Climate change denial is associated with diminished sensitivity in internalizing environmental externalities

espite a strong scientific consensus about the existence of anthropogenic climate change, widespread scepticism in the general population continues to exist. Past research has largely relied on self-reported behaviours or behavioural intentions when investigating downstream ‘behavioural’ consequences of climate change denial. As a consequence, there remains a large gap in the literature about how belief in climate change interacts with the pursuit of self-interested, environmentally harmful behaviours. To fill that gap, the present research uses a novel, experimental economic paradigm that allows to attach true environmental consequences to laboratory decisions. Based on ∼56 000 pollution decisions from 2273 participants in more than 30 countries, we find that belief in climate change meaningfully affects decision-making. Our results show that climate change scepticism predicts self-interested choices and showcases that sceptics have an insensitive acceptance of emissions, reaping benefits no matter how large the climate cost are or how small the personal benefits become. Therefore, our results critically augment meta-analytic evidence arguing that downstream behavioural consequences are small to medium in their effect size. We discuss the use of experimental economic paradigms as a crucial innovation tool for psychological research addressing people’s willingness to engage in climate action.


Introduction
Anthropogenic climate change has large, irreversible, and detrimental effects on humanity as well as on global ecosystems, making environmental sustainability our most pressing social, cultural, and policy challenge. Experts have established an overwhelming consensus about the rapid need to transform nearly all aspects of our economy, including the energy sector, the transportation sector, or the global food systems, in an effort to achieve a fundamental reduction of greenhouse gases (Hough-Guldberg et al 2018). Individual consumption significantly contributes to the total emissions (Ivanova et al 2016) and, as a result, experts call for an increased focus on individuals and their decision-making (Ethics Commission for a Safe Energy Supply 2011, Acatech et al 2017). Although successful climate change mitigation will involve political regulation in order to secure global cooperation (e.g. price or quantity regulation of emissions), how people respond to these incentives requires an accurate understanding of human beliefs and behaviours that alleviate or exacerbate emission levels (Creutzig et al 2016, Kaiser et al 2020. This results in an important role of psychological research in our mitigation efforts (Clayton et al 2015, Lange et al 2018.
Although intentions and behaviours are not always and fully in line with each other, many people in principle agree that climate change is real, dangerous, and that we have a moral obligation to combat the adverse effects thereof. Yet, a significant share of the population remains doubtful about the existence and anthropogenic origin of climate change. This share remains rather large and ranges between 35% and 46% depending on the study (Anderegg et al 2010, Hall et al 2018. It is therefore not surprising that the environmental social sciences and closely related disciplines have developed a distinct research interest in the causes and consequences of belief (and disbelief) in climate change. Hornsey et al (2016) conducted a large meta-analysis synthesizing the existing body of work. The most important results address antecedents of climate change belief, finding that traditional societal fault lines, among them gender, age, race, or income, seem only loosely connected to belief in climate change. In a similar vein, meta-analytical results suggest only a modest association of variables such as education, knowledge, or experience with (extreme) weather events and climate change beliefs. Antecedents with more predictive power are typically rooted in values, ideologies, or preferences for system justification (Hornsey et al 2016).
Next to the causes of climate change scepticism, the downstream consequences have similarly been addressed in previous research, also summarized in the extensive meta-study by Hornsey et al (2016). The body of research on downstream consequences is largely motivated by the fact that climate scepticism is regarded a prime inhibitor of pro-environmental behaviour and other collective climate action, therefore threatening successful mitigation efforts (Gifford 2011). As our mitigation efforts are urgent and require large-scale consensus about our joint need to act, an exact understanding of the behavioural consequences of climate scepticism is a paramount research task.
Surprisingly, the consequences of climate change scepticism in terms of actual decision-making remain largely unknown. This results from the fact that the research efforts on the downstream consequences of climate scepticism have almost exclusively relied on self-reports of behaviour or on intentions to engage in pro-environmental behaviours 3 . This seems partly driven by a lack of consequential behavioural tasks (Lange and Dewitte 2019). To illustrate, out of 196 studies investigated in the meta-analysis (Hornsey et al 2016), merely one investigated actual behaviour and none focused on controlled behaviour under varying incentives. This opens the door for potential measurement bias (e.g. social desirability, recall inaccuracy; Lange et al 2018, Lange and Dewitte 2019) and largely ignores the well-documented phenomenon that attitudes do not fully and accurately translate into behaviour, referred to as the environmental attitude-behaviour gap (Kollmuss and Agyeman 2002).
Thus, a large open spot in the research landscape persists, namely the investigation in how far deniers of versus believers in climate change are actually willing to internalize carbon-related externalities. The present research taps into that blankspot and employs a novel behavioural task with actual environmental consequences coined the carbon emission task (CET, Berger and Wyss 2021). Because the task uses a within-participants experimental manipulation of the personal and environmental consequences, we can observe to what degree people high versus low in belief in climate change are willing to internalize the associated climate costs of reaping personal benefits. We test if people who deny climate change are largely insensitive to the environmental harm caused by their action, meaning that people who believe in climate change take into account higher carbon cost and lower personal gains and respond with a higher propensity to engage in pro-environmental behaviour. In addition, we expect belief in climate change to correlate positively with average pro-environmental behaviour.

