SHAME AND GUILT IN CLIMATE CHANGE NEWS:
The effects of shame and guilt in climate change news on pro-environmental behavior intention and the role of framing, self-efficacy and response-efficacy
Noon MIA Fatah Elrahman Abdulqadir
Student ID: 11826819
Student Contact: noon.m.imad@gmail.com 2019-2020 Master’s Thesis
Graduate School of Communication Master’s Programme Communication Science
Research Masters
Supervisor: Dr. Andreas R.T. Schuck Date of Competition: 31/01/2020
ABSTRACT
As the discourse around climate change increasingly emphasizes the moral aspects of the issue, it becomes apparent that the emotions used in communication about climate change must complement these changes. The present study focuses on shame and guilt, two moral and self-conscious emotions that are often used interchangeably in both the public domain as well as scientific research. Specifically, we examined the effects of shame and guilt eliciting climate change news on pro-environmental behavior intention. Furthermore, we explore the differential effects of goal framing and the inclusion of response-efficacy-information, as well as the mediating effects of self and response efficacy. Results of a 2 (self-conscious emotions: shame versus guilt) x 2 (response-efficacy-information: not present versus present) x 3 (goal framing: neutral versus gain versus loss) experimental design (n = 807), showed that
participants exposed to guilt-eliciting articles had significantly higher pro-environmental behavior intention when compared to a control group, however, this effect did not occur for framed guilt-eliciting articles nor articles accompanied by response-efficacy-information. Moreover, a serial mediation effect was detected wherein guilt had a positive effect on self-efficacy, which in turn correlated positively with response-efficacy. Similarly, shame-eliciting articles demonstrated positive effects on pro-environmental behavior intention, however further analysis showed this effect is somewhat tenuous.
Keywords: shame, guilt, loss-framing, gain-framing, self-efficacy, response-efficacy, pro-environmental behavior intention
INTRODUCTION
We are like tenant farmers chopping down the fence around our house for fuel.
— Thomas Edison, to Henry Ford and Harvey Firestone about fossil fuel use As far back as the early 20th century, scientists have documented the devastating effects of carbon-dioxide emissions on the environment (Baum Sr., 2016), and just as long
communication practitioners have jostled to relay this danger and inspire action (Lorenzoni, Nicholson-Cole, & Whitmarsh, 2007; Pidgeon & Fischhoff, 2011; Spence & Pidgeon, 2009; J. Swim et al., 2009). Chapman, Lickel and Markowitz (2017) stressed the importance of developing authentic climate change communication strategies via a nuanced instrumentation of emotions. Indeed, Roeser (2012) stated: “[e]motions might be the missing link in effective communication about climate change” (p. 1033). The recent popularity of terms like ‘flight shame’ (Asquith, 2020) and ‘green guilt’ (Kotchen, 2009; Plevin, 1997) make apparent that the discourse around climate change nowadays revolves around the moral facet of the issue (Brönnimann, 2002; Höijer, 2010; Hulme, 2004). It seems natural, then, that emotions recognized as intrinsically morality-based would be germane in studying climate change communication. Two such emotions, the self-conscious emotions of shame and guilt, although both placed on the negative pole of valence, hold a wealth of potential in locating the very nuance for which Chapman, Lickel and Markowitz (2017) advocated.
The present paper explores this very premise. We begin our study by examining the singular effects of shame and guilt eliciting news about climate change on pro-environmental behavior intention. Likewise, we explore the role of three central concepts consistently underscored in the literature with regards to the effects of shame and guilt: framing, self-efficacy, and response-efficacy. Guided by appraisal theory, regulatory foci, approach and avoidance motivation model, and coping orientations, we identify the mechanisms by which shame and guilt exert their effects and, further, we follow the theoretical thread that binds the
two emotions to the concepts of interest. This endeavor culminates in a model that
encompasses all the stated concepts and explicates their expected interrelations (Table 1). Thus, this paper concerns itself with the following research question;
RQ: To what extent does the use of shame and guilt in news about climate change
affect pro-environmental behavior intention and what role do framing, self-efficacy and response-efficacy play in this relationship?
Figure 1. Conceptual model of the relationship between shame and guilt and pro-environmental behavior intention, and the expected effects of goal framing, self and response efficacy.
RELEVANCE OF STUDY
The contribution most substantial in this study is the departure from studying communicated emotions in their discreet or dual forms (e.g. anger, fear, positive versus negative, etc.), and rightly examining the communication of moral issues like climate change through the lens of moral emotions. The fields of psychology, sociology, anthropology, and marketing have long recognized and instrumentalized the distinction between the moral emotions of shame and guilt (Duhachek, Agrawal, & Han, 2012), yet political communication research has been reluctant to embrace their differences. Many studies in political communication confound the two by using them interchangeably, and doing so may be deprived of a wealth of knowledge. In borrowing the classification of shame and guilt from adjacent fields, we aim to situate the
two emotions firmly at the forefront of communication studies about moral issues. Moreover, our model provides insight to journalists and activists who aim to highlight the public’s inadequate response to morally challenging subjects without turning them away from addressing the issue.
THEORETICAL FRAMEWORK
DELINEATING SHAME AND GUILT
SIMILARITIES: NEGATIVE SELF-CONSCIOUS EMOTIONS
Shame and guilt are often used interchangeably in colloquial speech, and they do certainly bear many similarities. In the field of psychology, they are both under the larger umbrella of self-conscious emotions1 (Tangney, Stuewig, & Mashek, 2007; Tangney & Tracy, 2012). These are a class of emotions that require one to have developed a concept of the self as well as an ability to see the self as an object to be evaluated. Shame and guilt both entail negative affective states, usually prompted by perceived or anticipated failures or by transgressions against personal and societal standards (Tangney et al., 2007; Tignor & Colvin, 2017). Functionally and evolutionarily, they are moral emotions that promote social control and the attainment of social goals rather than the direct survival goals associated with discrete emotions (Breggin, 2015; Scheff, 1990; Tangney et al., 2007; Tangney & Tracy, 2012).
It is easy to see how shame and guilt can be mistaken for one another, but confounding them extends beyond just colloquial use; many studies on their use in
communication science do not make a clear distinction between them. A study by Rees and Bamberg (2014) regarding intention to participate in collective climate change mitigating actions had a central variable called group-based guilty conscience2, which was a collapsed measure of group-based guilt and shame. Guilty conscience was indeed significantly
1 Pride, envy and embarrassment also fall under the umbrella of self-conscious emotions.
2 Individual and collective shame and guilt differ along many axes. For a review of the literature, see Höijer
correlated with participation intention, but can we confidently say whether the results apply to guilt, shame, or both3? Similarly, Böhm’s (2003) found shame and guilt were elicited when individual contributions to environmental damage were highlighted. The measure used, however, was an index combining both shame and guilt: the self-related ethics-based
emotions index. In both studies, we see a tendency towards treating shame and guilt not only as interchangeable, but also as singular. Boudewyns, Turner and Paquin (2013) cite many studies that encounter this pitfall4 (Cotte, Coulter, & Moore, 2005; Heo & Park, 2017; Lwin & Phau, 2014; O’Keefe, 2012; Rees, Klug, & Bamberg, 2015; Turner et al., 2018).
DIFFERENCES: COGNITIVE ATTRIBUTIONAL THEORY
Notwithstanding, Neo-Freudian psychologists have empirically shown shame and guilt to be substantially distinct along cognitive, motivational, and behavioral dimensions (DaHee Han, Duhachek, & Agrawal, 2014; Keltner & Buswell, 1996; Tangney, 1998; Tangney, Miller, Flicker, & Barlow, 1996; Tracy & Robins, 2006; Zhu et al., 2019). On the cognitive dimension, H. B. Lewis (1971) posited a seminal approach for distinguishing the two emotions. She goes about this by pin-pointing the focus of the negative evaluation and the role of the self in such an evaluation. This view was dubbed the cognitive attributional theory by M. Lewis (1992) post factum and continues to be furthered and falsified by researchers across numerous fields.
