Can Climate Skeptics Be Convinced? The E ﬀ ect of Nature Videos on Environmental Concern

: Much research has demonstrated that videos can function as primers or nudges that inﬂuence attitudes and behaviors. Studies to date suggest that this includes inﬂuence over individual pro-environmental attitudes. However, the existing evidence all stems from samples comprised of university students. In this paper, we describe the results of a randomized online experiment in a sample of 468 climate skeptics. We presented 3-min nature documentary videos that highlighted either the beauty of nature, the endangerment of nature by humans, or a mixture of both. The results suggest that the mixed stimulus video, which shows ﬁrst the beauty of nature and then its endangerment by humans does indeed increase environmental concern by almost half a standard deviation. However, none of the video treatments increased donations to pro-environmental organizations. Still, the results suggest that nudging by video also works in samples of climate skeptics, which demonstrates the external validity of former ﬁndings.


Introduction
Climate skepticism is still an obstacle to more rigorous environmental policy, not only in the United States but also in Europe and many other countries. Even those where the environmental concern of the population is generally high have a critical number of conservative parties and politicians who prevent the implementation of more strict environmental policies. This is particularly obvious in the United States, where environmental skepticism is widespread among Republicans [1,2]. But the problem also exists in many European countries where environmental skepticism can be found both in political programs and in the slogans of leading political figures of conservative parties. This is also the case in Switzerland. International surveys show that the general population ranks very high in terms of pro-environmental attitudes [3]. At the same time, Switzerland's largest political party, the Swiss People's Party (SVP), reached a 25.6% share of the vote in the last elections, and still downplays the importance of protecting the environment. For example, one of the leading politicians of Switzerland's conservative party wrote recently "There is no scientific evidence that humans significantly influence climate" [4], a statement that is in obvious contradiction to what most climate research says. This raises the question of whether climate skeptics and, more generally, people holding low pro-environmental attitudes can be convinced that more environmental protection is necessary.
One obvious answer to the question lies in education. Much environmental research in sociology and psychology has consistently demonstrated that education is positively linked to pro-environmental attitudes [3,5,6]. However, increasing education among a population is a long-term endeavor and this makes the general nature of this answer unsatisfactory. Hence, research has started, to investigate whether environmentally related attitudes can also be strengthened using much simpler and faster methods such as videos. Generally, video-based experimental studies have been conducted in various presents the results that videos have on subjects' attitudes and donation behavior. Finally, the results and the limitations of our study are discussed in Section 4.

Materials and Methods
Former research suggests that videos can have an effect on individuals' donation behavior in student populations. In this study, we address the question of whether pro-environmental videos are also effective in a sample of climate skeptics. This is an important extension of the research in this area, since pro-environmentalists need no convincing, whereas those who are skeptical about the need for environmental protection and who often prevent more rigorous regulations are harder to persuade. There are several reasons why videos could have an effect, but also some reasons why they should not be expected to work or would have only a limited impact. First, videos might contain new information for viewers that induce a change in individuals' perceptions of the topic. However, environmental topics have been on the agenda in the media and in politics for some time. Therefore, it is unlikely that videos showing the beauty of nature or its endangerment would convey new information to participants. As a matter of fact, showing such videos might very well have the opposite effect due to fatigue. A further reason that videos can have an effect is because they function as a primer or a nudge [21,22]. Much psychological literature has demonstrated that primers can work. Taken together, it is hard to predict, theoretically, whether environmental videos would have an effect in a sample of conservative party members with weak pro-environmental attitudes, and in which direction they would work; hence, this is an empirical question. Given former evidence, we expect a positive effect from the video treatments, which should be stronger for the positive video treatment and weaker for the negative video.
We investigated this question via an online experiment among party members of Switzerland's largest conservative party. In the first step, we contacted subdivisions of various political parties in the canton of Bern in Switzerland and asked the parties' secretaries to forward an email to their members, which contained a link to a survey on the political attitudes of party members. All parties have several thousand members, and all agreed to forward our mail. The mail announced a study on the different political attitudes of party members and contained a link to an online questionnaire. Participation in the survey was voluntary. After completing the online questionnaire, the participants from the conservative party received an electronic voucher of 20 Swiss Francs that they could print and use in Switzerland's largest food retailer. The receipt of the voucher was announced in the introduction email for the survey. The survey for the other party members did not contain such a voucher. As a result, 468 or about 10% of the members of the conservative party did participate in the survey, but only very few took part from the other parties who did not receive an incentive. In the following analysis, we only take into consideration the sample of the conservative party, which is our primary target group.
