MSc Communication Science
Persuasive Communication
Master Thesis
What Prevails: Saving the Planet or Saving Money?
The Role of Intrinsic and Extrinsic Goal Frames in Environmental
Fear Appeals
by
Lina-Sophie Daller
ID: 11646454
Supervisor: Dr. Sandra Zwier
Submission Date: 29
thJune 2018
Abstract
Climate change constitutes a pressing problem for humankind and nature alike, and it is thus of critical importance that we reduce greenhouse gas emissions rather sooner than later. The present study examined the effect of exposure to fear appeals regarding the effects of climate change, on people's intention to engage in pro-environmental behaviours. Based on the Extended Parallel Process Model (Witte, 1992), it was proposed that appeals with high threat information would be more effective in raising pro-environmental action intentions than those without threatening information. Furthermore, it was expected that intrinsic versus extrinsic goal frames would moderate these effects. A web-based experiment among 179 participants was conducted to examine these predictions. In sum, participants in all conditions exhibited high levels of intentions to engage in pro-environmental actions, but there were no significant differences between individuals exposed to the high and low threat materials. When exposed to low threat appeals, participants in the intrinsic goal frame conditions indeed reported higher intentions to generally engage in pro-environmental behaviours than those in the extrinsic goal frame condition. However, this assumption could not be supported with respect to specific intentions of consuming less meat and household energy. In conclusion, the present study could not find convincing evidence for the role of intrinsic and extrinsic goal frames in fear appeals to further people’s intentions to combat climate change. This leaves a considerable need for additional research to communicate the threats of climate change as powerfully as possible to people and eventually save our planet.
Table of Contents Introduction ... 3 Theoretical Framework ... 5 Methods Section ... 15 Results ... 27 Discussion ... 35 Bibliography ... 43 Appendix ... 58 Appendix A: Measurements ... 58
Appendix B: Study Materials ... 62
Appendix C: Assumptions for Analysis of Covariance ... 63
Introduction
Climate change presents an urgent issue of concern and high relevance for society (Roeser, 2012). This is based on the fact that the consequences of climate change pose a risk for human health, animals and the environment (European Commission (EC), 2015). The main cause of climate change are greenhouse gases emitted by human actions (Arends, 2011; EC 2015). As postulated in the Paris Agreement 2015, the EU is required to reduce
greenhouse gas emissions to 80-95% below the concentration level in 1990 until the year 2050 for the purpose of mitigating the detrimental effects of climate change (EC, 2015). It is known that the production and consumption of energy, food and transport cause the most damage to the environment (Nita & Castellani, 2017).
In light of these facts, a change of lifestyle and consumption behaviours of the human population is inevitable in order to significantly decrease greenhouse gas emissions (Hidalgo, Casado, & García-Leiva, 2014). However, it has been shown that people understand climate change as a rather distant threat (Upham et al., 2009; Pidgeon, 2012) and perceive low efficacy of individual actions on reducing the consequences of climate change (Lorenzoni & Pidgeon, 2006; Defra, 2007; Upham et al., 2009). Therefore, creating awareness about the harmful impact of greenhouse gas emissions as well as promoting pro-environmental behaviours can be seen as crucial in combating climate change. That is why environmental risk communication gains increasing relevance in academia (Roeser, 2012; Hidalgo et al., 2014).
Early research has demonstrated that the perception of threat constitutes an important predictor of pro-environmental behaviour (Baldassare & Katz, 1992). The role of perceived threat in environmental communication has also been examined in studies on fear appeals (e.g. Hartmann et al., 2014; Hunter & Röös, 2016). Nevertheless, to date, only a few studies explored environmental fear appeals within the framework of the Extended Parallel Process Model (EPPM). The EPPM provides explanations for the potential failure and success of fear
appeals and has been effectively applied in respect to a number of health-risk behaviours (Witte, 1992; Maloney, Lapinski, & Witte, 2011).
The focus of the present research lies on strengthening individuals' motivation for consuming little meat and household energy as pro-environmental behaviours. Therewith, this study is one of the first projects that not only investigates the role of EPPM-based fear appeals in encouraging people to perform the abovementioned behaviours but also in promoting two behaviours at the same time. Furthermore, it is aimed at extending the limited body of research into variables that moderate the impact of the critical variables in the EPPM (Witte & Allen, 2000; Maloney et al., 2011) by scrutinizing the role of so-called 'goal framing'. Goal framing is based on the Self-Determination Theory, and postulates that human behaviour is guided by the motivation for intrinsic goals (e.g. community contribution) or extrinsic goals (e.g. wealth) (Deci & Ryan, 2000). Research has shown that messages which describe a behaviour as effective in achieving an intrinsic goal (i.e. intrinsic goal frame) are more persuasive than messages which describe a behaviour as effective in achieving an extrinsic goal (i.e. extrinsic goal frame) (e.g. Vansteenkiste, Soenens, Verstuyf, & Lens, 2009), and that pro-environmental behaviour is linked to the motivation for intrinsic goals (Thøgersen, 2003; Tabernero & Hernández, 2011). It has however also been demonstrated that people who experience threat are likely to shift their motivation from intrinsic to extrinsic goals (Sheldon & Kasser, 2008), which suggests that threatening messages stating that a behaviour leads to an extrinsic goal (i.e. extrinsic goal frame) are more persuasive. On these grounds, the present study specifically examines the effect of intrinsic and extrinsic goal framed fear appeal
messages and addresses the following research question:
To what extent do intrinsic and extrinsic goal frames moderate the effect of environmental fear appeals regarding climate change on intention towards pro-environmental behaviours?
Theoretical Framework Fear Appeals and The Extended Parallel Process Model
Fear appeals constitute messages which are designed to evoke fear by presenting the harmful consequences of a behaviour (Rogers, 1975; Witte, 1992, 1994). Fear comprises a form of negative emotion that can be articulated in physical and facial expressions or verbal self-reports and which is evoked by a threat that is understood as relevant and severe (Witte, 1992, 1994). Former studies illustrated that the influence of fear appeals in persuasive communication on behaviour change is contingent on different components (Rogers, 1975). These include the depicted effectiveness (Dabbs & Leventhal, 1966) and the specificity of action recommendations, the severity of the presented risk arising (Leventhal, Singer, & Jones, 1965) as well as the level of fear arousal, of which the latter led to contradicting results. While early results demonstrated that high levels of fear can induce defensive
reactions and avoidance (Janis & Feshbach, 1953), it also has been shown that high levels of fear can be more persuasive than low levels of fear (Leventhal et al., 1965; Dabbs &
Leventhal, 1966; Rogers, 1975).
There are several theoretical models that outline the persuasive impact of risk communication, for instance the Health Belief Model (Hochbaum, 1956), the Subjective Expected Utility Theory (Edwards, 1961), the Parallel Response Model (Leventhal, 1971) and the Protection Motivation Theory (Rogers, 1975). The Protection Motivation Theory (PMT) developed by Rogers (1975) explains how fear arousing messages lead to positive responses in the form of behaviour change and is based on the concepts of perceived vulnerability, severity, coping efficacy and self-efficacy (Rogers, 1975; Maddux & Rogers, 1983). The EPPM includes similar variables but expands on the PMT in the way that it further highlights the occurrence of negative message responses (Witte, 1992; Maloney et al., 2011). It must be mentioned that research on fear arousing communication often combines the PMT and the EPPM (e.g. Hartmann et al., 2014). Furthermore, it is important to note that within the
framework of the EPPM, fear arousing messages are defined as 'fear appeals'. However, more recent studies also use the term 'threat appeal' (e.g. Hartmann et al., 2014). On the grounds that this study specifically investigates the role of high versus low threat components in fear appeals, such appeals will also be referred to as high or low threat appeals in the present paper.