Open science practices
All materials, data, and code to replicate the statistical analyses are available on the Open Science Framework (osf.io/h2r5e). We confirm that we report all experimental conditions (i.e. these are uniquely within participants variations). All data exclusions follow the pre-registered protocol established in Berger and Wyss (2021).
We disclose below how we determined our sample size, including a power sensitivity analysis showing that the smallest detectable effect is smaller than the reported average effect in the meta-study by Hornsey et al (2016). For the present research, we collapsed all data we possess as of October 2020 that included the CET with identical payoffs, without an experimental manipulation preceding the assessment of behaviour, and for which we have a measurement of individual differences in belief in climate change. The final sample resulted in eight single experimental sessions involving a total of n = 2273 participants. The descriptive statistics for each session can be found in the supplementary material (available online at stacks.iop.org/ERL/16/074018/ mmedia). Study instructions were identical for each session. All participants were recruited through Amazon Mechanical Turk or Prolific.ac. To account for between-session heterogeneity, the respective session was included as a fixed effect in all statistical models.

Sample size determination and participants
Our study is grounded on a high-powered sample with data from 2273 participants (42% female, M age = 35 years, ranging from 18 to 79 years), involving roughly 56 000 decisions. The sampling decision followed budgetary constraints (Lakens 2021). As the meta-study published by Hornsey et al (2016) offers a good estimate of expected effect sizes-although these are grounded in self-reported behaviours rather than actual behaviour-we can compare our minimum detectable effect to the published average effect size of the correlation of climate change beliefs and self-reported behaviour. As our sample yields a minimum detectable effect of ρ = 0.07 based on an alpha error of 5% and high power of 95%, our study is sufficiently powered to detect a substantially smaller effect than in the metastudy. The 95% CI for the effect of climate change belief on private pro-environmental behaviour is around 0.28-0.35 per visual inspection (see figure  3 in Hornsey et al 2016). For each analysis, we used the maximum possible number of participants (i.e. implying that sample sizes vary between distinctive analyses). Throughout the data reported here, participants received a flat compensation plus the varying amount depending on their decisions in the CET. Our study received ethical approval and was conducted in line with the Declaration of Helsinki. All participants gave written informed consent prior to the study.

The carbon emission task
The CET is a validated behavioural experimental paradigm to measure consequential environmental behaviours, which has shown good internal consistency and construct validity (Berger and Wyss 2021). It taps into the individual trade-off between personal short-term gains and long-term environmental goals, therefore directly pitching financial rewards against people's motive to avoid carbon emissions. In the task, people are confronted with a series of decisions about choosing a financially rewarding Option A and a financially non-rewarding, but carbon-neutral Option B (see figure 1).
Potential bonus levels and carbon consequences are fully crossed and randomly presented to decisionmakers. The level of rewards were 0.20, 0.40, 0.60, 0.80, or 1.00 USD (Amazon Mechanical Turk) or GBP (Prolific.ac), the level of carbon emissions were 0, 0.23, 1.02, 4.46, or 19.85 lbs CO 2 . Participants received a lump sum payment for participation and the opportunity of an additional bonus, depending on their decisions. One trial is randomly selected for payoff, a method referred to as the 'pay one' method in experimental economics (Charness et al 2016). Crucially, the environmental externality attached to choices of Option A is realized through the purchase and retirement of emission certificates through the European Emission Trading System (see Berger and Wyss 2021, for details). Experimental economists increasingly use this method to attach actual environmental consequences to laboratory behaviour (e.g. Tavoni et al 2011).

Self-report measures
In addition to the CET, participants completed selfreport measures in the same fixed order. As the key interest of this paper, we had participants complete a three-item scale assessing 'Belief in Climate Change' (adapted from Heath and Gifford 2006, Cronbach's alpha >.73 for each individual session, see supplementary material). Particularly, we used three items tapping into the occurrence, causes, and consequences of global climate change. The items were 'Global warming is occurring now' , 'The main cause of global warming are human activities' , and 'Global warming will bring about some serious negative consequences' . Furthermore, we assessed a self-report measure of pro-environmental attitudes through the New-Environmental-Paradigm Scale-Revised (Dunlap et al 2000, Cronbach's alpha >.76 for each individual session, see supplementary material).
Participants answered items on both of these constructs using five-point scales ranging from 'strongly disagree' to 'strongly agree' . We formed composite measures for these two scales, as the underlying constructs are unidimensional. In addition, we collected various demographic variables, among them age, gender, education, income, religiousness, and political orientation on the liberal/conservative spectrum. Finally, additional self-reported scales that were not part of the present research question were gathered in some of the sessions. The entire materials are presented in the supplementary material.

