Energy Policy 167 (2022) 113051 Contents lists available at ScienceDirect Energy Policy journal homepage: www.elsevier.com/locate/enpol Biased perceptions of other people’s attitudes to carbon taxation Stefan Drews a, *, Ivan Savin a, b, Jeroen C.J.M. van den Bergh a, c, d a Institute of Environmental Science and Technology, Universitat Autònoma de Barcelona, Edifici Z, 08193, Cerdanyola del Vallès, Spain b Graduate School of Economics and Management, Ural Federal University, Ulitsa Mira 19, 620002, Yekaterinburg, Russian Federation c ICREA, Passeig de Lluís Companys 23, 08010, Barcelona, Spain d School of Business and Economics & Institute for Environmental Studies, VU University Amsterdam, De Boelelaan 1105, 1081 HV, Amsterdam, the Netherlands A R T I C L E I N F O A B S T R A C T Keywords: Beliefs about other people’s opinions on climate change influence one’s own opinion. Such beliefs can, however, Carbon pricing suffer from biases in perception. Using two nationally representative surveys, we examine this issue in a new False consensus context, namely of carbon-tax acceptance in Spain. We find that the more one expects the tax to be accepted by Pluralistic ignorance others, the more one accepts it personally. But opponents of a carbon tax tend to strongly overestimate the Policy support Second-order beliefs prevalence of their opinion, i.e. they exhibit a so-called false consensus effect. In contrast, despite holding the majority view, tax supporters somewhat underestimate the prevalence of their own view, which is known as pluralistic ignorance. We further test the role of information provision by providing participants with different percentages of people accepting the tax. Overall, we find little evidence that such information provision significantly increases tax acceptance. The impact of information provision on tax acceptance tends to be moderated by the degree of false consensus. 1. Introduction design issues like how to spend tax revenues (Maestre-Andrés et al., 2019), the framing and communication of the policy (Rhodes et al., Carbon pricing, in the form of taxation or cap-and-trade, is widely 2014; Hardisty et al., 2019), education and perceived knowledge about viewed as a cornerstone of an effective climate policy package to reach the carbon tax (Savin et al., 2020), and political identity (Van Boven the Paris Climate Agreement (Edenhofer et al., 2015; Baranzini et al., et al., 2018). Here we examine an issue which has only recently gained 2017; Best et al., 2020). Most countries around the world do, however, attention in the research on climate perceptions in general and to a lesser not have a carbon tax, while the stringency of the over 20 existing (sub) extent on policy acceptance, namely the role of social perceptions or national carbon taxes tends to be rather low (World Bank, 2019; OECD, so-called social second-order beliefs. 2018). A major factor explaining both facts is public opposition, trans­ A first-order belief is what an individual herself believes about an lating in insufficient political support (Jenkins, 2014; Klenert et al., issue, whereas a second-order belief is what a person believes what 2018; Rabe, 2018). For example, the emergence of the “Yellow Vest” others might think about it.2 In general, second-order beliefs can be movement led to a halt in the planned regular raise of the French carbon considered a special type of a social norm. There is a considerable tax, partly due to concerns about regressive tax effects (Douenne and amount of research regarding social norms and more general environ­ Fabre, 2020). mental behaviors (Kinzig et al., 2013; Nyborg et al., 2016; Farrow et al., A growing body of interdisciplinary research has emerged to un­ 2017; Bergquist et al., 2019), but this tends to focus on people’s thoughts derstand public attitudes1 to carbon pricing and underlying factors about what others do, not so much what others think. Arguably, it is (Carattini et al., 2018). These include, among others, personal and easier to observe whether other people litter (a well-known case in the distributional policy costs (Hammar and Jagers, 2007), other policy research on social norms, see Cialdini, 2003) than to know what other * Corresponding author. E-mail addresses: stefan.drews11@gmail.com, stefan.drews11@gmail.com (S. Drews). 1 Here we mainly employ the term “public acceptance”, but for stylistic reasons also use synonyms and antonyms such as “public support” and “public opposition”. However, it is worth noting that some authors see conceptual differences between these and other terms to describe public attitudes to climate policy (Kyselá et al., 2019). 2 Second-order beliefs can also be about one’s own beliefs (Baron, 1987). However, here we are concerned with beliefs about others’ beliefs. For reasons of brevity, we henceforth omit the term “social” and simply speak of “second-order beliefs”. https://doi.org/10.1016/j.enpol.2022.113051 Received 20 May 2021; Received in revised form 26 April 2022; Accepted 9 May 2022 Available online 1 June 2022 0301-4215/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by- nc/4.0/). S. Drews et al. Energy Policy 167 (2022) 113051 people think about a climate change and particularly specific climate observed for certain pro-climate views, e.g. people who believe in the policies. existence of climate change believe that climate change skepticism is Relatively few studies are available on social perceptions related to much more common than it actually is (Leviston et al., 2013). climate policy. For example, Schuldt et al. (2019) examine second-order The contribution of the present article is twofold. First, it aims to beliefs and public support for coal-to-gas policy in China and find that corroborate some of the earlier findings related to second-order beliefs beliefs about socially proximal reference groups in China (e.g., “My in a new context, namely of public acceptance of carbon taxation. Spe­ friends and family” or “Chinese people in general”) together accounted cifically, we examine in two nationally representative surveys for Spain for 18% of the variance in individual-level policy support. A study for whether perceived public attitudes to carbon taxes are associated with the US shows that perceived social consensus can bridge the notorious personal acceptance of the tax. At present, there is no economy-wide political divide between liberals and conservatives regarding climate carbon tax in Spain. In addition, we investigate to what extent our re­ change beliefs and policy attitudes (Goldberg et al., 2019). A study for spondents exhibit FCE regarding carbon taxation. Empirical findings the US and China finds that both American and Chinese respondents – reviewed above indicated that minorities have skeptical views about including the general public as well as political actors – underestimate climate change. If the same holds for views on climate policy, then we the prevalence of pro-climate views (Mildenberger and Tingley, 2019). expect that particularly carbon tax opponents show stronger FCE than An important question is how people respond when being confronted tax supporters. This expectation is also justified by more general with external information about other people’s beliefs. As mentioned research on consensus estimates (Krueger and Clement, 1997; Dvir-G­ before, social norm provision has generally been shown to have effects virsman, 2015). One explanation why minorities overestimate the on environmental behaviors, but little is known about this in the context prevalence of their own view is that this tendency helps to counteract of climate policy perceptions and attitudes. A US study involving uni­ the psychological threats of holding minority views. versity students examines norm interventions related to a “carbon A second contribution of this study is that we test the effects of emissions cap” (Bolsen et al., 2014). In one experimental condition, influencing second-order beliefs, and how they are moderated by FCE. participants were informed that 85% of Americans believe in climate There is a lack of experimental research on whether social perceptions change, are willing to engage in behavioral changes, and support a tax to related to climate change can be revised (Abeles et al., 2019). Some reduce greenhouse gas (GHG) emissions. This norm intervention, which related studies have experimentally manipulated perceptions of others’ mixes tax-related with other second-order climate beliefs, did not in­ policy attitudes (de Groot and Schuitema, 2012; Bolsen et al., 2014), but fluence policy acceptance. However, an additional experimental con­ they