1 Introduction

Due to its large population size and rapid industrial growth, China is one of the largest global emitters of carbon dioxide. Although per capita emission rates in China are still below most Western nations (Blanco et al. 2014), they are increasing and are expected to rise even further if the Chinese economy continues its dependence on energy from fossil fuels (Blanford, Rose, and Tavoni 2012). The latest report (AR5) of the Intergovernmental Panel on Climate Change (IPCC) (Pachauri et al. 2014) identified several significant threats to China linked to climate change, including: severe water shortages, decreased food security, significant losses of coastal ecosystems, species extinction, and increased disease risks due to flood contamination and mosquito proliferation. Education through raising individual awareness is recognized as an important tool to help ameliorate the impact of these changes on the Chinese population (Hijioka et al. 2014).

A recent report commissioned by Climate Asia (Copsey, Hoijtink, Shi, and Whitehead 2013), involving interviews with over 5,000 Chinese households, indicated that although respondents were confident that their government was taking action to combat climate change, they did not feel adequately informed about how they personally could take action. Given China’s large population and central role in the global economy, it is important to change this situation by encouraging pro-environmental behaviour amongst Chinese people. The current study employed a large urban sample, from Beijing, to explore how climate change communications could be framed to increase public engagement, and how pre-existing environmental worldviews may influence the effectiveness of such communications.

1.1 Extended Parallel Process Model of Risk Communication

Within psychology, there has been a long tradition of research on the effectiveness of fear appeals related to health (Janis and Feshbach 1953; Leventhal 1970; Maloney, Lapinski, and Witte 2011; Rogers 1983) and the environment (Hine and Gifford 1991; Poumadère, Mays, Pfeifle, and Vafeidis 2008). The Extended Parallel Process Model (EPPM; Witte 1992) was developed to explain why and when persuasive messages that incorporate fear will succeed or fail. According to the model, individuals exposed to fear appeals engage in two cognitive appraisals. They appraise the seriousness of the threat conveyed in the message and the efficacy of the recommended response to the threat (Witte, Meyer, and Martell 2001). Individuals are predicted to engage in fear control processing if they perceive a message as threatening and personally salient but simultaneously perceive its recommended action to be ineffective or infeasible (high threat – low efficacy messages). Fear control processing is proposed to lead to fear control responses, which involve attempting to manage emotions by rejecting the message or denying the threat, rather than taking action to eliminate or avoid the threat. In contrast, individuals will engage in danger control processing if they perceive a message as highly threatening but also perceive that the message has provided concrete advice about how to manage the threat (high threat – high efficacy messages). Danger control processing purportedly leads to danger control responses, which involve identifying and adopting possible solutions to reduce or avoid the threat (Witte 1992, 1994; Witte and Allen 2000).

Although the EPPM model of risk communication has been used extensively in the area of health communications, to date there have been relatively few applications to environmental and climate change communication (e.g., Hine et al. 2013; Phillips, Hine, and Marks 2015). In this study, we employed the EPPM to investigate the effectiveness of combining threat and efficacy messages to increase public engagement with climate change.

1.2 Environmental worldviews: New ecological paradigm

Environmental worldviews reflect people’s beliefs about the value of the natural world and their relationship to it, and influence how they assess and respond to natural and human generated hazards (Castro 2006; Dunlap and Van Liere 1978; Dunlap, Van Liere, Mertig, and Jones 2000). The most widely used environmental worldview measure is the revised New Ecological Paradigm (NEP-R) scale (Dunlap et al. 2000). The NEP-R is an updated version of Dunlap and Van Liere’s (1978) New Environmental Paradigm (NEP) scale which assesses beliefs about three facets of humanity’s relationship with nature: human dominance over nature (“humanity’s right to rule over the rest of nature”), limits to growth (“the existence of limits to growth for human societies”), and balance of nature (“humanity’s ability to maintain or upset the balance of nature”). The NEP-R was developed to address psychometric limitations of the original measure and assesses two additional facets: eco-crisis (“the likelihood of environmentally catastrophic changes”) and the rejection of human exemptionalism (“humans are exempt from the constraints of nature”).

