The Complex Effects of Social Influence and Efficacy Beliefs on Pro-Environmental Behaviour

1

The Complex Effects of Social Influence and Efficacy Beliefs on

Pro-Environmental Behaviour

Andrea V. Lind

University of Amsterdam

Brain and Cognitive Sciences

Literary Thesis

2

Abstract

Large scale behavioural change must be mobilised to mitigate the progress and potential

catastrophic impacts of anthropogenic environmental crises, like climate change. Efficacy beliefs and social influence have been firmly established as independent key predictors of

pro-environmental behaviour. Meanwhile, their theoretical underpinnings suggest that the two are highly interdependent. I conducted a systematic review of the empirical evidence (including 11 papers) on how the interplay between social influence and efficacy affects pro-environmental behaviour. Findings suggested that while social influence and efficacy both promote pro-environmental behaviour, they also affect each other, and in turn pro-pro-environmental behaviour, through feedback loops and nonlinear relationships, potentially obscuring the effects of

interventions. The evidence suggested that individuals’ personal efficacy beliefs are highly influential even in collective action. While larger scale action seems to arise from individual-level predictors, the existing literature was still characterised by methodological gaps between social levels of analysis. A complexity theoretical perspective was then introduced as an alternative paradigm to traditional statistical methods, through which to approach the complex interconnectedness and nonlinear dynamics identified in the review. Finally, it was exemplified how agent based modelling can supplement current efforts to understand the dynamics of these influential antecedents of pro-environmental behaviour.

3

Introduction

While climate change is fundamentally a biophysical phenomenon (manifested in e.g. temperature changes, biodiversity reduction, changes in ocean acidity and sea levels), its drastic currents rates are both rooted in and remains under the influence of human behaviour (Urry, 2015). Equally will the consequences of climate change potentially have immense and

asymmetrical consequences for humans across the world. In turn, some people may to a larger extent than others need to adapt their way of life to protect their livelihood (American

Psychological Association, 2009, p. 17). Due to the centrality of human behaviour, the role of the behavioural sciences has recently been recognised for its usefulness in researching the

antecedents of mitigation effective pro-environmental behaviour (Urry, 2015). This behaviour often constitutes a social dilemma in which individual- as well as collective level goals are at stake (Gupta & Ogden, 2009). That is, while, particularly Western, individuals may have to make immediate personal sacrifices in terms of habits and standards of living to mitigate the global changes, the consequences of such changes might be felt on a collective scale, and primarily by people far from the social context of the individual (Koletsou & Mancy, 2011).

This thesis seeks to broaden the understanding of the circumstances under which people decide to act in favour of the environment even in the absence of immediate reward. Specifically, this work is rooted in the research question: How do complex dynamics between efficacy beliefs and social influence affect pro-environmental behaviour? I will focus on these two factors as they have frequently been reported as key predictors of pro-environmental behaviour

independently of each other. It is outlined how individual pro-environmental behaviour is affected by social influences (Park & Ha, 2012) and efficacy beliefs (Al Mamun et al., 2018), that is, perceptions of being capable of performing an action. Through a systematic review, I then

4

tie together existing evidence that links social influences, efficacy beliefs and pro-environmental behaviour, to create a coherent picture of how these factors may further affect each other through complex causal patterns (Fritsche et al., 2018; Luszczynska & Schwarzer, 2005). Last, I will approach the characteristic patterns of the review findings from a complexity theoretical perspective to exemplify how these highly interrelated factors may be better understood as dynamic variables that unfold and interact over time.

Theoretical background

In Social Cognitive Theory, Bandura (1986) theorized efficacy beliefs as powerful determinants of behaviour change. Exhaustively, the term covers perceived capability of

performing actions and the effectiveness of those actions in achieving goals. The theory concerns self-efficacy, referring to one’s belief about the degree to which one can perform a certain act, and outcome expectancy (or response efficacy), referring to the perceived effectiveness of actually executing an action or handling an issue. Compelling theories have suggested that whether people cope with fear, risks, and challenges in a solution-oriented way depends on the belief of being capable of performing actions that will make a difference (e.g. Extended Parallel Processes Model; Witte, 1992). Without a sufficient sense of efficacy about an issue, fear can cause individuals to protect themselves against the issue rather than proactively confronting it (Van Zomeren, Spears & Leach, 2008). Indeed, efficacy constructs have been found predictive of various environmental behaviours, including private- and public-sphere

pro-environmental behaviour (Al Mamun et al., 2018; Doherty & Webler, 2016) like green tourism (Cheng et al., 2018), and youth online environmental activity (Allen, Wicks & Schulte, 2013).

5

However, since climate change is a collective issue, not only personal efficacy beliefs, but also collective efficacy (i.e., the perceived capability of the collective to perform an action) is of relevance, as the power to mitigate the threats of climate change depends on masses of people rather than a few individuals (e.g., Van Zomeren, Spears & Leach, 2010; Jugert et al, 2016; Fritsche et al., 2018). In fact, Van Zomeren et al (2010) argued that, contrary to the focus of many studies, collective efficacy beliefs are actually more influential than personal efficacy. This argument gains support from multiple studies showing the relative superiority of collective over personal efficacy beliefs in explaining pro-environmental behaviour, both correlationally and causally(Van Zomeren et al., 2010; Chen, 2015; Homburg & Stolberg, 2006). Meanwhile, some evidence suggests that manipulations of collective efficacy leading to pro-environmental

behaviour work by simultaneously increasing self-efficacy (Jugert et al., 2016). Consequently, efficacy beliefs at the individual as well as group level are of interest for understanding the antecedents of pro-environmental behaviour.

Social influences in pro-environmental behaviour

The inclination to partake in environmentally friendly behaviour has also been strongly linked to social influences, such as social norms and group membership. Social norms refer to the prevalence and approval of a given behaviour in a certain social setting by others, where the personal perception of these may be referred to as normative beliefs (Wang & Lin, 2017).

Injunctive normative beliefs describe the degree to which an action is thought to be perceived by others as acceptable or mandatory in the context. Descriptive normative beliefs, on the other hand, are beliefs about what others would do in a certain situation. Within the environmental domain, social normative beliefs have been found to both directly (e.g., in recycling; Fielding et al., 2016) and indirectly influence pro-environmental behaviour (Park & Ha, 2012; Jansson and

6

Dorrepaal, 2015). For example, the degree to which adolescents convey environmental

statements online is predicted by parental PEB, presumably because parental pro-environmental behaviour influences the descriptive normative beliefs of their children (Allen, Wicks & Schulte, 2013).

The degree to which someone identifies with a certain group, its members and prevailing attitudes constitutes another form of social influence (i.e., group membership). According to social identity theory (Tajfel & Turner, 1979), individuals act and understand themselves by their relations to the groups they belong to and those they do not. Sense of identity has been found to predict recycling behaviour (Mannetti, Pierro & Livi, 2004), leadership in biodiversity protection initiatives (Scopelliti et al, 2018), and commitment to environmental activism (Biddau, Armenti & Cottone, 2016). Notably, the link between social identity and public pro-environmental behaviour has been found to be fully mediated by social norms (Rees & Bamberg, 2014) Thus, while the two constructs have distinct meanings, social norms appears an important mechanism in for identity to exert its effects on behaviour. Through interpersonal observation, expectations and shared identity, social influence may thus affect pro-environmental behaviour from multiple angles.