do not examine how experimental information provision interacts dition which conveyed that only a 15% minority has pro-climate views with FCE. We fill this research gap by providing respondents with in­ led to lower policy acceptance. A study for the UK, involving 123 par­ formation about acceptance of a carbon tax among the Spanish popu­ ticipants, tested the effect of a strong (80% majority acceptance of lation to test the claim that people with high initial false consensus are policies) and a weak (20% acceptance) norm (de Groot and Schuitema, more resistant to change their own tax acceptance. This is based on 2012). It found that people informed about the strong norm were more general psychology research showing stability of FCE despite contrary supportive of two incentive-based policies, namely a car tax and a lit­ information (Krueger and Clement, 1994), as well as longitudinal evi­ tering fine. Furthermore, a study by Mildenberger and Tingley (2019) dence suggesting such a tendency for more general climate beliefs exposed US respondents to information about the true proportion of (Leviston et al., 2013). Finally, given that all of the prior research comes Chinese people’s pro-climate beliefs (98%), which on average slightly from Australia, China, the UK and the US, and at least in some cases increased support by US respondents for a global climate treaty. draws on non-representative samples, our study for Spain is a useful The effect of information provision about second-order beliefs might addition for generalizing results across countries and entire populations. be moderated by biases in social perceptions, in particular a so-called This paper reports results from two surveys conducted in Spain. false consensus effect (FCE, Ross et al., 1977). This refers to an indi­ Methods and results of the first and second survey are reported as two vidual overestimating how common their own opinion is in the wider separate studies in Section 2 and 3, respectively. The first study can be population. Several mechanisms can explain this effect. For example, viewed as rather exploratory, while the second study tried to improve people may lack exposure to dissimilar views; or people are motivated to upon some limitations of the first survey as well as confirm the initial appear normal. In general, understanding biases in social beliefs, such as findings. Section 4 provides a general discussion of the results. Section 5 FCE, and their implications for environmental policy, has been identi­ concludes. fied as an important research gap in environmental economics (Millner and Ollivier, 2016). 2. Study 1 Probably the first study examining second-order beliefs in the context of climate change (not climate policy) was Leviston et al. (2013). 2.1. Data and method It found that on average respondents overestimated the proportion of people who are skeptical about the existence of climate change, with the We draw on a survey conducted in August 2019 in Spain. The survey, strongest overestimations voiced by skeptics themselves. No attempt to including its experimental part, was approved by the University’s manipulate second-order beliefs was undertaken. The latter was done in Human Ethics Committee. It was conducted by the survey company a study by Geiger and Swim (2016) who led American university stu­ ‘Netquest’. Sampling was done by using quotas on age, gender and dents experimentally to believe that most other people are either con­ geographical distribution, making the survey sample representative of cerned or unconcerned about climate change (see “Study 2” in their the general population on these and other characteristics (see Table A1 paper). They found that those very concerned were more willing to talk in Appendix A for further information). The response rate was 59%, with about climate change if they were led to believe that most others are also the final sample consisting of 2004 complete responses. Respondents concerned. took on average about 15 min to finish. This is because the survey A related, sometimes contrasting phenomenon is so-called pluralistic included various additional questions on carbon taxation apart from the ignorance. This denotes people’s tendency to incorrectly believe that an ones described below (see Appendix B for full questionnaires). Partici­ opinion they reject is held by an absolute or relative majority of others pation and completion of the survey was encouraged through a non- (Leviston et al., 2013; Sokoloski et al., 2018). Such effects have been monetary gift voucher. 2 S. Drews et al. Energy Policy 167 (2022) 113051 Fig. 1. Actual and expected (non-)acceptance level of a carbon tax, depending on respondents’ own (non-) acceptance. The Y-axis shows how prevalent a certain level of (non-)acceptance actually is (in red) versus how people holding this opinion expect the preva­ lence of their opinion to be in the overall population (in green). Right panel is a disaggregation of left panel. Error bars indicate ± 2 standard errors. The green bars add up to more than 100% because they represent expected averages for each response option, whereas red bars denote actual shares of each acceptance level. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Our measure of carbon tax acceptance prior to any experimental consensus they exhibit. We operationalize FCE by estimating for each intervention is based on the following survey question: “How acceptable respondent the deviation of own expected (non-)acceptance from the do you find the carbon tax if its revenues are used to support the actual public (non-)acceptance level: development of climate projects?” (Q1). Participants could respond by false consensus biasi =expected (non)acceptancei – public (non)acceptance using a 5-point Likert scale with 1 (completely unacceptable), 2 (somewhat unacceptable), 3 (neither unacceptable nor acceptable), 4 (1) (somewhat acceptable) and 5 (completely acceptable). The frequencies While the former element on the right-hand side of Eq. (1) varies of responses to Q1 are depicted in Fig. 1 by the actual level of accep­ across the respondents collected by Q2, the latter is the same for all tance. After answering various other questions about carbon taxation subjects (and equals 69%) and is calculated as an average tax accep­ (used for other research projects), respondents were confronted with the tance, based on Q1. To distinguish between FCE for the two types of following question aimed at capturing second-order beliefs: “To the best opinions, we separately estimate Eq. (1) for those accepting and not of your knowledge, what percentage of the Spanish population would accepting a carbon tax.3 For example, for a person who accepts a carbon somewhat or completely accept a carbon tax? Type a number from 0 (no tax and expects a level of acceptance of 79%, while the actual level is one) to 100 (everyone).” (Q2). The frequencies of responses to Q2 are 69%, FCE will be equal to 10%, i.e. she overestimates the prevalence of presented in the Appendix in Figure A1 (left plot). Right afterwards, they her own view. Alternatively, for a person who does not accept a carbon read: “A recent public opinion survey in Spain demonstrated that 19% tax and expects a non-acceptance rate of 60% while the actual public [alternative: 67%] of people would accept a carbon tax if the revenues non-acceptance rate is of 31% (100%–69%), FCE will be equal to 29%, i. are used for climate projects. Given this information, how acceptable do e. she overestimates the prevalence of non-acceptance of a carbon tax. you find such a carbon tax?” (Q3). Approximately half of the sample Further note that a negative value of FCE may indicate pluralistic received at random the 19% and the other half the 67% version. Re­ ignorance. For example, a person who accepts the tax and expects sponses could again be given on a 5-point scale. Frequencies of responses prevalence of this view to be 51%, has negative FCE (51%–69% = to Q3 are again depicted in Figure A1 in Appendix. These fictitious − 18%), but does not show pluralistic ignorance. This is only the case if numbers were chosen to reflect two conditions of either minority or a expected prevalence of own opinion is below 50%. We will highlight majority acceptance. We decided to use fictitious numbers as there was cases of pluralistic ignorance at relevant places. Our analytical focus is, no recent real survey data available that we could draw on for our however, on FCE, given that this phenomenon is much more common, purposes. The European Social Survey for 2016 found that 27% of the particularly among the politically more relevant case of