The NEP and NEP-R have been used extensively across the world, in both Western and non-Western contexts (Hawcroft and Milfont 2010). The scales have been shown to be useful predictors of both self-reported and observed behaviour (e.g., Barr 2007; Casey and Scott 2006; Olli, Grendstad, and Wollebaek 2001; Vining and Ebreo 1992). NEP worldviews have also been identified as an important motivator of climate change action (Kellstedt, Zahran, and Vedlitz 2008; Xue and Zhao 2015).

In a recent study involving Chinese respondents, Xue et al. (2016) used exploratory and confirmatory factor analyses to conclude that a Chinese version of the NEP-R consisted of two moderately correlated factors: ecocentrism (the view that ecological problems are quite serious and need to be addressed) and anthropocentrism (the view that nature should serve human interests, and faith that science and technology can be effectively applied to overcome ecological limits). They found that respondents with higher levels of ecocentrism and lower levels of anthropocentrism perceived greater risks associated with climate change, which in turn (through a meditational model) predicted higher levels of self-reported mitigation behaviour.

Although numerous studies have demonstrated that NEP worldviews predict environmental risk perceptions and pro-environmental behaviour, we were unable to identify any studies that have investigated whether such worldviews moderate the impact of climate change messaging on public engagement with climate change, a key aim of the current research.

1.3 Current study

To date, research on climate change communication in China has focused primarily on documenting current levels of awareness and support for various policies related to climate change mitigation and adaptation (Copsey, Hoijtink, Shi, and Whitehead 2013; Zheng and Wang 2013; Wang and Li 2012). Little research has been conducted to assess the effectiveness of different types of climate change communications on levels of public engagement in Chinese audiences. In the current study, we had two main objectives.

First, we contrasted the effects of two types of climate change messages on the perceived efficacy of recipients in dealing with climate change threats. Based on the EPPM (Witte 1992), we predicted that messages incorporating concrete behavioral advice about how to combat climate change threats (compared to messages that emphasized only the threats) would increase recipients’ perceived efficacy for responding to climate change. In turn, higher efficacy was predicted to increase the likelihood that viewers would engage in danger control processing (i.e., accept the message and express an intention to act on the message) and decrease the likelihood that they would engage in fear control processing (i.e., reject the message and deny the threat).

A second aim of the study was to determine whether the effectiveness of efficacy-based messaging varied as a function of audience members’ environmental worldviews. Hine et al. (2016) demonstrated that audiences which were dismissive, uncommitted or alarmed about climate change varied in their receptiveness to different types of climate change messages. In the current study, we investigated whether environmental worldviews, as measured by the NEP-R scale, moderated the impact of climate change message content on an audience’s efficacy beliefs, which, in turn, would flow on to affect the likelihood of danger and fear control processing. We predicted that participants with pro-environmental (ecocentric) worldviews would be more receptive to receiving advice about how to combat climate change, thereby developing higher levels of perceived efficacy and increasing the likelihood of danger control processing. Conversely, we predicted that participants who viewed nature as a resource for human consumption (anthropocentric worldviews) would be less receptive to receiving climate change mitigation advice, thereby developing less efficacy and engaging in higher levels of fear control processing. The hypothesized mediation model is presented in Fig. 1.

Fig. 1
figure 1

Proposed mediation model in which climate change communications predict the efficacy of engaging in mitigation behaviour, which in turn predicts both respondents’ fear control and danger control process

Full size image

2 Method

2.1 Participants

The sample consisted of 515 Mandarin-speaking residents of Beijing, China. Respondents were aged from 18 to 75 years (M = 37.47, SD = 14.38), and just under half were female (48 %). In terms of education, 23 % had completed year 12 or less, 16 % had some university education, 51 % had completed a bachelor degree, and 10 % had completed a higher degree. A majority of respondents (81 %) participants were employed, 2 % were unemployed, 11 % were retired or disabled, and 5 % were students. The ratio of males to females in the sample was similar to the ratio reported in the Chinese census for Beijing (Yu 2011), but university educated respondents were over-represented in the sample.

2.2 Procedure

Translation and back translation

We followed the 4-step back translation procedures recommended by Harkness and Schoua-Glusberg (1998). The survey was initially translated into Chinese (Mandarin) by the first author, and then back-translated into English by the sixth author. The back translation version was compared with the original version. Correspondence between the two versions was very high (<5 % discrepancy). Differences were discussed by the two translators and revised through consensus. Prior to being administered in China, the survey was pilot tested by 16 native Mandarin speakers attending the University of New England, Australia. The pilot test resulted in minor re-wording of several items to increase clarity.