Theoretical link between efficacy and social influences

While social influences and efficacy beliefs have each been found to predict

pro-environmental behaviour, their theoretical underpinnings indicate that the two factors are likely to interact with each other. According to Bandura (1986), an individual will derive their self-efficacy for a given behaviour through four main sources of information: 1) previous personal experiences with the behaviour, namely enactive mastery experience; 2) achievements and previous observed behaviours of others, namely vicarious experience; 3) being convinced of

7

one’s capabilities through verbal persuasion by others; and 4) the current physiological state. Of these, empirical evidence has identified previous personal experience and vicarious experience as the most influential precursors of self-efficacy (Luszczynska & Schwarzer, 2005). This gives social factors a fundamental role in the development of efficacy.

Moreover, a sense of the goals, behaviour, and capabilities of other group members would rely on inferences about the norms and identity of the group (Malle, Moses, & Baldwin, 2001). To exemplify, Koletsou & Mancy (2011) argue that different efficacy measures are relevant, depending on whether goals held by the members of a collective, and the tasks at hand, are interdependent or independent. That is, sometimes one member of a group may have different reasons for their participation than other members. The success of the actions they perform may then depend on the behaviour of other group members or not. For instance, energy reduction motivated by personal financial saving (independent goal) may rely more on self-efficacy, whereas energy reduction with the intention of mitigating climate change (interdependent goal) may rely more on collective efficacy. Inferring a sense of collective efficacy would thus depend on one’s knowledge of other group members.

Efficacy beliefs can even act as social influences in themselves on close acquaintances; efficacy beliefs of parents have been found to affect efficacy beliefs of their adolescent children, indicating that efficacy in itself may act as a social influence (Mead et al., 2012). One might reason that the efficacy of others around you influence their behaviour which, in turn, eventually will be expressed through descriptive norms. In sum, there is good reason to believe that

8

Review of the literature

By now it has been confidently established that efficacy beliefs and social influences play important roles in promoting pro-environmental behaviour. Meanwhile, the theoretical

underpinnings of the factors suggest that the two are interdependent (Fritsche et al., 2018;

Luszczynska & Schwarzer, 2005). This interdependency may obscure the causal relations we can infer about how social influences and efficacy affect pro-environmental behaviour.

Understanding the causal patterns that lead to large scale pro-environmental behaviour is imperative for developing efficient interventions.

The following section therefore comprises the methodology and synthesis of the findings of a systematic review intended to provide an understanding of the subquestion ‘in which ways do efficacy beliefs and social influence interplay in affecting pro-environmental behaviour?’. The purpose of this review is to provide an elaborate characterisation of the current evidence, and to identify potential impasses to a deeper understanding of this relationship.

Conceptualisation

Pro-environmental behaviour. This thesis is concerned with the efforts people put into slowing down climate change and preserving the environment, in other words, mitigation behaviour. Mitigation behaviour can be described as ‘human interventions to reduce

anthropogenic drivers of climate systems’ (American Psychological Association, 2009), where adaptation refers to ‘adjustments in natural or human systems in response to actual or expected climatic stimuli […]’ (IPCC, 2007). Climate change related adaptation behaviour is often related to more instantaneous and local expected outcomes and are, as such, likely to rely on different psychological and social antecedents than mitigation behaviours. Accordingly, the

pro-9

environmental behaviour investigated here was restricted to mitigation and environmental preservation behaviour.

Pro-environmental behaviour may take place in different domains and organisational levels, and the antecedents of these behaviours may vary based on how demanding they are, and the social complexity involved (Koletsou & Mancy, 2011). Private-sphere behaviours may involve personal consumer behaviour, recycling, and habitual usage behaviour (Hensen et al., 2016). Furthermore, individuals may catalyse larger scale behavioural changes by engaging in public-sphere behaviour, which often involve environmental activism, support of public policies, joining environmental groups, or contacting government officials (Doherty & Webler, 2016). Both private-sphere and public-sphere behaviour, as well as intentions hereof, were considered in the review. Behavioural intentions have been found to correlate with actual behaviour, although only moderately, thus still leaving an intention-behaviour-gap (Bamberg, Rees & Seebauer, 2015). Nonetheless, measuring behavioural intentions is highly prevalent across the literature, due to the logistic demands of accurately measuring actual behaviour, and will therefore be considered in this review.

Social influence. To take an exploratory and inclusive approach to the question, the definition of social influences was not restricted to a single specific construct, but instead used a flexible definition of the term. That is, studies were deemed relevant at this stage if they

measured constructs concerning how an individual’s behaviour or characteristics are affected by other individuals (e.g. group-identity, social norms, sense of connection, or efficacy of others). While undeniably playing a leading role in guiding behaviour (Oishi, Kesebir & Snyder, 2009), structural and institutional factors affecting individual actions (e.g., legislation, recycling

10

infrastructure, economic incentives; American Psychological Association, 2009) were omitted here to focus on informal dynamics between individuals and groups.

Based on the review findings, four categories of social influence appear to describe the social influence which has been studied in relation to the research question: 1) Normative beliefs, covering perceived descriptive and injunctive norms (Cialdini, Reno & Kallgren, 1990), or other constructs concerning what an individual believes similar others do or expect from them; 2) social or group identity (Tajfel & Turner, 1979), which often concerns the degree of

identification with a certain social group; 3) the influence of others’ efficacy, such as family members or shared efficacy in the community (Tabernero et al., 2015); and 4) affinity with future generations (Hensen et al., 2016). This refers to the felt connectedness and empathy towards future others (Wade-Benzoni, 2008). While being distinct from the previous categories, empathic connections between people rely on inherently social processes (Decety & Lamm, 2006), and are consequently considered here as a fourth category of social influences.

Efficacy beliefs. While the body of research on efficacy beliefs within PEB has grown continuously for the last few decades, the field is characterised by inconsistency in the definitions and operationalisation of efficacy constructs (for review, see Koletsou & Mancy, 2011). Table 1 seeks to organise a number of efficacy terms, commonly used in the literature, in order to clarify their use in this thesis. Note that other authors may indeed define the constructs differently than here and appropriately operationalise them according to their definitions. Therefore the current organisation is primarily intended to state how the constructs will be interpreted and used in this thesis. For the sake of standardisation, the terms used throughout the review will be determined by the study operationalisation (as noted in column 2), and therefore

11

not necessarily the term used by the authors of the study. If studies refer to constructs which are not synonymous to the below, it will be stated.

Table 1 Efficacy definition

Term Definition Closely related constructs or synonyms

Self-efficacy Personal belief in one’s own capability of performing an action

• Perceived behavioural control

• Personal efficacy Response efficacy Personal belief that the actions

one performs can have the intended impact

• Outcome expectancy • Perceived effectiveness • Perceived consumer

effectiveness Collective efficacy Personal belief in the ingroup

capability of performing an action • Community efficacy • Group efficacy • Citizen efficacy Collective response efficacy Personal belief that the collective

actions of the ingroup can have the intended impact/ be effective in reaching collective goal

(often conflated with collective efficacy)

Methods

Search strategy. A literature search was conducted on February 14th _{2019 on three} search engines, Communication and Mass Media Complete, PsycINFO, and Web of Science. Searches were executed using search terms related to ‘efficacy’ and ‘pro-environmental behaviour’, while filtering for English language only (see Appendix 1 for complete, platform-specific search stings).

