Changing evangelical minds on climate change

One quarter of Americans self-identify as evangelical and a strong majority of these reject human-induced climate change. Can one lecture on climate science from a Christian perspective change their minds? Here, we examine the response of undergraduates at three evangelical institutions, located in upper state New York, Texas, and Canada, to a recorded lecture by an evangelical climate scientist. Pre/post-test surveys posed six questions from the Global Warming’s Six Americas instrument with additional questions on participants’ political and theological perspectives. All populations showed significant pre/post-test gains on almost every question, immediately after the intervention, with gains among the most conservative population being as large as the others. A one-month delayed post-survey showed that gains were still significant relative to pre-test values for most of the questions. Furthermore, gains did not decrease when the Christian frame was removed, except on one item related to Christian responsibility. These results suggest that even a limited exposure to accurate information can change minds, and that the context in which the information is presented (here, the evangelical college and faculty sponsor) may be more important than the content frame.


Introduction
In the United States, political affiliation is the primary predictor of the extent to which individuals accept that climate is changing due to human activities and are concerned over the potentially serious impacts (Leiserowitz et al 2018). This holds true for the general public as well as science teachers (Plutzer et al 2016), and grade 6-20 educators from regions across the country (McNeal et al 2014b). Even in Canada, there are significant differences regarding climate change acceptance between those who self-identify as Conservative versus Liberal, NDP, or Bloc Quebecois (Lachapelle et al 2014, Poushter 2016).
One quarter of Americans self-identify as evangelical (Lifeway Research 2017). Some evidence suggests that an overwhelming majority of them supported the Republican presidential candidate in the 2016 election (Bailey 2016), although this is contested by some (Carter 2016). At the same time, they overwhelmingly reject anthropogenic global warming (Chesnes and Joeckel 2013, Arbuckle and Konisky 2015, Branch et al 2016. This is seen in the Cornwall Alliance's 'Evangelical Declaration on Global Warming,' signed by many evangelical theologians, church leaders, and some scientists (Cornwall Allliance 2009), intended to move evangelicals away from worrying about climate change. Some researchers have found that the linkage between religious beliefs and dismissal of anthropogenic climate change is not necessarily causal (Schmitt 2014, Ecklund et al 2017. Rather, political ideology may lead to this rejection, which is then seemingly justified with Christian theology 4 . However, it is the strong correlation between US evangelical beliefs and rejection of climate science that forms the basis for this study evaluating the efficacy of targeted interventions with this group. Studies indicate that less than 60 min of intervention with undergraduate students can change their understanding of climate change (McCuin et al 2014, McNeal et al 2014a, Rannay and Clark 2016, Gil 2017. When the intervention was more intensive, significant gains persisted two years later (Burkholder et al 2017). One study examined the effect of a controlled intervention specifically on students at Houghton College, an evangelical institution in rural, western New York (Webb and Hayhoe 2017). After either attending a live lecture or viewing a nearly identical recorded lecture on climate change by an evangelical climate scientist who framed a discussion of the science, impacts, and solutions within the context of Christian stewardship and concern for our global neighbors, students displayed significant pre-post gains in their understanding and acceptance of human-induced climate change and the need for solutions. This last study, while illuminating, raised a number of questions. Would students at evangelical colleges with populations that were more politically conservative show similar pre-post gains after the same presentation? Would these gains persist? And to what extent were gains influenced by the Christian frame and identity of the presenter, as opposed to the content or the context of the presentation?

Method
To address the questions raised by our previous study, we conducted a series of experiments with undergraduate students at two other evangelical undergraduate institutions with similar theological positions to that of Houghton College, but one which we assumed might be more politically conservative (Dallas Baptist University (DBU) in Dallas, TX), and one in Canada (Tyndale University College in Toronto, Ontario), using the same survey and recorded lecture as used at Houghton College. The recorded lecture was very slightly revised for the colleges in Texas and Ontario, to make some of the impacts more relevant to their geographic areas. Three studies were conducted at DBU, giving us five groups to compare in total: Houghton (February 2015, 88 participants), Tyndale (October 2015, 47 ps), DBU1 (November 2015, 51 ps), DBU2 (March-April 2016, 67 ps), and DBU3 (March 2017, 109 ps). The DBU2 participants also completed a delayed post-survey a month later.

