Abstract
How can public support for science be encouraged? In early August 2016, a Zika vaccine entered its first human trial. Extensive media coverage followed. Using repeated cross-sectional surveys, we observed that, following this media coverage, survey respondents reported greater attention to Zika news and an increased trust in science as providing solutions to problems. Whereas the increase in attention was long-lived, lasting 6 weeks or more, the increase in trust was short-lived, lasting only 2 weeks. This result provides insight into the duration of media effects and suggests conditions that facilitate trust in science.
For over four decades, General Social Survey data have shown remarkably stable U.S. attitudes toward science (Smith & Son, 2013), a pattern consistent with the conclusion of the National Science Foundation’s 2016 Engineering and Science Indicators report that “Americans’ overall attitudes about science are either stable or becoming more positive” (National Science Board, 2016, Chap. 7). In 2014, the Engineering and Science Indicators found that 41% of Americans report “a great deal of confidence” in leaders of the scientific community, 49% say they have “only some confidence,” and 8% indicate “hardly any confidence at all,” results that “are nearly identical to 2012 and are similar to previous years.” Maintaining a high level of credibility in science is important because credibility predicts audiences’ levels of science-congruent beliefs (Bolsen, Druckman, & Lomax Cook, 2015).
However, public trust in at least some areas of science and some scientific agencies may be subject to short-term event-specific downdrafts. In particular, when scientific spokespersons offer assurances that are later shown to be false, the credibility of their institution is harmed, and such future assurances are less likely to be accepted. For example, mishandling of the Bovine Spongiform Encephalopathy (“Mad Cow Disease”) crisis in Britain led to a loss of trust in governmental institutions and health advice (Jasanoff, 1997). Similarly, attitudes toward the Centers for Disease Control and Prevention (CDC) dropped when the CDC made the inaccurate claim that U.S. hospitals were prepared to deal with the Ebola virus (Dutton, De Pinto, Salvanto, & Backus, 2014).
These findings show that false reassurance can harm the credibility of science and scientific agencies. But can the credibility of science be maintained or promoted, and, if so, how long do such influences last? Although a rich literature has considered public trust in science on a year-by-year basis, few surveys have had the necessary resolution to detect and observe short-term fluctuations in public trust. Often, reports take the form of case studies (e.g., Jasanoff, 1997) and do not have the necessary survey data to quantify changes in opinion.
In this study, we report weekly repeated cross-sectional data on a survey indicator of public trust in science. Observing this indicator’s behavior across time, we locate a sharp improvement following the announcement and media coverage of a Zika vaccine in human trials. This increase lasted 2 weeks, after which the indicator returned to its grand mean. This study suggests that news of scientific advances may improve public opinion of science, particularly when those advances address problems that the media cycle has placed on the public agenda.
When Might Communication Improve Attitudes Toward Science?
It is unclear under what conditions scientific knowledge improves public attitudes toward science. It was once assumed that greater scientific knowledge would lead to greater appreciation for scientific research, belief in scientific conclusions, and support for the scientific enterprise. Under this “knowledge deficit” model, the public would express greater approval of science if it knew more of what scientists knew. In contrast, recent research suggests instead that attitudes are determined by not just scientific knowledge but also cultural norms, value systems, and social identities (Allum, Sturgis, Tabourazi, & Brunton-Smith, 2008; Scheufele, Corley, Shih, Dalrymple, & Ho, 2008). Science may be evaluated positively when it provides beneficial applications and negatively when it is expensive or threatens personal values.
With this in mind, news of the Zika vaccine may improve attitudes toward science. A Zika vaccine has clear benefits, improving human health by preventing birth disorders, and vaccination does not conflict with mainstream public values or cultural norms. Additionally, media coverage of the Zika virus has established Zika prevention as a matter of public concern and general importance. These attributes make this news cycle a useful opportunity to observe the relationship between news of scientific progress and public attitudes toward science.
Zika Vaccine as Communication Stimulus
The context surrounding the Zika vaccine suggests that communication about the vaccine might improve public attitudes toward science. First, news of the vaccine arrived in the middle of the issue-attention cycle, at a time when media coverage had placed Zika and its prevention on the public agenda. Second, the announced vaccine was not controversial; the vaccine was not (yet) connected to conflicts in values or politics. This contrasted favorably to other proposed Zika remedies such as the aerial spraying of pesticides or the release of genetically modified mosquitos, which provoked environmental and bioethical concerns. In the presence of such agenda setting, and in the absence of such value-conflicts, it is possible that news of scientific advances could lead to greater credibility of science. Repeated cross-sectional data permit the necessary temporal resolution to observe such an effect and would allow estimation of its duration.