The Extended Parallel Process Model (EPPM) has been developed by Kim Witte (1992) and represents a theoretical framework that is derived from the abovementioned models (Witte, 1992; Maloney, Lapinski, & Witte, 2011). More precisely, the EPPM is based on two key variables, namely threat and efficacy, which both contain two sub-dimensions. The variable threat consists of perceived susceptibility which refers to one's perceived risk of being affected by the threat and severity which concerns the magnitude of the threat posed (Witte, 1992, 1994), (e.g. measured with items such as "I believe this condition is serious.") (Witte et al., 1998). Efficacy comprises of response efficacy which is defined as the
effectiveness of the recommended response behaviour in mitigating the potential threat and self-efficacy which describes an individual's ability to perform those recommended actions (Witte, 1992), (e.g. "Condoms work in preventing this condition.") (Witte et al., 1998).
The EPPM postulates that depending on the level of perceived threat and efficacy, fear appeals can either evoke fear control or danger control behaviour. When the perceived threat is low, there is no need to evaluate the remaining message content (Witte, 1992). In contrast, when a message is understood as highly threatening, individuals experience fear and are stimulated to further process the information presented (Witte, 1992). The combination of high threat and efficacy elicits danger control behaviour because individuals feel that they are able to effectively avoid the threat (Witte, Cameron, McKeon, & Berkowitz, 1996). They then develop a more positive attitude towards the recommended behavioural actions, have a higher intention to perform behaviour change and are subsequently more likely to engage in the recommended actions (Witte et al., 1996). However, when the threat is high but the efficacy is
low, individuals do not know how to counteract the threat and therefore only react to their fear. Fear control behaviour involves derogative or reactive responses as well as message avoidance, for example when the recipient does not want to reflect upon the content received (Witte, 1992; Witte, Cameron, McKeon, & Berkowitz, 1996).
The EPPM has been successfully applied in the context of health risk communication (Maloney et al., 2011) regarding a number of behaviours such as HIV/AIDS prevention (Witte, 1994; Murray-Johnson et al., 2001), prevention of genital warts (Witte, Berkowitz, Cameron, & McKeon, 1998) or smoking cessation (Wong & Cappella, 2009) in different age groups and countries (Maloney et al., 2011). A meta-analysis has also illustrated that fear appeals have an overall positive effect on attitude, intention and behaviour change that is stronger when susceptibility, severity and efficacy components are integrated (Tannenbaum et al., 2015). Furthermore, the EPPM is considered to be particularly valuable for practitioners since it gives specific advice for the design and targeting of health messages (Popova, 2012).
Nevertheless, there are certain boundary conditions that apply to the model. First, it only explains the persuasive process of fear appeals and does not relate to other positive or negative emotional appeals (Maloney et al., 2011). Second, when individuals already have high levels of fear, the model has little explanatory value (Maloney et al., 2011) and it has for instance been demonstrated that fear appeals concerning HIV/AIDS prevention do not affect recipients who were very frightened about the disease beforehand (Muthusamy, Levine, & Weber, 2009). Third, there are situations in which the perceived efficacy of a response behaviour is low and cannot be sufficiently risen in order to counteract the potential threat (Witte et al., 1998; Maloney et al., 2011). For example, women who had low pre-existing efficacy perceptions of condom use in preventing genital warts displayed high fear control behaviour after being exposed to fear appeal messages (Witte et al., 1998). Practitioners are advised to not use fear appeals in such cases since the opposite effect could be evoked (Witte et al., 1998; Maloney et al., 2011).
Environmental Fear Appeals. Fear is one of the important factors that has been investigated in the context of the promotion of pro-environmental behaviours. Not only has it been outlined that an individual's negative emotional reactions to environmental problems increase the likelihood of engaging in pro-environmental actions (Kollmuss & Agyeman, 2002), but it also has been demonstrated that perceived threat constitutes a significant influence factor for encouraging pro-environmental behaviour (Baldassare & Katz, 1992; Johnston & Scicchitano, 2000). Fear appeals have been examined in studies concerning the intention towards counteracting radon gas exposure (LaTour & Tanner, 2003) and nuclear energy (Hartmann, Apaolaza, D’Souza, Echebarria, & Barrutia, 2013), towards renewable energy consumption (Hartmann et al., 2014), electric car use (Bockarjova & Steg, 2014), lower meat intake (Hunter & Röös, 2016), energy saving and the usage of environment-friendly transport (Rainear & Christensen, 2017) as well as the engagement in a number of pro-environmental actions (Kim, Jeong, & Hwang, 2012; Li, 2014), which provide general support for the EPPM and the PMT.
Hartmann and colleagues (2013, 2014) applied environmental fear appeals within the theoretical frameworks of the EPPM and the PMT. In line with the premises of the EPPM, it was demonstrated that perceived threat and efficacy were significantly and positively
correlated with individuals' intention towards renewable energy consumption and opposition to nuclear power (Hartmann et al., 2013). It was also shown that perceived threat and fear arousal mediated the effect of manipulated threat on intentions to consume and support green electricity (Hartmann et al., 2014). In the latter study, efficacy was not identified as a
mediator which could be due to the fact that the efficacy manipulation was not successful in increasing the participants' perception of efficacy (Hartmann et al., 2014). In contrast, Li (2014) showed that, in comparison to the other variables of the EPPM, response efficacy exerted the strongest effect on individuals' attitudes and intentions towards pro-environmental behaviour.
Interestingly, Chen (2016) found that individuals who were exposed to a little threatening stimulus had higher intentions to take up environmental actions than those who received highly threatening information. The author argues that high levels of fear elicit defensive responses, which corroborates with early research on risk communication (Janis & Feshbach, 1953). However, this study did not test whether the presented fear appeal provided efficacy information and examined the variable collective efficacy instead of individual participants' perception of response efficacy. This could explain the contradicting findings to other fear appeal studies which indicate that messages with high threat components are more persuasive than low threat messages.
Pro-Environmental Behaviour. As mentioned above, the focus of the present study lies on two pro-environmental behaviours, namely eating less meat and saving household energy. This is based on the grounds that within the EU, expenditures in the areas of energy, transport and food, specifically animal-based products, constitute the consumption behaviours with the most harmful impact on the environment (Nita & Castellani, 2017). Since the usage of sustainable means of transport such as cycling is relatively high in the Netherlands (Pucher & Buehler, 2008, 2017), it was decided to focus on the reduction of meat eating and
household energy usage in particular. In contrast to one-time actions such as installing solar panels or choosing a green energy provider, low meat and energy consumption represent routine behaviours that have to be performed again and again over time. Although fear appeals have been proven to have a stronger impact when promoting one-time actions, they can still be considered as influential for routine behaviours (Tannenbaum et al., 2015).