Survey
The survey consists of two sections (appendix A). The first includes demographic questions (gender, year of study, major), religious questions (Christian denomination, beliefs about the Bible, beliefs about origins, influence of senior church pastor or youth leader), and political/economic questions (citizenship status, political beliefs, political party identification, views on the free market). Questions in the second section of the survey were taken from Global Warming's Six Americas surveys (Leiserowitz et al 2011). They include six questions about global warming (appendix A, #13-20, where answer categories for each question are also given): How sure are you that global warming is or is not happening?
If it is happening, is it mainly natural, mainly human caused, or a combination?
Do most scientists think that global warming is happening?
How worried are you about global warming? Do you think global warming is harmful? (to four different groups of people) What priority should be given to addressing global warming? (at four different levels) Both sections of the survey were included in the pre-test, but only the second was used in the post-test.
Participants were asked to create their own personal ID to write on the pre-test version, which they could easily remember to also use for the post-test version, so that the two could be matched in the analysis, and participants' identity kept anonymous.

Intervention
The 44 min recorded lecture was originally made for the study at Houghton College in February 2015 (see appendix B for an outline). Its introduction and conclusion present the difference between faith and science, and refer to Biblical verses on the stewardship of God's creation and on loving our neighbor as ourselves (i.e. caring for the global poor, likely to be most impacted by climate change). The rest of the lecture presents relevant scientific data, suitable graphs, and photos, around these four topics: For the Canadian experiment, some of the US-specific information on climate impacts was replaced with an equal amount of Canadian-specific information in the video. For the last study groups at DBU, to assess the impact of the Christian frame of the recorded lecture, the 44 min recorded lecture was edited down to 38 min by removing all Christian and Biblical references, found principally in the introduction and conclusion. We also removed the identification of the presenter as an evangelical. Since shortening the recorded lecture at Houghton College by a larger amount (10 min), by removing explanations of common misconceptions, had not significantly affected the prepost gain achieved there (Webb and Hayhoe 2017), we felt that the 6 min difference in length in the DBU3 study was not a critical factor, allowing us to focus on the effect of removing the Christian frame in this last study.

Population
At Tyndale, as at Houghton, participants were recruited with in-class and across-campus invitations, a promise of being anonymous, and a small incentive of winning a book coupon. At DBU, participants were enrolled in a non-science majors biology course, in three different semesters. No incentive for participation was offered, but class time was given, and anonymity was maintained. Very few participants at Tyndale (4%) and DBU (4%) were enrolled in a science or mathematics major, unlike at Houghton where more than a third were (36%). Almost all participants at Houghton (94%), Tyndale (98%), and DBU (99%) viewed the Bible as either 'the actual word of God and is to be taken literally,' or that 'it is the inspired word of God but not everything it should be taken literally' (appendix A). Three quarters of participants at Houghton (72%), Tyndale (77%), and DBU (71%) were female. Most participants at Tyndale were in their first year (70%), whereas at DBU and Houghton, participants were spread across the four undergraduate years (41% in their first year at DBU, 43% in their first year at Houghton). For some analyses, participants who were non-citizens of the country (<6%), or had left a question blank, such as the one on political party identification (<10%) or scientific consensus on climate change (<22%), were excluded. This somewhat reduced the number of participants for those analyses. (The tables include the number of participants in each analysis.)

Timeline
The four new studies took place between October 2015-March 2017. (The original Houghton study had taken place in February 2015.) All pre-test surveys were completed approximately a week before the recorded lecture presentation. The post-test survey was done immediately after the presentation, except at Houghton where participants were given until the next morning to complete it. (This potential problem is discussed below.) In the DBU2 study, a delayed postsurvey was done a month later. In the DBU3 study, participants were randomly divided into two groups of approximately equal size. Both groups answered the pre-test survey a week ahead of time, one watched the original 44 min recorded lecture while the other watched the 38 min recorded lecture with the Christian frame removed, and both answered the post-test survey immediately after.

Analysis
The six questions on global warming used Likert-scale responses that could be ordered, but were not in a ratio or interval scale (although it could be argued that the nine optional responses to the first question, ranging from extremely sure global warming is not happening to extremely sure it is, are paced roughly eight equal steps apart). Therefore, non-parametric analysis was necessary. The Wilcoxon Signed-Rank test was used to analyze the significance of the pre-post gains and delayed pre-post gain for the six global warming questions at the various colleges. The Mann-Whitney U test was used to compare pre-post gains from the two independent samples, in the DBU3 study, one group watching the Christian-framed video, and the other, the video with the Christian frame removed. (Online calculators for these tests were used, available at www.vassarstats.net or www.socscistatistics.com.) In the analyses one-tailed tests were used, as we hypothesized that post-test scores should be greater than pre-test scores, and the impact of a Christian frame in the recorded lecture should produce higher gains with Christian respondents than when the Christian frame is removed.