SHAME
Lewis (1971) stated that “shame is about the self” (p. 87). Feeling shame constitutes attributing both moral and non-moral transgressions to the global, specific or social self, so negative evaluations are focused on “a domain important to the self‐concept” (Gausel &
3 The authors made a point to single out guilt and shame as more apt for studying environmental action
compared to anger or fear – the emotions usually associated with non-environmentally oriented collective protest (van Zomeren, Spears, & Leach, 2010).
4 Some scholars have put forth the ‘shame-guilt state’ paradigm; an affective-cognitive hybrid state paradigm
wherein shame is an affect while guilt is a condition (Dobrenova, Grabner-Kräuter, Diehl, & Terlutter, 2017; Elison, 2005).
Leach, 2011, p. 475). Shamed individuals, thus, engage in what is termed characterological self-blame; “I am a bad person” (Niedenthal, Tangney, & Gavanski, 1994; Tangney, 1991; Tangney, Miller, et al., 1996). Shame’s inward focus of perceived negative evaluations by one’s own self or by the social others manifests in feelings of rejection, inferiority, failure, self-consciousness and self-directed anger5 due to not living up to an ideal; that is to say, there is a discrepancy between the actual and the idealized self (see Franks et al., 1992 for shame-rage spiral).
Shame is said to constitute self-focused personal distress (Tangney, 1991) wherein an individual experiences a great deal of pain regarding the unalterable and irreparable self, thus leading to feelings of powerlessness and a lowered degree of self-efficacy (Baldwin, Baldwin, & Ewald, 2006; Lazarus, 1984; Skinner & Brewer, 2002; Sujan, Sujan, Bettman, &
Verhallen, 1999; Weiner, 1985). A preoccupation with escaping the condemning others and hiding the ‘defective self’ to alleviate the felt distress – rather than making reparations – promotes a prevention regulatory focus wherein individuals are more sensitive to the presence or absence of said distress (Lee & Aaker, 2004; Pounders, Lee, & Royne, 2018). In this sense, individuals experiencing shame have an emotion-focused coping orientation as they feel no control over the transgression, yet they still aim to dissipate the negative emotion (de Hooge, Zeelenberg, & Breugelmans, 2007; Sujan et al., 1999). Shamed individuals achieve their coping goals through avoidant behaviors6 associated with the behavioral inhibition system (BIS) (Drummond, Hammond, Satlof-Bedrick, Waugh, & Brownell, 2017; Gausel & Leach, 2011; Pivetti, Camodeca, & Rapino, 2016; Schmader & Lickel, 2006; Sheikh &
Janoff-Bulman, 2010). Although not widely studied, Mikulincer and Florian (1997) found that shame could be differentiated from guilt, jealousy, and disgust on the basis of its threat appraisal
5 In case of an other-condemnation appraisal, other-directed anger and other-blame arise from feelings of
rejection (Gausel & Leach, 2011).
6 Avoidant behaviors include hiding one’s face, blushing, collapsing of the body, slumping, gaze aversion and,
because shame constitutes a threat to one’s self-image (Lickel, Schmader, Curtis, Scarnier, & Ames, 2005).
Shame has been observed as a trait as well. Trait-shame, dubbed shame-proneness, encompasses all the hallmarks of state-shame and has been linked to a number of maladaptive pathologies such as low self-esteem (Velotti, Garofalo, Bottazzi, & Caretti, 2017), anxiety disorders (Muris, Meesters, & van Asseldonk, 2018), depressive symptoms (Andrews, Qian, & Valentine, 2002), and a number of eating disorders7 (Grabhorn, Stenner, Stangier, & Kaufhold, 2006; Skårderud, 2007). Shame-prone individuals also have a higher tendency to blame others for personal wrongdoings (Tangney, Wagner, Fletcher, & Gramzow, 1992) and are more likely to interpret neutral or ambiguous messages as shaming (Planalp, Hafen, & Adkins, 2000).
Boudewyns, Turner and Paquin (2013) examined the persuasive effects of shame and guilt as distinct appeals. They found that shame appeals correlated significantly with
perceived manipulative intent leading participants to reject the action advocated by the message. Similarly, Bennett (1998) studied shame and guilt appeals’ effects on behavior change and action intention. They found that, although shame-proneness did not have a significant effect on action intention, shame appeals correlated significantly and negatively with it.
Given the literature on shame’s effects on pro-environmental behavior intention, we can expect;
H1a: Exposure to shame-eliciting news articles has a negative effect on
pro-environmental behavior intention when compared to neutral news articles.
7 It must be noted that shame should not be characterized as completed maladaptive or “bad”. Taihara and Malik
(2016) found shame to be adaptive in individuals with high levels of discomfort intolerance and Leach and Cidam’s (2015) meta-analysis of 90 published papers showed that when individuals perceived their failures and/or social image to be reparable, shame was positively associated with a constructive approach. A number of studies also sought to redefine it via the different appraisals and feelings associated with shame (Cibich, Woodyatt, & Wenzel, 2016; Gausel & Leach, 2011; Gausel, Vignoles, & Leach, 2016).
GUILT
In Lewis’ (1971) paradigm, guilt is “about something” (p. 87). This is widely thought to mean feeling guilt constitutes attributing wrongdoings to transient actions or states8 (Tangney, 1991). The negative evaluations are thus focused on behavior and guilt-ridden individuals engage in behavioral self-blame; “I did a bad thing” (Janoff-Bulman, 1979; Niedenthal et al., 1994; Tangney, 1998; Tangney & Tracy, 2012). The social others do not play the role of judge here, but rather the object towards which responsibility and reparations are oriented (Baumeister, 1994; Teroni & Deonna, 2008).
Broadly speaking, guilt constitutes feelings of regret regarding transient external behaviors (Teroni & Deonna, 2008). This orientation towards the external prompts individuals to experience other-focused empathy (Fontaine, Luyten, Estas, & Corveleyn, 2005; Tangney, 1998; Treeby, Prado, Rice, & Crowe, 2016) and fosters perspective-taking and a sense of responsibility towards the wronged social others (Basil, Ridgway, & Basil, 2006; Onwezen, Antonides, & Bartels, 2013; Tracy, Robins, & Tangney, 2007). Guilt is also associated with a relatively high degree of self-efficacy wherein correcting the faulty action or state seems feasible (Basil, Ridgway, & Basil, 2008; Blum, 2008; Keltner & Buswell, 1996). High-efficacy guilt-ridden individuals thus aim to correct their transgressions and are said to have problem-focused coping goals (Basil et al., 2006; Brown, González, Zagefka, Manzi, & Ćehajić, 2008; Sujan et al., 1999). This coping strategy is thought to originate from a
promotion regulatory focus wherein individuals are more sensitive to the presence or absence of positive outcomes (Lee & Aaker, 2004; Pounders et al., 2018). This focus is correlated
8 Gausel and Brown (2012) argue that Lewis saw the self as “active” (H. B. Lewis, 1971, p. 81) in the experience
of guilt just as much as shame, and that the self is indeed evaluated negatively. As self-conscious emotions require some evaluation of the self, they stated that “a shame-ideation tended to focus on the globally defective self [while] a guilt-ideation tended to focus on the ‘something’ that a ‘failure of the self’ had caused” (p. 549) by which they meant behavior.
with the behavioral approach system (BAS) (Schmader & Lickel, 2006) which manifests in reparative behaviors and general altruism (Basil et al., 2008; Hoffman, 1994).