Support for the persuasive impact of fear appeals in motivating the specific pro-environmental behaviours comes from two studies based on the PMT. More precisely, it was found that the factors of severity, vulnerability, response and self-efficacy positively predict intentions to engage in a reduction of meat consumption (Hunter & Röös, 2016) as well as the intention towards household energy saving (Rainear & Christensen, 2017). It was also found that the
perceived efficacy of eating less meat is particularly low (Hunter & Röös, 2016) and that the efficacy factors constitute the strongest determinants of intention (Hunter & Röös, 2016; Rainear & Christensen, 2017). This stresses the importance of further promoting simple action recommendations as well as stressing the effectiveness of low-meat diets and low household energy usage in mitigating the consequences of climate change.
Intention towards Pro-Environmental Behaviour. Danger control responses concern responses to fear arousing messages that are manifested in attitude, intention and behaviour change (Witte, 1992, 1994; Witte et al., 1996). This can also be related to the Theory of Planned Behaviour which postulates that attitude and intention constitute
antecedent variables of behaviour (Ajzen, 1991). Similar to former studies on environmental fear appeals (e.g. Hartmann et al., 2013, 2014), the present study will scrutinize behavioural intention as outcome variable. More specifically, the effect of high- and low-fear appeals on individuals' intention to engage in pro-environmental actions as well as their willingness to reduce meat and household energy consumption will be assessed. This is also based on the grounds that the measurement of the participants' actual reduction of meat consumption and household energy use would be outside the scope of this study.
In light of the arguments depicted and consistent with previous research on health-related and environmental fear appeals, it is proposed that individuals who are provided with highly threatening information about climate change will have greater behavioural intentions towards pro-environmental behaviour than those who receive little threatening information:
H1: Participants exposed to environmental appeals with high threat components will have a stronger intention towards pro-environmental behaviour than participants exposed to environmental appeals with low threat components.
Moderator Variables. Little is yet known about potential moderators that affect the relationship between the EPPM variables and individuals' attitudes and behavioural intentions (Witte & Allen, 2000; Maloney et al., 2011). Moderator variables that have been established
in previous research are sensation seeking, which refers to one's willingness to take risks (Morrison, 1995), one's tendency to be anxious (Witte & Allen, 2000), cultural orientation (Murray-Johnson et al., 2001) and progress in the process of behaviour change (Cho & Salmon, 2006; Wong & Cappella, 2009). Also, it has been found that fear appeals which emphasize the potential losses resulting from not performing the recommended actions are generally more effective than those which promote gains arising from behaviour change, known as gain-loss framing, (Ruiter, Kok, Verplanken, & Van Eersel, 2003) and that recipients who were reminded of their own values and identity are more likely to engage in behaviour change (Napper, Harris, & Klein, 2014). Research on climate change
communication outside the topic of fear appeals has also identified factors that can increase the effectiveness of such persuasive messages, for instance monetary rewards (Thøgersen, 2003; Hidalgo et al., 2014), the perceived geographical distance of the environmental threat and gain-loss framing (Spence & Pidgeon, 2010). Furthermore, Pelletier and Sharp (2008) have proposed that framing based on the Self-Determination Theory could raise individuals' adaptive responses towards environmental behaviour.
The Self-Determination Theory
The Self-Determination Theory (SDT) (Ryan, Mims, & Koestner, 1983; Deci & Ryan, 1985, 1990, 2000) focuses on personality, needs and motivation and is supported by
numerous studies regarding health care (Ng et al., 2012), physical activity (Owen, Smith, Lubans, Ng, & Lonsdale, 2014), work (Van den Broeck, Ferris, Chang, & Rosen, 2016), education (Vansteenkiste, Lens, & Deci, 2006), virtual entertainment (e.g. Ryan, 2017) and culture (e.g. Chirkov, Ryan, & Willness, 2005). During the 1970's and 1980's, the SDT evolved in the course of research about those traits and external factors that influence behavioural motivation (e.g. Fisher, 1978; Deci & Ryan, 1985). Over time, the theory was further developed and has been especially applied in practical research about behaviour change during the last two decades (to date 2018), (Deci & Ryan, 2017). The theory is built
upon six sub-theories (Deci & Ryan, 2017), namely the Cognitive Evaluation Theory (Ryan, Mims, & Koestner, 1983), Organismic Integration Theory (Deci & Ryan, 1990), Causality Orientations Theory (Deci & Ryan, 1985), Psychological Needs Theory (Ryan, 1995), Goal Contents Theory (Deci & Ryan, 2000), and Relationships Motivation Theory (Deci & Ryan, 2014). The basic principles of the SDT will be explained in the following paragraph.
First, it is postulated that human psychological well-being depends on the degree to which needs for autonomy, competence and relatedness are fulfilled. Second, people are assumed to be either impersonally, autonomy or control orientated (Deci & Ryan, 2000, 2015). When the abovementioned needs are not satisfied, people tend to be impersonally orientated which means that they do not feel competent to perform a behaviour (Deci & Ryan, 2015). The focus of autonomy orientated individuals lies on their own choice, while the behaviour of control orientated individuals is determined by controlling occurrences such as regulations (Deci & Ryan, 2000, 2015). Third, these orientations are linked to different motivations that underlie behaviour. A distinction is thereby made between behaviours motivated by intrinsic versus extrinsic goals (Deci & Ryan, 2000, 2015). Behaviours that are motivated by an intrinsic goal are based more on the desire to be autonomous (Vansteenkiste, Lens, & Deci, 2006), personal interest and enjoyment of the behaviour itself (Deci, 1972). In contrast, extrinsically driven behaviours are based more on rewards that are separated from the behaviour itself but arise from carrying out the behaviour (Deci, 1972; Vansteenkiste et al., 2006). Intrinsic goals are related to humans' basic needs and include personal growth, satisfying relationships, community contribution, and physical health. Extrinsic goals on the other hand include financial success, social popularity and physical attractiveness (Kasser & Ryan, 1996; Deci & Ryan, 2000, 2015). Research on the SDT tends to focus on whether messages that aim to motivate behaviour change are more effective when an intrinsic or extrinsic goal is emphasised (i.e. goal frame).
Goal Frames and Low Threat Appeals. Generally speaking, individuals are more prone to be motivated by intrinsic goal frames than extrinsic goal frames. For instance, in regard to study motivation (e.g. Vansteenkiste, Simons, Lens, Soenens, et al., 2004; Vansteenkiste, Simons, Lens, Sheldon, et al., 2004; Vansteenkiste, Timmermans, Lens, Soenens, & Van den Broeck, 2008) and physical exercise (e.g. Vansteenkiste, Simons, Lens, Sheldon, & Deci, 2004; Vansteenkiste, Matos, Lens, & Soenens, 2007), it has been found that framing a task as useful in achieving an intrinsic goal (e.g. "doing a little tae bo helps
you to remain physically fit and prevents you from becoming sick at a later age"), leads to
higher performance than when the behaviour is framed with an extrinsic goal (e.g. "doing a
little tae bo helps you to remain physically appealing to others and prevents you from gaining weight at a later age") (Vansteenkiste, Simons, Lens, Sheldon, et al., 2004, as cited
in Vansteenkiste, Matos, Lens, & Soenens, 2007).
