Abstract
Individual behavior change offers a large potential to curb greenhouse gas emissions. However, messaging promoting individual behavior change has been criticized as a strategy for addressing climate change due to its potential to diminish climate policy support. In a pre-registered study with a representative sample of American adults (n = 1069), we found that messages recommending the adoption of high-impact individual climate behaviors, such as flying less, eating less meat and reducing food waste, and highlighting their large impact do not affect support for a carbon tax. In an exploratory analysis, we found that this messaging results in a higher intention to adopt several climate behaviors. We recommend that interventions that advocate for individual climate action be best understood as complements, rather than undermining substitutes, to broad carbon regulatory policy.
INTRODUCTION
The Earth is facing adverse and unprecedented risks from human-caused climate change, which are predicted to disproportionately harm the world’s most disadvantaged communities [1, 2]. Thankfully, many solutions to mitigate climate change already exist [3, 4]. Yet, after technological improvements, increasing affordability of these solutions, and strong consensus on the threat of rising emissions, we are still far from a carbon-neutral future. The Global Warming of 1.5°C (SR15) report suggests that programs that encourage people to adopt behaviors that reduce their carbon footprint present a clear complement to proposed policy solutions [5]. For example, it is estimated that the UK’s path to Net Zero will be unlikely without societal and behavioral changes [6]. The rising popularity of behavioral solutions presents policymakers and behavior change program designers with new strategies that they can use to complement traditional climate regulation [7]. These solutions could inform the design of interventions that change what we eat, how we travel, what energy we use and what we do with waste, all of which have been shown to have a high impact on carbon emissions when scaled up to the population level [8].
Individual behavior change, however, has received skepticism from academics, climate scientists and the public due to its potential to diminish climate policy support [9–11]. The critics argue that prominently discussing behavior change can make it appear as a substitute for regulatory policy change, therefore undermining support for larger systemic policy efforts.
In the context of climate emergency, it is crucial for policymakers and program designers to know which strategies in their toolkit should be deployed to combat climate change. This includes understanding whether programs that rely on messaging to promote high-impact individual behavior change risk undermining policy support. The existing literature assessing such effects is mixed and largely indirect. Recent work has argued that messages recommending individual behaviors reduce willingness to support political candidates with a pro-climate agenda [9]. Similarly, messages that primed people of their past pro-environmental behaviors (PEBs) and included recommended individual actions to reduce energy consumption were found to reduce support for climate change policy [12]. Research in the related domain of policymaker decision-making found that policy support for a carbon tax is undermined when policymakers consider less intrusive ‘nudge’-based policies to encourage individual behavior change [13]. However, studies investigating the effect of inducting individuals to adopt PEBs have shown no effect on climate policy support [14–16]. Indeed, a recent meta-analysis reveals no aggregate effect of the adoption of pro-environmental individual behavior on climate policy support [17].
To inform whether policymakers and program designers should include behavioral solutions which highlight the impact of individual action, we posed the following research question: ‘Do messages that recommend the adoption of individual behaviors to address climate change affect individuals’ support for a carbon tax?’ As a secondary investigation, we also aimed to assess whether the messages affect the intention to adopt these behaviors.
MATERIALS AND METHODS
Participants
In March 2020, we collected responses from 1085 US participants via the Prolific.co platform. The sample was stratified by age, sex and ethnicity, based on quotas designed to be representative of the US Census adults. Past research with respondents from similar platforms has found that resulting demographic and empirical findings track well with US national probability sample benchmarks [18]. Per our pre-registration, we excluded 16 respondents from analyses due to incomplete survey responses leaving a final sample of 1069 respondents. The demographic composition of respondents is available in Supplementary Tables S2 and S3. Before beginning, all participants provided consent to participate. The study data, analyses code and other Supplementary Material are available in OSF (https://osf.io/m9ah3/).
Procedure
We randomly assigned participants to either the individual behavior messaging condition or a neutral control. To ensure that the impact of individual behavior on climate was not confounded with the saliency of the consequences of climate change, all participants were first shown a paragraph describing the Intergovernmental Panel on Climate Change finding that global temperatures are likely to rise in the future [1]. In the experimental condition, we presented participants with a message describing seven high-impact individual climate-mitigating behaviors that previous research has highlighted as highly impactful to adopt—reduce the amount of meat in your diet, switch to a green energy provider, install solar panels, reduce food waste at home, make an electric vehicle (EV) your next car purchase, reduce the amount you fly by at least one flight a year and purchase carbon offsets—that Americans could adopt to reduce their carbon emissions [8]. To highlight the impact of these behaviors, we informed participants that if each behavior was adopted by 10% of Americans, the USA would be back on track to meet its global commitments to reduce carbon emissions [8]. Participants in the control condition were not shown the message about the seven behaviors.