To explore the relationship between science news and public attitudes toward science, this study examines the behavior of a single general science attitude item across time in the context of an emerging science-relevant news cycle. Relationships with demographics are explored to determine which demographic populations, if any, are more (or less) trusting of science. Furthermore, when trust is observed to change across time, we test for potential mechanisms of such change.
Method
The study combines Google Trends data on search terms in the United States, content analyses of news coverage from Factiva and LexisNexis, and 31 weeks of a nationally representative cross-sectional survey.
Google Trends
Google Trends data were recovered for the search terms “Zika” and “Zika vaccine.” Google Trends searches were limited to a 90-day period from July 21 to November 18. Trends results were restricted to those Google searches performed in the United States.
Content Analysis of Media
News Reports on Government Agencies and the Zika Virus
LexisNexis searches for “Frieden AND Zika,” “Fauci AND Zika,” and “(‘Centers for Disease Control’ OR CDC) AND Zika” retrieved news stories between July 15 and September 20, 2016, on the Zika virus.
News Reports Mentioning a Zika Vaccine
A Factiva search retrieved broadcast, cable, print, and online media containing at least one mention of the search term “Zika.” Articles between July 1 and August 10 were coded as mentioning that there was no vaccine and/or that scientists were working on a vaccine. A full methodology is available in the supplement, hosted at https://osf.io/3wcz4/.
Survey Sample
Data for this study come from the Annenberg Science Knowledge survey, a national, weekly, dual-frame bilingual telephone survey. Each wave consists of 1,000 computer-assisted interviews, 600 of which are performed via cell phone, and about 35 Spanish-language interviews. Landline interviews are performed via random digit dialing, and cell phone interviews are performed via randomly generated cell numbers. Data are weighted to be nationally representative.
Following August 8, the survey began an additional oversample to better examine those most likely to be affected by Zika. For this oversample, an additional 500 surveys were performed with respondents from Florida.
The current study examines Waves 1 through 31 of this survey, yielding a sample of 34,266. Sample size in presented models will vary as a function of missingness in the data. See the supplement (https://osf.io/3wcz4/) for methodological details including weekly response rates.
Predictors
Time
Responses were examined as a function of the survey wave. Waves were collected Thursday through Monday of each week and are labeled by the last day of data collection.
Media Habits
Respondents were asked about the amount and partisanship of their news according to how much of their news they got from three broad categories: sources such as “Fox News, Rush Limbaugh, The Drudge Report, or Glenn Beck”; “MSNBC, NPR, Bill Maher, or Huffington Post”; and “ABC News, CBS News, NBC News, or CNN.”
Respondents were also asked about their attention to Zika news: “How familiar are you with news reports about the ZIKA virus?” (1 = very unfamiliar to 4 = very familiar)
Demographics
Demographics used for weighting and as covariates included age, sex, race (indicator codes for African American and Asian American), ethnicity (indicator code for Hispanic), religion (indicator codes for atheism, Judaism, and Islam), political ideology, political party identification, and education. Income information was collected, but an excess of missing responses made it unsuitable for use.
Outcomes
Confidence in Science
Respondents were asked, “Which comes closer to your view: science enables us to overcome almost any problem or that science creates unintended consequences and replaces older problems with new ones?” Responses were coded 1 = “Science enables us to overcome almost any problem” or 0 = “Science creates unintended consequences and replaces older problems with new ones.” Volunteered answers of “both” or “it depends” were scored as 0.5. This item, comparing the benefits and harms of science, is conceptually similar to that used by Bolsen et al. (2015) and the Science and Engineering indicators (National Science Board, 2016).
Opinion of Government Health Agencies
Respondents were asked, “Is your overall opinion of [Agency] very favorable, mostly favorable, mostly unfavorable, or very unfavorable?” Opinions were measured for the CDC and the National Institutes of Health (NIH).
Analysis
Results were analyzed by multiple regression using the svyglm function in the “survey” package for R (Lumley, 2016). This function performs linear models in survey designs with population weights.
Results
We present first our analyses of the general science indicator, followed by analyses of more specific feelings toward the CDC and NIH, as well as appraisals of the federal government’s ability to handle a Zika outbreak.
Confidence in Science
We examine, in turn, the behavior of this measure across demographics, the behavior of this measure across time, and finally, whether moderating variables help explain changes across time.
Who Endorses This Item?
A multiple regression was used to examine the relationships between demographics and the confidence-in-science outcome.
Results are given in Table 1. Older respondents, African Americans, women, and conservatives were more likely to express concerns that science has unintended consequences. Atheists and higher educated respondents were more likely to affirm that science enables us to overcome almost any problem.