Results
Survey responses confirmed that the respondents in the three studies at DBU were politically more conservative than those at Houghton or Tyndale (figure 1(a)). A Kruskal-Wallace analysis indicates that the difference is statistically significant (H=25.259; N=358; p<0.000). Another question examining Democratic or Republican party identification yielded similar results (figure 1(b)). A Kruskal-Wallace analysis for this showed that the difference between Houghton and DBU respondents was statistically significant (H=30.074; N=273; p<0.000). (The results for Tyndale are not shown here, due to differing variety of political parties in Canada.) The first two research questions were: Would students at evangelical colleges with populations that were more politically conservative show similar prepost gains, immediately after the presentation, and would these gains persist? Since we assumed yes to this question, we used a one-tailed Wilcoxon signed rank test. We see from table 1 that participants in all five studies showed pre-post gains immediately after the intervention that were statistically significant at the p < 0.01 level for all six questions related to global warming, with two exceptions for Q2 (Tyndale, p < 0.360; DBU2 p < 0.022). Furthermore, these gains persisted a month later for four of the six questions, in the DBU2 study.
One way of displaying these non-parametric results visually is to graph the direction of movement from pre-to post-test regarding participants' climate change beliefs. This was done in a previous paper for Houghton College (Webb and Hayhoe 2017). For each of the six questions, the number who moved to a more pro-climate belief were graphed against the number who did not change and the number who moved to a less pro-climate belief. The direction-of-movement graphs for the Tyndale and DBU studies looked quite similar to those for Houghton and are not presented here.
Another way to display the results is to show the percentage of the pre and post-scores that were supportive of climate change concerns. For Q1-Q4, this would mean that respondents believed global warming was happening (somewhat, very, or extremely sure), was caused mostly by human activities, was agreed upon by most scientists, and was causing participants to be somewhat or varied worried (appendix, #13-15, #16, #17, #18). Figure 2 displays these percentage scores for the first four questions for each of the five studies.
The graphs indicate that pre-test scores for DBU participants were lower than for Houghton and Tyndale participants. Using the Kruskal-Wallis test for ordinal data between independent samples, we find that these differences in pre-test scores were significant for Q1 (H=28.16, p<0.000), Q3 (H=10.75, p<0.030), and Q4 (H=27.71, p<0.000), but not for Q2 (H=4.91, p<0.30). The graphs also suggest that participants in the DBU studies experienced greater movement (preto post-test scores) towards acceptance of  (92). The figure shows the percentage who responded to each of the five answers to the question: 'generally speaking, do you tend to identify more with the Democratic Party or the Republican Party?' Between 12% and 16% of participants failed to respond in each study. Some were international students. global warming than those in the Houghton and Tyndale studies, although with ordinal data it is difficult to see if this greater movement is statistically significant. Also, since DBU scores were lower to begin with, they had more room to move up. The last research question was to what extent were gains influenced by the Christian frame and identity of the presenter, as opposed to the content or the context of the presentation? We assumed that framing the climate change lecture with Biblical teaching on stewardship and concern for loving our global neighbors (appendix B), and introducing the climate scientist as an evangelical, could be at least partly responsible for the significant gains at the three evangelical institutions. We tested this assumption by removing all Christian content from the recorded lecture and from the introduction of the lecturer and found that it made no difference in pre-post gains of the DBU3 participants (table 2(a)). Only on the fourth question, related to worry, was there a possible significant difference where the group watching the recorded lecture without the Christian frame was actually more worried (although with a p=0.048, this correlation is not sufficient to note, when doing a six-fold comparison).
These results took us by surprise. So, we conducted a further analysis of the sub-questions that made up the last two principal questions. The results (table 2(b)) suggest a possible difference between the two groups on one of the sub-questions. The group of participants who watched the recorded lecture with the Christian frame (i.e. Christian stewardship, love for global neighbors, etc) scored higher on this question than the group that watched the recorded lecture without the Christian frame (p<0.009, one-tailed test). We are cautions, however, in attributing statistical significance to this as we have now done an eleven-fold comparison of the two groups, and the highest significance level is only 0.009.