carbon tax Spanish population was in favor of “increasing taxes on fossil fuels” opponents. (Pohjolainen et al., 2018), which is somewhat similar to our wording The change in acceptance rate, in turn, is operationalized as a dif­ (“carbon tax”). This value is fairly close to the value for our minority ference between acceptance rates collected after (Q3) and before (Q1) group (19%). Study participants were debriefed at the end of the survey the fictitious information has been provided: about the nature and purpose of the provided information. Regarding change in acceptance = acceptance after − acceptance before (2) the use of fictitious information and potential ethical concerns, consider that disciplinary norms exist: It is typically permitted in social psy­ chology and sociology (Barrera and Simpson, 2012). In contrast, The constructed variable “change in acceptance” has a limited value experimental economics tends to proscribe it, but surveys of economists range: those who originally accept a carbon tax based on Q1 can at most show that even within this discipline there is disagreement on what is permissible or not (Colson et al., 2016; Krawczyk, 2019). To understand whether respondents change their tax acceptance, we 3 measure the impact of information provision on their tax acceptance Non-acceptance includes response options 1, 2 (“completely”/“somewhat unacceptable”) and 3 (“neither … nor …”). This is because when measuring level and assess if this is significantly associated with the level of false second-order beliefs through Q2 we distinguish only between two possibilities: either somewhat/completely accept a carbon tax (response options 4 and 5) or all the remaining options (1, 2 and 3). Excluding respondents who choose op­ tion 3 from non-acceptance would bias our results, as we cannot distinguish them from non-acceptance in Q2. In the next Section (Study 2) we use a revised format of Q2 to make such a distinction possible. 3 S. Drews et al. Energy Policy 167 (2022) 113051 increase their acceptance by one unit (i.e. from 4 to 5) and decrease most Table 1 by four units (i.e. from 5 to 1), whereas those who do not accept can Average tax acceptance before and after information provision. increase most by four (from 1 to 5) and decrease most by two (from 3 to Type of sample, Acceptance Acceptance after Pairwise 1) units. with sample size before information Mann- information manipulation Whitney manipulation test’s p 2.2. Results value 1 All respondents, n 3.886 3.455 First, we examine how perceptions of public opinion relate to in­ <0.0001 = 2004 (3.840–3.932) (3.399–3.511) dividuals’ own carbon tax (CT) acceptance. Conducting an OLS regres­ 2 Informed with 19% 3.884 3.332 <0.0001 sion of tax acceptance on perceived population-wide tax acceptance: acceptance, n = (3.817–3.950) (3.253–3.411) 1005 CT acceptance level = α + β∗expected acceptance level + ε (3) 3 Informed with 67% 3.888 3.579 <0.0001 acceptance, n = 999 (3.824–3.952) (3.500–3.657) 4 Informed with 19% 4.480 3.721 <0.0001 we find that the higher the perceived acceptance, the higher is the acceptance and (4.442–4.519) (3.636–3.807) respondent’s own acceptance (β = 0.012, SE = 0.001, p < 0.01, R2 = initially accepting 0.09).4 These results remain robust when controlling for socio- the tax, n = 689 5 Informed with 19% 2.582 2.484 0.0827 demographic characteristics like age, gender, education, household in­ acceptance and (2.504–2.660) (2.361–2.607) come and political orientation (see Table A2 in Appendix A). initially not Next, we look at how perceptions of public attitudes depend on one’s accepting the tax, n own tax acceptance. The left plot in Fig. 1 shows that people who accept = 316 6 Informed with 67% 4.461 3.977 a carbon tax, on average, estimate the proportion of the population <0.0001 acceptance and (4.423–4.499) (3.897–4.057) holding a similar attitude to be lower than what it actually is (36% vs. initially accepting 69%), whereas those not accepting a carbon tax tend to overestimate the the tax, n = 687 prevalence of their own attitude (77% vs. 31%). In other words, those 7 Informed with 67% 2.625 2.702 0.3528 critical of the tax on average exhibit FCE, whereas most tax supporters acceptance and (2.552–2.698) (2.571–2.833) initially non- display pluralistic ignorance. Note also that both groups underestimate accepting the tax, n the ‘true’ prevalence of carbon tax support (23% expected by those who = 312 do not accept and 36% expected from those who accept compared to the Note: Results are reported with±two standard errors in parentheses. actual 69%). The right plot in Fig. 1 gives a more detailed depiction of actual and expected acceptance depending on all levels of personal acceptance. In other words, for each group (separated on the X-axis) the public acceptance, while they most likely react to external information percentage of people who are in this group are depicted, versus the by revising their acceptance downwards (rows 4 and 6 in Table 1). In percentage of the population this specific subgroup believes to be in this contrast, negative FCE among tax opponents means they overestimate group. The figure shows that the expectation mismatch is largest for the acceptance of the tax, while they likely react to external information by extreme case that people find a carbon tax completely unacceptable. revising their acceptance upwards (rows 5 and 7 in Table 1). Therefore, In the following we compare own tax acceptance before and after by estimating Eq. (4) on the full sample, such nuances would be lost, as information provision. This is done for the overall sample of observa­ opposed to estimating it for separate subsamples. Our regression equa­ tions and by splitting it based on the information being provided and the tion for the four subsamples then has the following form: initial level of acceptance (Table 1). We use a pairwise Mann-Whitney false consensus effect = α + β∗change in acceptance + ε (4) test that explicitly compares responses from the same individuals. P- values below conventional significance levels indicate that acceptance has changed after information provision. We find that on average and tests if people with different FCE have a different propensity to acceptance went down for both communicated levels acceptance (19% change their acceptance level.5 We first estimate Eq. (4) using OLS.6 and 67%). This drop is a bit steeper if they are given the 19% compared Figure A3 in Appendix A depicts related scatter plots combined with 2D to the 67% treatment. There is a drop among people who accept and are density plots to illustrate where most observations are concentrated. The told 19%. There is a small (non-significant) drop among people who reported results are robust to alternative functional forms, namely or­ reject and are told 19%. There is a drop among people who accept and dered logit regression. Furthermore, to ensure robustness of our results are told 67%. And there is a slight (non-significant) increase among to outliers and allow for the possibility that people with different FCE those who reject and are told 67%. In sum, if we distinguish between may have changed their acceptance differently, we additionally con­ those who initially accepted the tax from those who did not, we find that ducted a quantile regression (QR, Koenker and Bassett, 1978) of Eq. (4) the effect is mostly explained by the drop in acceptance among the tax as it allows to differentiate the role of the explanatory variable supporters (row 4 and 6 versus 5 and 7 in Table 1). These results are throughout the distribution of the dependent variable. The latter moti­ further discussed in Section 4. vated us to put FCE on the left-hand side of Eq.(4), thus distinguishing As a next step, we test how FCE of the respondents is related to their change in tax acceptance. We do this in four subsamples, namely re­ 5 spondents who personally accept the tax or not, combined with the Note that on advice of one of the anonymous reviewers we also tested an message received, i.e. either the information of 19% or 67% population- equation regressing change in carbon tax acceptance on prior acceptance level, expected level of acceptance and their interaction effect. Results for Study 1 and wide tax acceptance. Testing results for separate samples is motivated by Study 2 are reported in the Appendix in Table A10 and do not show any sys­ the following consideration. FCE has been measured in different ways tematic pattern. for those accepting and those not accepting the carbon tax (see Eq. (1)). 