Multiple Regression Model of Demographic Variables’ Associations With the Confidence in Science Measure.
Note. Sex is coded 1 = male, 2 = female. Black, Hispanic, and Asian are indicator codes with White as the reference category. Atheist, Jewish, and Muslim are indicator codes with Christian as the reference category. Education is coded on an 8-point scale from 1 = less than high school to 8 = postgraduate or professional degree.
Was There Change in the Item Across Time?
Contrast codes were used to indicate weekly deviations from the grand average across the entire 30-week period. Because of the number of contrasts performed (27 contrasts), a more restrictive alpha threshold was used to evaluate statistical significance (α < .01). By this criterion, 4 weeks were significantly deviated from the grand average: The weeks of July 11 and July 18 were significantly below the average (p < .001, p = .005), and the weeks of August 8 and August 15 were significantly above the average (p < .001).
What Explains Change in This Item?
To improve our understanding of the changes across time in this measure, we first compared the confidence-in-science outcome with Google searches and news reports. The increase in trust followed sharp increases in Google searches for “Zika vaccine” and news reports featuring NIH director Anthony Fauci, CDC director Tom Frieden, and news related to a Zika vaccine (see Figure 1).
Media coverage, Google searches, and survey trust indicator across time.
Next, we applied multiple regression models to look for subgroups that may have driven the changes in the overall mean. Specifically, we tested whether certain predictors demonstrated interactions with the week of survey data. Significant moderation would indicate that a predictor’s relationship with the outcome grew stronger or weaker for that week. For example, one might expect that watching television news might have a slight overall association with the confidence-in-science outcome, but when the television news contains reports of progress on a Zika vaccine, watching television news might have a stronger, more positive association with the science outcome.
Attention to Zika news
We hypothesized that, insofar as the change in confidence in science was related to the news regarding the Zika vaccine, those who paid more attention to Zika news would demonstrate greater changes. Respondents’ attention to Zika-related news was entered into the regression model, adjusting for the demographic variables listed above. A main effect of attention to Zika news was significant, such that attention was positively related to endorsement of science as overcoming almost any problem (b = 0.01, t = 2.60, p = .009). However, attention to Zika news did not moderate the relationship (both p > .49). Rather, attention to Zika news demonstrated a significant increase that began August 8 and remained significant for the following weeks (b > 0.14, t > 3.6, p < .001). This result suggests that news attention did not moderate changes in science confidence; instead, the news cycle may have increased attention to Zika news through a prolonged agenda-setting effect.
Media habits
We entered our three measures of liberal, conservative, and moderate media use into the model, again looking for illuminating interactions. Liberal and moderate media diets were associated with greater endorsement of the science indicator (b = 0.01, t = 5.86, p < .001; b = 0.01, t = 2.91, p = .004, respectively). Conservative media diet was slightly and nonsignificantly associated with lower endorsement of the science indicator (b = −0.004, t = −1.88, p = .06).
Regarding interactions, however, none of the media measures interacted with time such that it explained the 2-week increase in confidence in science. An exploratory analysis suggested that a significant Week × Moderate media diet interaction may have contributed to the measure’s dip in mid-July. For the three consecutive weeks of July 3, July 11, and July 18, the relationship between moderate media diet and the science indicator was significantly lower than the average across all weeks, b = −0.03; t = −2.51, −2.75, and −2.44; p = .012, .006, and .015, respectively. However, no interaction term was statistically significant in August, and therefore media diet did not help explain the science indicator’s increase.
Opinion of Federal Agencies
As related measures, we examined respondents’ opinions of the CDC and the NIH. Insofar as news of a Zika vaccine was linked to these organizations, the news cycle may have warmed opinions toward these institutions. These opinions were fairly strongly correlated with each other (Spearman’s ρ = .50) but only modestly correlated with the science indicator (both Spearman’s ρs = .18).
National Institutes of Health
Opinion of the NIH was lower among conservatives and older respondents. Opinions were more favorable among women, Hispanics, Asian Americans, and higher educated respondents. Relative to Christians, atheist and Jewish respondents had higher opinions of the NIH, whereas Muslim respondents had lower opinions.
Unlike confidence in science, opinion of the NIH did not show significant change over the 2-week period. No week differed from the grand average by an amount significant at the .01 level (see Figure 2).
Opinion of the CDC, NIH, and federal preparedness to handle a Zika outbreak, and the trust indicator across time.