In addition, it has been shown that motivation for intrinsic goals mediated the effect of high self-efficacy beliefs about recycling on actual engagement with recycling (Tabernero & Hernández, 2011). On the other hand, the motivation for extrinsic goals did not explain the relationship between self-efficacy and recycling behaviour and was negatively correlated with recycling behaviour (Tabernero & Hernández, 2011). This is in line with former research which illustrated that pro-environmental actions such as recycling are particularly influenced by internal, moral motives (Stern, 2000; Chan & Bishop, 2013; Gatersleben, Murtagh, Cherry, & Watkins, 2017). Furthermore, it has been suggested that the extrinsic goal framing of environmental actions, for instance with monetary incentives, can stimulate dishonesty (Thøgersen, 2003). In general, it can therefore be assumed that low threat messages which present pro-environmental behaviour as promoting an intrinsic goal are more persuasive than low threat messages promoting an extrinsic goal. On these grounds it is hypothesized that intrinsic goal framing can enhance the effect of low threat
H2: Participants exposed to environmental appeals with low threat components combined with intrinsic goal framed information will have a stronger intention towards pro-environmental behaviour than participants exposed to low threat appeals combined with extrinsic framed information.
Goal Frames and High Threat Appeals. The motivation for extrinsic goals has been linked to lower mental health and well-being (Kasser & Ryan, 1993), states of anxiety (Kasser & Ahuvia, 2002), unhappiness (Sheldon, Gunz, Nichols, & Ferguson, 2010) and stronger engagement in health risk behaviours (Williams, Cox, Hedberg, & Deci, 2000). Also, the experience of threat seems to elicit extrinsic goal motivation. Early research has shown that threats, negative feedback or punishments decrease people's motivation for intrinsic goals (Deci & Wayne, 1972). Furthermore, Sheldon and Kasser (2008) have demonstrated in three studies that individuals' motivation significantly changed from intrinsic to more extrinsic after exposure to information about an existential, economic or interpersonal threat. It was also shown that calling attention to one's own mortality makes extrinsic goals more salient (Kasser & Sheldon, 2000), and that when presented with information about ethnic groups, individuals who are high in extrinsic motivation hold more negative attitudes and perceive such groups as more threatening than those high in intrinsic motivation (Duriez, Meeus, & Vansteenkiste, 2012).
Since no prior research has been found that examined goal framing in regard to EPPM-based messages, this study is the first one which proposes that highlighting the extrinsic outcome of pro-environmental behaviour positively moderates the effect of high threat appeals on behavioural intention:
H3: Participants exposed to environmental appeals with high threat components combined with goal framed information will have a stronger intention towards pro-environmental behaviour, but this effect will be more pronounced for extrinsic framed information than intrinsic framed information.
Methods Section Research Design
To test the hypotheses developed, a true and web-based experiment was conducted. The 2x2x2 experimental design consisted of three factors, namely Threat, Goal Frame and
Behaviour. Threat concerned the depicted level of danger from climate change, and
constituted a between-subjects factor with two levels: high or low. The variable Goal Frame was about the benefit the pro-environmental behaviours were portrayed to entail, and also represented a between-subjects factor with two levels: intrinsic or extrinsic goal frame. Finally, Behaviour, as the recommended action to combat climate change, was composed as within-subjects factor with two levels: reduction of meat and energy consumption. In total, this results in eight conditions with four different stimuli as shown in Table 1 below.
Table 1: Experimental design
Factor 1: Behaviour Factor 2: Threat Factor 3: Goal Frame
intrinsic extrinsic Meat consumption high
Energy consumption
Meat consumption low Energy consumption
Pilot-Study
Before the actual conduct of the study, a pilot-study was conducted with a sample of 32 undergraduate (n = 9) and graduate (n = 23) students, from which 81.3% were female and 18.8% male. Furthermore, 62.5% of the respondents were between 18 and 24 years old and 31.3% between 25 and 34 years.
The pilot-study served three purposes: a) selecting the best image that represents high threat versus low threat to the environment, b) determining the most important intrinsic and
extrinsic goal for performing the two pro-environmental behaviours, and c) assessing whether the textual manipulation of Threat is effective.
First, participants were exposed to ten images which depicted environmental scenes varying in the extent to which the environment seemed damaged. The participants were asked to rate their perception of threat for each image, with a scale from 'Extremely threatening' (1) to 'Not at all threatening' (7). Based on these findings, the most (M = 2.12, SD = .87) and least (M = 6.66, SD = 1.05) threatening image was selected in the manipulation materials of the two level factor Threat in the main study. These two images can be seen in Figure 1 and 2 further below.
Second, the participants were asked to imagine that they would reduce their meat and household energy consumption. They were then presented with three intrinsic and extrinsic reasons for energy saving and five intrinsic and extrinsic reasons related to a low meat diet. These intrinsic and extrinsic reasons were based on the Aspiration Index (Kasser & Ryan, 1993, 1996), and included reasons such as 'Because I can improve my own health.' (i.e. intrinsic reason) and 'Because I can lose weight and thereby improve my physical
appearance.' (i.e. extrinsic reason). The participants were asked to indicate how important each reason would be for them when performing the behaviour on Likert scales ranging from 'Extremely important' (1) to 'Not at all important' (5). For both of the behaviours, the intrinsic reason 'Because I can do something good for the environment.' (Energy: M = 1.47, SD = .67; Meat: M = 1.81, SD = 1.20) and the extrinsic reason related to saving money (Energy: M = 1.97, SD = .93; Meat: M = 2.75, SD = 1.34) were considered as the most important by the participants in the pilot study. Therefore, these two reasons were chosen for the manipulation of the two level factor Goal Frame in the main study.
Third, to check whether the manipulation of Threat would be perceived as intended and that there would be a difference between the high threat and low threat condition in regard to the perception of severity and susceptibility, the manipulation check of the textual
information in the main study was included in the pilot-study. The fear appeal components susceptibility and severity were assessed with items such as 'The text portrays severe
consequences resulting from climate change.' on a Likert scale ranging from 'Strongly agree' (1) to 'Strongly disagree' (7). A paired samples t-test indicated that the portrayed severity in the high threat condition (M = 2.06, SD = .98) was perceived as significantly higher than in the low threat condition (M = 3.75, SD = 1.93), t (31) = 4.51, p < .001, 95% CI [.92, 2.45], with a relatively strong effect of d = .79. Also, the portrayed susceptibility was perceived as significantly higher in the high threat (M = 2.28, SD = 1.05) than in the low threat condition (M = 3.81, SD = 1.99), with t (31) = 4.26, p < .001, 95% CI [.79, 2.26], and a moderately strong effect of d = .75. On the grounds that the Threat manipulation was successful in the pilot study, this manipulation was also used in the main study.
Finally, the pilot study checked whether respondents perceived that the text provided response efficacy and self-efficacy information, through items such as 'The text describes a reduction of meat consumption as effective in decreasing greenhouse gas emissions.' The scales here ranged from 'Strongly agree' (1) to 'Strongly disagree' (7). The descriptive
statistics demonstrated that portrayed response efficacy in the high threat condition (Energy: = 2, SD = 1.11; Meat: M = 1.97, SD = 1.18) and the low threat condition (Energy: M = 1.91, SD = .78; Meat: M = 1.75, SD = .84) was rated as high. Similarly, self-efficacy in the high
(Energy: M = 1.91, SD = .78; Meat: M = 1.84, SD = 1.08) and the low threat condition (Energy: M = 1.81, SD = .74; Meat: M = 1.78, SD = .83) was understood as relatively high. Furthermore, paired samples t-tests confirmed that the perception of efficacy between the two conditions was not different, which was expected since both of the presented texts contained the exact same efficacy information. On these grounds, it could be assumed that participants in the main study would understand the stimulus materials as containing clear response and self-efficacy information.
