which can, in turn, influence both perceptions of harm and policy support¹. This research, however, has typically conceptualized ideology only as a control variable and does not account for how political ideology may act as a moderator that amplifies or dampens the influence of media discourse^{1, 2}. In addition, research has often focused on the contrasting validity of the literacy model against the motivated reasoning model^{17, 18}, but rarely has examined how these two models may operate simultaneously. This is critical, as the partisan divide for politically polarizing issues such as climate change has been shown to fundamentally alter how individuals interpret issue-relevant information¹⁹, creating a belief gap in which ideology may influence differential knowledge gains^{20, 21}. Knowledge gains may still influence perceived harm and policy support, although the nature of this relationship may also be dependent on political ideology.

Previous knowledge and political ideology are also likely to interact with systematic differences¹² in how climate change is presented in different types of news story. Reporting in political news stories is more likely to engage in ‘false balance’, in which climate skeptics are provided equal time with individuals offering the scientific consensus on climate change, which results in an over-representation of contrasting ‘inaccurate’ information^{22, 23, 24}. Reporting in science and environmental stories, however, tends to offer more accurate information in describing climate change and less likely to engage in the false balance²³. On the basis of these differences, attention to political news is more likely to generate belief gaps based on political ideology; as the diversity of views on political news, which include more skeptical and scientifically inaccurate information, is likely to amplify polarization as individuals latch on to information that reinforces previously held beliefs.

The present study seeks to investigate the role of the motivated reasoning and literacy models in explaining the relationships between attention to science/technology news and political news, knowledge, perceived harm of climate change, and support for climate mitigation policies. Specifically, we propose that attention to scientific and political news will differentially influence policy support through the mediators of knowledge and harm, with moderation by political ideology such that attention to political news will have an overall negative association with policy support for conservatives but not for liberals, and attention to scientific news will have an overall positive association for liberals but not for conservatives. We also investigate how knowledge may directly impact policy support and indirectly impact policy support through harm, as moderated by political ideology (see Fig. 1 for conceptual map).

⁺p < 0.10,^∗p < 0.05,^∗∗∗p < 0.001. For all links, the top number is the unstandardized coefficient for the direct effect for strong liberals and the bottom number is the unstandardized coefficient for the direct effect for strong conservatives.

Data from this study are drawn from a nationally representative sample collected by GfK Knowledge Panel in August 2009 (N = 1,673, completion rate = 73.5%, cumulative response rate 1 = 9.1%; ref. 25).

Analysis.

Models were fitted using structural equation modelling in Mplus, version 7.

Measurement.

Variables in the model included attention to political news, attention to science news, knowledge, harm, and policy support. Indicators for harm included three items that measured: How much do you think global warming will harm: (1) plants and animals, (2) people in the United States, (3) you and your family. Responses ranged from 1 (not at all) to 4 (a great deal). There were eight indicators for policy support, including assessing support for a ‘carbon tax’, developing a new international treaty on global climate change, and several other climate policies (see Supplementary Methods for all items). Responses ranged from 1 (strongly oppose) to 7 (strongly support). Also included as observed variables were attention to political news—measured with an item that asked ‘Generally speaking, how much attention do you pay to news and information about the following topics?: News about National Politics’—and attention to science news that asked about ‘News about Environmental Issues’, and ‘News about Science and Technology’ (averaged). Responses ranged from 1 (none at all) to 4 (a great deal). Knowledge was measured with seven items that assessed knowledge about global warming, for example ‘The hole in the ozone layer is the primary cause of global warming’ (see Supplementary Methods for all items). Correct responses were summed to form an eight-point scale (0–7). Political ideology was measured on a seven-point scale from 0 (Extremely liberal) to 6 (Extremely conservative).

Controls included media use, environmentalist, evangelical, age, education, sex, income, and race (white). Indicators for media use included four items that measured: ‘During a typical week, how often do you use each of the following sources to get news and information?: (1) Newspapers (either print or online), (2) Internet news sites like Google or Yahoo, (3) Blogs or other individuals’ web sites, (4) TV News (local, national, cable, and so on).’ Responses ranged from never to 7 days a week. Environmentalist was measured with an item that asked: ‘Would you describe yourself as an environmentalist?’ with responses (1) Yes, definitely, (2) Yes, somewhat, and (3) No. Evangelical was measured with an item that asked: ‘Would you describe yourself as a born-again or evangelical Christian?’ (1) Yes, (2) No. Age, Education (from 1, no formal education to 14, professional or doctorate degree), sex, income, and racially white were also included as controls in the structural model.

Missing data were handled using hotdeck imputation²⁶ where respondents missing data on a given item were imputed a value randomly selected from among respondents who matched their age and sex. This method allows all cases (N = 1,673) to be retained and is known to be superior to listwise deletion²⁷.

First a measurement model was fitted, with all latent and observed model variables. Model fit was slightly subpar (root mean square error of approximation (RMSEA) = 0.072 {0.069, 0.076}; comparative fit index (CFI) = 0.914), and modification indices suggested that correlating the errors for media use on Internet news sites such as Google or Yahoo and on blogs or other individuals’ web sites; for harm to people in the United States and to you and your family; and for two of the policy support indicators (‘Do nothing to directly address global warming but rather limit government regulation, spending, and taxation to encourage economic growth’ and ‘Let the free market, economic competition, and technology attempt to address global warming without government action’) would substantially improve model fit. Once these correlations were included, model fit was adequate (RMSEA = 0.054 {0.050, 0.057}, CFI = 0.954).

Next, a structural model was fitted (Fig. 2), with all structural paths moderated by political ideology (accomplished by including multiplicative terms in the model, for example, attention to science news ^∗ ideology). As the interaction between harm and ideology was a latent interaction (using the XWITH command in Mplus), model fit for the structural model was not able to be determined (Mplus cannot determine model fit with latent interactions). However, given that the primary interest for this analysis is to assess relationships between variables, rather than model comparison, and that model fit for the measurement model was adequate, the inability to determine model fit was not a substantial barrier. Path coefficients at the specified levels of political ideology were determined by centring ideology at the specified level and re-estimating the model.

1. Zhao, X., Leiserowitz, A. A., Maibach, E. W. & Roser-Renouf, C. Attention to science/environment news positively predicts and attention to political news negatively predicts global warming risk perceptions and policy support. J. Commun. 61, 713–731 (2011).
   • ISI
   • Article
2. Ding, D., Maibach, E. W., Zhao, X., Roser-Renouf, C. & Leiserowitz, A. Support for climate policy and societal action are linked to perceptions about scientific agreement. Nature Clim. Change 1, 462–466 (2011).
   • ISI
   • Article
3. Sturgis, P. & Allum, N. Science in society: Re-evaluating the deficit model of public attitudes. Public Underst. Sci. 13, 55–74 (2004).
   • Article
4. Bauer, M. W., Allum, N. & Miller, S. What can we learn from 25 years of PUS survey research? Liberating and expanding the agenda. Public Underst. Sci. 16, 79–95 (2007).
   • Article
5. Brossard, D. & Lewenstein, B. V. in Communicating Science: New Agendas in Communication (eds Kahlor, L. & Stout, P.) 11–39 (Routledge, 2009).
6. Kunda, Z. The case for motivated reasoning. Psychol. Bull. 108, 480–498 (1990).
   • CAS
   • ISI
   • PubMed
   • Article
7. Taber, C. S. & Lodge, M. Motivated skepticism in the evaluation of political beliefs. Am. J. Polit. Sci. 50, 755–769 (2006).