6 One might argue that our estimates are biased because both FCE and change Thus, negative FCE among supporters indicates they underestimate in acceptability are constructed using own CT acceptance level. To test this, we estimate Eq. (4) by adding own CT acceptance level as a control and examining multicollinearity using Variance Inflation Factor (VIF). As one can see in 4 As a robustness check, we estimated an ordered logit model, since the Table A9 in Appendix A, our estimations remain robust, while VIF values vary dependent variable can be considered categorical, confirming that coefficient β between 1.02 and 1.06, which is far below the conservative benchmark of 5, from Eq. (3) is positive and significant (with a pseudo R2 of 9%). suggesting absence of multicollinearity. 4 S. Drews et al. Energy Policy 167 (2022) 113051 how the relationship between FCE and the change in acceptance level The sampling entailed first inviting respondents from Study 1. After varies depending on the extent of FCE. In other words, in contrast to reaching a certain threshold, the sample was extended by inviting new OLS, QR does not impose β to be constant but allows it to vary with the participants. This mixed sampling strategy is the reason for differences degree of FCE shown by the respondents. The results of QR for different in response rates between the two subsamples. Anyway, given that percentiles in FCE distribution are reported in Figure A4 and indicate a participation in Study 1 might somehow affect results of Study 2, we test nonlinear relationship between the degree of FCE and change in potential differences in key variables. Using the Kruskal-Wallis rank sum acceptance level. test with Bonferroni correction we tested if the level of acceptance Overall, the results show that people exhibiting a relatively high (asked initially and in a follow-up question after information manipu­ degree of FCE are less inclined to revise their own carbon-tax acceptance lation) differed between people who participated in Study 1 (n = 1172) in response to information about others’ opinions (Figure A3-A4). This or not (n = 1028). The results indicate that the two samples are not effect is stronger among what is arguably the politically most relevant significantly different in: any of the key variables (Table A4), the rela­ group, namely tax opponents. This type of analysis is repeated in a very tionship between own tax acceptance and expected acceptance of others similar way in Study 2 below. This is why we report detailed results only (Table A2), the extent of FCE (Figure A5), the effects of the experimental of Study 2 in the main text, while detailed results from Study 1 can be treatment (Table A5), and the role of FCE moderating the effect of in­ found in Appendix A. formation on changes in CT acceptability (Figure A6). Moreover, it is Study 1 has several limitations. One is the measure of the re­ worth considering that the respondents are members of a survey panel, spondents’ estimate of population-wide acceptance of a carbon tax, hence they participate in surveys of different types at a regular basis. We which is not perfectly comparable to the measure of respondents’ own therefore deem it unlikely that they have a strong memory of the first acceptance of carbon taxes. This is because we asked them to estimate survey taken almost one year before. In sum, we consider the two groups other people’s acceptance of a carbon tax in general, whereas we elicited of respondents as comparable and suitable for an integrated analysis. An personal acceptance of a carbon tax before and after information pro­ additional analysis shows that there are no significant differences in vision under the condition that the revenues will be spent on climate climate concern and tax acceptance between the participants who projects. The reason for this is that the latter survey question was part of participated in both surveys and the full samples of respondents in another research objective of the survey described in Maestre-Andrés Studies 1 and 2 (Table A6). Together with a low survey dropout rate et al. (2021). This wording might have affected our results, as public (6.6%), this indicates that there is little reason to be concerned about perceptions of carbon taxes can differ depending on the use of tax rev­ self-selection bias. enues. To improve upon this issue and to test the robustness of our in­ Attitudes to carbon taxation in this survey were measured by the sights was one of our motivations to conduct Study 2. question “How acceptable do you find a carbon tax?” (Q1), to which In addition, the conceptualization and measurement of “non-accep­ participants could respond by using the same Likert scale from 1 tance” in Study 1 is somewhat limited. Remember that tax opponents (completely unacceptable) to 5 (completely acceptable) as in Study 1. were not asked to directly estimate the prevalence of their own attitude, The tax with unspecified revenues was considered somewhat or but rather what percentage of the population they think accepts the tax. completely acceptable by 51% of respondents. Second-order beliefs Thus, we assumed that tax opponents’ beliefs of the prevalence of their were measured similarly as in Study 1: “In your opinion, what per­ own attitude can be measured by subtracting beliefs of tax acceptance centage of the Spanish population would accept or not accept a carbon from the total population (100%). However, previous research shows tax? Type a number from 0 (no one) to 100 (everyone). The total amount that people on average believe that about 20% of the population does should be equal to 100%.”. However, this time participants could enter not hold any substantive opinions (Leviston et al., 2013), i.e. they think three numbers for the “% of people who accept the tax”, “% of people others do not know, are indifferent or undecided about climate change. who do not accept the tax”, and the “% of people who are undecided or This means that when tax opponents are asked directly about the ex­ indifferent” (Q2). Afterwards participants answered a set of other pected prevalence of their own attitude, one might arrive at a lower questions which are not used for this article, such as on acceptance of number than in Study 1. In contrast, we added a third category of other carbon tax schemes and other climate policies, as well as on second-order beliefs to Study 2, namely of other people who are unde­ several experiences and perceptions related to the COVID-19 crisis (see cided about the tax. In this way, we can more precisely measure full questionnaire in Appendix B). Subsequently, a randomly drawn half aggregate second-order beliefs and individual-level FCE. This also re­ of the sample received the following information: “A recent public sponds to calls in the literature to avoid binary conceptualizations and opinion survey in Spain demonstrated that 43% of them would accept a measurements of second-order beliefs (Mildenberger and Tingley, carbon tax, 38% would not accept it, and 19% is undecided.” A control 2019). group received no such information. In contrast to Study 1, the provided information is not fictitious but derived from the results of Study 1. Note 3. Study 2 that these numbers are calculated from a question on carbon taxes without any specified use of the tax revenues. Respondents in both the 3.1. Data and method treatment and control group were then asked again the question from before, that is, “How acceptable do you find a carbon tax?” (Q3).7 The To apply lessons from the previous study and overcome some of its logic of the control group was to capture potential attitude changes due limitations we undertook a second study. The survey was conducted to other random noise in the survey. between June–July 2020 in Spain. It was approved by the University’s Here we operationalize FCE by estimating for each respondent the Human Ethics Committee. The same survey company and techniques as deviation of own expected attitude from the actual level (Eq. (1)), i.e. in Study 1 were used. Sampling was again done by using quotas on age, the same as in Study 1 with the only difference that here we also gender and geographical distribution, making the survey sample measured FCE separately for those who were undecided (response op­ representative of the general population on these characteristics (see tion “3”) on Q1. Table A3 in Appendix A). The sample included 2200 respondents who took on average 19 min to finish. The response rate was 68%. About half of the respondents participated also in the survey of Study 1. These participants had a response rate of 82%, compared to 57% among the remaining ones. The reason for partially inviting the same respondents is a related research project on temporal changes of atti­ 7 Frequencies of responses to Q2 and Q3 are reported in the Appendix in tudes and behaviors, which however is not part of the present analysis. Figure A2. 