Guilt-proneness, or trait-like guilt, is thought to be adaptive and nonpathological within Tangney’s (1991) interpretation of the cognitive attributional theory9 (Tangney, 1998; Tangney, Burggraf, & Wagner, 1995; Tangney & Tracy, 2012). Guilt-prone individuals are also less likely to blame others for transgressions committed by themselves, however, they are more likely to interpret neutral or ambiguous messages as guilt eliciting (Planalp et al., 2000).
A number of studies have been conducted on the effects of guilt on pro-environmental behavior intention. In fact, such studies on guilt have become ubiquitous enough to warrant its own nomenclature: green guilt (Kotchen, 2009; Plevin, 1997). Bamberg and Möser (2007) conducted a meta-analysis of 57 published papers on pro-environmental behavior and
determined that individual green guilt plays a vital role in influencing pro-environmental behavior through intention. Harth, Leach and Kessler (2013) found that when in-group aligned individuals were assigned responsibility for environmental damage, guilt predicted reparative behavior intention. Ferguson and Branscombe (2010) studied the role of green guilt in the relationship between beliefs about the causes-and-effects of climate change and
willingness to engage in pro-environmental behavior. They found that guilt mediated this relationship, and that it was most effective when individuals believed in global warming as a whole and believed it to be anthropogenic, i.e., caused by human activity in particular10.
In light of the findings presented in this section, we can expect that;
9 At times, the literature showed inconsistent findings regarding the adaptiveness of guilt. The only consensus
seems to be that over-generalized guilt, contextual-maladaptive guilt, (proneness to) neurotic guilt, and guilt paired with high levels of discomfort intolerance are indeed maladaptive (Alexander, Brewin, Vearnals, Wolff, & Leff, 1999; Barr, 2004; Bryan, Morrow, Etienne, & Ray-Sannerud, 2013; Burney & Irwin, 2000; Tilghman-Osborne, Cole, & Felton, 2010).
10 Ferguson and Branscombe’s (2010) findings are in line with Böhm’s (2003) regarding the self-related
ethics-based emotions measure. It was additionally shown that highlighting the minor effects of climate change – effects that are perceived to be reparable – led to more willingness to engage in pro-environmental behavior when compared to highlighting the more disastrous effects.
H1b: Exposure to guilt-eliciting news articles has a positive effect on
pro-environmental behavior intention when compared to neutral news articles. CLIMATE CHANGE FRAMES
In order to highlight the ‘social other’ towards whom the experience of shame and guilt would be oriented, a suitable climate change frame must be employed. Drawing from Nisbet’s (2009) typology of science-related issue frames11, Myers, Nisbet, Maibach and Leiserowitz (2012) found public health framing to be the most effective in making climate change threats salient through highlighting health consequences to individuals and their communities. Public health framing also allows for establishing the social others in an unassuming manner while emphasizing the impact of personal environmental habits on their well-being12.
GOAL FRAMING: GAIN AND LOSS FRAMES
Outside of issue-specific climate change frames, messages can be framed in terms of their outcome within prospect theory (Tversky & Kahneman, 1981). For the purpose of this paper, we will examine goal framing13 (Levin, Schneider, & Gaeth, 1998; Meyerowitz & Chaiken, 1987). Messages that utilize goal framing emphasize either the beneficial outcomes of performing a recommended behavior, i.e., gain-framing, or the detrimental outcomes of not doing so, i.e., loss-framing14 (Levin et al., 1998; McCormick & Seta, 2016; Tversky & Kahneman, 1981).
Framing studies are indeed numerous however findings continue to affirm that the effectiveness of gain and loss frames is context-dependent. For instance, health
11 Nisbet’s (2009) typology includes eight frames. These are social progress, economic development and
competitiveness, morality and ethics, scientific and technical uncertainty, runaway science, public accountability and governance, middle way/alternative path, and conflict and strategy.
12 Public health framing is linked to hope (Poels & Dewitte, 2008), which also relies on the regulatory foci (De
Mello & Macinnis, 2005; Kim, Kang, & Mattila, 2012).
13 Levin, Schneider and Gaeth (1998) differentiated goal framing from risky choice framing and attribute
framing. Goal framing is suitable here as it is concerned with a message advocating a behavior rather than one that deals with a gamble or an evaluation of characteristics.
14 Gain-frames can be positive, wherein engaging in a recommended behavior produces positive outcomes, or
negative, wherein abstaining from a derided behavior safeguards against negative outcomes (Bilandzic et al., 2017).
communication studies find that loss-framing was most effective when advocating the adoption of specific health behaviors (Cho & Boster, 2008; Meyerowitz & Chaiken, 1987). Contrary to this, Spence and Pidgeon (2010) found that gain-framing was most effective in increasing positive attitudes towards pro-environmental behavior. Morton, Rabinovich, Marshall and Bretschneider (2011) similarly found that, in times of uncertainty, negative gain-framing increased pro-environmental behavior intention whereas loss-framing decreased it.
FRAMING AND THE SELF-CONSCIOUS EMOTIONS
For shame and guilt, the expected effects of framing are best understood with the help of the appraisal model, the approach and avoidance motivation model, and the regulatory foci. Within these models, gain-framing stimulates a challenge appraisal (Lazarus, 1991b), the reward-seeking, behavioral approach system (BAS), and is most persuasive when individuals have a promotion regulatory focus (Carver, Sutton, & Scheier, 2000; Elliot & Covington, 2001). Conversely, loss-framing is associated with a threat appraisal (Skinner & Brewer, 2002; Symes & Putwain, 2016), the punishment-averse behavioral inhibition system (BIS) (Updegraff, Sherman, Luyster, & Mann, 2007), and is most effective when individuals have a prevention regulatory focus (Crowe & Higgins, 1997; Higgins, 1997). Higgins (2000) called the persuasive effects of frames when employed with the appropriate regulatory foci the regulatory fit effect, wherein individuals ‘feel right’ about adopting a recommended behavior if a frame fits an individuals’ regulatory focus, i.e. if goal compatibility is achieved.
Nabi et al. (2019) asserted that an antecedent emotional state invariably dictates the effectiveness of a frame, and the frames’ outcome direction must be congruent with the emotion experienced in order for messages to be persuasive15. We can, thus, expect a goal
15 In their study, Nabi et al. (2019) advocated for a valence-based approach to determine congruency, however,
compatibility effect when pairing guilt with gain-framed messages and shame with loss-framed messages. Indeed, Baek & Yoon (2017) found individuals primed with guilt expressed a higher pro-environmental action intention when presented with a gain-framed ad than when presented with a loss-framed ad. Conversely, individuals primed with shame expressed a higher pro-environmental action intention when presented with a loss-framed ad than when presented with a gain-framed ad. Duhachek, Agrawal and Han’s (2012) proposed an
appraisal-coping fit occurred wherein gain-framing activates challenge appraisal favored by guilt’s problem-focused coping orientation while loss-framing activates threat appraisal which fits with shame’s emotion-focused coping orientation. This appraisal-coping fit activates the emotions’ corresponding coping strategies leading to less resistance to a message’s persuasive effects and to ‘feeling right’ about engaging in said message’s recommended response16.
Taking into account the parallels between the effects of appraisal-coping fit discussed above, the following two hypotheses can be made;
H2a: Exposure to shame-eliciting loss-framed news articles will have a positive effect
on pro-environmental behavior intention when compared to (1) shame-eliciting neutrally-framed news articles and (2) shame-eliciting gain-framed news articles.
H2b: Exposure to guilt-eliciting gain-framed news articles will have a positive effect
on pro-environmental behavior intention when compared to (1) guilt-eliciting neutrally-framed news articles and (2) guilt-eliciting loss-framed news articles. ENVIRONMENTAL SELF-EFFICACY
As we discussed above in the section pertaining to shame and guilt, self-efficacy is deeply intertwined with shame and guilt (Lazarus, 1984; Skinner & Brewer, 2002; Sujan et al., 1999; Tangney, Miller, et al., 1996). Self-efficacy is defined by Bandura (1994) as an individual’s
negative affect are neither anchors of a single bipolar dimension nor uniformly devoted to approach and avoidance” (p. 140).