The 468 participants were randomized into four groups. All four groups received an identical survey with about 100 questions on various political attitudes. The surveys only differed with respect to the videos the different groups received. The average completion time was 27 min. The questionnaire started with 30 questions about general political attitudes, followed by either a 3-min positive video showing the beauty of nature, a 3-min negative video, or a 3-min positive and negative nature video. The control group saw no video. The films were compiled from YouTube videos of scenes provided by the BBC (mostly from the program "The Blue Planet"), National Geographic ("Before the Flood"), and Al Gore's film "An Inconvenient Sequel". The positive video was accompanied by positive and enthusiastic instrumental music, and subtitled with text describing the scenes: "Our blue planet emerged about four billion years ago-ice, wind, and water shaped nature during millions of years -glaciers store about 70% of the earth's freshwater reserves-and influence the climate as forests do -plants produce oxygen which is so important for humans-about 70% of the oxygen emerges from the oceans-oceans are the birthplace of life-there exist about 8.7 million different species of plants and animals-let us save our planet".
The second group saw a 3-min video of scenes that demonstrate environmental pollution and destruction, e.g., pollution in cities, the burning of rainforests, large scale monoculture farming, and the accumulation of plastic garbage in the ocean. The negative video was accompanied by calm and sad instrumental music, and subtitled with the following text: "Our blue planet emerged about four billion years ago-but the environment is in danger-by air pollution-by the exploitation of non-renewable resources-by deforestation-by intensive agriculture-by too many CO 2 emissions-having disastrous consequences-millions of tons of plastic end up in the oceans-the ice caps melt and seawater level is rising-floods are increasing-with the rise of temperature weather extremes increase-let us save our planet". The third group saw a combination of both, first the positive video and second the negative video, lasting 1.5 min each. Finally, the fourth group saw no video and functioned as the control group. The videos were followed by two questions about them, by items that measure environmental concern, and by further constructs that are related to environmental concern. Figure 1 displays the distribution into the four experimental groups. Overall, 468 members of the conservative party participated in the survey. However, we excluded 79 respondents because they answered that they belong to a different party than the one the emails were sent to. This could be due to different reasons. Either some respondents obscured their party membership or the survey was answered by other respondents in the household. The questionnaire also contained a test question, asking respondents to leave this question unanswered. Overall, 28 participants did give an answer to this question, either from ignoring the instruction or because they did not read the question carefully. These respondents were also excluded from further analysis, as was one participant who took an extremely long time to answer the survey. Excluding these cases left us with 360 valid cases. However, the results were not essentially affected by these exclusions. renewable resources-by deforestation-by intensive agriculture-by too many CO2 emissionshaving disastrous consequences-millions of tons of plastic end up in the oceans-the ice caps melt and seawater level is rising-floods are increasing-with the rise of temperature weather extremes increase-let us save our planet". The third group saw a combination of both, first the positive video and second the negative video, lasting 1.5 min each. Finally, the fourth group saw no video and functioned as the control group. The videos were followed by two questions about them, by items that measure environmental concern, and by further constructs that are related to environmental concern. Figure 1 displays the distribution into the four experimental groups. Overall, 468 members of the conservative party participated in the survey. However, we excluded 79 respondents because they answered that they belong to a different party than the one the emails were sent to. This could be due to different reasons. Either some respondents obscured their party membership or the survey was answered by other respondents in the household. The questionnaire also contained a test question, asking respondents to leave this question unanswered. Overall, 28 participants did give an answer to this question, either from ignoring the instruction or because they did not read the question carefully. These respondents were also excluded from further analysis, as was one participant who took an extremely long time to answer the survey. Excluding these cases left us with 360 valid cases. However, the results were not essentially affected by these exclusions.  Figure 1 demonstrates that the randomization worked well and left us with a roughly equal number of cases in each condition (the differences are statistically not significant). Furthermore, Table  1 displays further randomization checks. In particular, we compared the age, sex, and income distributions between groups, as well as the distributions of some key political attitudes. To summarize, the sample contained mostly men (81%), with an average age of 57. All respondents were asked to self-rate themselves on a left-right continuum with an 11-point answering scale ranging from 0 (very left) to 10 (very right). The average in our sample was, as expected, rather high, with a mean value of 8.04. A similar result is obtained if the answer was slightly reformulated (conservative versus liberal). Here, the mean value was 4.42. These results suggest that we were indeed dealing with a  Figure 1 demonstrates that the randomization worked well and left