Selection flow. Academic articles were selected for inclusion through a selection flow inspired by the PRISMA guidelines for systematic reviews (Moher et al., 2009; see Appendix 2

12

for flow diagram). In total, 509 articles were yielded during the search. Titles and abstracts of these were screened to exclude articles which did not concern efficacy and pro-environmental behaviour, resulting in 307 articles across the three platforms, and 239 after removing duplicates. These were screened by the criteria: a) due to the rapid changes in availability and nature of climate change information over the last two decades (Moser, 2016) papers had to be published from year 2000 onwards, b) they had to concern climate change mitigation and environmental conservation behaviour rather than adaptation behaviour, c) being empirical and quantitative, and d) include social dynamics.On this basis, 200 papers were excluded, and one paper was

identified through backward citation, resulting in 40 articles for full-text assessment. Three of these had to be omitted due to inaccessibility. Following from the research question, articles were included in the review if they reported a statistical assessment of how the interplay between social dynamics and efficacy constructs affect pro-environmental behaviour as a dependent variable. Specifically, either 1) efficacy as a moderating or mediating variable on the relationship between social dynamics and pro-environmental behaviour, or 2) social dynamics as a

moderating or mediating variable on the relationship between efficacy and pro-environmental behaviour. Studies which only tested the independent effects on either social influence or efficacy on pro-environmental behaviour were therefore excluded. Twenty-nine papers were excluded, resulting in 11 research papers included in the review (marked in the reference list with *).

Findings

Methodological Considerations. Before synthesising the findings, this section will underline three methodological considerations that characterise the reviewed articles, namely a)

13

the age of the studies, b) the operationalisation of efficacy, and c) the causality assumptions of mediation models.

Age of studies. The papers included in the review span from 2009 to 2018, but skew

towards the latter. This serves as an indication of a growing attention to the subject matter. Yet, even within these 10 years, the general opinions and feelings towards climate change have been subject to much interference. The way climate change has been portrayed in the media, both in terms of quality and quantity, has increased drastically even within this timespan (Moser, 2016). Similarly, the urgency of action is continuously growing (IPCC, 2018). A substantial body of evidence further suggests that recent biased media coverage and elite political discourse has made climate change opinions a central component of ideological divides (Jacquet, Dietrich & Jost, 2014). Consequently, people’s opinions, risk perception, coping-mechanisms and

behavioural habits may have also changed on both on a micro and macro level in these years, which could affect the comparability of the results.

Operationalisation of efficacy. As previously mentioned, research within efficacy beliefs

is characterised by inconsistency in the definitions and operationalisation of efficacy constructs (Koletsou & Mancy, 2011). This was also the case for the reviewed studies. For example, personal and collective response efficacy was sometimes operationally conflated (Wang, 2018) and even included in composite measures with awareness and obligation constructs (Koustova, 2018). Other studies conflated response efficacy and self-efficacy (Gupta & Ogden, 2009; Allen, Wicks & Schulte, 2013). Nonetheless, different efficacy types have been found to be factorially distinct (e.g., Doherty & Webler, 2016), making rigid distinctions relevant when discerning different effects of different constructs. As a consequence of inconsistent operationalisation, some studies may make explicit conclusions about different constructs than those that were

14

actually measured, which may compromise the validity of integrating the results (Roser-Renouf & Nisbet, 2008). This illustrates the importance integrating results according to terms

standardised from their operationalisation rather than definition.

Causality assumptions in mediation. The majority of the studies (9 out of 11) reported

mediating relationships, thus inferring causality between the variables. Determining the directional influences at play is at the heart of understanding whether efficacy and social influences promote or attenuate pro-environmental behaviour as well as the underlying

mechanism behind this. Still, cross-sectional survey data, which was used in all of the reported mediation results, cannot capture the necessary temporal aspect of causality between two variables. Through mediation, path analysis and structural equation modelling, researchers may test hypothesised directionality from which causality assumptions can be inferred, but never proven. (Bullock, Harlow, & Mulaik, 1994).

Patterns Characterising the Literature. In the following section, the findings of the review are presented by bringing forth four features, which characterise the reviewed evidence. Of these, two reflect how social influence, efficacy beliefs, and pro-environmental behaviour seem to affect each other via complex dynamics, namely through a) multi-directional influences, and b) context dependent effects of efficacy beliefs. Furthermore, two features suggest that current research on the topic faces methodological issues in explaining how exactly individual action relates to collective action. Specifically, these features are c) relative centrality of personal efficacies, and d) a methodological gap from the individual to society. The implications of these characteristics will be discussed as the thesis progresses.

The following two sections therefore concern how social influence, efficacy beliefs, and pro-environmental behaviour are reported to affect each other via seemingly complex dynamics.

15

Multi-directional influences. In accordance with their theoretical underpinnings, the

reviewed evidence suggests that efficacy and social influence reciprocally affect one another. That is, one does not without exception precede and cause the other. Rather, social influences can cause changes in efficacy and efficacy can cause changes in social influence. The most frequently reported relationship between the variables (reported in seven studies) was that the effect exerted by social influence on pro-environmental behaviour was mediated by efficacy beliefs (Wang & Lin, 2017; McDonald, Fielding & Louis, 2014; Rees & Bamberg, 2014; Wang, 2018; Allen, Wicks & Schulte, 2013; Doherty and Webler, 2016; Bamberg, Rees & Seebauer, 2015). That is, the relationship is explained by social influence affecting efficacy beliefs which, in turn, affects pro-environmental behaviour. For example, positive normative beliefs appear to promote private-sphere behaviour by first increasing both personal (Wang & Lin, 2017) and collective (Wang, 2018) response efficacy. Even being exposed to conflicting norms by being a member of multiple contrasting groups shows the same effect (McDonald, Fielding & Louis, 2014), indicating that even norm conflicts may have motivating rather than demotivating effects on pro-environmental behaviour.

Public-sphere behaviour seems to be promoted by similar mechanisms. That is, descriptive norms promote public-sphere behaviour through self-, response, and collective response efficacy – but not collective efficacy (Doherty & Webler, 2016). Potentially operating through social norms (Rees & Bamberg, 2014), identification with an environmental group has been also been found to increase participation intention (public-sphere), through an increase of collective efficacy (Bamberg, Rees, & Seebauer, 2017). Conversely, one study found that a direct path from social norms to public and private behaviour explained the variance better than when mediated by efficacy beliefs (Koustova, 2018). It should, however, be noted that this study used

16

a pooled measure of efficacy types (also including awareness of consequences and obligation to act), which may have affected the results. Overall, the evidence aligns with the theoretical assumption that efficacy is partially affected by vicarious experience (Bandura, 1986). It appears that perceiving that others are being successfully pro-environmental, and that they expect the same from you, will make you more confident in the utility of pro-environmental actions and thereby promote this behaviour. The current evidence suggests that this occurs across private and public-sphere behaviour.

