Personality Traits and Climate Change Denial, Concern, and Proactivity: a Systematic Review and Meta-Analysis

Abstract


Introduction
Climate change (CC) and its anthropogenic origin are held in a virtually unanimous consensus in the scienti c community (1,2).Despite the overwhelming scienti c evidence warning about this phenomenon and its tangible effects on the weather, the reaction (or even the acknowledgment) of the general population has so far been insu cient to achieve a substantial change in net greenhouse gas emissions (3).This lack of widespread action sits in surprising contrast against other past environmental catastrophes and public health crises -such as lead compound pollution due to additives in fuel (4) -which historically incited a reaction in the general population followed by a resolutive intergovernmental action.While the previous crises could be solved through the implementation of simple policies (e.g. the ban of fuel additives), climate change has no de nitive univocal solution; it is the product of multiple interconnected systems, and involves several stakeholders, each with con icting motivations.These characteristics de ne climate change as a "wicked problem" (5) exceedingly challenging for the human mind to grasp and tackle (6).
Interpersonal and sociocultural dynamics in uence societal attitudes and reactivity towards wicked problems.Environmental sensitization campaigns which might gain a positive reaction in a particular group of individuals might also obtain a contrarian reaction from another; this "boomerang effect" (7) might thwart many institutional attempts to inform and stimulate collective action towards climate change or other issues (8).On a more optimistic note, research shows encouraging evidence that reframing an issue in a way which better aligns to the characteristics of an individual (e.g. its political values) might elicit support even from sections of the population which frequently tend to be opposed to it (9).Thus, many of the strategies implemented to approach wicked problems also seek to bridge the gap between different individuals by taking into account their different beliefs, attitudes, and goals (10).
For their generalizability over radically different populations and cultures, personality traits, that is the "[...] relatively enduring patterns of thoughts, feelings, and behaviours that distinguish individuals from each other" (11), are valid psychological constructs on which to base tailored climate campaigns (12)(13)(14).Previous research on environmental issues has shown that there is a frequent link between personality traits, environmental attitudes, and pro-environmental behaviour (15)(16)(17)(18)(19).Moreover, there have been attempts to identify a speci c "Green personality" inclined towards pro-environmental behaviour (20,21).However, how personality traits in uence attitudes and behaviours in the speci c issue of climate change still remains unclear.
To ll this gap, herein we report a systematic review and meta-analysis aimed at aggregating extant literature on the involvement of personality traits in constructs and processes related to climate change; including, but not limiting to: climate change denial, concern for climate change, and intention to act towards climate change.When su cient data are gathered, a quantitative synthesis of the literature is also provided by performing a meta-analysis.
Unlike previous meta-analyses (e.g.Soutter et al., 2020), we will examine this relationship in isolation, detached from other environmental issues or attitudes.

Search strategy and selection criteria
This systematic review and meta-analysis was conducted in conformity with the PRISMA 2020 guidelines (22) and it has been pre-registered on the PROSPERO platform (ID: CRD42022364726).
In order to highlight relationships between personality traits and measures pertaining climate change, we included only articles which: 1. Provided at least one quantitative measure of a personality trait.
2. Provided at least one quantitative measure which explicitly pertains to climate change attitudes and behaviours.
3. Reported empirical research (thus excluding literature reviews, conference proceedings, and other types of papers).

Had been published in peer-reviewed journals.
The electronic databases selected for this research were Elsevier's Scopus®, and Clarivate's Web Of Science™.A number of keywords relating to personality traits and climate change were selected a priori based on current literature in the eld of personality psychology.We included keywords referring to models and frameworks of personality for which a sizable body of research exists, such as the Five Factor Model (FFM) (23), Eysenck's personality questionnaire (EPQ) (24), Cloninger's Temperament and character inventory (TCI) (25), and Grey's reinforcement sensitivity theory (RST) (26).After a limited exploratory research on the topic, three more keywords representing relevant personality traits were added to the initial list (i.e.Social Dominance Orientation (SDO), Right Wing Authoritarianism (RWA), and Consideration for Future Consequences (CFC).Keywords pertaining to climate change were selected a priori following other meta-analyses on general environmental attitudes (e.g. the Meta-Analysis by Soutter and colleagues) (18).The keyword "global warming" was added to better include literature published before the phasing out of this terminology in favour of "climate change" (27,28).Table S1.0 in Supplementary Materials displays the full list of keywords from which search queries were constructed for Scopus® and Web Of Science™ search engines.Both searches were performed on the 26th of October, 2022 and included articles published since database inception.Full search queries are available in Supplementary Material (S2).
Collected articles were screened and selected independently by two of the authors (EC & DM).After electronically removing duplicates, a rst screening was conducted by assessing titles and the abstracts.A second screening was then performed by evaluating full-text articles.The inclusion, or the exclusion, of dubious articles was determined on agreement after discussion between two of the authors (EC & DM).In order to include articles originally not available in English, we sent an email to the corresponding author asking to provide us with a full translation of the paper.
Due to the majority of the gathered studies being cross-sectional surveys (number of studies: k = 67; 88.2% of the total), we assessed study quality by employing the "Quality Assessment Checklist for Survey Studies in Psychology" (Q-SSP) (29).For non-survey studies, we employed the "Effective Public Health Practice Project Quality Assessment Tool (EPHPP) (30).All the checklists and tools were compiled by EC & SF.