Discussion
We assumed that the participants volunteering at DBU would be more politically conservative than those at Houghton and Tyndale, and our results confirm this. Both the political beliefs (figure 1(a)) and party identification ( figure 1(b)) of the participants at DBU were more conservative than those at the other college(s).  Table 2. (a) Tests of significance differences in pre-post gains for the Christian-framed versus non-Christian-framed recorded lecture groups for DBU3, on the six questions about global warming, using the Mann-Whitney U test for ordinal data. One-tailed tests were used (p<0.05 * ; p<0.01 ** ). For the last two questions, the mean of the scores on the sub-questions was used. (b) Tests of significance differences in pre-post gains for the Christian-framed versus non-Christianframed recorded lecture groups for DBU3, on the sub-questions for the last two questions about global warming, using the Mann-Whitney U test for ordinal data. One-tailed tests were used (p<0.05 * ; p<0.01 ** ). For the second last question, sub-question (c) was unintentionally omitted from the survey, and thus is not included in this Since we assumed this to be the case, we also expected that the views of the DBU participants concerning anthropogenic climate change would be less receptive, and that perhaps even the pre-posttest gains, immediately after the intervention, would be less than at Houghton and Tyndale. However, a Wilcoxon signed rank test indicated statistical significance at the p<0.01 level for each of the six questions in all five studies, with only two exceptions (table 1). Figure 2 shows these gains for the first four questions. Although DBU participants were initially more skeptical about global warming than those at Houghton and Tyndale, after the intervention their beliefs were almost as high as at the other two universities (figure 1). This is especially true of their acceptance that global warming is happening (Q1, figure 2(a)), is mostly caused by humans (Q2, figure 2(b)), and is something that most scientists agree on (Q3, figure 2(c)). Of course, those who are initially most negative have the greatest room available for gain; but it cannot be assumed that therefore they will achieve the greatest gain, especially if they are more politically conservative. Yet, these results suggest that they did achieve a gain as great or greater than that of the less politically conservative respondents. These results are encouraging for those who give presentations to politically conservative populations of college students. Whether they can be extrapolated to politically conservative adults in general is a question for further research. However, it is a hopeful sign that some groups of politically conservative evangelical adults are not as immune to changing their minds about climate change as perhaps previously thought.
The question on whether global warming was mainly caused by natural or human activities (Q2) resulted in the lowest initial scores and lowest gains ( figure 2(b)). Canadian participants (Tyndale) showed almost no gain at all (34%-37%) in their acceptance of this science-based belief, even though it was clearly emphasized in the video lecture. The other four groups showed a significant gain, but DBU2 study participants, who on other questions maintained their gain for at least one month after the study was over, largely reverted to their earlier understanding on this question, a month later. This suggests that participants in at least two of the five studies found this question difficult to understand or the presentation less convincing than on the other points. More work apparently needs to be done in explaining, to this level of recipients, how scientists know that climate change is mostly caused by human activities.
It is encouraging that respondents in all five studies made significant gains in agreeing that most scientists think global warming is happening, and that these gains were preserved one month later, in the DBU2 study (table 1; figure 2(c)). This is a critical finding, as previous research has identified the scientific consensus as a 'gateway belief' that increases acceptance that climate is changing and humans are responsible (Van Der Linden et al  2015). Similarly, respondents in all five studies made significant gains in being worried about global warming (table 1; figure 2(c)), which suggests that the intervention was effective in shifting perceptions regarding the personal relevance of this issue to the students.
Before discussing results concerning the last research question, we need to mention a potential bias in the Houghton sample compared with that of the other colleges that could impact the conclusions presented above. Respondents at Houghton had a longer time in which to complete the post-test survey. Although these surveys were emailed to participants immediately after they watched the presentation, they had until 11:00 am the next morning to complete them. In the other four studies, post-test surveys were administered immediately after the lecture. Other aspects of the Houghton study give us some confidence that this may not have produced a significant bias. The fact that the three different experimental groups at Houghton experienced similar pre-post gains, after being exposed to different forms and lengths of the presentation, may suggest that there was some stability to their changing opinions which would not be greatly influenced by whether they responded immediately after the presentation, the same evening, or early the next morning.
The most unexpected result of our research was the null effect when the Christian frame was removed (table 2). The only student response that was affected was in regard to Christian responsibility. Several previous studies had suggested that using Christian experts to communicate climate change to evangelicals may be effective at influencing their climate beliefs (Wilkinson 2010, Carr et al 2012. In those studies, however, participants were members of churches in the Southwest who participated in focus groups or individual interviews, not university students participating in a controlled experiment. The present study disentangles the strength of the impact of a Christian frame on a scientific, factual presentation on climate change for students at an evangelical college. The findings are consistent with research showing the minimal effect of different frames around teaching about climate change (McCright et al 2016), but may also be the result of other factors involved in the experiment. The first is the quality of the lecturer, who has received awards for outstanding climate science communication, and who adopts best practices advocated by social science researchers including emphasizing the use of scientific consensus in addressing climate skeptics and providing detailed information on regional climate impacts relevant to each audience (Kahan et al 2011, Lewandowsky et al 2013, Rosenau 2015. The second is the fact that the DBU studies were sponsored and introduced by a professor who was well regarded by the students, and who had won the Faculty of the Year award. It is likely that participants considered this professor, their course instructors, and the university to be endorsing the climate change lecture. This may have had the greatest influence on participants' willingness to consider and accept the information provided; this is consistent with what other researchers have noted (Kahan 2010).
To further test the possibility that the evangelical context of the study, and not the inclusion of a Christian frame to the lecture, has the most influence on the acceptance of global warming by evangelical students, we suggest the following. Students could be solicited for a study at a large secular university, where there is a relatively high concentration of evangelicals, but where the course instructor(s), study sponsor, and many of the students are not evangelical. The participants (both evangelical and non-evangelical) would be randomly divided into two groups, one watching the lecture with a Christian frame, and the other without. Results would show whether evangelical participants in the group with the Christian-framed lecture experienced the same gain as those in the other group. Similarly, with non-evangelical students. Furthermore, if this were done early in the fall, in a full year course, a post-test survey six months later would enable us to look at longitudinal gains by all groups of students.