As previously suggested, not only can social influences affect efficacy beliefs, efficacy also appears to alter perceptions of social influence. Only a single study reported investigating this causal relationship. Bamberg, Rees, & Seebauer (2017) tested a second model, which provided a significant indication that group identification mediates the path from collective efficacy to participation intention in an environmental group. In other words, perceiving the collective as being very capable of acting pro-environmentally can potentially make individuals identify more strongly with the group and as a result decide to join or support their work. This suggests that social influence might not only be a cause of efficacy changes, but also an effect of them. Furthermore, while pro-environmental behaviour was only investigated as a dependent variable in the reviewed studies, there is reason to believe that the effects can be cyclical (Lindsley, Brass & Thomas, 1995). For instance, efficacy beliefs are partially derived from behaviour in the form of previous personal and vicarious experience, the latter of which strongly related to descriptive social norms in referring to the observed behaviours of others (Bandura, 1986).

Together these findings indicate that the relations between efficacy beliefs, social influence and pro-environmental behaviour cannot be comprehensively described in simple cause-effect

17

terms. Rather, the reciprocal connections can cause feedback loops, obscuring their exact impact on each other. For instance, if an intervention effectively increases self-efficacy in a community, the effects may not be limited to how efficacy increases pro-environmental behaviour or social norms. These outcome variable changes may cause further unexpected changes in efficacy, thus sustaining the loop (Lindsley, Brass & Thomas, 1995).

Context dependent effects of efficacy beliefs. Above, it was illustrated how social

influence and efficacy can be highly interconnected, and that their relationship affects pro-environmental behaviour in a way consistent with their theoretical foundation (Bandura, 1986; Al Mamun et al., 2018; Park & Ha, 2012). That is, on the basis of causal inferences, positive environmental social influence and high efficacy generally act as promoting factors of pro-environmental behaviour. Studies assessing the moderating effects of efficacy and social influence, on the other hand, paint a more ambiguous picture of how the variables interact in affecting behaviour.

Firstly, the evidence about how the state of efficacy interacts with the link between social influence and behaviour is inconclusive. One study found that the positive relationship between descriptive norms and pro-environmental behaviour was stronger when response efficacy is low (Gupta & Odgen, 2009). Putatively, this suggests that one’s engagement in private-sphere behaviour will rely more strongly on knowing what others around them are doing if one is uncertain about the effectiveness of one’s actions. Similarly, when firmly believing that individuals can make a difference, norms perhaps seem less relevant in the decision of

performing pro-environmental behaviour. Hensen et al. (2016), on the other hand, manipulated the response efficacy and the feeling of affinity with future generations (i.e., connectedness and empathy towards future others, such as grandchildren) across three experimental studies. They

18

used the outcome measure, consumer environmental stewardship (i.e., the willingness to set aside personal needs for the sake of promoting long term environmental benefits), which was found strongly related to pro-environmental behaviour. In stark contrast to the findings by Gupta et al. (2009), all three studies found that increasing the social variable, affinity with future generations, could increase consumer environmental stewardship. Critically, this was only the case when response efficacy was also experimentally increased.In the low response efficacy condition, manipulations of affinity with future generations seemed to backfire, now causing a significant decrease in stewardship. These results appear to contrast those of Gupta &

Ogden(2009), in implying that social influence promotes pro-environmental behaviour most powerfully, when people feel that their actions make a difference, and causes people to abstain even more from such behaviour if they do not. As such, it cannot be concluded that response efficacy unconditionally amplifies or dampens social influence’s positive relation to pro-environmental behaviour. Nonetheless, the ambiguity may be explained by the difference of social influence. Where descriptive norms allow people to relate to the actions and abilities of others, affinity with future generations is a sense of closeness with other’s that does not relate to behaviour. As such, it is not surprising that the two interact with efficacy beliefs in distinct ways.

Furthermore, two studies reported that social influence moderates the relationship between efficacy and pro-environmental behaviour (Tabernero et al., 2015; Bamberg, Rees, & Seebauer, 2017). For example, Tabernero et al. (2015) found an interaction between self-efficacy and community efficacy on private-sphere recycling behaviour. Community efficacy was measured as the aggregated efficacies of the individuals of a certain area, that is, the general extent to which people in the community believe that recycling efforts can be done and be effective. As such, this efficacy construct is a feature of the community, not the individual, and consequently,

19

it is here considered a social influence. While always significantly predicting recycling, the predicting power of self-efficacy became greater when community efficacy was low.This finding indicates that people rely more on their own self-efficacy for recycling, when their social context (such as general community efficacy) is not particularly stimulating of the behaviour. As previously discussed, Gupta & Ogden (2009), showed that when response efficacy is low, the predictive power of social norms on being a green consumer (i.e., deliberately purchasing environmentally friendly products) strengthens. Jointly, these findings indicate that either variable may take on a larger importance in promoting pro-environmental behaviour, when the other variable is low (e.g., relying more on descriptive norms when one’s own efficacy is insufficient). Further, the evidence together exemplifies a potentially critical intersection, when neither social influences nor personal efficacy beliefs are strong enough to support

pro-environmental behaviour.

So, while the notion that social influence and efficacy are both independently important promoters of pro-environmental behaviour, once interacting, they do not necessarily reinforce the effect of one another (Gupta & Ogden, 2009; Tabernero et al., 2015, Bamberg, Rees, & Seebauer, 2017). A certain state of one variable can cause unexpected effects of another. In addition to feasible feedback loops,this suggest that the dynamics between efficacy, social influence, and pro-environmental behaviour may often be characterised by nonlinear relations where a change in one variable does not simply lead to a proportional change in the other.

The following two sections concern how the reviewed research faces methodological issues in explaining how exactly individual action relates to collective action. Relative centrality

of personal efficacy beliefs. The results of two studies indicated that different efficacy beliefs

20

Bamberg, Rees, & Seebauer, 2017). Specifically, they suggested a relative importance of personal efficacy beliefs in collective action. As previously mentioned, many researchers have argued the contrary, namely, that collective action against climate change is determined more by collective efficacy beliefs, because sufficient change necessitates the action of multiple people (Van Zomeren, Spears & Leach, 2010; Jugert et al, 2016; Fritsche et al., 2018). Indeed, meta-analyses have shown that collective efficacy is a powerful predictor of general collective action (Van Zomeren et al., 2008), not only concerning environmentalism.