5 S. Drews et al. Energy Policy 167 (2022) 113051 4. Results underestimation. Undecided people are accurate about the prevalence of those with undecided views, while both rejectors and supporters un­ We start again by examining how perceptions of public opinion derestimate the prevalence of undecided views. People who accept a relate to individuals’ tax acceptance. Conducting an OLS regression of carbon tax estimate the proportion of the population holding a similar tax acceptance on estimated acceptance (Eq. (3)), we find that the higher attitude to be lower than what it actually is (36% vs. 51%), while at the is the perceived population-wide tax acceptance, the higher is the re­ same time overestimating the prevalence of tax opponents (48%). This is spondent’s own acceptance (β = 0.021, SE = 0.001, p < 0.01, R2 = again a form of pluralistic ignorance, although this time only to the 0.13).8 As in the first study, these results here are robust when con­ extent of incorrectly assuming a relative, but not absolute, majority. trolling for socio-demographic characteristics (see Table A2). As one can Those rejecting a carbon tax strongly overestimate the prevalence of see, the explanatory power of expected acceptance of carbon taxation in their own attitude (67% vs. 26%) and underestimate prevalence of other Study 2 is higher than in Study 1. opinions (19% and 14% for acceptance and indecisiveness). In other Next, we investigate how perceptions of public attitudes depend on words, the tax opponents on average again exhibit FCE. Finally, the one’s own tax acceptance. First of all, the upper panel in Fig. 2 shows undecided respondents tend to overestimate only the prevalence of tax that everyone overestimates tax rejection, but the opponents themselves opponents (54%). The right panel in Fig. 2 gives a more detailed have a stronger overestimation. Everyone underestimates acceptance, depiction of actual and expected acceptance depending on all levels of but opponents and those who are undecided, again, exhibit stronger personal acceptance. It shows that the expectation mismatch is largest at the extreme where people completely reject a carbon tax. In addition, there is generally more overestimation of minority opinions than un­ derestimations of majority opinions (further see Table A7). Taken together, despite somewhat different question wordings, these results are largely consistent with Study 1. Now we compare acceptance before and after information manipu­ lation for the overall sample of observations and by splitting it depending on the information being provided and the initial level of acceptance (Table 2). This is done with a pairwise Mann-Whitney test that explicitly compares responses from the same individuals. We find that acceptance is reduced irrespective of whether any information on the level of acceptance in Spain is provided. If we distinguish between Table 2 Average tax acceptance before and after information provision. Sample, with Acceptance rate Acceptance rate Pairwise sample size before after information Mann- information manipulation Whitney manipulation test’s p value 1 All respondents, n 3.338 3.243 <0.0001 = 2200 (3.284–3.392) (3.189–3.297) 2 Informed with 43/ 3.335 3.247 0.0007 38/19% acceptance (3.256–3.412) (3.170–3.324) rates, n = 1110 3 No additional 3.342 3.239 <0.0001 information, n = (3.266–3.417) (3.163–3.315) 1100 4 Informed with 43/ 4.382 4.058 <0.0001 38/19% acceptance (4.342–4.423) (3.987–4.129) rate and initially accepting the tax, n = 570 5 Informed with 43/ 3.000 2.914 0.0899 38/19% acceptance (3.000–3.000) (2.812–3.016) Fig. 2. Actual and expected (non-)acceptance of a carbon tax, depending on rate and initially respondents’ own (non-)acceptance. The Y-axis shows how prevalent a certain undecided, n = 245 6 Informed with 43/ 1.526 1.912 <0.0001 level of (non-)acceptance actually is (in red) versus how people holding this 38/19% acceptance (1.467–1.586) (1.788–2.036) opinion expect prevalence of other opinions to be in the overall population rate and initially (other colors). The lower panel is a disaggregation of the upper panel. Actual rejecting the tax, n acceptance in the lower panel is given for each specific response unit, while = 285 expected acceptance is an average of (non-)acceptance (i.e. 1-2 and 4–5). Error 7 No additional 4.396 4.081 <0.0001 bars indicate ± 2 standard errors. The red bars denote actual shares of each information and (4.355–4.438) (4.014–4.148) acceptance level and add up to 100%, while other bars add up to more than initially accepting 100% because they represent expected averages for each response option. (For the tax, n = 555 8 No additional 3.000 2.907 0.0625 interpretation of the references to color in this figure legend, the reader is information and (3.000–3.000) (2.803–3.010) referred to the Web version of this article.) initially undecided, n = 268 9 No additional 1.560 1.874 <0.0001 information and (1.500–1.619) (1.755–1.992) initially rejecting 8 the tax, n = 277 Again, as a robustness check, we estimated an ordered logit model, con­ firming that β is positive and significant (with a pseudo R2 of 14%). Note: Results are reported with±two standard errors in parentheses. 6 S. Drews et al. Energy Policy 167 (2022) 113051 those who initially accepted the tax from those who did not, we find that in the lower percentiles of the FCE, i.e. those who overestimated the those who accepted the tax reduced their acceptance on average, while prevalence of carbon tax opposition by no more than 50%. Results are those who originally rejected the tax revised their acceptance upwards. similar for the control group of tax opponents that did not receive in­ People who were initially undecided in their tax attitudes are the only formation (β = − 4.989, p < 0.01, R2 = 0.04, n = 277). However, at group for who we do not observe a significant revision in tax acceptance. lower levels of FCE the β coefficients of the QR estimation are consid­ Surprisingly, the same effects are identified among the cohort of people erably higher in the treatment group, suggesting that FCE moderated the who received no additional information about population-wide accep­ effect of information. Overall, the negative association implies that tance. These results remain robust if we exclude respondents who people who had lower FCE revised their acceptance more strongly up­ originally reported very high support or opposition to the tax (values 1 wards, while those who had a very high FCE revised them less, which is and 5 on the Likert scale) on Q1 (see Table A5). Therefore, we can in line with our expectations and findings in Study 1. conclude that our findings cannot be explained by the regression to the If we now look at the people who initially accepted the carbon tax mean effect (Yu and Chen, 2015). Although our analytical focus here is and received additional information (n = 570, upper-left panel in Fig. 3), on within-subject comparison, an additional between-subject analysis of we find a positive relationship between FCE and change in acceptance levels of acceptance after the information provision confirms no signif­ (β = 3.3047, p < 0.01, R2 = 0.02). This suggests that those with lower icant treatment effects (see Table A8). We discuss these findings further FCE are more likely to revise their acceptance downwards than those in Section 4. with high FCE. This intuition is supported by the results of the QR in We proceed by testing how FCE of the respondents is related to their upper left panel of Fig. 4, where we see a significant association between change in tax acceptance. We do this for six subsamples, namely re­ the two variables only for upper percentiles of FCE corresponding to spondents who personally accept the tax, are undecided or reject it, absolute values of FCE in the interval (− 11, − 1), i.e. those with rela­ combined with information manipulation received, i.e. either the in­ tively accurate expectations of carbon tax acceptance. Thus, people who formation on 43/38/19% acceptance levels or no additional informa­ accepted the tax but slightly underestimated the prevalence of their own tion. As in Study 1, we measured FCE separately for those accepting or opinion revised their acceptance downwards, while those who under­ not the carbon tax (see Eq. (1)), with the difference that here we addi­ estimated public acceptance a lot did not change their acceptance tionally measured FCE for those who are undecided