Data analysis
The measurements pertaining to CC attitudes and behaviours were largely heterogeneous, due to the instruments ranging from single questionnaire items (e.g."I think climate change is a hoax"), to complex multidimensional scales; and measuring constructs which -although overlapping -are not perfectly equivalent.In order to manage this, measurements were coded in three "meta-measures": "Denial" (e.g.CC scepticism); "Concern" (e.g.CC risk perception); and "Proactivity" (e.g.CC policy support) upon which the meta analyses were calculated.A full list of the measures coded in each meta-measure is displayed in Table 1.

[INSERT TABLE 1 "META-MEASURES" HERE]
We provide a ux diagram of the meta-analytic stages in Supplementary Materials (Figure S3.0).Below, to better explain the meta-analytic process we refer to the individual steps in the diagram (steps S3.1 to S3. 19).
For each study gathered with our systematic search (S3.1), two of the authors (EC & DM) searched for correlation coe cients pertaining to relationships between quantitative personality measures and quantitative measures of CC attitudes, which were the effect size indices to compute the meta-analysis.In case a study didn't provide this information, we contacted the corresponding author and asked to provide them (S3.3) in order to include his/her study in meta-analysis (S3.5).The extracted effect sizes were then written in a table (S3.4).This process resulted in a raw effect size table for each "personality trait -CC attitude" relationships (S3.6).
Next, the raw table was subjected to operations in order to make it suitable for meta-analysis calculations: if two or more effect sizes in a table were obtained from the same sample (S3.8),only the most recently published one was retained (S3.7).If two or more effect sizes in a table were sub-dimensions of the same measure (S3.9), they were averaged (S3.10).Then, effect sizes which were thematically coded in the same metameasure but were directionally opposite (such as belief in CC in the "CC Denial" meta-measure) were sign-changed (S3.11).These operations resulted in the nal effect size tables (S3.12).
Subsequently, we performed random-effect meta-analysis model calculations (31) using the metafor package version 3.4-0 (32) for R Studio 2022.02.3 (33), obtaining Fisher's Z summary statistic (Z r ) and heterogeneity indices (, I 2 , and Q test of heterogeneity) for each effect size table (S3.13).In the case of meta-analytic combinations of personality measures and CC measures with very few studies (k < 4; S3.15), we computed its achieved power using equations reported by Valentine and colleagues (34) (S3.14).If the meta-analytic model did not reach a satisfactory statistical power (), we opted to exclude it (S3.16).Next, we generated forest plots and funnel plots (S3.17).Finally, we assessed possible publication bias by visually inspecting generated funnel plots.This assessment was aided by performing an Egger's regression test (S3.19),calculated using the function "regtest()" in metafor (35).Following the recommendations in the handbook by Higgins and colleagues (36), we performed moderator analyses by meta-regression (S3.19) only for meta-analyses which reached a satisfactory number of studies (k ≥ 10).We decided a priori to test separately the moderating effects of average age, gender, and country of origin of these samples.The rationale behind this choice was the frequent association that age and gender have with CC measures (37,38), and the extant signi cant difference in CC attitudes across different countries and cultures (39,40).Following a similar meta-analysis by Soutter and colleagues (18), we coded each sample in four larger groups corresponding to their geographic region: "Europe", "North America", "Australia and New Zealand", and "Mixed".
All data and the code script used to perform this meta-analyses can be retrieved in the Open Science Framework by consulting the following link: https://osf.io/erbfg/?view_only=004faac07ec64ca1aeaf81b829113be4.