Nevertheless, a structural equation model by Doherty and Webler (2016) advocates self-efficacy as the most central self-efficacy mechanism, through which descriptive norms influence public-sphere pro-environmental behaviour. Their model showed significant path coefficients from descriptive norms to self-efficacy, personal response efficacy, collective response efficacy, and collective efficacy, in that descending order of strength. Overall, climate related activism and citizenship action (voting, contacting government officials, signing petitions) was best explained by descriptive norms, followed by self-efficacy, response efficacy, and collective response efficacy. Interestingly, the path from collective efficacy to behaviour was not significant, and as such, does not appear to mediate the link between norms and behaviour. Another study found that collective efficacy only significantly mediated the path between social identity

(identification with local climate group) and behavioural intentions (to participate in the group), as long as participative efficacy was omitted from the model (Bamberg, Rees, & Seebauer, 2017). Participative efficacy addresses the belief that one’s own participation will meaningfully contribute to the success of collective action (Van Zomeren, Saguy, & Schellhaas., 2013).When participative efficacy was included in the model, this became the only significant mediator. Collective efficacy entails a higher confidence that the collective as a whole is able to perform

21

the necessary actions. From a cost-benefit perspective, rather than incentivising contribution, collective efficacy could therefore indeed motivate individuals to act as free-riders (Olson, 1965). Participative efficacy necessarily refers to the belief in the capabilities of the collective specifically given one’s own contribution (Bamberg, Seebauer, & Rees, 2017). As such, the terms creates a bridge between the individual and the collective, which could explain why a personal efficacy belief is found the strongest predictor of even collective action.

In further support of this, the same study found two interactions, suggesting that the degree of group identification moderates the importance of different efficacies (i.e., personal and

collective) on participation intention (in the aforementioned climate group; Bamberg, Rees & Seebauer, 2017). When identifying strongly with the group, participation intention was best explained by participative efficacy. However, when subjects only identified weakly with the initiative, self-efficacy predicted participation intentions, and the predictive power of

participative efficacy became non-significant. In spite of including collective efficacy in the model, it was not a significant predictor. Hence, it might be more salient, whether one can contribute to the effectiveness of collective action, if one feels a strong connection with a group. Conversely, a weaker connection might make one’s own capabilities and actions more relevant for action.

In sum, by testing several efficacy types against each other, the reviewed studies together suggest that the most central efficacy beliefs for engaging in public-sphere pro-environmental behaviour are those relating to personal capabilities and contributions, rather than those of the collective as a whole. Participative efficacy appears to bridge the gap between personal and collective efficacy and action. Thus, pro-environmental social influence may incentivise people to engage in pro-environmental behaviour with others, in a way that heavily relies on the

22

individual’s belief in the feasibility and utility of their own actions. That individuals must be convinced that their own action is possible and useful, in order for them to take part in collective action, suggests that bottom-up processes are important for the mobilisation of large scale action.

Methodological gaps between individual and society. This section addresses what can and

cannot be concluded from the current evidence about the relation between the variables on micro- (focus on individual level phenomena) and macro-levels of analysis (focus on phenomena in larger populations; Barbour, 2017). Based on their results discussed in the previous section, Bamberg, Rees, & Seebauer (2017) argued that collective action within the pro-environmental domain was best explained by the interplay between individual cost-benefit considerations and social identification. Consequently, they proposed a definition of such collective action as “the joint activities by a wide group of actors on the basis of mutual interest in improving community liveability and local ecological and economic resilience” (p. 161). While sufficient climate change mitigation involves the coordinated actions of a great number of people, the reviewed evidence promotes the idea that social influence and efficacy beliefs’ effect on

pro-environmental behaviour, can meaningfully be investigated in a bottom-up fashion. In other words, large scale changes in pro-environmental behaviour can be seen as arising already from the individual level, and spread through groups via social influence.

Nonetheless, the reviewed evidence also illustrated a clear methodological division between individuals and collectives, both in terms of efficacies (personal/collective) and behaviour (private-/public-sphere). Only three studies used mixed measures of private- and public-sphere pro-environmental behaviour (McDonald, Fielding & Louis, 2014; Hensen et al., 2016; Koustova, 2018), yet these measures still refer to separate levels of analysis (e.g., personal

23

behaviour and group behaviour), which were subsequently conflated. This cannot capture the continuity between the levels.

A clear division between individual and larger scale phenomena is of course not arbitrary. As Doherty & Webler (2016) point out, public-sphere action is certainly of interest because it has large potential for catalysing political action to catalyse structural changes. Such structural changes may exert effects on behaviour on a much larger scale than the effort put in by the individual. At the same time, the same authors also found personal and collective efficacy beliefs to be factorially distinct. Both points warrant the investigation of whether individual and

collective behaviour could have different antecedents, and whether these too are associated with individual or collective factors. Investigating these questions would potentially demand a clear separation of different levels of social and psychological analysis, i.e., the individual, the group, the community and so forth.

Meanwhile, this division neglects that collectives and societies are constituted of individuals (Lindsley, Brass, & Thomas, 1995), just as collective action is individual agents acting based on a mutual interest to improve common circumstances (Bamberg, Seebauer & Rees, 2017). According to Lindsley, Brass & Thomas (1995), collective behavioural change may be seen as efficacy-performance spirals, where efficacy and performance cyclically reinforce one another in a positive or negative direction. These spirals are not only rooted in the efficacy-performance feedback loops, but also the cyclical relations between social influence and

efficacy. That is, they develop across individuals through to macro-level changes in efficacy and behaviour. This perspective recognises that individuals, groups, and societies are interconnected as a whole, rather than being separate categories. Tabernero et al. (2015) tried to integrate this view in their study through a multilevel approach. That is, they inferred the connections between

24

the individual efficacy, aggregated community efficacy, and community recycling behaviour, from the interactions between them. Nonetheless, a multilevel approach like this remains cross-sectional, and still cannot offer direct insight into how these factors unfold symbiotically over time through their interrelatedness. From the current literature, it is therefore only possible to infer the connections between different levels of social organisation through analysis of cross-sectional data. It does, however, not allow for observation of how they arise from and influence each other.

Collectively, this review indicates that the motivational underpinnings of

pro-environmental behaviour, both in micro and macro-scale, span multiple levels of organisation. That is, social influence affects individuals in a top-down fashion, while personal cost-benefit considerations also play a role in individual and, by extension, collective action bottom-up (Bamberg, Rees, & Seebauer, 2017). Yet, the predominant investigations of this topic seem to be characterised by methodological caveats between the organisational levels of individuals, groups, communities, and society. Consequently, it is remains implicit and unclear how exactly social influence, efficacy and pro-environmental behaviour affect each other across different levels of analysis over time.

Sub-conclusion

The evidence at hand suggests that social influence, efficacy beliefs and pro-environmental behaviour are characterised by nonlinear interdependencies, making it a problematic subject matter for traditional reductionist research methods, that is, the explanation of a phenomena by decomposition into smaller and smaller parts (Sayer, 2010). Such methods are useful for deriving linear relationships between the variables. Meanwhile, synthesising the evidence aptly illustrates that the specific state of each variable may have intricate effects on the others. Furthermore, the

25

feedback mechanisms identified, as well as the social components of the topic, imply that the dynamics between pro-environmental behaviour, efficacy and social influences are deeply rooted in temporal and spatial relations. Context is thus essential to understand the behaviour of these variables. For methodological convenience, however, behavioural research traditionally operates primarily with cross-sectional data when investigating personal and interpersonal phenomena (Bullock, Harlow, & Mulaik, 1994), as is the case in the reviewed evidence. As such, researchers “essentially stop the clock to isolate static features of mind and action that are tethered to a small number of external causes” (Wiese, Vallacher & Strawinska, 2010, p. 1018). This creates an impasse for understanding how large-scale climate change mitigation action arises in societies over time from the dynamic interplay between heterogeneous individuals embedded in an complex social context.