us with a roughly equal number of cases in each condition (the differences are statistically not significant). Furthermore, Table 1 displays further randomization checks. In particular, we compared the age, sex, and income distributions between groups, as well as the distributions of some key political attitudes. To summarize, the sample contained mostly men (81%), with an average age of 57. All respondents were asked to self-rate themselves on a left-right continuum with an 11-point answering scale ranging from 0 (very left) to 10 (very right). The average in our sample was, as expected, rather high, with a mean value of 8.04. A similar result is obtained if the answer was slightly reformulated (conservative versus liberal). Here, the mean value was 4.42. These results suggest that we were indeed dealing with a very conservative sample, as targeted. Moreover, the randomization worked well with respect to these political attitudes since the differences are very small and do not differ statistically significantly between groups. After a few questions on other political issues, respondents in the three treatment groups watched the 3-min nature videos. The videos had no voice, but came with music. In the case of the positive video, the music was very enthusiastic and optimistic, and was very slow and sad in the negative treatment. The same music was also used in the mixed video. Furthermore, the videos were accompanied by written text describing the pictures, and ended with the words "let us save our planet". All texts had exactly the same length of 74 words. The video was followed by a few questions about it and by the measurement of environmental concern. Environmental concern is usually defined as the extent of an individual's insight that humans endanger the natural environment, combined with their willingness to protect nature. This definition consists of two components, the cognitive component of having the rational insight, and the conative component of being willing to do something about it. We used a measure suggested by Franzen and Vogl (2013) [3] that consists of nine items and reflects these two components (see Table 2). These items are also included in the International Social Survey Programme (ISSP). All items can be answered on a five-point answering scale varying from (1) "not at all willing" to (5) "very willing" (for the first three items), and from (1) "do not agree at all" to (5) "agree strongly" for items 4 through 9. The sum index ranges from 9 to 45. The quality of the measurement with respect to its test-retest reliability and its congruent, discriminant, and external validities has been rigorously tested using a multi-trait multi-method design in another study [23]. The test results show that the scale has a high reliability and high external validity, and outperforms single item measures of environmental concern. Also, the test results in this sample show that the scale has a high internal consistency (Cronbach's alpha = 0.78). A principle component analysis (PCA) extracts two components, which are termed "conative" (items 1 through 4) and "cognitive/affective" (items 5 through 9) which confirms -former study results. Table 2. The measurement of environmental concern (ISSP9). (1) How willing would you be to pay much higher prices in order to protect the environment? 0.80 (2) How willing would you be to pay much higher taxes in order to protect the environment? 0.75 (3) How willing would you be to accept cuts in your standard of living in order to protect the environment? 0.74 (4) I do what is right for the environment, even when it costs more money or takes more time. 0.77 (5) Modern science will solve our environmental problems with little change to our way of living. 0.71 (6) In order to protect the environment, the country needs economic growth. 0.68 (7) It is just too difficult for someone like me to do much about the environment. 0.38 (8) We worry too much about the future of the environment and not enough about prices and jobs. 0.63 (9) People worry too much about human progress harming the environment. 0.69 Note: n = 360, Cronbach's α = 0.78, mean (ISSP9) = 27.03, sd = 5.87, min. = 11, max. = 44. Each item contains five answer categories ranging from 1 = "very unwilling" to 5 = "very willing" for items 1-3, 1 = "disagree strongly" to 5 = "agree strongly" for item 4, and 1 = "agree strongly" to 5 = "disagree strongly" for items 5-9. Items 1 through 4 denote the conative component. Items 5 through 9 denote the cognitive/affective component. Numbers indicate factor loadings after varimax-rotated principal component analysis (PCA). Figure 2 depicts the result of the experiment with respect to changes in environmental attitudes. As expected, the average environmental score is lowest in the control group (25.9), which did not see any video. The largest value can be observed for the mixed video treatment, in which we measure a score of 28.4. This difference is statistically significant at the 1% level (t = 2.98, p = 0.003, two-sided t-test). The result remains highly statistically significant after correcting for multiple tests according to the Holm-Bonferroni method [24]. In terms of standard deviations, the participants of the mixed video treatment show 0.44 standard deviations more environmental concern as compared to the control group. The difference is not big but is visible. Additionally, the positive video seems to have a small effect of 0.28 standard deviations. However, this difference is statistically only significant at the 10% level. By comparison, the values for the negative video do not differ statistically from the values measured in the control group. The mean values of environmental concern (ISSP9) by experimental group including 95% confidence intervals. The numbers in the bars indicate the mean in each group. The numbers on top of the bars stand for the differences in means between the respective groups in standard deviations (sd). Two-sided t-tests: + = p < 0.10, * = p < 0.05, ** = p < 0.01.