Procedure
After giving informed consent, participants first completed the 25 trials of the CET in randomized order. Each participant provided his or her answer within a time limit of 15 s. The time limit was introduced in order to assure a similar duration for each participant in the task. That said, this timing constraint did not restrict any participants. After completion of the CET, participants provided answers to the self-report scales as well as the demographic questions. Finally, participants were thanked and received the notification that their payment would be processed through the platforms (Amazon Mechanical Turk or Prolific.ac).

Diminished sensitivity of climate change deniers
The results show strong behavioural differences in line with the hypothesis. The more sceptic people are, the less they take into account the environmental externality associated with their choices, and the less sensitive they are to the prospective bonus payments. In contrast, people who believe in climate change are highly reactive to incentives and forego the opportunity to gain financially when the environmental cost are high or when the bonus is low.  Table 3 shows that this result is robust to various statistical controls, among them date and source of data collection (Model 1), gender, education, income (Model 2), as well as political ideology and pro-environmental attitudes (Model 3).

Discussion
Past meta-analytical research on the downstream behavioural consequences of (dis-)belief in climate change has concluded that 'in terms of the consequences, a salient message from the data is that climate change beliefs have only a modest impact on the extent to which people are willing to act in climate-friendly ways' (Hornsey et al 2016, p 625).   Our research re-assesses this conclusion using a behavioural game attaching actual environmental consequences to laboratory behaviour. In contrast to the findings based on self-reports, our study shows strong behavioural effects in a high-powered sample. People high in scepticism about climate change are largely unresponsive to the environmental harms attached to their behaviour, no matter how small the personal financial benefits are. People who believe in climate change react to environmental harms with lower propensity to accept financial bonuses. This finding is robust to demographic controls. Our results suggest that the effects particularly emerge under specific incentives, an effect that is often masked when using self-reported behaviours that tap into the average propensity to act pro-environmentally.
Our results further show that climate change beliefs are weakly correlated with pro-environmental behaviour when environmental costs are low or personal benefits are high. Thus, similar to research on the environmental attitude-behaviour gap (Kollmuss and Agyemann 2002), we find that belief in climate change are not always and fully reflected in pro-environmental behaviour. One reason for this gap may stem from the lack of cognitive resources, such as self-control capacity (Langenbach et al 2020, Nielsen 2017, which refers to people's ability to align their behaviour with their long-term goals. Another potential barrier inhibiting the translation of climate change beliefs into behaviour may result from pessimistic second-order beliefs. Even people who in principle belief in anthropogenic climate change may not act upon their individual belief if they have reasonable doubt that others join their efforts. Climate change mitigation, as a global public good, requires collective action and second-order beliefs may thus undermine individual willingness to act (Jachimowicz et al 2018, van der Linden 2021. Taken together, our results make a conceptual and methodological contribution to the literature on downstream, behavioural consequences of climate scepticism. On the conceptual side, our results call for a further integration of the full spectrum of the social sciences. Using behavioural economic paradigms can be leveraged to study social scientific hypotheses without using self-reports of behaviours while maintaining control over decision-making parameters. Ample evidence showcases that it is not sufficient to draw conclusion based on hypothetical answers or self-reported recalls of past behaviour as much variance remains unexplained (Kormos and Gifford 2014).
On the methodological side, research has thus far established only a weak link of climate change denial and behaviour through an excessive reliance on behavioural intentions or self-reports. Our results show that within-participant manipulation of personal benefits and environmental harms can unravel decision-making patterns that previously remained hidden, as highlighted by the fact the strongest behavioural difference occurred under strong incentives. Merely observing average pro-environmental behaviour in the CET shows similar effects as presented in the meta-study (Hornsey et al 2016).
On the practical side of psychological research on climate change mitigation, our study shows that simple behavioural games may help to move forward environmental social science, similar to the use of games in other fields studying social behaviour (Camerer 2003). However, typical of experimental tasks (Levitt and List 2007), the CET also provides first and foremost qualitative insights into the motivation underlying pro-environmental behaviour. As other games, it does not necessarily allow accurate parameter estimates of real-world trade-offs. Nevertheless, the insights derived from experimental games can substantially contribute to our understanding of the various motives governing pro-environmental behaviour.
Finally, understanding the relationship between climate change beliefs and behaviour may help us to design optimal policy responses. There is a strong consensus that climate change mitigation requires institutions that 'govern the commons' and experimental economics has delivered crucial insights into how beliefs and cooperative behaviour are causally linked and leveraged by institutions (Ockenfels et al 2020).

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https:// osf.io/c5rde/.