This leads me to the second sub-question of the thesis: ‘is there basis for assessing the dynamics between efficacy beliefs, social influences and pro-environmental behaviour through a complexity theoretical perspective?’. Consequently, the following section is a discussion of the potential utility of complexity theory as a complementary approach to the predominant statistical methods currently used to assess the topic. Complexity theory is presented as a scientific

paradigm. Its contrasts to traditional reductionist methods are highlighted, and its relevance to the characteristics of the reviewed evidence is discussed. Finally, an outline of an agent based model is presented as a concrete example of how the subject matter can be methodologically approached from a complexity theoretical perspective.

26

A complexity theoretical perspective

Traditional statistical approaches. As a point of reference, I will briefly clarify some salient features of traditional statistical approaches. First, they are reductionist (explain phenomena by decomposition ever smaller parts), and thereby focus on isolated

objects, individual hierarchical causal links, and individual parameter estimates such as isolated effect sizes (Sayer, 2010). Second, they often assume linear relationships among variables (where a change in one variable causes predictable and proportionate changes in another variable; Luke & Stamakatis, 2012). Third, they provide static information about the variables' relationships, even in longitudinal designs, which typically also rely on cross-sectional data (Wiese, Vallacher, & Strawinska, 2010). Finally, they are mostly limited to a single level of analysis, due to the measurements they are based on (Luke & Stamakatis, 2012). Traditional statistical approaches therefore provide great potential for inferring information about the average person and linear relationships between variables. Yet, they largely do so by

neglecting the vast heterogeneity and interconnectivity between humans, and the nonlinearities that govern their interactions (Wiese, Vallacher, & Strawinska, 2010).

Pro-environmental behaviour in complex system. Within behavioural sciences, complexity theory can be seen as a scientific paradigm, through which subject matters are approached as intimately connected parts of dynamical systems (Guastello & Liebovitch, 2009). One definition of a complex system is that its “properties are not fully explained by an

understanding of its component parts” (Gallagher and Appeneller, 1999). As such, complexity theoretical approaches take a holistic (as opposed to reductionist) perspective, to capture how exactly the interacting system components over time allow unexpected properties to emerge on higher levels of analysis (Luke & Stamakatis, 2012). These features will now be concretised

27

using the findings on the complex relationship between social influence, efficacy beliefs and pro-environmental behaviour as identified in the literature review.

From this holistic view, the functioning of any part of a complex system can only be understood through their relationships to other parts and to the system as a whole

(Wiese, Vallacher, & Strawinska, 2010). The parts or agents in a system can refer to many different levels of complexity, such as neurons, molecular components, predators or

organisations (Joyce, Laurenti, and Hayasaka, 2012; An, 2015; Colon, Claessen, & Ghil, 2015, Hughes et al., 2012). In the social sciences, agents often represent people in a society (Jackson et al., 2017). From a complex systems perspective, the actions and characteristics of a

person therefore cannot be understood in isolation, but only via the context the person lives in. Families, neighbours, and other social groups are examples of such social contexts. For similar reasons, networks are often a central methodological notion within the paradigm, as they

conceptualise the connections between the parts of the system (Wiese, Vallacher, & Strawinska, 2010). Furthermore, agents of a system may be heterogeneous (Luke & Stamakatis, 2012). In human agents this means, for instance, that they may vary along psychological parameters such as efficacy beliefs or group identification. These parameters, in turn, determine how each agent or part of the system acts, and influences the characteristics of other parts of the system when they interact. The local context becomes important, because an agent will be more likely to interact with some agents rather than others, like people in a community. Because

the members in one group are different than the members in another, it is highly determining of an individual’s characteristics, (like efficacy) which group he or she is embedded in

28

dynamics of a system. When zooming out to macroscale, these local patterns of asymmetry may be observed.

To understand how a system organises itself in such patterns, it useful to visit the terms system dynamics and emergence. Temporally dependent dynamics are a fundamental aspect of the investigation of complex systems. The parts of a dynamical system evolve over time by continuously locally influencing each other in accordance with their respective properties and relationships (Guastello, Koopmans, & Pincus, 2009). Through these interactive local dynamics, the system is self-organising (Wiese, Vallacher, & Strawinska, 2010). Imagine a global transition to sustainable energy, and see this as a higher-level property of a system. The notion of self-organisation means that, rather than having this transition imposed on the system from above or by an external force, the transition arises from the internal workings of the system. More

specifically, these internal workings could refer to the environmental actions and interactions between efficacious people in local neighbourhoods, leading to snowballing effects in other areas. This example captures an essential notion in complexity theory, namely emergence. Emergence occurs when the interactions between elements at a lower organisational level give rise to higher-order phenomena, in a bottom-up fashion (Guastello, Koopmans, & Pincus, 2009). Still, these emergent properties may eventually grow to constrain the elements that gave rise to them (Wiese, Vallacher, & Strawinska, 2010), like a society imposes top-down constraints on its citizens. As such, emergent properties of a system, such as a global transition to sustainable energy, are higher order phenomena, which cannot be reduced to the properties of their

constituting elements in isolation. In other words, the macro-level change cannot necessarily be understood by extrapolating what is known about the individual agents, because it emerged from the specific interactions between them (Hughes et al., 2012).

29

Complexity theoretical approaches assume nonlinearity, meaning that a change in one variable is not proportional to a change its dependent variable (Luke & Stamakatis, 2012). For instance, this entails feedback loops and tipping points, where the relationship between variables may change rapidly due to the system state. The multi-directional relationships and interactions between efficacy, social influence and pro-environmental behaviour, identified in the literature review, exemplify such nonlinearities. If a system is in an equilibrium, it is in a stable state without change (Guastello, Koopmans, & Pincus, 2009). But due to the interconnectedness of the system parts, and because their relations are often nonlinear, a small local change in a variable may cause seemingly disproportionate macro-level changes. This can be seen as analogous to the large-scale behavioural changes that are needed for climate change mitigation. Because this perspective captures the continuity between micro- and macro-scale dynamics, it can provide insight into the multi-level dynamics leading to both stability and change in emergent

phenomena (Wiese, Vallacher, & Strawinska, 2010). Such insights are valuable for identifying leverage points that could inform interventions on how to mobilise large-scale behaviour change most efficiently.

Agent based modelling as a methodological approach

A variety of methodologies are compatible with complexity theoretical assumptions. The previous section explained the compatibility of the complex effects of social influence and efficacy beliefs on pro-environmental behaviour with the complexity paradigm. A brief example of a hypothetical agent based model (ABM) will now be used to concretise how the subject matter could be approached methodologically from this paradigm.