as a difference be­ significantly. The percentage of acceptance we have communicated was tween expected fraction of undecided respondents and actual fraction of relatively low (43%, which is lower than the acceptance observed in such survey participants. The results are plotted in Fig. 3. The reported Study 2, namely 51%), which might explain that people with low FCE results are robust to alternative functional forms, namely ordered logit who were originally more positive about the carbon tax reduced their regression. In addition, just like in Study 1, results of QR for different acceptance. In addition, it should be noted that results look fairly similar percentiles in FCE distribution are reported in Fig. 4. in the control group of tax supporters without information (β = 3.665, p Starting with the case of respondents who initially rejected the < 0.01, R2 = 0.02, n = 555). proposal of a carbon tax and who received information (n = 285), we If we look on people who were undecided and received additional find the association to be negative and significant (β = − 6.684, p < 0.01, information (n = 245, top central chart in Fig. 3), we find no statistically R2 = 0.07, upper right panel in Fig. 3). According to QR analysis in the significant relationship between FCE and change in acceptance (β = upper right panel of Fig. 4, this negative association is more pronounced − 0.11, p = 0.95, R2 = <0.001). A similar result is obtained for Fig. 3. The relationships between FCE and change of carbon tax acceptance after receiving or not information about population-wide tax acceptance, based on OLS estimation of Eq. (4). The six sub-samples are defined by combinations of information messages and personal carbon tax acceptance, indifference or non-acceptance. Higher density indicates where observations are concentrated. 7 S. Drews et al. Energy Policy 167 (2022) 113051 Fig. 4. Quantile regression of Eq. (4). On the X-axis the percentile in the distribution of FCE is given, while the Y-axis shows the value of the β-coefficient from Eq. (4) with 95% confidence interval. Since observations on FCE are distributed unevenly, next to the percentile we report in parentheses the actual FCE value corresponding to it. With red color the estimation provided by OLS is given for comparison. The six sub-samples are defined by combinations of information messages and personal carbon tax acceptance, indifference or rejection. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) undecided individuals who received no additional information (n = 268, the prevalence of their own attitude, although to a somewhat lesser bottom central chart in Fig. 3; β = − 1.18, p = 0.49, R2 = <0.002). This is extent than tax opponents overestimate (particularly in Study 2). Based not surprising given high accuracy of respondents in these groups in on insights from related research (Geiger and Swim, 2016), these find­ predicting popularity of their own opinion (Fig. 1) and low propensity of ings suggest that tax supporters may self-silence their (majority) view, changing their carbon tax acceptance level (see Table 2). while opponents may be more inclined to voice their (minority) posi­ tion. These findings are in line with non-climate research showing that 5. Discussion majorities and minorities tend to under- and overestimate, respectively, their own views (Krueger and Clement, 1997). Using two nationally representative surveys from Spain, we derived Our research further contributes to the literature by experimentally several insights about perceptions regarding other people’s acceptance examining whether influencing social (mis)perceptions triggers a of a carbon tax, a topic that has so far received little attention in the change in people’ own carbon tax acceptance, depending on FCE. In literature on public opinion about carbon pricing (Carattini et al., 2018; general, our findings suggest that a relatively high FCE leads people to Maestre-Andrés et al., 2019; Levi, 2021). In both studies we found that a be fairly unresponsive to information provision about others’ tax person’s own acceptance of a carbon tax is positively associated with acceptance. The politically perhaps most relevant case of respondents their beliefs about others’ acceptance. This result replicates previous who initially reject the carbon tax shows that those with low FCE can findings on related climate perceptions from the US and China (Schuldt become slightly more supportive of a carbon tax when they are informed et al., 2019; Goldberg et al., 2019) for the context of carbon tax about others’ attitudes. These findings are largely consistent with Lev­ acceptance in Spain. It is worth noting at this point that the causality of iston et al. (2013) who found that people with high FCE showed more this relationship is unclear. People’s social beliefs can shape their own stability of climate perceptions over time. Our experimentally derived beliefs, but also vice versa, which can result from processes of social insights on change of policy attitudes complement their longitudinal contagion or homophily (Shalizi and Thomas, 2011). research design examining opinion stability. We further found that everyone underestimates public acceptance of The results also indicate that low FCE can work in the other, unde­ a carbon tax. This general tendency may be explained by considerable sired direction: Those initially accepting the carbon tax reduced their publicity given to climate contrarian views in traditional mass media acceptance after receiving information that either a minority or a ma­ (Kjeldahl and Hendricks, 2018), which could be the result of journalistic jority of Spanish citizens support a carbon tax. Overall, such ‘downward’ norms, or even of organized political campaigns (Boykoff, 2013). More attitude change is more prevalent in our data than ‘upward’ revision. specifically, however, we found that carbon tax opponents strongly These diverging effects, which depend on FCE, might explain why some overestimate the prevalence of their own attitude in the population. On prior research for the US has found zero overall effects on support for average, opponents believe that there are more than twice as many climate policy after communicating a strong social norm of 80% pro- people thinking like them than what the numbers actually show. In both climate sentiment (Bolsen et al., 2014; but see de Groot and Schui­ studies, overestimations of tax opposition were higher than 40%. This tema, 2012 for an effective norm intervention). In other words, some can be considered as relatively high when compared with the few people might become more favorable about climate policy, while in­ existing other, related estimates. For example, people who believed that formation may reduce acceptance for others. The latter effect possibly climate change is not happening overestimated their opinion by 25% in occurs because some types of messages trigger the well-known psycho­ the US study by Mildenberger and Tingley (2019). However, such logical phenomenon of reactance (Rains, 2013). In any case, the climate denialism represents arguably a more marginal opinion than experimental results need to be interpreted with caution, as will be opposition to a carbon tax, which makes our FCE estimates appear as discussed in the following. fairly sizeable. In contrast, carbon tax supporters tend to underestimate This research has some limitations. First, Study 1 lacks a control 8 S. Drews et al. Energy Policy 167 (2022) 113051 group, so one cannot compare with the case of no information provision. 