16 Duhachek, Agrawal and Han’s (2012) also found that the congruency observed allows for greater fluency
beliefs about their personal capabilities in performing a certain action. Challenge appraisal, promotion regulatory focus, problem-focused coping goals and approach behavior motivation are all empirically and positively correlated with self-efficacy (Alloy et al., 2009; Jin, 2012; Keller, 2006; Putwain & Symes, 2014; Trouillet, Gana, Lourel, & Fort, 2009). In guilt, the other-focused empathy and the reparative approach behaviors promote a heightened
perception of self-efficacy (Lazarus, 1984; Pounders et al., 2018), whereas the self-focused personal distress associated with shame leads to relatively lowered self-efficacy (Alessandri, Caprara, Eisenberg, & Steca, 2009). Additionally, Keller (2006) found that self-efficacy is more oriented towards a promotion regulatory focus, and leads to persuasion when this orientation is satisfied. As guilt stimulates a promotion-focus, we can expect it to have a positive effect on self-efficacy.
Given these orientation effects, we can expect that (see Error! Reference source not
found. Figure 4. Conceptual model showing the relationship between framed
emotion-eliciting articles and pro-environmental behavior intention as mediated by self-efficacy (PROCESS Model 2)., Figure 6. Shame, self-efficacy and pro-environmental behavior intention path plot from SEM-based mediation path analysis. H1 represents exposure to neutral articles versus shame articles, SI represents self-efficacy, and DV represents pro-environmental behavior intention. and Figure 7. Guilt, self-efficacy and pro-pro-environmental behavior intention path plot from SEM-based mediation path analysis. H1 represents
exposure to neutral articles versus guilt articles, SI represents self-efficacy, and DV represents pro-environmental behavior intention.);
H3ai: Exposure to shame-eliciting news articles has a negative effect on
environmental self-efficacy when compared to neutral news articles.
H3aii: and this decrease in self-efficacy has a negative effect on
H3bi: Exposure to guilt-eliciting news articles has a positive effect on environmental
self-efficacy when compared to neutral news articles;
H3bii: and this increase in self-efficacy has a positive effect on
pro-environmental behavior intention. ENVIRONMENTAL RESPONSE-EFFICACY
As per Witte’s (1992) extended parallel process model (EPPM)17, response-efficacy goes hand-in-hand with self-efficacy. Response-efficacy18 is the extent to which an individual believes a recommended behavior is effective in bringing about a desired outcome, i.e. the perceived effectiveness of said action (Bandura, 1982; Maddux & Rogers, 1983). Prevention regulatory focus, emotion-focused coping goals, and avoidance behavior motivation are all consistently linked to response-efficacy19 (Dahee Han, Duhachek, & Agrawal, 2016; Keller, 2006; Ludolph & Schulz, 2015).
Lindsey (2005) found that stimulating both self and response efficacy led to
anticipated guilt and facilitated the reparative behaviors associated with it. Moreover, a causal sequence occurred wherein an increase in perceived self-efficacy led to an increase in
response-efficacy (Sharma & Morwitz, 2016). Although the literature on response-efficacy does not point to it arising from shame or guilt, they point to response-efficacy arising independently as a by-product of self-efficacy. This serial mediation effect has not been thoroughly examined in the context of emotional appeals20, so expectations regarding the effects of shame and guilt on response-efficacy can be drawn from the regulatory foci of the two emotions. Keller (2006) stated: “people with a prevention-focus will be more concerned with whether [a message’s] proposed action will be effective than with how easy it is to move
17 Based on Rogers’ (1983) protection motivation theory (PMT).
18 Response-efficacy is also called response-outcome expectancy in Social Cognitive Theory (SCT; Umeh,
2003).
19 Low self-efficacy – and not high response-efficacy – has been linked to threat appraisal.
to the desired end state” (p. 110)21. Thus, including response-efficacy-information22 in a message reconciles shame’s prevention-focus leading to a regulatory fit effect which in turn serves to bolster efficacy. The question remains, however, regarding the induction of self-efficacy from response-self-efficacy-information (H5).
Following the lead of Sharma and Morwitz (2016), we can expect that the inclusion of response-efficacy-information in shame-eliciting news articles would bolster self-efficacy, and in turn pro-environmental behavior intention, such that (see Error! Reference source
not found. Figure 5. Conceptual model showing the relationship between emotion-eliciting
articles, with and without accompanying response-efficacy-information, and
pro-environmental behavior intention as mediated by self and response-efficacy (PROCESS Model 6).);
H4ai: Exposure to shame-eliciting news articles accompanied by
response-efficacy-information has a positive effect on environmental self-efficacy when compared to shame-eliciting news articles not accompanied by response-efficacy-information;
H4aii: and this increase in self-efficacy has a positive effect on
pro-environmental behavior intention.
Furthermore, we may hypothesize about the mediating effects of emergent response-efficacy with regards to guilt-eliciting news articles and the role of self-response-efficacy in the relationship (see Error! Reference source not found. Figure 5. Conceptual model showing the relationship between emotion-eliciting articles, with and without accompanying response-efficacy-information, and pro-environmental behavior intention as mediated by self and response-efficacy (PROCESS Model 6)., Figure 8. Shame, self-efficacy, response-efficacy,
21 The rationale for prevention-focus fitting response-efficacy is that individuals with a prevention-focus seek to
reduce losses – not increase gains – through engaging in behaviors they are guaranteed would not result in unwarranted depletion of resources.
22 Response-efficacy-information is information includes factual recommendation and success statistics that give
and pro-environmental behavior intention path plot from SEM-based mediation path analysis. H1 represents exposure to neutral articles versus shame articles, SI represents self-efficacy, Rs represents response-efficacy, and DV represents pro-environmental behavior intention. and Figure 9. Guilt, self-efficacy, response-efficacy, and pro-environmental behavior intention path plot from SEM-based mediation path analysis. H1 represents exposure to neutral articles versus guilt articles, SI represents self-efficacy, Rs represents response-efficacy, and DV represents pro-environmental behavior intention.);
H4bi: Exposure to guilt-eliciting news articles has a positive effect on environmental
self-efficacy when compared to shame-eliciting news articles;
H4bii: this increase in self-efficacy has a positive effect on response-efficacy; H4biii: and the increase in response-efficacy has a positive effect on
pro-environmental behavior intention.
METHOD
SAMPLING AND PARTICIPANTS
The present study was conducted online and was hosted on Qualtrics. Participants included in this study were 18 years of age or older and residing in the US. All participants were recruited through convenience non-probability sampling from Amazon’s Mechanical Turk (MTurk)23. Participants were compensated $1 per assignment and a total of individual 817 24 accepted the assignment. Of the 817, 10 participants were dropped as they did not attempt all items. The final sample included n = 807 participants of mean age 42.49, SD = 12.99, min = 18, max =
23 MTurk has been repeatedly shown to provide quality data from demographically diverse participants
(Buhrmester, Kwang, & Gosling, 2011).
24 In order to determine the minimum sample size required for this study, a power analysis was conducted with
G*Power 3 software(Faul, Erdfelder, Lang, & Buchner, 2007). Previous literature on the framing of shame and guilt messages showed medium effect sizes, e.g. 2 =.07 for Baek & Yoon (2017), 2 =.16 for Pounders et al.
(2018) and 2 =.44 for Duhachek et al. (2012). Assuming a desired effect size of .16, an alpha of .05, and beta of
78. The sample had a fairly even distribution of gender with 48% female and 50.7% male25. The political ideology of the participants on the left-right continuum was also fairly evenly distributed, M = 5.43, SD = 3 on an 11-point scale. Within the US party system, 37.2% said they were Democrats and 19% said they were Republicans.