Next, all participants stated their support for a carbon tax (‘Impose a carbon tax on companies and products based on how much emissions they create’, adapted from Ref. [13]). Their response (1 = Yes or 0 = No) formed our first dependent variable. To reduce possible demand effects, the carbon tax policy question was embedded in a matrix with a set of eight other questions focused on other government policies unrelated to climate (e.g. ‘Introduce government-mandated paid maternity leave’) [19].
For six of the seven behaviors, we asked participants ‘If you had to guess, what is the percent chance you will take each of the actions in the next 12 months?’. Due to the infrequency of purchasing a vehicle in the USA, intention to purchase an EV was measured through the following item: ‘If you had to guess, what is the percent chance you will make an EV your next car purchase?’. Participants responded to all behavioral intention measures on a 100-pt slider scale (0 = No chance at all to 100 = Certain). To create a single behavioral intention index variable representing a respondent’s overall propensity to take individual action, we standardized each behavioral intention variable (subtracted the mean, divided by the standard deviation), and then took the average of all seven values for every respondent. This individual-level index formed our second dependent variable.
For exploratory reasons, we collected several demographic measures, such as gender, age, ethnicity, education, political viewpoint and household income.
Analytic strategy
To evaluate whether our messaging affected policy support and whether the difference between conditions was equivalent, we conducted a combination of pre-registered traditional null hypothesis significance tests (NHSTs) and equivalence tests. To test for this equivalence, we relied on two one-sided tests (TOST) procedure that allows us to specify a lower (ΔL) and upper bound (ΔU) based on the smallest effect size of interest. The results falling within the bounds are deemed equivalent to the absence of a meaningful effect [20].
In the TOST procedure, the null hypothesis is the presence of a true effect of ΔL or ΔU. The alternative hypothesis is that the effect is smaller than the smallest effect size of interest and it falls within the equivalence bounds. The observed effect is compared against ΔL and ΔU in TOSTs, one testing if the effect is larger than the lower bound of the equivalence range and one testing whether the effect is smaller than the upper bound of the equivalence range. When both these one-sided tests can be statistically rejected, we can conclude that the observed effect falls within the equivalence bounds and is close enough to zero to be practically equivalent. Previous research found that messaging on the effect of a nudge policy reduced policy support for a carbon tax by 18% [13]. Based on this finding, we established the smallest effect size of interest of 10% change in carbon tax support both for traditional NHST of difference as well as the TOST of equivalence.
To test our secondary hypothesis of whether exposure to messaging affects individuals’ intention to adopt high-impact climate behaviors, we followed the same pre-registered analytical procedure of testing for both difference and equivalence, evaluating whether the differences in the mean reported intention to adopt behaviors were statistically equivalent. Based on literature standards for a small effect, we set d = 0.2 as our smallest effect size of interest [21]. Therefore, the equivalence range goes from d = −0.2 to d = 0.2.
RESULTS
Following the analytic procedure of testing for both difference and equivalence, we find no evidence that messaging on the importance of individual action reduces support for a carbon tax. Specifically, those in the individual behavior condition supported a carbon tax 0.2% less than those in the control condition, which is statistically equivalent (z = 4.169, P < 0.001) and not statistically different (z = −0.0938, P = 0.925) from zero.
To test our secondary hypothesis of whether messaging affects individuals’ intention to adopt high-impact climate behaviors, we followed the same analytic procedure of testing for both difference and equivalence. We find the differences in the mean intention to adopt behaviors (d = 0.127), while small, are statistically distinguishable from zero (t(1060.29) = 2.081, P = 0.0377) and not statistically equivalent (t(1060.29) = −1.186, P = 0.118). That is, the messaging statistically significantly increased the average intention to adopt high-impact climate behaviors. Supplementary Fig. S1 is the graph of the observed mean difference in mean intention to adopt the behaviors. To evaluate the degree to which the messaging affected intention to adopt each behavior highlighted in the messaging, we conducted an exploratory analysis of the effect of the messaging on each behavior (Fig. 2).