Experimental Stimulus
Following previous research on fear appeals, the stimulus material in all conditions was composed of three parts, namely a) imagery of nature (Hartmann et al., 2014), b) textual information about climate change, and c) action recommendations (Hunter & Röös, 2016). The full stimulus material, including the references, can be found in Appendix B.1
Threat. The participants in the high threat condition saw an image which depicted a flooded road and this image was evaluated as highly threatening in the pilot study (see Figure 1). In contrast, participants in the low threat condition were exposed to a landscape image that was evaluated as little threatening in the pilot-study (see Figure 2).
1.
1.
The textual information in the low and high threat condition was the same, apart from one paragraph in the high threat condition that described the serious consequences resulting from
1
The information's form and wording was intended to comply with a fear appeal that was created by Hunter and Röös (2016). The facts about the severity and susceptibility regarding climate change were derived from the Climate Change Factsheet 2015 which was published by the European Commission (EC, 2015), a paper about the consequences of climate change in the Netherlands by Arends (2011) and information from the PBL Netherlands
Environmental Assessment Agency (2012). In addition, the efficacy components are based on a recent report on behalf of the European Commission from Nita and Castellani (2017), findings from Garnett (2014), Hallström et al. (2015), Hidalgo et al. (2014) and a factsheet from the WHO (2008).
Figure 2. Low Threat Condition Figure 1. High Threat Condition
climate change (i.e. severity) and the way in which the population in North-Western European countries will be affected by climate change (i.e. susceptibility). The textual information can be seen in Figure 3.
Pro-Environmental Behaviour. On the grounds that Behaviour is treated as within-subjects factor, all participants were provided with information that promoted both a
reduction of meat, and a reduction of energy consumption as pro-environmental behaviours that can counteract climate change (see Figure 3). Two different behaviours were selected for the purpose of avoiding bias arising due to one's existing attitude towards a certain behaviour.
Goal Frame. Based on the findings from the pilot-study, the intrinsic goal frame presented the pro-environmental behaviours as a way to do something good. The extrinsic goal frame on the other hand however referred to monetary benefits (see Figure 3).
Climate experts agree that climate change is caused by greenhouse gas emissions that originate in the main from the areas of energy, food and transport. The amount of greenhouse gases must be reduced now to mitigate the future consequences of climate change.
[Threatening Information]
Particularly in North-Western Europe, climate change will bring rising sea levels, floods and extreme weather conditions which can lead to severe changes of ecosystems, a damage of agricultural crops, animal diseases, plagues and a displacement of animal species.
Climate change is expected to put the health of millions of people in danger. In North-Western European countries, this includes an increase in allergies, infections and health conditions related to heat, high humidity and air pollution.
[Efficacy Information]
By cutting down on the consumption of meat products, greenhouse gas emissions can be reduced up to 35% which is about 540 kg of CO2 per person per year. There are a few simple steps you can take to effectively decrease your meat consumption:
Cut your number of meals with meat to half.
Plan your meals to reduce food waste.
Eat more foods with a low carbon footprint such as legumes, whole grains, and fruits instead of animal-based products.
Domestic energy use is responsible for 40% of the total energy consumption within the EU. There are simple steps you can take to effectively decrease your household's energy consumption:
Choose energy efficient light bulbs which produce up to 400 kg less CO2 than normal bulbs.
Turn off the lights when not in use which can save up to 400 kg of CO2 per year.
Disconnect electrical appliances and chargers from the plugs when not using them.
[Intrinsic Goal Frame]
Reducing your environmental impact is a way to make this world a better place. If you eat less meat products and save energy, you will do something good and help to save the environment!
[Extrinsic Goal Frame]
In comparison to meat products, plant-based foods such as legumes and vegetables are less costly. Similarly, reducing your energy consumption decreases your electricity costs. If you consume less meat and energy, you will therefore lower your household expenses and save money.
Selection of Research Units
For the present study, a convenience sample was chosen. The data were collected within a time period of one week in April/May 2018. The link to the online-experiment, which was conducted via the survey platform Qualtrics, was shared via Facebook. All
participants were debriefed about the purpose of the study after completing the questionnaire. In order to attain a statistical power of about 80 percent (Von Voorhis & Morgan, 2007), it was intended to reach at least 30 participants for each of the experimental groups which led to a minimum sample size of 120 participants.
Characteristics of Research Units
The sample comprised of 198 respondents who completed the online questionnaire and gave their informed consent to participate in the study. In an exposure check, the participants were asked which topic they remember having read information about and they could either choose between the answers 'Climate change, meat consumption and energy consumption.' or 'Climate change, transport usage and plastic consumption.' Nineteen
participants were excluded from the sample because they selected the wrong answer, showing that they had paid little or no attention at all to the stimulus materials. The final sample therefore consisted of 179 participants. It must be noted that the exclusion of participants led to unequal sizes of the four experimental groups (n1 = 41, n2 = 41, n3 = 51, n4 = 46).
The demographic statistics showed that the age range of participants lay between 18 and 74 years, with an average age of almost 28 years (MAge = 27.77, SD = 10.99). The sample
consisted of 75.4% female, 22.9% male and 1.7% participants who considered their gender to be non-binary. The countries in which the participants lived the longest were Austria (40.8%), Germany (15.1%) and the Netherlands (11.2%), followed by 23 other countries. Furthermore, most of the participants held a master's (26.3%), a bachelor's (43%), or a high school degree
(22.3%) and were enrolled as graduate (50.1%) or undergraduate student (21.8%) at the time of the study.
Randomization Checks. In order to check whether variables concerning the
participants' demographics and relevant pro-environmental habits, were randomly distributed between the four experimental groups, randomization checks were conducted. One-way analyses of variance (Anova) showed that there was no significant difference between the conditions in regard to age, F (3,175) = .55, p = .657, participants' current meat consumption,
F (3,175) = .26, p = .853, their existing efforts to save household energy, F (3,175) = .18, p =
.906, and their perceived relevance of the topic climate change, F (3,175) = .65, p = .582. However, the variable gender was not equally distributed between the experimental groups, F (3,175) = 3.08, p = .029, η2 = .050. Therefore, gender was included as covariate in further analyses.
Manipulation Checks. The same items as in the pilot-study were used for the manipulation check of the factor Threat with scales ranging from 'Strongly disagree' (1) to 'Strongly agree' (7) (see Appendix A for a full list of measurement items). An independent samples t-test was performed which indicated that the participants in the groups which were exposed to the high threat stimulus (M = 5.78, SD = 1.28) rated the portrayed susceptibility as significantly higher than the groups exposed to the low threat stimulus (M = 4.05, SD =1.89),
t (169.68) = 7.24, p < .001, 95% CI [1.26, 2.2], with a moderate effect of r = .39. Similarly,
portrayed severity was perceived as significantly higher in the high threat groups (M = 5.5,
SD = 1.39), compared to the low threat groups (M = 4, SD = 1.85), t (174.72) = 6.19, p < .001,
95% CI [1.02, 1.98], with a strong effect of r = .49. These results imply that the manipulation of the two level factor Threat was successful.