6. Conclusion and policy implications This was improved upon in Study 2. Second, the experimental findings of Study 2 are, surprisingly, very similar to the control group. This could This study has examined several aspects of social perceptions related mean that attitude changes happened for other reasons than informa­ to the acceptance of a carbon tax. It finds a positive association between tion. After responding to the first question on personal acceptance of people’s own tax acceptance and their beliefs about others’ acceptance. carbon tax, and to the question eliciting beliefs about others’ tax This is in line with prior research on more general climate change per­ acceptance, participants were asked questions about other climate ceptions in other contexts, which finds that opinions are consistent with mitigation policies, carbon tax schemes under distinct revenue uses (e.g. perceived consensus. We not only identify a general tendency of with progressive redistribution), as well as various questions about underestimating others’ tax acceptance, but demonstrate that opponents health and economic effects of COVID-19. Such questions might have of a carbon tax underestimate public acceptance of the policy, thus reduced acceptance of a carbon tax, for example, because preferences for overestimating the prevalence of their own view. That is, they show a other policies, or other carbon tax designs, became more salient. This false consensus effect. To a lesser extent, carbon tax supporters under­ may explain the overall decrease of tax acceptance with and without estimate the prevalence of their own majority position, i.e. many of information provision. Yet, some questions may also have increased them show pluralistic ignorance. Such effects may lead to a vocal mi­ acceptance of those who initially opposed the tax. For example, several nority of tax opponents, and a silent majority of tax supporters. We tried questions referred to the use of the revenues of carbon taxes. People who to influence these social perceptions by communicating hypothetical have been made aware that carbon taxes create revenues which can be and actual information about what percentage of the population accepts used for various purposes (e.g. environmental, social) may have become or not a carbon tax. We obtain tentative evidence that people who more favorable of the policy in general. One detail of our survey design initially reject the tax slightly increase their acceptance when they are may have contributed to the generally more moderate responses when informed that a majority is in favor of the tax. Such an attitude change questioned the second time, namely the communicated percentages tends to take place when the false consensus effect is relatively small. others’ tax (non-)acceptance. The numbers (43/38/19%) can be inter­ There is some indication that communicating others’ tax acceptance can preted as if there are still many varying opinions on the topic, which may also reduce tax acceptance. Overall, our study suggests that perceptions have increased participants’ reflection on the policy. Future studies of others’ tax opinions can be very inaccurate and that influencing these could thus test attitude change using other research designs. Moreover, perceptions with the goal of improving tax acceptance will not be easy. although our single item measure of carbon tax acceptance is fairly Further research using other methods and types of communication is common in the literature (Maestre-Andrés et al., 2019), future studies needed to get a better understanding of this challenge. might use multi-item indexes of carbon tax acceptance to measure attitude change more reliably (Kyselá et al., 2019). Finally, another Final note possibility is that attitude changes at least to some extent resulted from respondents paying insufficient attention to questions, but a statistical Data and code for this study are available at https://github.com/Iva test did not provide any evidence for this proposition.9 nVSavin/FalseConsensusProject. Future research also needs to dive deeper into the processes under­ lying the observed patterns. For example, in line with the idea of (dis) CRediT authorship contribution statement confirmation bias one could examine whether respondents with high FCE are more likely to discard information contradicting their prior Stefan Drews: Conceptualization, Investigation, Writing – original beliefs, so as to protect their opinion. On the other hand, it is also draft, Writing – review & editing. Ivan Savin: Data curation, Formal possible that people do not update their beliefs because they consider analysis, Writing – review & editing, Visualization. Jeroen van den the provided message as unreliable or untrustworthy (Druckman and Bergh: Writing – review & editing, Funding acquisition. McGrath, 2019). Here one could test whether other forms of information provision, such as higher consensus messages or more trusted message Declaration of competing interest sources, can lower resistance to climate policy. Such information pro­ vision may also depend on people’s (un)certainty about opinion prev­ The authors declare that they have no known competing financial alences, exemplified by simple mental shortcuts such as “fifty-fifty” rules interests or personal relationships that could have appeared to influence (de Bruin et al., 2000). In addition, one could extend our work by testing the work reported in this paper. other than population-level beliefs, such as about friends, family or other social reference groups, including populations from other coun­ Acknowledgements tries. This can build on earlier work (Schuldt et al., 2019; Goldberg et al., 2019; Rohring and Akerlof, 2020). Furthermore, it might be worth This work was funded by an ERC Advanced Grant from the European investigating how social media use and social networks (Konc and Savin, Research Council (ERC) under the European Union’s Horizon 2020 2019) shape a person’s social perceptions. Many people nowadays get Research and Innovation Programme [grant agreement n◦ 741087] and their news and information in general and specifically on climate change by a María de Maeztu Program for Units of Excellence awarded to ICTA- from like-minded peers in online environments (Williams et al., 2015), UAB by the Spanish Ministry of Science and Innovation. I.S. acknowl­ which may further amplify the identified biased perceptions. The latter edges financial support from the Russian Science Foundation [RSF grant does not have to be the case, though, as proposals are emerging to number 19-18-00262]. establish behavioral online interventions aimed at counteracting false consensus and other biased perceptions (Lorenz-Spreen et al., 2020). Appendix A and B. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.enpol.2022.113051. 9 Specifically, in Survey 2 we find no statistically significant correlation be­ References tween change in acceptance and survey completion time (which we consider a proxy of paying attention). The Pearson correlation coefficient was 0.004 (p = Abeles, A.T., Howe, L.C., Krosnick, J.A., MacInnis, B., 2019. Perception of public opinion 0.85) for the whole sample and − 0.018 (p = 0.55) for those who received no on global warming and the role of opinion deviance. J. Environ. Psychol. 63, additional information. 118–129. https://doi.org/10.1016/j.jenvp.2019.04.001. 9 S. Drews et al. Energy Policy 167 (2022) 113051 Baranzini, A., van den Bergh, J.C.J.M., Carattini, S., Howarth, R.B., Padilla, E., Roca, J., Krueger, J., Clement, R.W., 1994. The truly false consensus effect: an ineradicable and 2017. Carbon pricing in climate policy: seven reasons, complementary instruments, egocentric bias in social perception. J. Pers. Soc. Psychol. 67, 596–610. https://doi. and political economy considerations. Wiley Interdiscip. Rev.: Clim. Change 8 (4), org/10.1037/0022-3514.67.4.596. e462. https://doi.org/10.1002/wcc.462. Krueger, J., Clement, R.W., 1997. Estimates of social consensus by majorities and Baron, J., 1987. Second-order probabilities and belief functions. Theor. Decis. 23, 25–36. minorities: the case for social projection. Pers. Soc. Psychol. Rev. 1, 299–313. https://doi.org/10.1007/BF00127335. https://doi.org/10.1207/s15327957pspr0104_2. Barrera, D., Simpson, B., 2012. Much ado about deception: consequences of deceiving Kyselá, E., Ščasný, M., Zvěřinová, I., 2019. Attitudes toward climate change mitigation research participants in the social sciences. Socio. Methods Res. 41, 383–413. policies: a review of measures and a construct of policy attitudes. Clim. Pol. 19, https://doi.org/10.1177/0049124112452526. 878–892. https://doi.org/10.1080/14693062.2019.1611534. Bergquist, M., Nilsson, A., Schultz, W.P., 2019. A meta-analysis of field-experiments Levi, S., 2021. Why hate carbon taxes? Machine learning evidence on the roles of using social norms to promote pro-environmental behaviors. Global Environ. Change personal responsibility, trust, revenue recycling, and other factors across 23 59, 101941. https://doi.org/10.1016/j.gloenvcha.2019.101941. European countries. Energy Res. Social Sci. 73, 101883 https://doi.org/10.1016/j. Best, R., Burke, P.J., Jotzo, F., 2020. Carbon pricing efficacy: cross-country evidence. erss.2020.101883. Environ. Resour. Econ. 77, 69–94. https://doi.org/10.1007/s10640-020-00436-x. Leviston, Z., Walker, I., Morwinski, S., 2013. Your opinion on climate change might not Bolsen, T., Leeper, T.J., Shapiro, M.A., 2014. Doing what others do: norms, science, and be as common as you think. Nat. Clim. Change 3, 334–337. https://doi.org/ collective action on global warming. Am. Polit. Res. 42, 65–89. https://doi.org/ 10.1038/nclimate1743. 