Results
The electronic database search was performed on 26/10/2022, and yielded a total of 1805 articles, after removing duplicates.After title and abstract screening, the literature search yielded a total of 105 articles eligible for full text examination.Of these, 56 were excluded due to not conforming to the inclusion criteria.Five relevant papers were also retrieved from the reference list of other articles during the full text reading phase, and thus added to the pool, resulting in a nal group of 54 included articles.A owchart of the full screening process is displayed in Figure 1.
A synthetic table of the studies included in our review is displayed in Table 2 including: numerosity of the sample, mean age of the sample, country of origin of the sample, percentage of females in the sample, personality traits studied, CC measures studied, main ndings, and risk of bias assessment score.12 studies scored < 75% on the Q-SSP, indicating questionable quality.
An extended summary of included articles is available in Supplementary Materials (S6).From the included articles we extracted 76 datasets encompassing a total of n = 205988 individual data points, sampled from 19 different countries and locations from all continents.The single country which was most frequently sampled was the U.S.A. (k = 25; 33%), while the geographic region which was most represented in our sample of studies was Europe (k = 28; 37%).

[INSERT TABLE 2 "STUDY SYNTHESIS" HERE] [INSERT FIGURE 1 "PRISMA FLOWCHART" HERE]
By applying the criteria we set a priori, we performed a meta-analysis of 16 discrete correlations between personality constructs and CC metameasures.A summary table of our meta-analyses results is displayed in Table 3. Full forest and funnel plots of each meta-analysis are available in the Supplementary Materials (S4 and S5).
A forest plot showing summary statistics of meta-analyses is displayed in Figure 2.
Our moderation analysis through a meta-regression model of the SDO and CC Denial relationship (k = 23) did not show any signi cant effect for both age, and percent of females in the sample.Besides, we found a signi cant effect for the geographical region factor, accounting for 31% of heterogeneity between samples.Samples gathered from Europe (b = 0.36; p = 0.015*) or North America (b = 0.41; p = 0.006**) show a signi cant effect, meaning that samples obtained from these geographical regions vary considerably from those gathered from other sites.Further, our moderation analysis of the RWA and CC Denial relationship (k = 15) did not yield signi cant effects for country, and age.However, it did a signi cant effect for percent of females (b = 0.57; p = 0.015*), accounting for 29% of between-sample heterogeneity.A summary table of the heterogeneity measures and achieved power for each meta-analysis is displayed in Table 4.

[INSERT TABLE 3 "META-ANALYSIS RESULTS" HERE]
Examining values of Q, τ 2 , and I 2 , suggests substantial levels of heterogeneity across the majority of the meta-analyses we performed, with the notable exception of the meta-analyses of the correlations between Conscientiousness, Neuroticism, Extraversion, and climate change denial.Large values of heterogeneity are expected in the eld of personality and social psychology (41).In the case of our study, this large variability might be attributable to the wide breadth of different instruments through which climate change is measured.However, the small quantity of studies we gathered for most of our meta-analyses (k < 10) renders unfeasible the computation of moderator analyses which are capable of examining the impact of measure instruments, and thus we cannot ascertain this effect.
Publication bias was assessed with visual inspection of funnel plots, supported by the Egger's test.Egger's test was not feasible for metaanalyses with k = 2 (Openness to experience and CC concern; RWA and CC proactivity, AOT and CC denial).We found little evidence of consistent publication bias, although sporadic patterns of asymmetry were detected.These asymmetries are likely to be imputable to the high levels of heterogeneity of the models.We found the meta-analysis of the correlation between CFC and CC denial, and the correlation between SDO and CC concern to be signi cant to the Egger's test (see Table 4).