30

An agent-based model is a computational simulation of interdependent agents who, in their interaction with each other and their environment, constitute a larger complex dynamic system (Jackson et al., 2017). The model is designed on the micro-level, where interesting macro-level phenomena may emerge as output. For ecological validity the macro-level can therefore be specified according to an empirical foundation (Smaldino et al., 2012; Schröder & Wolf, 2016). As such, the agents can be computational representations of people in a large community, whose likelihood of interacting with certain kinds of agents can be based on for instance similarity or neighbourhood (Schröder & Wolf, 2017).For each round (or iteration), agents can interact and exchange information with another agent, and subsequently decide whether or not to engage in pro-environmental behaviour. The agents can be heterogeneous by varying in a number of variables, such as efficacy and perceived norms. These dynamic agent characteristics in turn determine the probability of being pro-environmental. Empirically based effect sizes can be used to specify the degree to which efficacy and perceived norms affect pro-environmentalism. Based on the theoretical underpinnings of descriptive norms (Wang & Lin, 2017), the perceived norms of an agent may be specified to be directly influenced by the actions of the agents it interacts with. Similarly, others’ actions may, as well as the agent’s own, directly affect their efficacy beliefs, thus representing personal and vicarious experience (Bandura, 1986). As such, the model can be specified to represent the different relationships identified between efficacy, social influence and pro-environmental behaviour. This entails feedback loops between the variables leading to nonlinear system dynamics.

The interaction and action decisions among agents constitute the micro-level dynamics, which over time may yield unexpected emergent dynamics on the macro level, such as patterns of strong or weak mitigation action (Greeven et al., 2016). Because the rules that determine the

31

micro-level dynamics (agent action and interaction) are highly specified, there is no ambiguity about the bottom-up mechanisms leading to these exact emergent patterns (Scholl, 2001). A model, which has been validated against empirical data, may therefore offer powerful insights into what happens to the macro-level dynamics if an “intervention” (e.g. modifying efficacy) is introduced on selected agents or groups.

It is important to note that real-world systems are constituted by the dynamics of a

virtually inexhaustible list of interacting factors. Meanwhile, a useful model or theory cannot and should not include every factor which potentially influences the phenomenon of interest (Box, 1976). As Jackson et al., (2017) point out, the most useful ABM is one, which explains how complex macro-level phenomena emerge from simple micro-level rules of interaction.

Fortunately, the growing body of evidence pro-environmental behaviour provide valuable insight into the most important predictors, which can then be included in complex models. This

exemplifies how traditional reductionist approaches can support complexity theoretical methods by providing the empirical foundation for micro-level specifications of the model (Smaldino et al., 2012; Schröder & Wolf, 2016). Researcher should also remember that ABM is simulation and not observation of naturally occurring dynamics, and as such offers a highly formalised insight into theoretically founded, rather than real world dynamics (Grune-Yanoff, 2009). As such, one should strive for strong empirical validation (Jackson et al., 2017).ABM can therefore be seen as complimentary to empirical studies through accounting for complex system dynamics in prediction, explanation of mechanism, or hypothesis development.

In sum, AMB is a simulation of complex systems, which offers strong potential for testing and improving theories or guiding interventions (Guastello & Liebovitch, 2009) concerning pro-environmental behaviour change by: 1) formalising and specifying the exact dynamics behind

32

the theory; 2) allowing for nonlinear dynamics and heterogeneous interacting agents; 3) illustrating through simulation precisely how macro-level dynamics emerge from micro-level, leaving no ambiguity about mechanisms unlike theories developed post-hoc from data (Scholl, 2001); 4) identifying tipping points where the system dynamics change rapidly.

Evaluation and future directions

Above it was shown how a complexity theoretical perspective through its core principles (e.g., nonlinear dynamics, interdependence, and emergence), can address the four identified characteristics of the complex effects of social influence and efficacy beliefs on pro-environmental behaviour, which challenge the limitations of traditional statistical approaches. That is, the complexity theoretical paradigm may very well offer additional insights through its core assumptions. Nonlinearity is well suited for addressing for the a) multi-directional

influences between the variables (causing potential feedback loops), and the possible tipping points associated with the b) context dependent effects of efficacy beliefs. The concept of emergence may capture how c) the relative centrality of personal efficacies can relate to bottom-up dynamics, where collective action arises partly from individual processes, which cannot be observed due to the d) methodological gap between the individual and society of traditional statistical approaches.

Thus, approaching pro-environmental behaviour from a complexity theoretical paradigm can equip researchers with the tools to better understand how it, in concert with its key

antecedents, develops over time and spreads between individuals and communities. Agent based modelling is one tool, which allows researchers to explore nonlinearity and emergence. Through simulation, it allows for formalised testing of precise theories regarding the mechanism of how local changes in efficacy, social influence, and behaviour can cause different patterns of

pro-33

environmental behaviour to emerge on a macro-scale. Furthermore, ABMs can be used in the field in the future to guide the development of new, highly specified hypotheses. Models with strong empirical traction can, through identification of leverage points, have great implications for policymakers and environmental groups designing interventions (Scholl, 2001).

Due to the focus of this thesis, much remains unexplored about the distinct relationships between different efficacy and social influence constructs. Future efforts should be directed at clarifying how exactly personal and collective efficacy respectively interact with different forms of social influence to affect pro-environmental behaviour. This includes whether personal and collective efficacy beliefs affect private- and public-sphere behaviour asymmetrically, and how social influence in involved in this process. Future research should also address the pro-environmental dynamics in social networks, where some people have a lot of social influence (e.g., experts, activists, or bloggers). Additionally, both traditional and complexity theoretical research should investigate how top-down influences (e.g., legislation, media, or interventions) govern individual behaviour through efficacy and social influence.

Conclusion

Efficient interventions, aimed at mobilising large scale behavioural changes to mitigate anthropogenic environmental crises, rely on a comprehensive understanding of the relevant social and psychological antecedents. While pro-environmental efficacy beliefs and various social influences in isolation have been found to positively predict pro-environmental behaviour, this thesis has suggested that important information may be lost by investigating the two in isolation. Efficacy beliefs and social influence appear to exert causal effects on each other. These feedback loops, along with differing effects of efficacy depending on the state of social factors, indicate nonlinear dynamics between the variables. As such, a manipulation of social influence

34

will not necessarily have a proportional effect on efficacy and in turn on pro-environmental behaviour (and vice versa). Addressing these dynamics from a complexity theoretical perspective may allow researchers to better capture how these interdependent variables develop together over time. Complexity compatible methodologies like agent based modelling could supplement

traditional statistical methods with important insights into how macro-level pro-environmental behaviour change can emerge bottom-up through individual level dynamics. This could

ultimately provide imperative information about social and psychological leverage points to be integrated in future interventions.