Survey administration

Respondents were recruited from a Qualtrics online panel between June and July 2013. To be eligible, panel members had to be over the age of 18 years, Mandarin speakers, and residing in Beijing. All respondents read an information sheet describing the study prior to providing informed consent to participate. Responses were anonymous, and all survey items and experimental materials were presented in Mandarin.

Experimental manipulation

Prior to the experimental manipulation, respondents completed questions relating to demographics and environmental worldviews. They were then randomly assigned to one of two experimental conditions: (1) high threat - low efficacy message, and (2) high threat - high efficacy climate change message. Participants were randomly assigned to one of two experimental conditions through the Qualtrics online system: (1) high threat – high efficacy, and (2) high threat – low efficacy. Participants in the high efficacy conditions viewed a video, excerpted from China’s Greenpeace website, highlighting potential negative impacts of climate change in China http://www.greenpeace.org/china/zh/multimedia/video/climate-energy/2009/climate-sounds/ Footnote 1.

The video comprised young Chinese describing the causes of climate change and impacts of severe weather events, sea level rise, and glacial melting on humans and ecosystems. It also described strategies to reduce climate change threats, and how to apply those strategies in daily life. In contrast, participants in the low efficacy condition were presented with a shortened version of the video that contained no information about strategies to reduce climate change threats. All information was presented in Mandarin. After viewing the climate change messages, respondents completed several questions assessing the extent to which the messages caused them to engage or disengage with the topic.

2.3 Measures

Gender, age and education levels were all assessed using single items.

Environmental worldviews

Environmental worldviews were assessed using the revised New Ecological Paradigm scale (Dunlap et al. 2000). The NEP-R consists of 15 items assessing five facets of environmental worldview: balance of nature, eco-crisis, rejection of exemptionalism, limits to growth, and anti-anthropocentrism. The NEP-R consists of both positively and negatively worded items, all assessed on a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree). Preliminary psychometric work involving both exploratory and confirmatory factor analyses (Xue et al. 2014, 2016) indicated that the Chinese NEP-R assesses two reliable environmental worldview dimensions: ecocentrism (reflecting beliefs that ecological problems are quite serious and need to be addressed, α = .77) and anthropocentrism (reflecting beliefs that nature should serve human interests, and faith that science and technology can be effectively applied to overcome ecological limits, α = .73). Chinese translations of the NEP items can be found in Xue et al. (2014, 2016 ), and in Appendix A.

Message assessment variables

After viewing the climate change messages, respondents evaluated them in terms of the extent to which they increased perceived efficacy, and elicited danger control and fear control responses. All responses were assessed on 7-point scales. Perceived efficacy was assessed by four items that measured respondents’ perceptions about how easily the messages recommendations could be put into action (response efficacy) and the extent to which the message made them feel equipped to effectively deal with climate change (self-efficacy). Given that the two types of efficacy were highly correlated (r = .64) and individually exhibited lower-than-ideal levels of reliability, they were combined into a single index of efficacy (a = .85). Danger control response was assessed using four items gauging the extent to which respondents perceived the climate change messages to be important and valuable, and motivated respondents to take action and seek out more information about climate change (a = .91). Fear control response was assessed by three items measuring the extent to which respondents perceived the climate change messages to be manipulative, exaggerated and encourage denial (a = .83). All message assessment items were based on similar items used in previous studies (e.g., McMahan et al. 1998; Smalec and Klingle 2000). The items and their Chinese translations are presented in Appendix A.

3 Results

3.1 Mediation analysis

Path analysis, using AMOS v21, was used to assess the mediation model presented in Fig. 1. The analysis enabled us to determine whether (1) the climate change message manipulation increased perceived efficacy to combat climate change; (2) the perceived efficacy mediated the impact of the experimental manipulation and worldviews on danger and fear control responses; and (3) whether this impact was moderated by respondents’ environmental worldviews.

Descriptive statistics together with zero order correlations for all variables considered for the path model are presented in Table 1. Four indices were used to determine model fit based on Kline (2005): χ 2/df < 3, CFI > .90, RMSEA < .08 (90 % CIs [.05, .10]), and SRMR < .10. All significance tests were based on 1,000 bias-corrected bootstrapped samples. The hypothesized model (see Fig. 1), in which all video and environmental worldview (NEP) effects on fear and danger control responses were mediated through efficacy failed to reach the criteria for adequate fit on two fit indices, χ 2(11) = 68.75, p < .001; χ 2/df = 6.25; CFI = .99; RMSEA = .10, 90 % CIs [.08, .13]; SRMR = .05.

Table 1 Zero-order correlations and descriptive statistics for all variables
Full size table

Based on conceptual considerations that worldviews may influence danger and fear control responses directly (as well as indirectly through self-efficacy), and also AMOS modification indices (which suggested two such direct paths), we constructed an alternative mediation path model (see Fig. 2), in which anthropocentrism directly predicted fear control and ecocentrism directly predicted danger control. This revised model provided an excellent fit to the data: χ 2 (9) = 14.99, p = .09; χ 2/df = 1.67; CFI = 1.00; RMSEA = .04, 90 % CIs [.00, .07], SRMR = .02. The model explained 24 % of the variance in fear control responses and 61 % of the variance in danger control responses (both ps < .01).

Fig. 2
figure 2

Path model showing efficacy mediating the effects of climate change message type (1 = high threat – low efficacy, 2 = high threat – high efficacy) on fear and danger control response. Values on pathways represent standardized regression weights. Model fit indices: χ 2 (9, N = 515) = 14.99, p = .09, CFI = 1.00, GFI = .99, RMSEA = .04. *p < .05,**p < .01, ***p < .001. The Johnson-Neyman analysis regions of significance analysis for the interaction effects indicated that message type significantly predicted efficacy for respondents who scored 3.9 or below on ecocentrism or 2.6 or below on anthropocentrism

Full size image

Examination of the path coefficients revealed that participants who scored higher ecocentrism reported higher levels of perceived efficacy to respond to climate change. The direct effect of the message type on efficacy was not significant, but was qualified by two significant interactions that indicated the message effect on efficacy was moderated by both NEP worldview dimensions. We probed both interactions using the Johnson-Neyman conditional analysis option in the Process moderation/mediation macro for SPSS (Hayes 2013). The Johnson-Neyman analysis identifies regions of significance for an independent variable across a range of values for a designated moderator.

For the interaction involving message type and ecocentrism, the analysis indicated that providing specific advice about stopping climate change produced significant increases in efficacy for respondents who had ecocentrism scores that ranged from 1.0 to 3.9 (out of 5). Message type had no effect on efficacy for respondents who scored very high (above 3.9) on ecocentrism.Footnote 2 For the interaction involving message type and anthropocentrism, the conditional analysis indicated that providing specific advice about how to stop climate change produced higher efficacy for respondents who scored between 1 and 2.62 on anthropocentrism. For respondents who scored above 2.62 (out of 5) on anthropocentrism, message type had no impact on efficacy. Overall, this pattern of results partially supported our hypothesis about efficacy-building communications. Providing concrete, actionable advice about how to combat climate change increased perceived efficacy in some viewers, but not those who scored moderate to high on anthropocentrism or very high on ecocentrism.

As predicted, increased perceived efficacy was significantly associated with higher danger control responses (message acceptance) and reduced fear control responses (message rejection). The standardized indirect effects of the message-type by anthropocentrism interaction on danger control responses (β = −.08, 95 % CIs [−.15, −.02], p < .01) and fear control responses (β = −.05, 95 % CIs [−.10, −.01], p < .01) were both statistically significant. The indirect effects of the message-type by ecocentrism interaction approached significance for both danger control responses (β = −.06, 95 % CIs [−.13, .021, p = .10) and fear control responses (β = −.04, 95 % CIs [−.09, .01], p = .09).

In addition to the meditation and moderation effects described above, there were two significant direct paths from the environmental worldview measures and danger/fear control: (1) Respondents who scored higher on ecocentrism were more likely to engage in danger control processes, and (2) respondents who scored higher on anthropocentrism were more likely to engage in fear control processing.

4 Discussion

In this experimental study, we contrasted the effects of two types of climate change messages on the perceived efficacy of recipients in dealing with climate change threats. Both messages contained information about climate change threats, but only one incorporated efficacy information; providing specific advice about what Chinese citizens can do to combat climate change. We predicted that messages would be more likely to increase efficacy if they included information about how to deal with the threat. In addition, we tested whether the effect of this manipulation varied as a function of participants’ ecological worldviews. We predicted that the influence of self-efficacy information would be amplified for those high in ecocentrism and attenuated for those high in anthropocentrism. We further predicted that increased self-efficacy would lead to stronger danger control responses and weaker fear control responses, in accordance with Witte’s EPPM (Witte 1992; Witte and Allen 2000).