Furthermore, an independent samples t-test was conducted to assess whether the manipulation of the two level factor Goal Frame was effective. This was measured with two items, one of which was 'The information portrayed a reduction of meat and energy
consumption as financially beneficial' (see Appendix A for both items). The two items did not significantly correlate which each other, r = .12, p = .101, meaning that the relationship between these variables was positive but that there was only a low degree of correlation. This suggests that the two measures were independent from one another and concerned different constructs.
The results showed that the participants in the extrinsic goal frame condition (M = 5.08, SD = 1.91) had a significantly higher perception that the pro-environmental behaviour was presented as financially beneficial (i.e. extrinsic goal frame) compared to the intrinsic goal frame condition (M = 4.05, SD = 1.89) t (177) = 7.04, p < .001, 95% CI [1.38, 2.45], with a strong effect of r = .47. However, there was no difference in participants' perception that the pro-environmental behaviours were presented as a way to do something good for the environment (i.e. intrinsic goal frame) between the two groups provided with the extrinsic goal framed information (M = 6.16, SD = 1.22) and the groups presented with the intrinsic goal framed information (M = 6.16, SD = 1.05), t (177) = .01, p = .990, 95% CI [-.32, .32]. This indicates that the manipulation of intrinsic versus extrinsic goal frames was not fully successful: participants in the extrinsic goal frame conditions felt they were exposed to information framed with both the extrinsic goal of receiving financial benefits but also the intrinsic goal of doing something good for the environment.
Measures
Control Variables. Three variables that were expected to influence the dependent variables were included in the study. This concerned participants' meat consumption (Meat
Eating: M = 2.94, SD = 1.31), their efforts to save energy (Energy Saving: M = 3.96, SD =
.85), and their perceived relevance of climate change (Climate Change Relevance: M = 4.45,
SD = .74) (See Appendix A for measurement items and answer scales).
Dependent Variable - Intention. The participants' intention towards pro-environmental behaviour was assessed through the measurement of three sub-variables,
namely participants' intention to generally act more environment-friendly as well as their intention to eat less meat and to save energy, and these are discussed in more detail below.
Pro-Environmental Intention. The participants' willingness to generally engage in
environment-friendly actions was assessed with four items created by Lee (2011) which were thematically based on Bang et al. (2000). The participants were asked to rate their intention to act more sustainably with 5-point Likert scales (see Appendix A for items and scales). An exploratory factor analysis with axis factoring as extraction method and orthogonal rotation (varimax) was conducted on the four items. The Kaiser-Meyer-Olkin measure, KMO = .83, confirmed the sampling adequacy for this analysis with a value above the limit of .5 (Field, 2009). Bartlett's test of sphericity, χ2 (6) = 397.48, p < .001, showed that the correlations between the items were suitable for the factor analysis. The items loaded on 1 factor with an Eigenvalue of 2.98, explaining 74.48% of the variance. The scree plot depicted 1 factor before the inflection point and the correlation matrix demonstrated that all items were positively correlated. Furthermore, the reliability analysis showed that the reliability of the scale was high, Cronbach's ⍺ = .89. On these grounds, the four items were aggregated into the new variable Pro-Environmental Intention (M = 4.12, SD = .64).
Meat Reduction Intention. Participants' intention towards eating less meat was
measured with four items and 5-point Likert scales (see Appendix A for items and scales).2 The items were examined in a factor analysis using axis factoring as extraction with orthogonal rotation (varimax). The Kaiser-Meyer-Olkin measure above .5, KMO = .73, provided support for the adequacy of the sample and Bartlett's test of sphericity, χ2 (6) = 312.99, p < .001, demonstrated that the correlations between the items were reliable. Furthermore, the items loaded on 1 factor with an Eigenvalue of 2.53 which accounts for 63.28% of the variance. The scree plot demonstrated 1 factor before the inflection point and
2
The measurement items of the intention to engage in the two behaviours were formulated by the author of this study and are similar to previous measurements of intention towards pro-environmental actions based on de Boer, Witt and Aiking (2016).
the correlation matrix showed that the items correlate positively. One item was removed on the grounds that the factor loading of .414 was below the recommended loading of .512 for a similar sample size (Field, 2009)3. The reliability analysis indicated that the reliability of the scale covering the remaining three items was high, Cronbach's ⍺ = .88. Therefore, the new variable Meat Reduction Intention (M = 4.08, SD = .98) was computed.
Energy Reduction Intention. The intention towards consuming less household energy
was measured with four items and 5-point Likert scales (see Appendix A for items and scales). An exploratory factor analysis with axis factoring as extraction method and
orthogonal rotation (varimax) was performed and the Kaiser-Meyer-Olkin measure, KMO = .67, confirmed the sampling adequacy for this analysis. Bartlett's test of sphericity, χ2 (6) = 140.66, p < .001, indicates that the correlations between the items were sufficiently high. The items loaded on 1 factor with an Eigenvalue of 2.1, explaining 52.48 % of the variance. The scree plot displayed 1 factor before the inflection point and the correlation matrix
demonstrated that the items were positively correlated. However, the reliability of the scale was low, Cronbach's ⍺ = .63. Since the value of alpha would not significantly improve, no item was removed and the new variable Energy Reduction Intention (M = 4.49, SD = .50) was created. Information concerning the normal distribution of the variables can be found in Appendix C.
Relationship Between Intention Variables. Relationships between the three scales measuring pro-environmental intentions outlined above, were assessed through correlations. As shown in Table 2 below, the variable Meat Reduction Intention correlated highly and positively with the more general variable Pro-Environmental Intention, r = .51, p < .001. Similarly, Energy Reduction Intention correlated moderately and positively with
Pro-Environmental Intention, r = .27, p < .001. In contrast, the two variables Meat and Energy Reduction Intention did not correlate significantly with each other r = .12, p = .110. This
3
indicates that the measure of the intention to generally perform environment-friendly actions can be viewed as overarching concept, encompassing participants' intention to eat less meat and their intention to use less household energy which constituted more specific measures that were also independent from one another.
Table 2 also displays the correlations between the three dependent and the control variables, perceived relevance of climate change (Climate Change Relevance), participants' current meat consumption (Meat Eating) and their efforts to save household energy (Energy
Saving) (see section about the Control variables above) and Gender. The control variables that
showed to correlate highly with the dependent variables were controlled for in the analyses. This will be further explained in the results section below.
Table 2: Means, Standard Deviations, and Correlations of Dependent and Control Variables
N = 179, ** p < .001. Measure M SD 1 2 3 4 5 6 7 1. Pro-Env. Intention 4.11 .64 — 2. Meat Intention 4.08 .98 .51** — 3. Energy Intention 4.49 .50 .27** .12 — 4. Climate Relevance 4.45 .74 .65** .44** .35** — 5. Meat Eating 2.94 1.31 -.32** -.65** -.07 -.30** — 6. Energy Saving 3.96 .85 .31** .12 .52** .29** -.08 — 7. Gender 1.26 .48 -.03 -.07 -.00 .03 .23** -.09 —
Results
In order to test Hypothesis 1, 2 and 3, two-way analyses of covariance (factorial Ancova) were conducted. The below sections will first provide an explanation and
justification of the analyses performed. A discussion of the results will follow in the sections thereafter.
Explanation and Justification of Analyses
The first Ancova was conducted to evaluate whether the two level factor Goal Frame moderates the effect of Threat on participants' Pro-Environmental Intention. Since Gender was not equally distributed between the experimental groups, the variable was added as a covariate in the analysis. In addition, participants' perceived Climate Change Relevance was included as a covariate on the grounds that this variable highly correlated with the dependent variable Pro-Environmental Intention, r = .65, p < .001 (see Table 2), and could therefore account for variance in the dependent variable that is not caused by the manipulated factors (Field, 2009). The assumptions for Ancova were checked before conducting the actual analysis and the results of these checks are reported in Appendix C. In summary, the
assumptions of independence and homogeneity of regression slopes were met, apart from the fact that the variable Gender was not independent from the manipulated factors Threat and
Goal Frame. A violation of this assumption indicates that the effects of the covariates
imbricate the effects of the manipulated factors and thereby confound the results (Field, 2009). On the grounds that this assumption was partly violated, the results should be interpreted with caution. The full results of the Ancova are listed in Table 3 in Appendix D and will be further discussed in the next section.
Furthermore, a two-way analysis of covariance was conducted to test the effect of
Threat and Goal Frame on participants' Meat Reduction Intention, controlling for Gender and Meat Eating. Meat Eating was included because this variable highly correlated with Meat Reduction Intention, r = -.65, p < .001 (see Table 2), and could therefore cause some of the
variance in this variable. As mentioned above, Gender was not independent from Threat and
Goal Frame. Apart from that, the assumptions of independence and homogeneity of
regression slopes were met. Further details about these assumptions are explained in
Appendix C. The full results of the Ancova can be seen in Table 4 in Appendix D and will be outlined in the next section.
Moreover, a two-way analysis of covariance examined the effect of Threat and Goal
Frame on Energy Reduction Intention, including the covariates Gender and Energy Saving. Energy Saving was included as covariate because it correlated highly with Energy Reduction Intention, r = .52, p < .001 (see Table 2), and could cause variance in the dependent variable.
The assumption of independence was partly violated since Gender was not independent from the factors Threat and Goal Frame and because Energy Saving was not independent from
Threat. As aforementioned, a violation of this assumption can confound the results (Field,
2009). The assumption of homogeneity of regression slopes was fulfilled, apart from the non-homogenous regression slopes between Energy Saving and the factor Threat. This means that the covariate was different between the experimental conditions and that the model was not representative of all conditions (Field, 2009, 2016). More details concerning the assumptions for Ancova are attached in Appendix C. On the grounds that the assumptions for Ancova were not fully met, findings must be treated with caution. The full results of the Ancova can be seen in Table 5 in Appendix D and will be described in the next section.
The Main Effect of Threat
Hypothesis 1: Effect of Threat Appeals on Pro-Environmental Intentions. Hypothesis 1 proposes that individuals in the high threat condition would have greater intentions to engage in pro-environmental behaviour than those in the low threat condition. Firstly, it was tested whether participants' Pro-Environmental Intention differed as a function of the level of threat in the message. The results showed that participants' Pro-Environmental
and the low threat condition, F (3,175) = .53, p = .469. The full results of the Ancova
described are listed in Table 3 in Appendix D and the means are presented in Table 6 below. Secondly, respondents' Meat Reduction Intention was examined. Similar to the aforementioned result, participants' willingness to eat less meat was relatively high, but there was no significant difference between participants' Meat Reduction Intention in the high and the low threat condition, F (3,175) = .21, p = .650. The full results of the Ancova are depicted in Table 4 in Appendix D. Table 7 below shows the means of Meat Reduction Intention in the two conditions.
Thirdly, the effect of Threat on Energy Reduction Intention was investigated. The results again indicated that Energy Reduction Intention was high and that there was no
significant difference between the high and the low threat condition, F (3,175) = .34, p = .560. The full results of the Ancova can be found in Table 5 in Appendix D. The means of Energy
Reduction Intention of the high and low threat condition are depicted in Table 8 below.
In all, on the grounds that no main effects of Threat on any of the Intention measures were found, Hypothesis 1 could not be supported: level of threat in a climate change message that participants were exposed to did not impact participants' willingness to engage in more environment-friendly behaviours. This result will be further elaborated upon in the discussion section.
Table 6: Means & Standard Deviations of Pro-Environmental
Intention by Threat
Condition n M SD Ma
High Threat 82 4.13 .64 4.15
Low Threat 97 4.10 .65 4.10
Notes. N = 179, a The means depicted are adjusted to the covariates Climate Change Relevance and Gender.
Table 7: Means & Standard Deviations of Meat Reduction
Intention by Threat
Condition n M SD Ma
High Threat 82 4.02 1.03 4.05
Low Threat 97 4.13 .94 4.10
Notes. N = 179, a The means depicted are adjusted to the covariates Meat Eating and Gender.
Table 8: Means & Standard Deviations of Energy Reduction
Intention by Threat
Condition n M SD Ma
High Threat 82 4.48 .45 4.47
Low Threat 97 4.51 .54 4.51
Notes. N = 179, a The means depicted are adjusted to the covariates Energy Saving and Gender.
The Moderating Role of Goal Frames in Shaping the Effects of Threat Appeals A next step in the analyses targeted to what extent the goal frames presented in the stimulus materials shaped the effects of threat appeals on participants' pro-environmental intentions.
Hypothesis 2: Moderating Role of Goal Frames in the Effects of Low Threat Appeals. Hypothesis 2 concerned the prediction that low threat appeals in combination with intrinsic goal frames would lead to higher intentions than in combination with extrinsic goal frames. The relevant Ancova showed a significant, but small interaction effect between
.045. The full results of the Ancova are illustrated in Table 3 in Appendix D. In order to interpret the significant interaction effect between Threat and Goal Frame in greater detail, a second Ancova of Pro-Environmental Intention by the four experimental conditions was conducted. The pairwise comparison of means with Bonferroni correction illustrated the means adjusted to the covariates. As shown in Table 9 below, individuals in the low threat condition had significantly higher Pro-Environmental Intentions when the behaviour was framed with an intrinsic (M = 4.26) than an extrinsic goal (M = 3.94) (Mdifference = .323, p < .001). Also, the combination of low threat and intrinsic goal frame was the most effective in comparison to the other three conditions. In regard to participants' willingness to generally engage in pro-environmental actions, these results provide support for Hypothesis 2.
Table 9: Means & Standard Deviations of Pro-Environmental Intention in low Threat
conditions.
Condition n M SD Ma
Low Threat & Extrinsic Goal Frame 51 4.00 .65 3.94
Low Threat & Intrinsic Goal Frame 46 4.21 .63 4.26
Notes. N = 97, a The means depicted are adjusted to the covariates Climate Change
Relevance and Gender.
Furthermore, an Ancova demonstrated that there was no significant interaction effect between level of Threat and Goal Frames regarding Meat Reduction Intention, F (3,175) = .04, p = .852. The findings do not support the propositions of Hypothesis 2 in respect to participants' willingness to consume less meat. The full results of the Ancova described are listed in Table 4 in Appendix D. The means of intention within the low threat conditions are shown in Table 10 below.