10.1177/1532673X13484173. Lorenz-Spreen, P., Lewandowsky, S., Sunstein, C.R., Hertwig, R., 2020. How behavioural Boykoff, M.T., 2013. Public enemy no. 1? Understanding media representations of outlier sciences can promote truth, autonomy and democratic discourse online. Nat. Human views on climate change. Am. Behav. Sci. 57 (6), 796–817. Behav. 1–8. https://doi.org/10.1038/s41562-020-0889-7. Carattini, S., Carvalho, M., Fankhauser, S., 2018. Overcoming public resistance to carbon Maestre-Andrés, S., Drews, S., van den Bergh, J., 2019. Perceived fairness and public taxes. Wiley Interdiscip. Rev.: Clim. Change 9, e531. https://doi.org/10.1002/ acceptability of carbon pricing: a review of the literature. Clim. Pol. 19, 1186–1204. wcc.531. https://doi.org/10.1080/14693062.2019.1639490. Cialdini, R.B., 2003. Crafting normative messages to protect the environment. Curr. Dir. Maestre-Andrés, S., Drews, S., Savin, I., van den Bergh, J., 2021. Carbon tax acceptability Psychol. Sci. 12, 105–109. with information provision and mixed revenue uses. Nat. Commun. 12, 7017. Colson, G., Corrigan, J.R., Grebitus, C., Loureiro, M.L., Rousu, M.C., 2016. Which https://doi.org/10.1038/s41467-021-27380-8. deceptive practices, if any, should Be allowed in experimental economics research? Mildenberger, M., Tingley, D., 2019. Beliefs about climate beliefs: the importance of Results from surveys of applied experimental economists and students. Am. J. Agric. second-order opinions for climate politics. Br. J. Polit. Sci. 1–29. https://doi.org/ Econ. 98, 610–621. https://doi.org/10.1093/ajae/aav067. 10.1017/S0007123417000321. de Bruin, W.B., Fischhoff, B., Millstein, S.G., Halpern-Felsher, B.L., 2000. Verbal and Millner, A., Ollivier, H., 2016. Beliefs, politics, and environmental policy. Review of. numerical expressions of probability: “it’s a fifty–fifty chance. Organ. Behav. Hum. Environ. Econ. Pol. 10, 226–244. https://doi.org/10.1093/reep/rew010. Decis. Process. 81, 115–131. https://doi.org/10.1006/obhd.1999.2868. Nyborg, K., Anderies, J.M., Dannenberg, A., Lindahl, T., Schill, C., Schlüter, M., de Groot, J.I.M., Schuitema, G., 2012. How to make the unpopular popular? Policy Adger, W.N., Arrow, K.J., Barrett, S., Carpenter, S., Chapin, F.S., Crépin, A.-S., characteristics, social norms and the acceptability of environmental policies. Daily, G., Ehrlich, P., Folke, C., Jager, W., Kautsky, N., Levin, S.A., Madsen, O.J., de Environ. Sci. Pol. 19–20, 100–107. https://doi.org/10.1016/j.envsci.2012.03.004. Zeeuw, A., 2016. Social norms as solutions. Science 354 (6308), 42–43. https://doi. Douenne, T., Fabre, A., 2020. French attitudes on climate change, carbon taxation and org/10.1126/science.aaf8317. other climate policies. Ecol. Econ. 169, 106496 https://doi.org/10.1016/j. OECD, 2018. Effective Carbon Rates 2018: Pricing Carbon Emissions through Taxes and ecolecon.2019.106496. Emissions Trading. OECD. https://doi.org/10.1787/9789264305304-en. Druckman, J.N., McGrath, M.C., 2019. The evidence for motivated reasoning in climate Pohjolainen, P., Kukkonen, I., Jokinen, P., Poortinga, W., Umit, R., 2018. Public change preference formation. Nat. Clim. Change 9, 111–119. https://doi.org/ Perception on Climate Change and Energy in Europe and Russia: Evidence from 10.1038/s41558-018-0360-1. Round 8 of the European Social Survey. Dvir-Gvirsman, S., 2015. Size matters: the effects of political orientation, majority status, Rabe, B.G., 2018. Can We Price Carbon? American and Comparative Environmental and majority size on misperceptions of public opinion. Publ. Opin. Q. 79, 1–27. Policy. MIT Press, Cambridge, MA. https://doi.org/10.1093/poq/nfu061. Rains, S.A., 2013. The nature of psychological reactance revisited: a meta-analytic Edenhofer, O., Jakob, M., Creutzig, F., Flachsland, C., Fuss, S., Kowarsch, M., review. Hum. Commun. Res. 39, 47–73. https://doi.org/10.1111/j.1468- Lessmann, K., Mattauch, L., Siegmeier, J., Steckel, J.C., 2015. Closing the emission 2958.2012.01443.x. price gap. Global Environ. Change 31, 132–143. https://doi.org/10.1016/j. Rhodes, E., Axsen, J., Jaccard, M., 2014. Does effective climate policy require well- gloenvcha.2015.01.003. informed citizen support? Global Environ. Change 29, 92–104. https://doi.org/ Farrow, K., Grolleau, G., Ibanez, L., 2017. Social norms and pro-environmental behavior: 10.1016/j.gloenvcha.2014.09.001. a review of the evidence. Ecol. Econ. 140, 1–13. https://doi.org/10.1016/j. Rohring, E.J., Akerlof, K.L., 2020. Perceptions of social consensus at the regional level ecolecon.2017.04.017. relate to prioritization and support of climate policy in Maryland, USA. Reg. Environ. Geiger, N., Swim, J.K., 2016. Climate of silence: pluralistic ignorance as a barrier to Change 20, 72. https://doi.org/10.1007/s10113-020-01652-3. climate change discussion. J. Environ. Psychol. 47, 79–90. https://doi.org/10.1016/ Ross, L., Greene, D., House, P., 1977. The “false consensus effect”: an egocentric bias in j.jenvp.2016.05.002. social perception and attribution processes. J. Exp. Soc. Psychol. 13, 279–301. Goldberg, M.H., van der Linden, S., Leiserowitz, A., Maibach, E., 2019. Perceived social https://doi.org/10.1016/0022-1031(77)90049-X. consensus can reduce ideological biases on climate change. Environ. Behav. 52 (5), Savin, I., Drews, S., Maestre-Andrés, S., et al., 2020. Public views on carbon taxation and 495–517. https://doi.org/10.1177/0013916519853302. its fairness: a computational-linguistics analysis. Climatic Change 162, 2107–2138. Hammar, H., Jagers, S.C., 2007. What is a fair CO2 tax increase? On fair emission https://doi.org/10.1007/s10584-020-02842-y. reductions in the transport sector. Ecol. Econ. 61, 377–387. https://doi.org/ Schuldt, J.P., Yuan, Y.C., Song, Y., Liu, K., 2019. Beliefs about whose beliefs? Second- 10.1016/j.ecolecon.2006.03.004. order beliefs and support for China’s coal-to-gas policy. J. Environ. Psychol. 66, Hardisty, D.J., Beall, A.T., Lubowski, R., Petsonk, A., Romero-Canyas, R., 2019. A carbon 101367 https://doi.org/10.1016/j.jenvp.2019.101367. price by another name may seem sweeter: consumers prefer upstream offsets to Shalizi, C.R., Thomas, A.C., 2011. Homophily and contagion are generically confounded downstream taxes. J. Environ. Psychol. 66, 101342 https://doi.org/10.1016/j. in observational social network studies. Socio. Methods Res. 40, 211–239. https:// jenvp.2019.101342. doi.org/10.1177/0049124111404820. Jenkins, J.D., 2014. Political economy constraints on carbon pricing policies: what are Sokoloski, R., Markowitz, E.M., Bidwell, D., 2018. Public estimates of support for the implications for economic efficiency, environmental efficacy, and climate policy offshore wind energy: false consensus, pluralistic ignorance, and partisan effects. design? Energy Pol. 69, 467–477. https://doi.org/10.1016/j.enpol.2014.02.003. Energy Pol. 112, 45–55. https://doi.org/10.1016/j.enpol.2017.10.005. Kinzig, A.P., Arrow, K.J., Ehrlich, P.R., 2013. Social norms and global environmental Van Boven, L., Ehret, P.J., Sherman, D.K., 2018. Psychological barriers to bipartisan challenges: the complex interaction of behaviors, values, and policy. Bioscience 63, public support for climate policy. Perspect. Psychol. Sci. 13, 492–507. https://doi. 164–175. https://doi.org/10.1525/bio.2013.63.3.5. org/10.1177/1745691617748966. Kjeldahl, E.M., Hendricks, V.F., 2018. The sense of social influence: pluralistic ignorance Williams, H.T.P., McMurray, J.R., Kurz, T., Hugo Lambert, F., 2015. Network analysis in climate change. EMBO Rep. 19 (11), e47185. reveals open forums and echo chambers in social media discussions of climate Klenert, D., Mattauch, L., Combet, E., Edenhofer, O., Hepburn, C., Rafaty, R., Stern, N., change. Global Environ. Change 32, 126–138. https://doi.org/10.1016/j. 2018. Making carbon pricing work for citizens. Nat. Clim. Change 8, 669–677. gloenvcha.2015.03.006. https://doi.org/10.1038/s41558-018-0201-2. World Bank, 2019. Carbon Pricing Dashboard [WWW Document]. URL. World Bank, Koenker, R., Bassett, G., 1978. Regression quantiles. Econometrica 46, 33–50. 7.1.19. https://carbonpricingdashboard.worldbank.org/. Konc, T., Savin, I., 2019. Social reinforcement with weighted interactions. Phys. Rev. E Yu, R., Chen, L., 2015. The need to control for regression to the mean in social 100, 22305. https://doi.org/10.1103/PhysRevE.100.022305. psychology studies. Front. Psychol. 5, 1574. https://doi.org/10.3389/ Krawczyk, M., 2019. What should be regarded as deception in experimental economics? fpsyg.2014.01574. Evidence from a survey of researchers and subjects. J. Behav. Exp. Econ. 79, 110–118. https://doi.org/10.1016/j.socec.2019.01.008. 10