Discussion
The systematic review and meta-analysis of the literature on the associations between personality traits and climate change (CC) attitudes and behaviours yielded a nal selection of 54 papers, including samples from the whole world.Most of these studies indicate that some personality traits show signi cant correlations with attitudes towards CC.
From the systematic review and the meta-analyses emerged innovative conclusions, not evident from single studies.CC denial was not signi cantly correlated to personality traits which had been previously found to be correlated with general environmental attitudes (i.e.Agreeableness, Conscientiousness, and Extraversion) (18).In actuality, the higher complexity of CC makes it a wicked problem, thus different from any other environmental issue.As an example, the preservation of endangered species can be solved just by reducing a single behaviour (the hunting) carried on by a restricted group of people.On the contrary, a full understanding of the complex causes and consequences of CC does not suggest an immediate solution depending on the mere willingness to solve the problem: this higher complexity and abstraction paves the way for a stronger in uence of socio-cultural and political in uences on pre-existing individual speci cities.
The meta-analysed factors that showed a signi cant impact on CC attitudes are summarised as follows: Most of the selected studies found a correlation between CC denial and personality traits associated with conservative political identi cation (42), in particular Social Dominance Orientation (SDO; trait tendency to espouse non-egalitarian social structures based on dominance) (43); and Right Wing Authoritarianism (RWA; trait tendency to follow and maintain authoritarian hierarchies) (44).These correlations were meta-analytically con rmed in a paper by Stanley and Wilson (19).In our meta-analyses (see Table 3), these effects might be interpreted as an indirect measurement of political identi cation.Indeed, some studies suggest a central role of sociopolitical dynamics in the nexus between SDO, RWA, and CC attitudes: CC denial is stronger in individuals high in SDO who are interested in politics (45), and feeling threatened by climate mitigation measures or environmentalist groups moderates the relationship between SDO, RWA, and CC denial (46,47).Coherently with these results, our meta-analysis found that the geographic origin of the sample was a signi cant positive moderator of the correlation between SDO and CC denial, suggesting that the socio-political context of each country plays a dominant role in determining attitudes towards CC.
However, this boosting effect of socio-political context does not imply that SDO could not predict denial towards CC independently from political orientation.Indeed, many personality traits are reported to play a signi cant role in attitudes towards CC: both SDO and RWA predicted negative CC attitudes even better than political identi cation (47)(48)(49) and system justi cation (50); Neuroticism achieves a very small, albeit signi cant, negative correlation with CC denial; trait Perceived Vulnerability to Disease induced conservatives to support CC policies regardless of the endorsing party, contrary to author's prediction that being higher in this trait would make individuals less open-minded to outgroup ideas; individuals high in SDO appear to be less willing to support CC policies endorsed by the opposing political party (51,52).
To this regard, our meta-analysis con rmed the relevance of open-mindedness and cognitive exibility in polarising the debate on CC: the Openness factor of the Five Factor Model of personality (FFM) (23) and trait tendency to engage in Actively Open-Minded Thinking (AOT) (53) were negatively correlated with CC denial; trait Consideration for Future Consequences (CFC) shows a transversal involvement in all three metameasures of our meta-analysis and especially with proactivity towards CC, indicating that individuals high in this trait might possess an increased salience of the threat which CC constitutes, and thus act to mitigate it.
Finally, personality traits involved in the affective aspects of intergroup dynamics also appear to be involved: trait tendency to empathy (i.e.Empathic concern) was reported to be negatively correlated -although not signi cantly, in our meta-analysis -with CC denial (50), while Psychopathy and Machiavellianism -personality traits characterised by callousness and lack of empathy -showed a positive correlation with CC denial (54,55).
Taken together, these ndings depict a complex relationship between personality traits and attitudes toward CC, justifying the need to investigate it separately from other environmental issues.
The integration of cultural, political, social, and individual dynamics calls for a greater understanding of how to better tailor CC interventions and communication campaigns to reach each person in the most effective way.Based on the outcomes of the present systematic review and metaanalysis, CC communication campaigns should aim at: 1) defusing or de-escalating polarisation and politicisation of the debate on CC; 2) piercing intergroup barriers (e.g.employing testimonials belonging to groups tendentially opposed to CC action might increase their trust in the communication campaign); 3) considering the di culty of accepting harsh truths as those revealed with growing alarm by the scienti c community.
Our work contains a number of limitations: 1) Personality traits are a broad class of constructs with di cult operational de nition.2) The majority of studies gathered for our review employed a cross-sectional survey design, and thus causal relationships between variables cannot be established.3) Although explicitly seeking them, our literature search did not yield any articles relating to Eysenck's EPQ, Cloninger's TCI, or Gray's RST.Although articles analysing the relationship of these models with general environmental attitudes exist (e.g. the paper by Wiseman & Bogner (56)), it seems that their speci c correlates with the psychology of climate change still remain to be investigated.This might be an interesting future avenue of research to pursue, especially in relation to the relevance of these models for the eld of psychophysiology, and thus their focus on the biological architecture of personality.
To conclude, adopting a theoretical framework which takes into account these exceeding levels of social complexity -i.e., treating CC as a wicked problem -might be advisable.Future policies and communication campaigns should take into account these dynamics and their possible complex effects, while researching targeted methods to reach out to different segments of the population.Indeed, wicked problems might require wicked solutions.

Declarations
Contributors EC and DM ideated the research.AG supervised and advised the research process.EC performed the initial literature search in the online databases.EC and DM screened the articles independently for inclusion.EC and SF scored the articles using bias assessment tools.EC performed the statistical analyses and generated the graphs.EC and SF wrote the rst draft of the manuscript with input from DM. DM nalised the manuscript.

Figure 1 :
Figure 1: PRISMA owchart of all the phases of the study selection process."CC" = Climate Change.

Table 4 :
summary of heterogeneity parameters, publication bias, and achieved statistical power.