DESIGN AND PROCEDURE
This study employed a 2 (self-conscious emotions: shame versus guilt) x 2
(response-efficacy-information: not present versus present) x 3 (goal framing: neutral versus gain versus loss) between-subjects experimental approach embedded within an online survey. Participants were assigned to one of 13 conditions. In addition to 62 (7.7%) participants assigned to a control condition neutral in emotion-elicitation and framing, participants were assigned to one of the groups illustrated in Table 1.
Table 1
Number and percentages of participant assigned to 12 of the 13 groups in 3 by 2 by 2 Experimental Design, n = 807
STIMULI: EMOTION ELICITING NEWS ARTICLES, FRAMING AND
RESPONSE-EFFICACY-INFORMATION
The stimuli consisted of 13 fictitious hard news articles about the United Nations Climate Change Conference (COP25) and the publication of a report by the World Health
Organization (WHO) about the health effects of climate change. All articles used the issue of
25 7 participants (.9%) identified as non-binary and 4 (.5%) did not disclose their gender.
Response-Efficacy Goal Framing Shame Self-Conscious Emotions Guilt
Not Present Neutral 62 (7.7%) 63 (7.8%)
Gain 62 (7.7%) 62 (7.7%)
Loss 59 (7.3%) 64 (7.9%)
Present Neutral 63 (7.8%) 63 (7.8%)
Gain 63 (7.8%) 61 (7.6%)
climate change’s effects on public health as an overarching frame26. The articles were constructed in accordance with the guidelines from Boudewyns, Turner and Paquin (2013).
Shame-eliciting articles emphasized the readers’ ‘failure of the self’ in mitigating climate change and guilt-eliciting articles emphasized the readers’ transgressive behaviors against the environment. Consequences of these failures for children was also emphasized so as to foster other-focused empathy. The articles employing attributed affect, wherein affect in an article came from quoted actors (Stenvall, 2014), so as to maintain objectivity. The quotes were constructed as per the cognitive attributional27, and used second-person pronouns in the shame-eliciting news articles. Goal framing was established by emphasizing the benefits or costs to the readers’ health, children’s health, and to the healthcare system. Response-efficacy-information was added at the end of designated articles (APPENDIX B) as a report on a study28 that identified individual action most impactful on the environment and the collective health benefits of reducing each29. Lastly, all the articles were written with balance in mind, thus the argument from climate change deniers was also included (APPENDIX B). MANIPULATION CHECKS
The success of the shame and guilt manipulation was established using a 7-point semantic differential scale item that asked participants to indicated whether the article was focused on personal shortcomings (1 = shame) or personal responsibility (7 = guilt). Goal framing manipulation was ascertained using three items adapted from Baek & Yoon (2017). On a 7-point semantic differential scale, the items asked whether the stimulus article focused on (1) costs or benefits, (2) gains or losses, and (3) negative or positive outcomes, Cronbach’s alpha
26 All the articles were undated, had no discernible source, and were attributed to a news release from Madrid,
Spain, the city where the COP25 summit was held. All articles were titled “Health and climate change” and the word count of all the articles ranged from 350 to 480 words and the same presentation style was used
throughout. In all the articles, the same speakers were quoted in the same order, however, quotes varied depending on the emotion to be induced. The control article had no quotes.
27 The main emphasis was on the cognitive attributional theory’s ‘self-versus-behavior’ distinction. 28 The study was from Lund University titled ‘The climate mitigation gap: education and government
recommendations miss the most effective individual actions’ (Wynes & Nicholas, 2017).
= .87, M = 3.51, SD = 1.69. The last item checked the response-efficacy manipulation on a 5-point scale by asking participants about the amount of factual information the article gave. MEASURES
MEDIATOR: ENVIRONMENTAL SELF-EFFICACY
Self-efficacy was measured using a 3-item scale adapted from Abraham, Pane and Chairiyani (2015). The scale measured local and global self-efficacy, and efficacy in terms of influencing policy-makers. All items were measured at a 7-point agreement scale and a factor analysis with oblimin rotation extracted a single component, Cronbach’s alpha = .88, M = 5.06, SD = 1.41, KMO = .70, Bartlett’s test at 2 (3) = 1431.73, p = .000.
MEDIATOR: ENVIRONMENTAL RESPONSE-EFFICACY
This study employed 5 items from the Environmental Action Internal Control Index (Smith-Serasto & Fortner, 1994)30. Additionally, two items specific to the response-efficacy-information used in the stimuli were included31 (APPENDIX B). All items were measured on a 7-point agreement scale and a factor analysis with oblimin rotation extracted a single component, Cronbach’s alpha = .92, M = 5.24, SD = 1.4, KMO = .92, Bartlett’s test at 2 (21) = 3939.11, p = .000.
DEPENDENT VARIABLE: PRO-ENVIRONMENTAL BEHAVIOR INTENTION
Pro-environmental behavior intention (Pro-EBI) was measured using an 8-item scale32 adapted from Broomell, Budescu and Por (2015). The items asked individuals if they planned to engage in actions to mitigate climate change, their intention to seek information more about
30 Ahn, Fox, Dale and Avant (2015) identified these items as measuring environmental response-efficacy. 31 The included items asked participants about their belief in the effectiveness of eating a plant-based diet and
reducing air travel.
32 The scale was originally a 9-item scale used by Brody, Grover and Vedlitz (2012) and Chen (2016) based on a
scale devised by O’Connor, Bord and Fisher (1999). Broomell, Budescu and Por (2015) adjusted the scale by adding a measure regarding whether concerns about climate change may guide respondents voting behavior.
the issue and to discuss it with significant others (APPENDIX B). All items were measured on a 7-point agreement scale and a factor analysis with oblimin rotation extracted a single component, Cronbach’s alpha = .92, M = 4.98, SD = 1.44, KMO = .93, Bartlett’s test at 2
(28) = 4215.66, p = .000.
COVARIATES AND CONTROL VARIABLES
In addition to age, gender, and general 6-item measure that asked about the perception of the journalistic standard of the stimuli (APPENDIX B), the below variables were also measured. CLIMATE CHANGE ISSUE OPINION
The scale used to measure climate change issue opinion consisted of ten items, four of which were taken from a national survey of UK residents on climate change perceptions used by Spence, Poortinga and Pidgeon (2012) and Wolsko, Ariceaga and Seiden (2016). These items assessed climate change certainty (reverse coded), preparedness to act, concern, and human causation. Three items asked participants about their general attention to the issue of climate change: conviction that climate change is a problem, importance of the issue to them
personally, and strength of their opinions about the issue. The last three items asked
participants whether they thought climate change was a threat to the US, whether they thought climate change affects them personally (proximal vulnerability - 5-point scale), and whether they thought it would affect future generations (distal vulnerability - 5-point scale). All items except the latter two were measured in a 7-point Likert scale. A mean scale was not calculated as the items were not measured on the same scale.
CLIMATE CHANGE KNOWLEDGE
Participants’ knowledge about climate change was measured using a 4-item scale adapted from O’Connor, Bord and Fisher (1999). The current paper used the three most contemporary causes of the 9 original scale items (2 major causes and 1 minor). We also
included ‘The earth’s natural climate cycles’ as a fourth factor33. 3 answer options were provided, i.e. ‘not a cause at all’, ‘minor or secondary cause’, and ‘major or primary cause’. A final climate change knowledge score was acquired by summing the correct answers, Mdn = 2.00, SD = 1.90.
SHAME-PRONENESS AND GUILT-PRONENESS
Shame or guilt prone individuals are more likely to interpret vague or ambiguous messages as shame or guilt eliciting respectively (Planalp et al., 2000), thus we control for both traits. Shame and guilt proneness were measured using a modified version of the Test of Self-Conscious Affect - 3 (TOSCA-3) devised by Tangney, Dearing, Wagner and Gramzow (2000). The present study used 6 scenarios34 found by Luyten, Fontaine and Corveleyn (2002) to load substantially and independently on either shame or guilt (APPENDIX B). Each
scenario was accompanied by 2 reaction statements, one corresponding to shame-proneness, and the other to guilt-proneness. Participants indicated their likelihood of experiencing each reaction on a 7-point scale. The scores for all answers were summed and the thresholds for the level of shame and guilt proneness for each participant were determined based on their gender (Tangney et al., 2000), see Table 6 for thresholds used in this study and Table 2 below for distribution of shame and guilt proneness35.
Table 2
Distribution of high and low guilt and shame proneness, n = 807
33 Referred to as ‘solar cycles’ in climate-change-skeptic community, earth’s natural climate cycle is a
talking-point used within this community to discredit the consensus regarding carbon dioxide’s role in climate change (Idso, 1998; Lupo, 2008).
34 The original TOSCA-3 consists of 11 scenarios, each with three accompanying statements that are rated by
respondents on a 5-point likelihood scale.
35 The remaining participants were average in shame and guilt proneness, average shame-proneness =
FELT ANGER
Felt anger was measured using 3 separate items inspired by Boudewyns, Turner and Paquin’s (2013) measure of felt anger. The items measured anger directed at the self, at non-specific others, and at the author of the stimuli article36. All items were answered on a 7-point Likert scale but were not averaged as they measured different forms of anger.
GENERAL EMOTIONAL REACTION
Six general emotions that literature showed were most commonly experienced in relation to climate change-related messages were measured (Bilandzic, Kalch, & Soentgen, 2017; Höijer, 2010; Krznaric, 2010; J. K. Swim & Bloodhart, 2015). These emotions were: happiness, sadness, disgust, fear, empathy and hope, all measured on 7-point Likert scale.
HOSTILE MEDIA PERCEPTION
Hostile media perception (HMP) was measured using two items adapted from Feldman, Hart, Leiserowitz, Maibach and Roser-Renouf (2017). The items ask the participants whether they agreed that media coverage of climate change (1) was biased against their views, and (2) was distorted by powerful interest groups. Both items are measured on a 7-point Likert scale, however, since the scale comprised of only two congeneric items, reliability was calculated using the Brown coefficient (Eisinga, Grotenhuis, & Pelzer, 2013), Spearman-Brown coefficient = .82, M = 4.23, SD = 1.70.
36 Boudewyns, Turner and Paquin (2013) found that the experience of shame is often accompanied by some
anger, particularly anger directed at the self that can be later directed at others (Elison, Garofalo, & Velotti, 2014; Franks et al., 1992; Tangney, Wagner, Hill-Barlow, Marschall, & Gramzow, 1996).
Shame-Proneness Guilt-Proneness
Low 93 (11.5%) 99 (12.3%)
POLITICAL IDEOLOGY
Political ideology was measured using two scales. Firstly, the 11-point Left-Right
Self-Placement scale (ALLBUS) designed by Breyer (2015) asked participants to place themselves along a sliding scale ranging from 1 (left) to 11 (right), M = 5.43, SD = 3. Secondly, a US-specific scale asked participants about the political party with which they identified37. See Table 3 below for answer options and distribution.
Table 3
Distribution of participants by party identification, n = 807
RESULTS
A randomization check showed that there were no statistically significant differences among participants on age (2 = 714.06, p = .127), gender (2 = 28.59, p = .806), left-right political placement (2 = 105.79, p = .819), and party identification (2 = 64.84, p = .312).
Manipulation checks were successful for loss-framing (M = 3.05, SD = 1.71) and gain-framing (M = 4.30, SD = 1.60), t(492) = -8.36, p = .000, and for news articles with information (M = 2.92, SD = .953) and without accompanying response-efficacy-information (M = 2.07, SD = 1.14), t(804.50) = -11.50, p = .000. For the shame and guilt manipulation check, controlling for all levels of shame and guilt proneness gave statistically
37 This variable was contrast (dummy) coded to include those identifying as Democrat and leaning Democrat
into one group and those Republican and leaning Republican into one group.
Political party Participants
Republican 153 (19%) Leaning Republican 77 (9.5%) Democrat 300 (37.2%) Leaning Democrat 107 (13.3%) Independent 140 (17.3%) No preference 18 (2.2%) Not applicable 12 (1.5%)
significant results in the predicted direction, F(1,738) = 3.93, p = .048, MShame = 4.74, SD = 1.88, MGuilt = 4.98, SD = 1.80.
To test hypothesis H1a, which states that exposure to shame articles versus neutral articles will have a negative effect on Pro-EBI, an independent samples t-test was conducted. The results achieved statistical significance at t(122) = -2.06, p = .042, 95%CI[-1.11,-.02], d = .3738. Participants in the shame condition exhibited significantly higher Pro-EBI (M
Shame = 5.17, SD = 1.44) compared to participants in the control condition (MNeutral = 4.60, SD = 1.63). Although the results were significant, the direction of the effect runs opposite of that
predicted by the hypothesis, thus H1a is rejected. For hypothesis H1b, which states that exposure to guilt articles versus neutral articles will have a positive effect on Pro-EBI a similar independent samples t-test was conducted. Levene’s test of equality of variance proved significant, p = .001, so the Welch-Satterthwaite formula was applied (Delacre, Leys, Mora, & Lakens, 2019; Moder & Moder, 2010). The analysis showed a statistically
significant difference existed between the two groups, t(101.23) = -2.13, p = .036, 95%CI[-1.00,-.04], d = .3839 (Table 4). Participants in the guilt condition exhibited significantly higher Pro-EBI (MGuilt = 5.12, SD = 1.00) compared to participants in the control condition (MNeutral = 4.60, SD = 1.63). Thus, H1b is retained.
38 An OLS regression that includes all the above-mentioned covariates was also conducted. The model was
statistically significant, F(34,87) = 10.57, p = .000, R2 = .80, however, exposure to shame articles did not
significantly predict Pro-EBI, b = .13, t(87) = .83, p = .409. Similarly, in a t-test comparing neutral articles to all shame articles, and not just shame-eliciting neutrally-framed articles (no covariates) was conducted. The results were not statistically significant, t(430) = -1.89, p = .059, 95%CI[-.78,.01], with equal variances assumed, Levene’s F = 1.47, p = .226.
39 The results remained statistically significant when including all the above covariates in an OLS regression.
The regression model was statistically significant, F(34,89) = 9.58, p = .000, R2 = .78, and exposure to guilt
articles significantly predicted an increase in Pro-EBI, b = .35, t(89) = 2.22, p = .029. Similarly, in a t-test comparing neutral articles to all guilt-eliciting articles, and not just guilt-eliciting neutrally-framed articles (no covariates), was conducted. The results remained significant, t(435) = -2.19, p = .029, 95%CI[-.81,-.04], d = .28, with equal variances assumed, Levene’s F = 2.77, p = .097.
Table 4
Mean pro-environmental behavior intention, self-efficacy and response-efficacy scores of participants in the neutral, shame and guilt conditions, n = 807
Note. The p-values indicate mean differences when compared to the control condition. * p <.05. ** p <.01. *** p <.001.
Next, we tested hypothesis H2a, which states that exposure to shame-loss articles will have a positive effect on Pro-EBI when compared to (1) shame articles and (2) shame-gain articles. A one-way ANOVA was conducted and the test was not statistically significant40, F(2,180) = .644, p = .526, η2 = .00741 (see for mean scores). Thus, H2a is rejected. For H2b, which states that exposure to guilt-gain articles will have a positive effect on Pro-EBI when compared to (1) guilt articles and (2) guilt-loss articles, was tested using a similar one-way ANOVA. Levene’s test of equality of variance proved significant at p = .001 so Welch’s F is reported. The ANOVA showed no statistically significant difference between the three groups42 (Table 5), Welch’s F(2,119.71) = .452, p = .638, η2 = .006. Thus, H2b is also rejected43.
40 Post-hoc tests (Bonferroni, Tukey and Scheffe) further confirmed that the differences between the groups were
not statistically significant.
41 A univariate analysis that includes all the above-mentioned covariates was also conducted. The test was not
statistically significant, F(2,146) = 1.85, p = .161, partial η2 = .025, and the three shame-eliciting stimuli did not
differ in their effects on Pro-EBI. Furthermore, they did not differ in their effects on self-efficacy in a one-way ANOVA test, F(2,180) = .13, p = .876, nor in a univariate analysis with covariates included, F(2,145) = .07, p = .932.
42 A univariate analysis that includes all the above-mentioned covariates was also conducted. The test was not
statistically significant, F(2,152) = .06, p = .938, partial η2 = .001, and the three guilt-eliciting stimuli did not
differ in their effects on Pro-EBI. Furthermore, they did not differ in their effects on self-efficacy in a one-way ANOVA test, F(2,186) = .78, p = .462, nor in a univariate analysis with covariates included, F(2,151) = .18, p = .834.
43 Further analysis showed that, of all the gain and loss framed shame and guilt eliciting articles (no
response-efficacy-information included), only loss-framed guilt articles had a marginally statistically significant effect on Pro-EBI when compared to neutral articles, t(124) = -1.98, p = .050, 95%CI[-1.04,-.00], d = .35 MNeutral = 4.60,
SD = 1.63, MGuiltxLoss = 5.12, SD = 1.30.
Green behavior intention Self-efficacy
Neutral Shame Guilt Neutral Shame Guilt
Mean 4.60 5.17* 5.12* 3.96 4.31 4.62*
Std. Deviation 1.63 1.44 1.00 1.51 1.58 1.42
Note. Excluded 623 rows from the analysis that correspond to the missing values of the split-by
Table 5
Mean pro-environmental behavior intention scores of participants in the loss, shame-gain, guilt-loss, and guilt-gain conditions, n = 807. The p-value indicates difference when compared to control condition.
Note. The p-values indicate mean differences when compared to the control condition. * p <.05. ** p <.01. *** p <.001.
Error! Reference source not found.As hypotheses H3a and H3b deal with the
mediating effects of self-efficacy, the SPSS macro PROCESS - version 3.4 (Hayes, 2018) was used to conduct an OLS regression mediation analysis (Model 4, bootstrap = 5000). Hypothesis H3ai, which states that exposure to shame articles has a negative effect on self-efficacy when compared to neutral articles, is represented by the a-path of the regression model. The a-path was not statistically significant, F(1,122) = 1.58, p = .2109, R2 = .01, so exposure to shame articles has no significant effect on self-efficacy44, b = .35, t(122) = 1.26, p = .2109, (MShame = 4.31, SD = 1.59; MNeutral = 3.96, SD = 1.51). Thus, H3ai is rejected.
Although the regression equation for the b-path and the c’-path was significant, F(2,121) = 55.79, p = .0000, R2 = .48, as well as the total effects model, F(1,122) = 4.24, p = .0416, R2 = .03, exposure to shame articles was not a predictor in the mediation model45, thus H3aii must be rejected.
To test hypothesis H3bi, PROCESS’ Model 4 was again utilized. H3bi, which states that exposure to guilt articles has a positive effect on self-efficacy when compared to neutral
44 A similar regression analysis was conducted with all the above-mentioned covariates included. The a-path was
statistically significant, F(33,88) = 3.82, p = .0000, R2 = .59, however, shame articles did not significantly
predict efficacy, b = .17, t(88) = .73, p = .4656. Only general efficacy significantly predicted self-efficacy, b = .28, t(81) = 2.10, p = .0387.
45 On the b-path and the c’-path, self-efficacy was the sole significant positive predictor of Pro-EBI, b = .67,
t(122) = 10.19, p = .0000.
Green behavior intention
Shame-Gain Shame-Loss Guilt-Gain Guilt-Loss
Mean 4.87 5.04 4.91 5.11*
Std. Deviation 1.59 1.42 1.56 1.31
Note. Excluded 564 rows from the analysis that correspond to the missing values of the split-by
articles, is represented by the a-path of the mediation model. The a-path proved statistically significant, F(1,123) = 6.49, p = .0121, R2 = .05. Exposure to guilt articles versus neutral articles significantly predicted self-efficacy, b = .67, t(123) = 2.55, p = .0121 (MGuilt = 4.62, SD = 1.42; MNeutral = 3.96, SD = 1.51). Thus, H3bi is retained (Table 4). H3bii, which states that exposure to guilt message has a positive effect on Pro-EBI via guilt’s positive effects on self-efficacy, is represented by a combination of the a-path, the b-path and the total effect of the mediation model. The b-path and the c’-path, which tested the effects of self-efficacy and guilt articles on Pro-EBI respectively, had a statistically significant regression equation, F(2,122) = 37.78, p = .0000, R2 = .38, however, self-efficacy was the only significant predictor of Pro-EBI, b = .55, t(122) = 8.27, p = .0000. The total effect model achieved statistical significance46, F(1,123) = 4.57, p = .0345, R2 = .03, and the indirect effect output showed that mediation has occurred, EffectIndirect = .37, SE = .16, 95%CI[.08,.71]. Exposure to guilt articles versus neutral articles had a significantly positive effect on Pro-EBI, b = .52, t(123) = 2.14, p = .0345, and this effect was fully mediated by self-efficacy (see Table 4, Figure 2. Unstandardized regression coefficients for the relationship between exposure to guilt-eliciting articles and pro-environmental behavior intention as mediated by
environmental efficacy., and Error! Reference source not found. Figure 7. Guilt, self-efficacy and pro-environmental behavior intention path plot from SEM-based mediation path analysis. H1 represents exposure to neutral articles versus guilt articles, SI represents self-efficacy, and DV represents pro-environmental behavior intention. for SEM-based path plot). Moreover, additional moderation analysis showed that climate change certainty and the
46 The total effects model was also significant when controlling for all the relevant variables, F(34,89) = 9.58, p
degree of felt fear moderated the effects of exposure on self-efficacy47, (Error! Reference source not found. Table 8). Thus, H3bii is also retained.
Note. * p < .05 ** p < .01, *** p < .001
Figure 2. Unstandardized regression coefficients for the relationship between exposure to guilt-eliciting articles and pro-environmental behavior intention as mediated by
environmental self-efficacy.
Hypotheses H4ai and H4aii, which deals with mediation through self-efficacy, were also tested via PROCESS’ Model 4. Hypothesis H4ai, which states that exposure to shame articles accompanied by response-efficacy-information has a positive effect on self-efficacy when compared to shame-eliciting articles not accompanied by
response-efficacy-information, is represented by the a-path of the mediation model. The a-path was not statistically significant, F(1,123) = .05, p = .8274, R2 = .0004. Thus, H4ai is rejected. Although the regression equation for the b-path and the c’-path was significant, F(2,122) = 57.92, p = .0000, R2 = .48, as well as the total effects model, F(1,123) = 1.27, p = .2621, R2 = .10, exposure to guilt articles compared to shame articles was not a significant predictor in the mediation model48, thus H4aii must be rejected.
47 When controlling for all the relevant variables, a significant a-path was found, F(35,88) = 3.39, p = .0000, R2
= .57, however, only the interaction effect for climate change certainty was significant, b = -.31, t(88) = -2.80, p = .0062.
48 On the b-path and the c’-path, self-efficacy was the sole significant positive predictor of Pro-EBI, b = .62,
Note. * p < .05 ** p < .01, *** p < .001
Figure 3. Unstandardized regression coefficients for the relationship between exposure to guilt-eliciting articles (vs shame-eliciting articles) and pro-environmental behavior intention as mediated by environmental self-efficacy and response-efficacy.
Hypotheses H4b deal with serial mediation, that is, the effects of guilt articles on Pro-EBI as mediated through self-efficacy firstly, followed by response-efficacy. PROCESS’ Model 6 was used to compare the relevant groups (Figure 3. Unstandardized regression coefficients for the relationship between exposure to guilt-eliciting articles (vs shame-eliciting articles) and pro-environmental behavior intention as mediated by environmental self-efficacy and response-efficacy. and Error! Reference source not found. Figure 9. Guilt,
self-efficacy, response-self-efficacy, and pro-environmental behavior intention path plot from SEM-based mediation path analysis. H1 represents exposure to neutral articles versus guilt articles, SI represents self-efficacy, Rs represents response-efficacy, and DV represents
pro-environmental behavior intention. for SEM-based path plot). Hypotheses H4bi, which states that exposure to guilt articles versus shame articles has a positive effect on response-efficacy mediated through self-efficacy, is represented by the a1-path. The regression equation for the a1-path and the c’-path was not statistically significant, F(1,123) = 1.40, p = .2396, R2 = .01, thus H4bi is rejected 49. Hypothesis H4bii, which states that guilt articles will have a positive
49 A similar serial mediation analysis was conducted to compare guilt articles to neutral articles (Figure 9. Guilt,
self-efficacy, response-efficacy, and pro-environmental behavior intention path plot from SEM-based mediation path analysis. H1 represents exposure to neutral articles versus guilt articles, SI represents self-efficacy, Rs
effect on response-efficacy via self-efficacy, is represented by the d-path of the model (Hayes, 2013). The regression equation for the d-path and the a2-path was statistically significant, F(2,122) = 34.46, p = .0000, R2 = .36, however, only self-efficacy positively predicted
response-efficacy, b = .51, t(122) = 8.25, p = .0000. As exposure to the stimuli did not predict neither self-efficacy nor response efficacy50, despite a statistically significant b
1-path and b2 -path, F(3,121) = 116.57, p = .0000, R2 = .74, as well as a significant total effect model, F(1,123) = .06, p = .8139, R2 = .02, we must reject H4biii.
DISCUSSION
In conducting this study, we set out to examine the effects of shame and guilt in climate change news by delved into the underlying mechanisms by which the two emotions influenced pro-environmental behavior intention. We also attempted to justify, as deliberately as possible, their exploration in tandem with concepts of relevance to those latent
mechanisms.
We began by investigating the direct effects of shame and guilt on pro-environmental behavior intention. We expected that exposure to guilt-eliciting articles would lead
individuals to make amends owing to guilt’s approach behaviors and problem-focused coping orientation. Conversely, shame’s association with avoidance behaviors and emotion-focused coping orientation was hypothesized to result in a significant decrease in pro-environmental behavior intention. Results showed that neutrally-framed guilt indeed had a positive and
represents response-efficacy, and DV represents pro-environmental behavior intention.Error! Reference source
not found. Figure 10. Unstandardized regression coefficients for the relationship between exposure to
guilt-eliciting articles (vs neutral articles) and pro-environmental behavior intention as mediated by environmental self-efficacy and response-efficacy. and Figure 11 for SEM-based path plot). The a1-path of the model was
statistically significant, F(1,123) = 6.49, p = .0121, R2 = .05, and exposure to guilt significantly and positively
predicted self-efficacy, b = .67, t(123) = 2.55, p = .0121. The d-path and the b1-path were also significant,
d-path: F(1,122) = 31.44, p = .0000, R2 = .34, b
1-path: F(1,121) = 80.96, p = .0000, R2 = .67, and self-efficacy fully
mediated the effects of exposure to guilt articles on response-efficacy, b = .50, t(122) = 7.17, p = .0000. In turn, response-efficacy partially mediated the effects of response-efficacy on Pro-EBI, b = .65, t(121) = 10.18, p = .0000.
50 Both self-efficacy (b = .19, t(121) = 4.09, p = .0001) and response-efficacy (b = .68, t(121) = 12.17, p = .0000)
significant effect on behavior intention, and this effect persisted when controlling for several variables. Moreover, self and response efficacy positively mediated the relationship between guilt and behavior intention. These findings attest to the robustness of not only the manifest effects of guilt as per the cognitive attributional theory (H. B. Lewis, 1971), but also the latent assumptions of the model.
Exposure to shame-eliciting articles, against our expectations, had a positive effect on pro-environmental behavior intention. This effect, however, was not significant when
accounting for the above-mentioned covariates, and shame did not exert any significant effect on self nor response efficacy. One reason our sample may have responded positively to shame is that incidental shame and guilt are often experienced simultaneously (Boudewyns et al., 2013; Tangney & Dearing, 2002). This is especially true for individuals high in guilt-proneness, or individuals part of a guilt-based culture51 (Creighton, 2004). Among such audiences, shame and guilt may not be as antithetical as we predict, and the two emotions are most likely symbiotic in such instances.
The expected framing effects due to appraisal-coping fit (Duhachek et al., 2012) were not detected for guilt nor shame. Morton, Rabinovich, Marshall and Bretschneider (2011) suggested that the risk features inherent to the issue of climate change lend themselves to risk-framing52 more so than other typologies of framing. Although the objective risk posed by climate change is substantial, the issue remains, for most, psychologically, temporally, and spatially distant (Broomell et al., 2015; Ratter, Philipp, & Von Storch, 2012; Rickard, Yang, & Schuldt, 2016). Consequently, varying the risk perception of climate change is more likely than goal framing to yield significant results as goal framing only addresses the outcomes of an event whereas risk-framing may convey the magnitude of a transgression.
51 North-American and western countries are classified as guilt-cultures.
52 In risky choice framing, gain-framing promotes risk-averse behaviors wherein individuals are willing to make
choices that safeguard their gains. Loss-framing, on the other hand, stimulates risk-taking behaviors as individuals feel as though they have “nothing to lose”.
With regards to induced behavior intention, our initial findings imply that, given the right audience, self-conscious moral emotions may be beneficial when negative evaluations are directed both at the self (i.e. shame) and the behavior (i.e. guilt). This implication should be heeded cautiously however, as the positive effects of shame detected in this study may attenuate, or even backfire, due to any number of factors. Instead, the larger picture still advocates for sparse use of shame-eliciting news, and further, it advocates for a better understanding of the use of shame and guilt in news. When it comes to moral issues,
communication practitioners, activists, and journalists alike must be able to use the language most suitable to convey the severity of the subject without inadvertently discouraging the audiences from addressing it. Guilt is indeed the more apt emotion for this purpose than shame, evident by its effects on not just behavior intention, but also the underlying concepts that drive it.
LIMITATIONS AND RECOMMENDATIONS
As with any research endeavor, this study has a few limitations that are of note. Although this study aimed to direct shame and guilt at the personal self, the increasing polarization and politicization of the issue of climate change (Mccright & Dunlap, 2011; Unsworth & Fielding, 2014) might have brought to the participants’ mind their ideological group identification, thus participants may have experienced shame or guilt as directed towards their ingroup self rather than their personal self. Gausel and Brown (2012) found that the experience of shame and guilt differs significantly when transgressions are attributed to the ingroup self than to the personal self53. For this reason, we controlled for strength of ideological identification, climate change issue opinion, and hostile media perceptions, however, future researchers are invited to examine both collective and individual shame and guilt in a manner that allows the effects of both to be isolated.
53 Gausel and Brown (2012) found that ingroup, or group-based, shame motivated a change of the ingroup self