Conclusion
This study finds that a single lecture can significantly alter acceptance of climate science, concern regarding its impacts, and support for action among undergraduate students at evangelical institutions in the US and Canada-both immediately following the lecture, as well as up to a month afterwards. Other studies have found similar changes, but this study's participants were all evangelicals, committed to deeply religious views about the Bible, and yet still able to be moved on climate change. Further, the net gains of the most politically conservative student population, DBU, were as high or higher than those at the other institutions, even though initial beliefs about climate change were significantly lower, implying that those with the lowest initial agreement may have the potential to be moved the furthest. This is significant given the large percentage of evangelicals in the United States, most of whom are politically conservative.
These results suggest that if this approach were used in schools, colleges, and churches across America, significant changes might occur in the opinions of evangelicals about climate change. If Christian leaders in these institutions encouraged their followers to spend only 35-45 min watching a video lecture like this, it could help swing the momentum of the country towards accepting the threat of climate change and to taking concrete action to reduce the potential future global damage caused by it. Given the place that America occupies in the world, both in its economic impact and contribution to global warming through its production of fossil fuels, this change in direction for the country as a whole could have a very significant effect on global attitudes towards climate change.
Although a significant number of evangelical organizations already exist that support concern about climate change, they are apparently not having the impact of some other conservative evangelical organizations, such as the Cornwall Alliance, which resolutely opposes this view. This study suggests a need for a new, broadly evangelical alliance to be formed, strongly supported by academics, business people, and religious and church leaders, to take the lead and focus on this urgent issue, to speak with one voice especially to the 25% of evangelicals in the United States. 10. A free market system is an economic system characterized by limited government intervention, private property rights, and where economic activity, such as wages and prices, is determined solely by the forces of supply and demand. How much do you agree or disagree with the following statements?
a. An economic system based on free markets unrestrained by government interference automatically works best to meet human needs.
i. Strongly Agree 1 ii. Somewhat Agree 2 iii. Unsure 3 iv. Somewhat Disagree 4 v. Strongly Disagree 5 b. The free market system may be efficient for resource allocation but it is limited in its capacity to promote social justice.
i. Strongly Agree 1 ii. Somewhat Agree 2 iii. Unsure 3 iv. Somewhat Disagree 4 v. Strongly Disagree 5 c. The preservation of the free market system is more important than localized environmental concerns. Weather is short-term, what happens in a certain place at a certain time. Climate is the longterm average of weather over at least 20-30 years. We cannot just look at one place to see if climate is changing; we have to look around the whole world. And we cannot just look at temperatures trends for a few years, but over climate scales of 20-30 years One experiment placed people randomly in a warm or cold room. Those in the warmer room were more likely to say that the Earth's climate was changing than those in the colder room. This shows how sensitive we are to our environment When we look at the US, we see that not only are global temperatures going up, we have longer growing seasons and more frequent heat waves, like the one in Australia a few years ago. We see increases in precipitation, because the warming results in more evaporation of water into our atmosphere, which then falls as rain or snow.
We also see higher ocean temperatures resulting in more energetic hurricanes. Other examples: the center of blueberry production has shifted north from Maine to Quebec; Cherry trees in Kyoto, Japan, and Washington, DC, are flowering weeks earlier than they have for a thousand years. We also see increasing crop failures around the world, as well as melting glaciers. In fact, responses to warming temperatures are seen in more than 26 500 physical and biological systems around the world.