Exposure to the message emphasizing the large climate-mitigating impact of individual behaviors does not affect support for the carbon tax. Proportion difference, P = −0.002 (black square), plotted with 90% TOST confidence intervals (CIs) (thick horizontal lines) and 95% NHST CIs (thin horizontal lines) with equivalence bounds ΔL = −0.1 and ΔU = 0.1 for a test result that is statistically equivalent and not statistically different from zero. A 90% TOST CI (1 − 2α) is used instead of a 95% CI (1 − α) because TOSTs (each with an α of 5%) are performed.
Exposure to the message positively affects the reported intention to adopt several high-impact individual behaviors. Percentage point difference in the intention to adopt seven high-impact climate behaviors as the result of exposure to the messaging. Error bars represent 95% CIs.
We found that those exposed to the messaging on the impact of adopting individual behaviors reported statistically significantly higher intention to eat less meat, fly less and reduce food waste. We found no statistically significant effect of messaging on the intention to adopt the remaining four behaviors. We hypothesize that this may be due to the three behaviors which show a significant change all lacking a monetary cost and therefore being perceived as easier to adopt, whereas those that showed non-significant effects all have monetary costs.
For exploratory reasons, we constructed a linear regression model regressing support for a carbon tax on demographic variables, political viewpoint and the experimental condition. Political viewpoint was the greatest unique predictor of the support for a carbon tax (Supplementary Table S1). Other predictors, including condition assignment, showed largely insignificant or small effects.
DISCUSSION AND IMPLICATIONS
Previous discourse has expressed concern that supporting individual actions for the environment can diminish support for climate policy. As a result, programs that recommend individual behavior change have been discouraged in favor of regulatory policy [9–11]. However, despite growing public concern about climate change in the USA, policies that aim at lowering greenhouse gas emissions and helping to mitigate climate change, such as a carbon tax, face political gridlock [22]. In part due to these political challenges, these much-needed regulatory policies can be difficult to implement in the near term. Complementary to regulatory initiatives, behavioral policy solutions to climate change offer high mitigation potential as well as the opportunity to create demand for greater technological and regulatory shifts while being more politically palatable in the near term. In addition, behavioral solutions can offer positive impacts on a range of challenges at lower costs than those associated with traditional economic policies [23]. For example, the Department of Energy supported Solarize campaigns apply social influences to drive the adoption of household solar with a direct program cost of $21 per ton of CO2 reduced, a cost below the social cost of emissions [24, 25]. However, due to the critical need for comprehensive policy reform, programs encouraging individual action may not be worth pursuing if they undermine broader support for regulatory approaches like a carbon tax. Based on our results, policymakers and program designers working to address climate change should view behavioral solutions and messages as potential complements driving shifts in large-scale climate policy.
The study design results in several limitations for its interpretation. The study evaluates stated carbon tax support and intention to adopt the behaviors. Actual adoption and voting behavior can diverge from expressed intentions and preferences. Furthermore, it is possible that the carbon tax was framed too broadly (‘carbon tax on companies and products…’) and perceived as insular to the individual behaviors in the messaging. Future work may extend our findings by measuring the effect of messaging on support for behavior-specific policies, such as federal tax credits for EVs or taxes that are costly to the individual.
SUPPLEMENTARY DATA
Supplementary data are available at Oxford Open Climate Change online.
ACKNOWLEDGMENTS
We thank members of the Center for Behavior and the Environment at Rare for providing insightful comments and discussion. We also thank the reviewers for their thoughtful recommendations and suggestions. Finally, we thank the study respondents who provided their time to support this research.
STUDY FUNDING
This work was conducted with support from the Arthur Vining Davis Foundations, the Grantham Environmental Trust and Mr. Samuel G. Rose.
CONFLICT OF INTEREST
None declared.
AUTHORS’ CONTRIBUTIONS
Abdurakhim Rakhimov (Data curation [equal], Formal analysis [lead], Writing—original draft [lead]), Erik Thulin (Conceptualization [lead], Data curation [equal], Formal analysis [supporting], Funding acquisition [lead], Supervision [lead], Writing—review and editing [lead]). Erik Thulin conceptualized the research, and reviewed and edited the manuscript. Abdurakhim Rakhimov and Erik Thulin designed the research. Abdurakhim Rakhimov analyzed the results and wrote the manuscript.
Data Availability
The data and other study materials are available in a public repository at https://osf.io/m9ah3/.