Centers for Disease Control and Prevention
Opinion of the CDC was similarly lower among conservatives and higher among Hispanics, atheists, and higher educated respondents. Unlike the science indicator, opinion of the CDC was very stable across weeks; no week differed from the grand average (all p > .05). Again, see Figure 2.
Federal Preparedness for a Zika Outbreak
If the public were led to feel more trusting in science by news of a Zika vaccine, they might also feel more trusting toward science-related governmental functions. In particular, progress on a Zika vaccine could reflect a broader federal preparation to deal with the Zika virus. Thus, we also examined whether respondents felt the federal government was prepared to deal with a Zika outbreak in their area. Unlike the science indicator, this measure did not significantly increase following the news cycle (for August 8, b = .04, t = 1.09, p = .278; for August 15, b = −.04, t = −1.08, p = .282; both coefficients relative to the grand average). Again, this stability is evident in Figure 2. However, ratings of federal Zika preparedness were only modestly correlated with confidence in science (r = .20), suggesting that the two measures were related to relatively distinct attitudes.
Discussion
Drawing on nationally representative survey data, this study found that widespread news coverage of the first human trial of the Zika vaccine was associated with a significant, albeit short-term, increase in an otherwise stable indicator of confidence in science.
These results suggest that it is possible to bolster the credibility of science when science provides uncontroversial solutions to problems covered by the media. We note, however, that the duration of this bolstered credibility is brief compared to increase in attention to Zika news; the credibility measure had returned to baseline after 2 or 3 weeks, whereas the attention measure remained elevated for at least 6 weeks. These results provide insight into the duration of such media effects. The change in trust we observed had a duration consistent with media effects produced by political campaign stimuli, for example, effects of political messages in ads (Hill, Lo, Vavreck, & Zaller, 2013; Kenski, Hardy, & Jamieson, 2010).
This finding opens the possibility that confidence in science could be bolstered in a more sustained fashion by regularized communication about advances made by science. These communications may be particularly effective when they provide potential solutions to problems placed by media on the national agenda. However, such effects are likely to be relatively brief even under the best of circumstances. Additionally, such communications may backfire if the public feels that a problem or its solution is overstated for the personal benefit of scientists, politicians, or the media, although further empirical research is needed (see Master & Resnik, 2013).
Results also suggest the limits of such increases in public opinion. Curiously, opinions toward the CDC or the NIH did not warm despite the considerable increase in media representation by the directors of those agencies, nor did respondents report feeling that the federal government was more prepared to manage a Zika outbreak. We note that although there was a substantial increase in Zika-related news and that Fauci and Frieden are often quoted in those reports, they are generally not the central focus of the article. A more thorough content analysis would be necessary to determine the frames used in these articles.
Limitations
There are several limitations to this research. First, the confidence in science item is double-barreled—shifts in the item may reflect increased agreement that “science enables us to overcome almost any problem,” or it could reflect decreased agreement that “science creates unintended consequences and replaces older problems with new ones.” Additionally, it is a single item taken from a survey of several dozen questions, and this item is explored post hoc because of its sharp movement. The research conclusions are therefore less firm than those that would be drawn from an a priori research hypothesis tested with a battery of items with good psychometric properties.
Second, related governmental items did not change in the same time span. On the one hand, it is possible that opinions of the CDC, NIH, and science should be linked, and therefore the inactivity of these other measures indicates that the change in science attitudes is a fluke. This possibility is particularly salient given that previous reports have found evidence for news events affecting the credibility of specific institutions (e.g., Dutton et al., 2014; Jasanoff, 1997). On the other hand, respondents may consider and evaluate science and governmental institutions as separate entities. One could have positive attitudes toward science but not toward organizations such as the CDC or NIH, as indicated by the relatively weak correlations between the science indicator and the institutional opinion measures.
Third, we did not ask respondents whether they had heard about the progress on a Zika vaccine. This limitation prevents us from estimating the difference in science attitudes associated with (un)awareness of the vaccine news. Our conclusions are correspondingly more circumspect.
Finally, we did not observe significant moderation of the relationship between media diet and the science indicator across time. Such a moderation would have helped identify and explain the increase as a media effect.
Conclusion
When members of the public evaluate whether science is a positive and trustworthy force in society, they may consider cognitively accessible scientific news. When recent and salient news reports describe scientific progress toward solutions for salient problems, publics may form more positive appraisals of science. Unless reinforced, such increases in science trust are likely to be short-lived even under the best of conditions.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Annenberg Science Knowledge survey was funded by the Science of Science Communication endowment of the Annenberg Public Policy Center of the University of Pennsylvania. The endowment was created from revenue generated by the APPC endowment funded by the Annenberg Foundation and managed by the University of Pennsylvania.