Table 10: Means & Standard Deviations of Meat Reduction Intention in low Threat
conditions.
Condition n M SD Ma
Low Threat & Extrinsic Goal Frame 51 4.16 .90 4.14
Low Threat & Intrinsic Goal Frame 46 4.10 .99 4.07
Notes. N = 97, a The means depicted are adjusted to the covariates Meat Eating and
Gender.
Moreover, the results showed that there was no interaction effect between level of
Threat and Goal Frames regarding Energy Reduction Intention between the participants, F
(3,175) = .40, p = .526. The full results can be seen in Table 5 in Appendix D, and the means of intention within the low threat conditions are illustrated in Table 11 below.
Table 11: Means & Standard Deviations of Energy Reduction Intention in low Threat
conditions.
Condition n M SD Ma
Low Threat & Extrinsic Goal Frame 51 4.50 .55 4.49
Low Threat & Intrinsic Goal Frame 46 4.51 .54 4.53
Notes. N = 97, a The means depicted are adjusted to the covariates Energy Saving
and Gender.
In light of these findings, it must be concluded that Hypothesis 2 was partly supported. Under condition of low threat, intrinsic goal frames indeed led to higher intentions to engage in pro-environmental behaviour than extrinsic goal frames. However, this concerned only
participants' intention to generally engage more in pro-environmental actions and did not concern the specific targeted pro-environmental behaviours of intention to consume less meat or household energy.
Hypothesis 3: Moderating Role of Goal Frames in the Effects of High Threat Appeals. Hypothesis 3 proposed that high threat appeals combined with extrinsic goal frames would result in greater intentions than high threat appeals with intrinsic goal frames. In regard to Pro-Environmental Intention, a significant interaction effect between Threat and Goal
Frame was found (see Table 3 in Appendix D for results). The post hoc comparison of means
demonstrated that the difference between the low and high threat appeal was not significant (p = .408). Nevertheless, the direction of the effect was as expected, since the high threat appeal led to somewhat higher Pro-Environmental Intentions when the behaviour was framed with an extrinsic (M = 4.19) than an intrinsic goal (M = 4.11) (Mdifference = .088) (see Table 12 below). As there was no significant difference among the two groups, the proposition that high threat appeals in combination with extrinsic goal frames lead to higher willingness to engage in pro-environmental actions can therefore not be proven.
Table 12: Means & Standard Deviations of Pro Environmental in high Threat conditions.
Condition n M SD Ma
High Threat & Extrinsic Goal Frame 41 4.15 .67 4.19
High Threat & Intrinsic Goal Frame 41 4.13 .61 4.11
Notes. N = 82, a The means depicted are adjusted to the covariates Climate Change
As elaborated upon in the paragraphs above, there was no significant interaction effect between Threat and Goal Frame in regard to participants' Meat or Energy Reduction Intention found (see Table 4 and 5 in Appendix D for results). The means of intention are depicted in Table 13 and 14. It can be concluded that participants' intentions towards consuming less meat and energy were not significantly different between the experimental groups. Therefore, Hypothesis 3 cannot be supported.
Table 13: Means & Standard Deviations of Meat Reduction Intention in high Threat
conditions.
Condition n M SD Ma
High Threat & Extrinsic Goal Frame 41 4.06 .97 4.10
High Threat & Intrinsic Goal Frame 41 3.97 1.10 4.00
Notes. N = 82, a The means depicted are adjusted to the covariates Meat Eating and
Gender.
Table 14: Means & Standard Deviations of Energy Reduction Intention in high Threat
conditions.
Condition n M SD Ma
High Threat & Extrinsic Goal Frame 41 4.43 .46 4.41
High Threat & Intrinsic Goal Frame 41 4.53 .44 4.53
Notes. N = 82, a The means depicted are adjusted to the covariates Energy Saving
Discussion
Hypothesis 1: Effect of Threat Appeals on Pro-Environmental Intentions
Based on the Extended Parallel Process Model (EPPM) (Witte, 1992), explained in the theoretical framework, it was proposed that high threat appeals would be more effective in increasing participants' willingness to engage in pro-environmental behaviour than low threat appeals. However, as outlined in the results section, participants exposed to the high threat stimulus materials in the present study did not exhibit higher levels of intention than those in the low threat condition. Therefore, Hypothesis 1 could not be supported regarding
participants' intention to generally perform pro-environmental actions, neither regarding the intentions to carry out the specific pro-environmental behaviours of consuming less meat and household energy.
Although there was no significant difference in these measures of intention between the high and low threat conditions, it must be noted that participants' pro-environmental intentions were all relatively high (see Tables 6, 7, 8) with means ranging from 4.02 to 4.51 measured through scales ranging from 'Certainly not willing' (1) to 'Certainly willing' (5). On the one hand, this could suggest that both high and low threat appeals were fairly effective in raising participants' willingness to perform more environment-friendly actions. Contrary to the propositions of the EPPM, this would also mean that the low threat appeal was not ineffective and did not have an opposite effect (Witte & Allen, 2000). On the other hand, since there was no control group and the experiment did not constitute a repeated measures design, the results could not be compared to participants who were not exposed to any of the stimulus materials. It could therefore also be possible that the participants had already high intentions to behave environmentally-friendly before taking part in the experiment. This would be in line with relevant control variables, which indicated that participants' meat consumption was moderately low, and that they already engaged in actions to save energy at home (for further details, see section about Measures). Also, the participants perceived
climate change as fairly important issue, suggesting that their levels of fear regarding climate change were already rather high as well. Tannenbaum and colleagues (2015) argued in their meta-analysis that fear appeals which address issues that already raise a high amount of fear will not become more persuasive by adding more fear, which indicates that there is a
"maximum effective value" of fear (Tannenbaum et al., 2015). At the same time, including more fear will not degrade the positive effect of a high fear appeal (Tannenbaum et al., 2015). In terms of this argument, it might be that overall levels of intentions were high because by solely addressing the topic of climate change, participants in both conditions experienced fear. Consequently, both types of appeals might have constituted high fear appeals.
In a somewhat similar vein, it could also be the case that the low threat appeals
achieved a similar result as high threat appeals because participants attached high relevance to the topic of climate change and were therefore particularly motivated to act upon the
information received (Ruiter et al., 2001). This corroborates the Elaboration Likelihood Model (Petty & Cacioppo, 1986), which postulates that individuals who are involved with the topic presented are more likely to actively process the message. In the field of health risk communication, it has however also been found that high personal relevance of the issue can increase participants' defensive responses when being exposed to a high threat appeal
(Liberman & Chaiken, 1992; Ruiter et al., 2001). This finding stands in contrast to the high intentions of participants in the high threat conditions.
In regard to Hypothesis 1, it can be concluded that there was no significant difference in intentions between participants exposed to the high and the low threat stimulus materials. Furthermore, it cannot be deduced whether the manipulations in both conditions were effective or whether participants' high intentions were caused by other factors. The Moderating Role of Goal Frames in Shaping the Effects of Threat Appeals
Hypothesis 2: Moderating Role of Goal Frames in the Effects of Low Threat Appeals. On the grounds that former studies on extrinsic and intrinsic goal frames