Consistent with our hypotheses, higher efficacy related to climate change action was associated with a greater likelihood that participants would engage in danger control processing (i.e., accept the message and express an intention to act on the message) and a decreased likelihood they would engage in fear control processing (i.e., reject the message and deny the threat). These findings provide further support for the EPPM model, and extend it to the area of climate change communication. Our results are also consistent with previous climate change studies linking perceived efficacy to climate change mitigation responses in Asian respondents (Kim et al. 2013).

Perhaps our most interesting findings related to the significant moderating effects of environmental worldviews on our message-type manipulation. These moderation effects suggest that messages providing concrete behavioral advice about how to combat climate change (compared to messages that emphasized only climate change threats) may not be uniformly effective in boosting efficacy for all audience types. In particular, the provision of behavioural advice failed to increase efficacy in participants who scored just above the mid-point on anthropocentrism or very high on ecocentrism. The moderating effect of anthropocentrism on efficacy boosting messaging suggests that anthropocentrics, who generally exhibit lower levels of environmental concern, may tune out from climate change messaging, leading them fail to absorb or actively reject practical advice about how to reduce their carbon footprint. Contrary to our predictions, the efficacy-boosting message was also ineffective for participants who scored very high on ecocentrism. Research on the confirmation bias suggests that, in general, viewers tend to be receptive to messages that are consistent with their values and pre-existing beliefs (Whitmarsh 2011). However, it is possible that highly ecocentric viewers may fail to engage with value-consistent messaging if they believe that message is offering nothing new over and above what they already know. Another related possibility is that most high ecocentrics already possessed high levels of climate-change-mitigation efficacy prior to the study, resulting in a possible ceiling effect for this subgroup, and limiting the amount of possible upward movement in response to the experimental manipulation.

Finally, anthropocentric worldviews were directly positively associated with fear control processing and ecocentric worldviews were similarly associated with danger control processing. This is consistent with previous findings regarding involving the NEP-R, which suggest that individuals subscribing to the new ecological paradigm (reflecting ecocentrism) are more likely be concerned about the environment and engage in pro-environmental behaviour, whereas those who subscribe to the dominant social paradigm (anthropocentrism) are less likely to do so (e.g., Barr 2007; Casey and Scott 2006; Olli, Grendstad, and Wollebaek 2001; Vining and Ebreo 1992; Xue et al. 2014, 2016). Thus, environmental worldviews not only moderated the impact of message content on viewers’ perceived efficacy to combat climate change, but also appeared to directly influence participants’ propensity to accept or reject climate change messages.

4.1 Limitations

Although we collected a large, heterogeneous online sample from Beijing, our sample contained a higher proportion of college educated that respondents relative to Chinese census data. Additional research is required to determine whether our findings would generalize to the Chinese population as a whole and to other settings and other respondents. In particular, more research is required on worldviews and climate change beliefs and behaviours of rural Chinese residents. Rural residents comprise a substantial proportion of the Chinese population, and due to their elevated environmental and economic exposure are likely to be particularly vulnerable to climate change impacts (Copsey et al. 2013).

In addition, the noted relationships between perceived efficacy and fear/danger control responses are based on cross-sectional, correlational data. Although the direction of these pathways is intuitively plausible, we are unable to draw strong causal inferences between these secondary links in the mediation model.

4.2 Conclusion

The results of this study provide support for Witte’s (1992) Extended Parallel Process Model in a non-Western culture and highlight the value of incorporating efficacy-building information into climate change communications. Our results also suggest that efficacy building in climate change communications may not be uniformly effective for all audiences and that it may be beneficial to tailor messages for audiences with different worldviews (Hine et al. 2016). Overall, the results of this study support the IPCC view that awareness-raising in Asian countries is an important adaptive tool in meeting the challenges of climate change (Hijioka et al. 2014), and demonstrates practical ways of doing this in an effective manner. These findings should be of interest to risk communication researchers, as well as those involved in developing and delivering climate change communications.