Results
The results are unchanged if we conduct multiple OLS regression analysis, controlling for various other variables. First, model 1 in Figure 3 mirrors the result when we compare the environmental concern score between the treatment groups. The group that saw the mixed video has Figure 2. The results of the video treatment on environmental concern. The mean values of environmental concern (ISSP9) by experimental group including 95% confidence intervals. The numbers in the bars indicate the mean in each group. The numbers on top of the bars stand for the differences in means between the respective groups in standard deviations (sd). Two-sided t-tests: + = p < 0.10, * = p < 0.05, ** = p < 0.01.
The results are unchanged if we conduct multiple OLS regression analysis, controlling for various other variables. First, model 1 in Figure 3 mirrors the result when we compare the environmental concern score between the treatment groups. The group that saw the mixed video has an environmental concern score elevated by 0.44 standard deviations. Next, model 2 takes some sociodemographic information into account such as sex, age, education, and household equivalence income. Besides sex, none of these control variables has a statistically significant influence on environmental concern. Sustainability 2020, 12, x FOR PEER REVIEW 8 of 15 Figure 3. The results of the regression of environmental concern on various other variables. Coefficient plots of the linear ordinary least squares regressions of environmental concern (ISSP9) on various other variables. Depicted are coefficients in standard deviations (the outcome and all metric independent variables are z-standardized) with 95% confidence intervals of robust standard errors. See Table A1 in the Appendix information for more details.
Next, we supplemented the study with the measurement of some further constructs that are presumably associated with environmental concern such as empathy, the need for social approval, and the importance of money. We measured empathy by 12 items suggested by Davis (1983) [25]. This scale (see Table 3) has high internal (Cronbach's alpha = 0.80), as well as test-retest, reliability [26]. As can be seen from the results (see Figure 3), empathy is positively related to environmental concern. Furthermore, the finding that females have higher concern disappears if we control for empathy (see model 3), suggesting that the sex effect is explained by the higher empathy levels of women.  The results of the regression of environmental concern on various other variables. Coefficient plots of the linear ordinary least squares regressions of environmental concern (ISSP9) on various other variables. Depicted are coefficients in standard deviations (the outcome and all metric independent variables are z-standardized) with 95% confidence intervals of robust standard errors. See Table A1 in the Appendix A information for more details.
Next, we supplemented the study with the measurement of some further constructs that are presumably associated with environmental concern such as empathy, the need for social approval, and the importance of money. We measured empathy by 12 items suggested by Davis (1983) [25]. This scale (see Table 3) has high internal (Cronbach's alpha = 0.80), as well as test-retest, reliability [26]. As can be seen from the results (see Figure 3), empathy is positively related to environmental concern. Furthermore, the finding that females have higher concern disappears if we control for empathy (see model 3), suggesting that the sex effect is explained by the higher empathy levels of women. After seeing a play or movie, I have felt as though I were one of the characters. 0.80 (7) When I watch a good movie, I can very easily put myself in the place of a leading character. 0.81 (8) When I am reading an interesting story or novel, I imagine how I would feel if the events in the story were happening to me. 0.70 Empathetic concern (9) I often have tender, concerned feelings for people less fortunate than me. 0.75 (10) When I see someone being taken advantage of, I feel kind of protective towards them. 0.75 (11) I am often quite touched by things that I see happen. 0.64 (12) I would describe myself as a rather soft-hearted person. 0.69 Note: n = 360, Cronbach's α = 0.80, mean (IRI) = 43.72, sd = 6.09, min. = 16, max. = 60. Each item contains five answer categories ranging from 1 = "disagree strongly" to 5 = "agree strongly". Numbers indicate factor loadings after varimax-rotated principal component analysis.
The need for social approval is measured by the eight items displayed in Table 4. The index consists of two factors which we tagged "public approval" and "private approval", and also displays a satisfactory internal consistency of Cronbach's alpha of 0.67. As is shown by the results (see Figure 3), respondents with a high need for social approval by colleagues and friends also show higher levels of environmental concern. Note: n = 356, Cronbach's α = 0.67, mean = 27.08, sd = 4.34, min. = 12, max. = 38. Each item contains five answer categories ranging from 1 = "disagree strongly" to 5 = "agree strongly" for items 1-4 and 6, and 1 = "agree strongly" to 5 = "disagree strongly" for items 5, 7, and 8. Numbers indicate factor loadings after varimax-rotated principal component analysis.
Next, the questionnaire also contained a scale intended to measure the importance of money (see Table 5). This eight-item scale is one-dimensional and has high internal consistency as indicated by a Cronbach's alpha of 0.81. The results show that people for whom money and material things are important show less environmental concern as compared to respondents for whom money is less important. To make more money, I would work more immediately. 0.52 (5) If in doubt, I prefer to have more than less money. 0.55 (6) Financial security is important for my wellbeing. 0.72 (7) Material wealth is important for me. 0.75 (8) One can only have a decent life with a lot of money. 0.50 Note: n = 359, Cronbach's α = 0.81, mean = 26.98, sd = 5.00, min. = 10, max. = 39. Each item contains five answer categories ranging from 1 = "disagree strongly" to 5 = "agree strongly". Numbers indicate factor loadings after varimax-rotated principal component analysis.
Furthermore, since an increase in environmental concern after the video treatment might have been due to social desirability, we incorporated a measurement of social desirability according to 10 items suggested by Crowne and Marlowe (1960) [27][28][29]. These ten items are displayed in Table 6. In our sample of climate skeptics, the additive index is negatively associated with environmental concern (see Figure 3), suggesting that in this sample, it is socially desirable to show less environmental concern.
Taking social-demographic information (sex, age, and education) or the measures of social desirability, empathy, the need for social approval, and the importance of money into consideration does not change our fundamental finding that the mixed video did increase respondents' environmental concern (see Figure 3).
Finally, we turn to the question of whether the video treatment also enhances subjects' donation behavior. For this purpose, we gave the participants the opportunity to donate some or all of their 20 Swiss Franc voucher to an environmental organization. Participants were first asked whether they want to donate, and thereafter received a list of three well-known pro-environmental organizations such as the World Wildlife Fund (WWF), Greenpeace, and Pro Natura. Furthermore, we added an open field in which subjects could name any organization to which they wanted to donate. Overall, 13% of our participants donated some money to a pro-environmental organization. However, donating was not related to any treatment group or any other variable we took into consideration in any model displayed in Table 7.

Discussion
Our results show that videos about the beauty and endangerment of nature can indeed elevate environmental concern even among members of conservative parties that are traditionally very skeptical about environmental issues. Thus, we confirm former findings that were mainly generated with samples of university students. However, only the mixed video-which started with positive messages (beauty of nature) followed by scenes that show the destruction of nature-was effective. The group that saw this video showed concern values that were higher by 0.44 standard deviations. Hence, the effect is visible but modest. However, those who saw the video did not donate more than participants in the control group. This result underlines the notion that environmental behavior cannot be easily altered simply by providing information trying to persuade individuals to act pro-environmentally. Additionally, our experiment only shows that there is a modest short-term effect. Whether, and for how long, it lasts is another question left for further research. Thus, the good news is that pro-environmental videos do have, at least, a positive short-term effect, even among very skeptical individuals. Additionally, the results suggest that such videos do not cause reactivity since none of the treatment groups showed lower values as compared to the control groups. The bad news is, however, that the effect of such videos is limited and that they do not seem to elevate pro-environmental behavior, at least not in form of donations to pro-environmental organizations.
At first sight, one possible argument against our result is that it could have been driven by social desirability or the experimenter demand effect. However, if that were the case then the effect should also appear in the negative and positive video treatment groups. The fact that this is not the case speaks against a demand effect in our experiment. Moreover, we measured social desirability according to the well-known Crowne-Marlowe scale and incorporated it into the model (Figure 3). The results show that the effect is negative. This suggests that in our sample, the socially desirable response is to show less, not more, environmental concern. This reconfirms the notion that we are dealing with a highly environmentally skeptical group for which it is socially desirable to not be concerned with environmental issues.
Finally, one might wonder why only the mixed video consisting of both a positive and a negative video sequence is effective. One possible interpretation is that the mixed video does contain the most information, whereas the other two treatment groups received only information on either positive or negative things. This result also contradicts the interpretation that the videos simply work as a primer by simply activating already-existing information. If priming were the reason, one would expect that the positive or negative video would perform equally well.
Taken together, our results suggest that videos promoting environmental concern do have a small effect which is due to new information they convey to subjects. Hence, educational curricula and environmental organizations are well advised to use videos to convince skeptical subjects. The problem, probably, is how to get skeptical groups to watch such movies.
Author Contributions: Both authors contributed equally to the manuscript. All authors have read and agreed to the published version of the manuscript. Appendix A Table A1. The results of the linear ordinary least squares regression of environmental concern on various other variables (see also Figure 3).