35

References

Al Mamun, A., Mohamad, M., Yaacob, M., & Mohiuddin, M. (2018). Intention and behavior towards green consumption among low-income households. Journal Of Environmental Management, 227, 73-86. doi: 10.1016/j.jenvman.2018.08.061

*Allen, M., Wicks, R., & Schulte, S. (2013). Online Environmental Engagement Among Youth: Influence of Parents, Attitudes and Demographics. Mass Communication And Society, 16(5), 661-686. doi: 10.1080/15205436.2013.770032

American Psychological Association [Task Force on the Interface Between Psychology & Global Climate Change]. (2009). Psychology & Global Climate Change: addressing a multifaceted phenomenon and set of challenges. Washington: American Psychology Association. Retrieved from https://www.apa.org/science/about/publications/climate-change

An, G. (2015). Introduction of a Framework for Dynamic Knowledge Representation of the Control Structure of Transplant Immunology: Employing the Power of Abstraction with a Solid Organ Transplant Agent-Based Model. Frontiers In Immunology, 6. doi:

10.3389/fimmu.2015.00561

*Bamberg, S., Rees, J., & Seebauer, S. (2015). Collective climate action: Determinants of participation intention in community-based pro-environmental initiatives. Journal Of Environmental Psychology, 43, 155-165. doi: 10.1016/j.jenvp.2015.06.006

Bandura, A. (1986) Social Foundations of Thought and Action: A Social Cognitive Theory. Englewood Cliffs, NJ: Prentice-Hall

36

Barbour, J. (2017). Micro/Meso/Macrolevels of Analysis. The International Encyclopedia Of Organizational Communication, 1-15. doi: 10.1002/9781118955567.wbieoc140

Biddau, F., Armenti, A., & Cottone, P. (2016). Socio-psychological aspects of grassroots participation in the Transition Movement: An Italian case study. Journal Of Social And Political Psychology, 4(1), 142-165. doi: 10.5964/jspp.v4i1.518

Box, G. E. P. (1976). Science and Statistics. Journal of the American Statistical Association,

71(356), 791-799. doi:10.2307/2286841

Bullock, H., Harlow, L., & Mulaik, S. (1994). Causation issues in structural equation modeling research. Structural Equation Modeling: A Multidisciplinary Journal, 1(3), 253-267. doi: 10.1080/10705519409539977

Chen, M. (2015). Self-efficacy or collective efficacy within the cognitive theory of stress model: Which more effectively explains people's self-reported proenvironmental behavior?. Journal Of Environmental Psychology, 42, 66-75. doi: 10.1016/j.jenvp.2015.02.002

Cheng, J., Chiang, A., Yuan, Y., & Huang, M. (2018). Exploring Antecedents of Green Tourism Behaviors: A Case Study in Suburban Areas of Taipei, Taiwan. Sustainability, 10(6), 1928. doi: 10.3390/su10061928

Cialdini, R., Reno, R., & Kallgren, C. (1990). A focus theory of normative conduct: Recycling the concept of norms to reduce littering in public places. Journal Of Personality And Social Psychology, 58(6), 1015-1026. doi: 10.1037//0022-3514.58.6.1015

37

Colon, C., Claessen, D., & Ghil, M. (2015). Bifurcation analysis of an agent-based model for predator–prey interactions. Ecological Modelling, 317, 93-106. doi:

10.1016/j.ecolmodel.2015.09.004

Grüne-Yanoff, T. (2008). The explanatory potential of artificial societies. Synthese, 169(3), 539-555. doi: 10.1007/s11229-008-9429-0

*Doherty, K., & Webler, T. (2016). Social norms and efficacy beliefs drive the Alarmed segment’s public-sphere climate actions. Nature Climate Change, 6(9), 879-884. doi: 10.1038/nclimate3025

Fielding, K., van Kasteren, Y., Louis, W., McKenna, B., Russell, S., & Spinks, A. (2016). Using individual householder survey responses to predict household environmental outcomes: The cases of recycling and water conservation. Resources, Conservation And Recycling, 106, 90-97. doi: 10.1016/j.resconrec.2015.11.009

Fritsche, I., Barth, M., Jugert, P., Masson, T., & Reese, G. (2018). A Social Identity Model of Pro-Environmental Action (SIMPEA). Psychological Review, 125(2), 245-269. doi: 10.1037/rev0000090

Gallagher, R., & Appenzeller, T. (1999). Beyond Reductionism. Science, 284(5411), 79-79. doi: 10.1126/science.284.5411.79

Guastello, S. J., & Liebovitch, L. S. (2009). Introduction to nonlinear dynamics and complexity. In S. J. Guastello, M. Koopmans, & D. Pincus (Eds.), Chaos and complexity in

psychology: The theory of nonlinear dynamical systems (pp. 1-40). New York, NY, US:

38

*Gupta, S., & Ogden, D. (2009). To buy or not to buy? A social dilemma perspective on green buying. Journal Of Consumer Marketing, 26(6), 376-391. doi:

10.1108/07363760910988201

Greeven, S., Kraan, O., Chappin, É., & Kwakkel, J. (2016). The Emergence of Climate Change Mitigation Action by Society: An Agent-Based Scenario Discovery Study. Journal Of Artificial Societies And Social Simulation, 19(3). doi: 10.18564/jasss.3134

Grüne-Yanoff, T. (2008). The explanatory potential of artificial societies. Synthese, 169(3), 539-555. doi: 10.1007/s11229-008-9429-0

*Hensen, N., Keeling, D., de Ruyter, K., & Wetzels, M. (2016). Me, Myself, and Future Generations: The Role of Affinity and Effectiveness in the Creation of Consumer Environmental Stewardship (CENS). Psychology & Marketing, 33(5), 389-406. doi: 10.1002/mar.20882

Homburg, A., & Stolberg, A. (2006). Explaining pro-environmental behavior with a cognitive theory of stress. Journal Of Environmental Psychology, 26(1), 1-14. doi:

10.1016/j.jenvp.2006.03.003

Hughes, H. P. N.; Clegg, C. W.; Robinson, M. A.; Crowder, R. M. (2012). "Agent-based modelling and simulation: The potential contribution to organizational psychology".

Journal of Occupational and Organizational Psychology. 85 (3): 487–502. doi:10.1111/j.2044-8325.2012.02053.x.

Intergovernmental Panel on Climate Change [IPCC] (2007). Glossary of terms used in the IPCC Fourth Assessment report: Working Group II (WG2). Retrieved from

39

Intergovernmental Panel on Climate Change [IPCC]. (2018). Summary for Policymakers (pp. 1-32). Geneva: World Meteorological Organization.

Jackson, J., Rand, D., Lewis, K., Norton, M., & Gray, K. (2017). Agent-Based Modeling: a Guide for Social Psychologists. Social Psychological And Personality Science, 8(4), 387-395. doi: 10.1177/1948550617691100

Jacquet, J., Dietrich, M., & Jost, J. (2014). The ideological divide and climate change opinion: "top-down" and "bottom-up" approaches. Frontiers In Psychology, 5. doi:

10.3389/fpsyg.2014.01458

Jansson, J., & Dorrepaal, E. (2015). Personal Norms for Dealing with Climate Change: Results from a Survey Using Moral Foundations Theory. Sustainable Development, 23(6), 381-395. doi: 10.1002/sd.1598

Joyce, K., Laurienti, P., & Hayasaka, S. (2012). Complexity in a brain-inspired agent-based model. Neural Networks, 33, 275-290. doi: 10.1016/j.neunet.2012.05.012

Jugert, P., Greenaway, K., Barth, M., Büchner, R., Eisentraut, S., & Fritsche, I. (2016). Collective efficacy increases pro-environmental intentions through increasing self-efficacy. Journal Of Environmental Psychology, 48, 12-23. doi:

10.1016/j.jenvp.2016.08.003

Koletsou, A., & Mancy, R. (2011). Which efficacy constructs for large-scale social dilemma problems? Individual and collective forms of efficacy and outcome expectancies in the context of climate change mitigation. Risk Management, 13(4), 184-208. doi: