Energy Economics 125 (2023) 106815 Contents lists available at ScienceDirect Energy Economics journal homepage: www.elsevier.com/locate/eneeco A tragedy of the horizons? An intertemporal perspective on public support for carbon taxes Mariateresa Silvi a, Emilio Padilla Rosa b, * a Department of Economics, Ghent University, 9000 Ghent, Belgium b Department of Applied Economics, Autonomous University of Barcelona, 08193 Bellaterra, Spain A R T I C L E I N F O A B S T R A C T Keywords: Most collective dilemmas—that is, situations in which private interests contrast with collective interests—have Carbon tax an embedded intertemporal component in that they often imply that the rewards from defection are immediate Public support but the rewards from cooperation are delayed and often accrue to people in the future. This also applies to carbon Temporal preferences taxes since they imply additional individual costs for benefits which will mostly be enjoyed by future genera­ Discounting Pre-commitment mechanisms tions, which undermines their political support. In an experiment on a representative sample of 1000 United States adults, we presented individuals with twelve alternative carbon tax formulations with varying start dates, temporal horizons of carbon abatement objectives and revenue uses. We find that public support is highest when individuals can pre-commit to policies that start a few years into the future and for policies that express their emission cuts objectives in more distant and ambitious terms —i.e. achieving carbon neutrality by 2050 as opposed to halving emissions by 2030. Individual temporal discounting, exogenously measured, account for large part of these preferences. These preferences are in contrast with the most efficient policy, which is the one that starts immediately and distributes equitably mitigation costs across time. We find two ways to realign preferences with it. First, the most efficient policy becomes politically feasible when the tax includes a dividend that is redistributed to citizens. Delivering an economic compensation at the same time of the individual costs of the tax neutralizes the effect of individual discounting and of the policy’s temporal context on tax support. Secondly, when the price of the carbon tax is adjusted upward to compensate for the opportunity cost of delaying its introduction, individuals start trading off its delay with avoidance of tax increases. 1. Introduction mostly be enjoyed by future generations, which undermines political support in favor of climate regulation (Gollier, 2020). The temporal Sixty years ago, Garret Hardin connoted the term the Tragedy of the distribution of the environmental future benefits and of the costs it Commons to indicate that exploiting a common resource beyond its imposes naturally influences choices and the likelihood of achieving regeneration capacity provides private payoffs whereas the benefits of successful cooperation through individual climate mitigation efforts in preserving the resource are shared (Hardin, 1968), which leads to its real-world social dilemmas. To achieve successful cooperation in inter­ overexploitation. While Ostrom (1990) demonstrated that with moni­ temporal environments, it is crucial to understand: a) the influence that toring and sanctioning systems in place the commons can be protected, the temporal context of the decision has on decision outcomes; b) the the term has stayed to characterize open-access common-pool resources, role individual time preferences and altruism play in the decision to such as the atmosphere or many fisheries (Libecap, 2008). But there is cooperate; and c) the interplay between individual temporal preferences also a Tragedy of the Horizons hampering the protection of the com­ and the temporal context. In this paper, we focus on the public accept­ mons (Gollier, 2020). The fruits from the sacrifices individuals and or­ ability of a carbon tax and explore a new dimension in the literature: ganizations could undertake today to protect resources will mostly be What is the temporal formulation of a carbon tax that maximizes its enjoyed in the very long term. This is particularly relevant in the context acceptability? of natural resources preservation and to the climate commons since Carbon pricing is commonly regarded by economists as the most individuals must sacrifice current wellbeing for benefits which will efficient and effective climate mitigation policy (either through carbon * Corresponding author. E-mail addresses: mariateresa.silvi@ugent.be (M. Silvi), emilio.padilla@uab.es (E. Padilla Rosa). https://doi.org/10.1016/j.eneco.2023.106815 Received 9 January 2023; Received in revised form 6 June 2023; Accepted 16 June 2023 Available online 22 June 2023 0140-9883/© 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/). M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 taxes or emission trading systems; Baranzini et al., 2017). However, shifting preferences however comes at a future welfare loss, as the costs most world emissions are still not affected by a carbon pricing instru­ associated with climate mitigation increase with each year of inaction. ment. The reluctance of politicians to adopt carbon taxes has been This highlights a potential trade-off between the distribution of miti­ linked, among other things, to their unpopularity among voters, who gation costs that is most efficient and the one that is most politically notably perceive them as too coercive in comparison to “pull measures” feasible. This contrast can however be addressed with appropriate policy such as subsidies. This paper investigates whether the temporal con­ design and communication. When the opportunity cost of delaying the textualization of carbon taxes can be adjusted to make them more tax is reflected within the policy proposal —i.e. the carbon price is popular. While seeking to optimize a policy’s temporal context to the adjusted upward to compensate for the delayed introduction— in­ public’s temporal preferences, one must not forget the temporal horizon dividuals start trading-off the postponement of the tax with its increased of biophysical processes. The ambition of maintaining the temperature price and they no longer prefer to indefinitely postpone the tax. We also objective of 2◦ C depends on a fixed carbon budget constraint. The find that redistributing revenues as dividends to citizens counteracts this optimal timing to consume this remaining carbon budget in an inter­ preference for postponing the policy, as this diminishes its immediate temporally efficient manner has been analyzed by Gollier (2022). Ac­ personal cost. More impatient individuals are less likely to support the cording to the author, the distribution of abatement costs recommended tax and favor policies that start 6 years in the future. Receiving cash- by the 5th IPCC report models shifts too much of the effort to the future, back payments from the tax proceeds is also effective in neutralizing through carbon pricing proposals that start too cheap and increase too the negative effect of temporal preferences on tax acceptance. In line steeply. He attributes this intertemporal imbalance to the political with previous research, we find that tax support decreases with its cost. unfeasibility of a high carbon price and to the unverifiable assumption Our analysis shows that tax levels compatible with a high probability of that low mitigation technologies will be available in the future. reaching the Paris’ agreement climate targets —i.e. US$80/tCO2— Reportedly this causes high welfare losses. Thus, after investigating the become politically feasible if the tax includes cash-back payments. temporal contextualization that maximizes public support, we then Promising cash-back payments to citizens is also effective in softening addressed the parallel question: what makes the most efficient policy, the opposition from citizens with conservative political views, climate- also politically feasible? change deniers and citizens who distrust public institutions. We undertook a hypothetical choice experiment in which a nation­ The rest of the paper is articulated as follows. Section 2 introduces ally representative sample of 1000 United States (US) adults voted in the state of the relevant literature on intertemporal choice and on the favor or against the introduction of alternative carbon tax designs which acceptability of carbon taxes. It also provides context to legislation and were characterized by different introduction dates (immediate, three or public opinion on carbon taxes in the US. Section 3 describes the six years delay) and different temporal horizons (emission cuts are experimental design and how the survey and the analysis were struc­ expressed either in 2030 or 2050 objectives). Additional policy attri­ tured. Section 4 presents the results through qualitative and econometric butes included revenue use, percentage of expected CO2 emission cuts, analyses. Section 5 summarizes our key findings and conclusions. taxation levels, and the estimated cost and revenues the tax may impose on households. To increase the salience of the experiment and its in­ 2. Background ternal validity, we customized choice cards within the survey, rather than just showing average cost estimates, with ad-hoc calculations of the 2.1. Literature review economic impact of the policy based on each participant’s declared in­ come and the consumption profile of the average person in her/his in­ 2.1.1. Intertemporal choice come quintile. We separately measured individual temporal preferences The temporal dimensions of collective dilemmas and of the accept­ using a multi-pricelist experiment and analyzed whether or not more ability of carbon pricing are surprisingly unexplored.1 Individual dis­ present-biased individuals are more likely to accept the tax in any given count rates and a resource pool’s temporal horizon have been identified scenario. This allows us to disentangle the effects of temporal strategy as temporal dimensions that should theoretically influence cooperation from individual temporal preferences; for example, impatient in­ in resource pool dilemmas (Hendrickx et al., 2001). However, empirical dividuals may prefer earlier rewards and to postpone sacrifices, whereas evidence to date on the influence of these factors on the likelihood of more future-oriented individuals or altruistic individuals may be more cooperation is fragmentary. Earlier research has highlighted that indi­ eager to commit to larger later rewards which are shared in the future vidual discounting correlates negatively with cooperation, suggesting between themselves and the next generations. With policy relevance in that the temptation to free ride may be stronger for more impatient mind, we include carbon taxation prices that are deemed to be in the individuals (Mannix, 1991; Curry et al., 2008; Fehr and Leibbrandt, range of compatibility with the Paris Agreement objectives of main­ 2011). Research that has analyzed the influence of the temporal horizon taining global warming well below 2◦ C compared to pre-industrial of choice highlights that the longer people feel invested in a community, levels. Additionally, we look at the impact of different temporal con­ the more they will be willing to cooperate for its common good. For textualizations on population groups that are found to be more opposed instance, Mannix and Loewenstein (1993) found that managers who to the tax and that have at least one of the following characteristics: anticipated that they would stay longer in the same role would climate change deniers, individuals with conservative political values, contribute more to a public organizational fund. Additional experi­ individuals who distrust public institutions and who reject any personal mental research has shown that cooperation becomes more unlikely as responsibility to care for the environment. the gains from cooperation are shifted further away in time (Jacquet We find that it is possible to optimize the temporal distribution of the et al., 2013). Arora et al. (2012) found that when the consequences of a carbon tax’s costs and benefits to the public’s preferences. Public sup­ decision are shifted to the future—such as whether or not to donate to an port for a carbon tax increases substantially if its introduction and its investment cooperative if the payment is delayed by six month­ associated economic costs are shifted forward by 3 and 6 years. Ceteris s—individuals tend to show higher cooperation compared to when they paribus, approval rates increase by 7 and 9 p.p. respectively, compared to when the tax is introduced immediately. Expressing the emission cuts objectives of the policy in more distant and ambitious terms —i.e. 1 We acknowledge that, while accepting to pay an environmental tax achieving carbon neutrality by 2050 as opposed to halving emissions by certainly implies a cooperative attitude, we cannot exclude that tax opposition 2030— increases public support, however this policy attribute is sta­ may not necessarily relate to unwillingness to avert climate change. Tax op­ tistically insignificant once other policy attributes are controlled for. position may be motivated by tax avoidance more in general or skepticism This suggests it may increase support in combination with other policy towards the effectiveness of carbon taxes as a mitigation strategy. To control for attributes but not on its own. A policy that caters to the public’s cost this possibility we calculate an index of individual trust in political institutions. 2 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 face the immediate consequences of their choices. In addition, pre­ forms of carbon pricing programs have been independently adopted by commitment mechanisms are one of the most known mechanisms to 13 of its 50 States, starting with the Regional Greenhouse Gas Initiative induce individuals to avoid present bias and invest in their future in 2005 (which involved 11 states on the northeast coast) and followed wellbeing both academically and in the established practice, for by California’s cap and trade program in 2013 and the State of Wash­ instance, by enrolling in saving schemes (Thaler and Benartzi, 2004; ington’s Climate Commitment Act, which started in 2023 (Center for Benartzi and Thaler, 2013). Climate and Energy Solutions, 2021). The available evidence thus suggests that the default temporal At the federal level, five concurrent carbon pricing designs have been setting in which the costs of cooperation are immediate, and gains are introduced in the 117th Congress between 2019 and the first half of delayed undermines the possibility of cooperation. Can asking people to 2021. They vary in terms of emissions covered, taxation level, and commit to a carbon tax that starts a few years from now increase its revenue use. The increasing presence of carbon pricing in the political public support by shifting the costs it imposes to households in the debate is mirrored by an increase in public debate. Although wording future? Are individuals more likely to accept it if the gains from coop­ differs between opinion polls, one can observe a steady increase in eration—in terms of future carbon emissions cuts—are expressed by a support for these policies. The percentage of Americans who reportedly more proximal date? Do present- and future-focused individuals have support the introduction of a carbon tax stood at 36% in 2009 and it has similar preferences regarding the temporal distribution of the sacrifices gradually increased to 50% in 2016 (Puskin and Mills, 2017). The and rewards of a carbon tax? The extent to which the temporal distri­ highest support has been registered in 2020, with 68% of registered bution of costs and gains can be manipulated to increase cooperation has voters in support of requiring fossil fuel companies to pay carbon taxes not been systematically studied; nor are the interactions between indi­ (Leiserowitz et al., 2021). In addition, Kotchen et al. (2017) analyze vidual time preferences and the contextual temporal setting known. American households’ willingness to pay for a carbon tax through a series of nationally representative surveys: they find a net increase from 2.1.2. Carbon tax acceptance an average US$ 85 per year in 2011, to US$ 177 in 2016. Despite these The literature on the acceptability of carbon taxes has analyzed the percentages carbon taxes in the US are still controversial, and polls impact of individual characteristics, contextual factors, and different seemingly overestimate actual support for these policies (Anderson policy designs on various geographical settings, mainly through stated et al., 2023). Support for a carbon tax is a politically divisive topic, with preferences approaches (for a review, see Drews and van den Bergh, two-thirds of Democrats in favor of it and just 30% of Republicans in 2016). Common findings that are relevant to our study indicate that support of it when the revenue use is not mentioned. The political people tend to overestimate carbon tax costs and underestimate their divide, however, becomes narrower and overall support increases when environmental effects, that tax acceptability decreases with the personal tax revenues are earmarked to support further mitigation activities or to costs associated with it, and that tax resistance decreases after people lower other taxes (to 45% of Republicans and up to 77% of Democrats; have gained experience with the tax (Cherry et al., 2014; Bergquist et al., Leiserowitz et al., 2021). 2022; Douenne and Fabre, 2022). Several studies and opinion polls have Given the increasing political debate about carbon taxes and the indicated a marked preference for the earmarking of revenues: in­ consequent relevance and familiarity of the subject to Americans, the US dividuals are more favorable to carbon taxes when revenue use is offers an ideal landscape in which to test the acceptability of different destined to finance other environmental mitigation activities, to reduce carbon tax formulations. The concrete carbon tax designs to be discussed other taxes, or for redistributions in the form of cash-back payments as in the current US Congress enable us to replicate some key features of opposed to when the use of the revenues is not specified (Bachus et al., some of these designs. Concurrently with our core stated objective of 2019). Support for carbon taxes also depends on the level of trust in testing the relevance of the temporal context in the acceptance of a political institutions (Savin et al., 2020) and knowledge about climate carbon tax, this enables us to provide information on the political change (Douenne and Fabre, 2022). feasibility of some of the current policy designs in the US. Furthermore, To date, only two experiments have looked into the relationship the availability of yearly polling data on this topic allows us to make between acceptability and the temporal context of an environmental tax. comparisons with our experimental results. While not all features of the In a laboratory experiment with 218 students, Tiezzi and Xiao (2016) US population may be generalizable to the rest of the world, the ten­ undertook an intertemporal market experiment based on auctions of an dency to discount the future and hyperbolic discounting are a charac­ externality-producing good. They surveyed how support for the intro­ teristic human trait that has been observed worldwide (Wang et al., duction of an environmental tax to account for the externality changed 2016). Furthermore, temporal preferences in the US are highly proximal as a function of the timing of the externality. They found that environ­ to the rest of the Anglo/American world but also to other cultural mental taxes were less likely to be approved when the externality ma­ clusters, the German/Nordic countries, the Middle East, Asia and terializes after the good that produces them has been traded and Eastern Europe (Wang et al., 2016). As such, the US may offer useful consumed. They also concluded that in real life, individuals are equally insights on how the temporal framing of a policy and temporal prefer­ less likely to support environmental taxes when the externality they aim ences could affect other contexts. to reduce is not readily visible to them; for example, the benefit of consuming gasoline is experienced immediately whereas its externality, 3. Methods that is, pollution, accumulates over time. Bechtel et al. (2020) find experimentally that when given the option of choosing between carbon 3.1. Experimental design tax rates that are constant or that increase over time, individuals prefer constant rates. Our paper relates to this literature and integrates it, in The experiment is designed as an online survey in which a nationally that we show how political support for a carbon tax is affected by the representative sample of the US population expresses how they would perspective of a delayed introduction that incorporates the carbon vote on a national referendum in support of the introduction of a tax on budget constraint, through higher starting prices. CO2 emissions. The full survey is available in the Annex. Our survey was developed using the survey software Qualtrics and it was administered 2.2. The national context in the United States through Prolific Academic. Prolific Academic is a widely used panel for academic research that allows sampling of representative members of The US is a signatory of the Paris Climate Agreement and has the US population. It has also been recommended for delivering higher committed to reaching net zero emissions economy-wide by 2050. In quality data compared to alternative platforms (Palan and Schitter, addition, it has pledged to cut its national greenhouse gas emissions by 2018; Peer et al., 2017). The experiment aims to investigate whether 50–52% by 2030 compared to 2005 (The White House, 2021). Different support for this policy is sensitive to changes in the start year of the 3 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 Table 1 (Bishop and Boyle, 2019). As Arrow et al. (1993) observed, referenda on Treatments (combinations). the provision of public goods are not uncommon, and various US states Avoided CO2 emissions —including Massachusetts, Vermont, New Hampshire and Washington By 2030 By 2050 State— have cast referendums or included non-binding questions on the CB BC CB BC introduction of carbon taxes in their electoral ballots. This context also Immediate (2021) 1 2 7 8 makes the study incentive compatible, meaning it is a context in which Policy start date In three years (2024) 3 4 9 10 the optimal strategy for the participant is to give a true representation of In six years (2027) 5 6 11 12 her/his preferences. In particular, participants have an incentive to CB=Cashback, revenues as payments to citizens, BC=Basecase, revenues to the general budget. truthfully reflect their preferences in the survey if (a) they see the survey results as an opportunity to influence business or government actions and (b) they care about the outcomes of those actions (Carson and policy and its temporal horizon. We thus created six temporal scenarios, Groves, 2007). To recall this incentive, and similarly to Carattini et al. identified by all possible combinations between three alternative policy (2019), we informed participants that the survey results could poten­ start dates and two alternative temporal horizons by which the CO2 tially generate insights for the government with respect to the viability emissions cuts are expressed. These six scenarios are tested twice, once of adopting a CO2 tax. We assumed that most adults care about either or with unspecified revenue use and once with revenue use earmarked as both of the issues of global warming and an increase in their annual cashback payments to constituents. The resulting 12 possible combina­ costs. tions define the treatment groups described in Table 1. A novel feature of our survey design is that it introduced a person­ In summary, the proposed experiment follows a 3x2x2 factorial be­ alized estimate of the potential cost and eventual carbon dividend tween subject design which identifies 12 treatments across 3 policy payment to which the individual household would be entitled based on attributes. the income declared by the participant, the consumption profile of the average person in her/his income quintile and the tax level.5 This design i) One factor pertains to the immediacy of costs and of the eventual feature enabled us to elicit preferences that were based on more accurate cashback payments (policy start date set to 2021, 2024, or 2027). economic expectations and, thus, were closer to predicting the actual ii) One factor pertains to the immediacy with which the environ­ reactions that individuals would have in real life when facing the eco­ mental rewards are presented (emissions abatements targets are nomic consequences of the tax. To increase the policy relevance of our expressed by either a 2030 or a 2050 goal). estimations we model the revenue distributed to individual households iii) One factor pertains to whether revenue use is either unspecified using as a reference an actual carbon fee and dividend policy, currently (Basecase) or earmarked for redistribution (Cashback).2 deposited in the US Congress, namely the Energy Innovation and Carbon Dividend Act of 2021 (H.R. 2307). It is the one which currently enjoys The experiment follows a split sample design to avoid anchoring the biggest support from legislators (highest number of co-sponsors) (US effects.3 One respondent is assigned to just one out of the 12 combina­ Congress, 2021). tions available. Participants answer a sequence of questions consisting of Concerns about internal validity also guided the number of choice four choice tasks4 framed within the same temporal context but with options shown in each choice set. In similar, related studies, some re­ progressively increasing tax levels which are reflected not only in higher searchers have chosen to include three or more choice alternatives as a costs imposed on the average household but also in higher carbon way to maximize the amount of information gathered from a small emissions abatement targets to be achieved by the tax. To minimize sample; that is, to elicit both the likelihood that a policy would be sequencing effects, we disclose at the beginning of the survey the accepted against the status quo and the relative preference between number of choice sets with which participants will be presented (Bate­ different policy alternatives. In contrast, we chose to present partici­ man et al., 2004). pants with just two choice options—the status quo and one policy alternative—for three reasons: 3.1.1. Internal validity and hypothetical bias The survey design takes several steps to ensure its internal validity. a) According to the mechanism design theory by Gibbard (1973) and In line with the literature (see for example Carattini et al., 2019; Kotchen Satterthwaite (1975), only binary response formats can be incentive et al., 2017), we use a referendum format. Practitioners, including the compatible and thus induce participants to realistically represent National Oceanic and Atmospheric Administration (NOAA) panel on their choices.6 contingent valuation, recommend the use of this format to make the b) The number of alternative options included in a choice set has been choice more realistic (Bishop and Boyle, 2019). Five studies which found to affect answers in several convergent validity studies (see a compared the outcome of contingent valuation surveys formulated as review in Weng et al., 2021). Measured willingness to pay in referenda found that they performed well in comparison to actual voting contingent valuation surveys and response rates vary depending on whether respondents reply to a binary-choice question in which one option is the status quo and the alternative is the introduction of an 2 In this paper we include only two of the potential revenue uses discussed in environmental policy; or they reply to a multiple-choice question the literature: revenue which is conferred back to the State’s general budget and revenue which is redistributed to citizens. Preferences for the different types of revenue recycling schemes have been extensively addressed in the literature 5 (Drews & Van den Bergh, 2015; Dechezleprêtre et al., 2022; Carattini et al., Similarly, Douenne and Fabre (2022) calculated the estimated economic 2019; Douenne and Fabre, 2022). In this experiment we choose to restrict the impact on each household, however their approach differs from ours in that uses of revenues to just two of the available alternatives to contain the already rather than communicating the specific amounts to respondents, they inform high number of treatments. each household that they may have an 80% chance of losing (or winning) from 3 Anchoring effects refer to the possibility that cognitive processes may be the policy based on their own estimations. 6 biased by the first pieces of information displayed to individuals. Exposing the For example, when participants are presented with more policy alternatives, same respondent to several temporal frames in a sequential order, would have they may infer that there is uncertainty about the cost of a given policy, and potentially introduced an influence of the first temporal contextualization on they may be induced not to reveal their true preferences (e.g. they may try to the following choice sets. induce the government to supply the good at the lowest price, even if their 4 Based on Louviere et al. (2000) who recommend not to exceed choices to 4 actual willingness to pay was higher than they indicate) (Carson and Groves, questions. 2007). 4 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 format in which status quo and a number n > 1 of alternative policies Table 2 are considered. Recent empirical (Weng et al., 2021) and theoretical Policy attributes and levels. (Carson and Groves, 2007) studies and the NOAA panel on contin­ Attributes Status Quo Levels gent valuation recommend a binary elicitation format to reduce 1. Policy start date never 2021, 2024, 2027 anchoring and other sources of bias typical of multiple-choice 2. Tax level 0 US Tax levels for policy start in 2021: 15, questions (Arrow et al., 1993). (not shown on choice cards) $/tCO2 40, 60, 80 US$/tCO2 (three are c) A survey that reproposes the choice elicitation method of the context compatible with the Paris Agreement it wants to mimic is more relatable to participants, and their results objectives, although with differing degrees of probability, and one is should be more realistic and credible (Carson and Groves, 2007). potentially below). Since actual referenda use a binary format, it seems logical to Tax levels adjusted for policy start reproduce it in our survey. delayed to 2024: 50, 67, 75, 100 US $/tCO2 Tax levels adjusted for policy start We, therefore, expect the binary elicitation format to be the one that delayed to 2027: 62, 93, 123, 130 US records stated preferences which are closer to real-world choices. To $/tCO2 further minimize the risk of hypothetical bias and in accordance with 3. Cost of the tax for the 0 US The value is calculated within the best practices, we added a follow-up question after the experiment, household $/year survey and it depends on the tax level asking respondents to rate how certain they were of their responses as a and the household income declared by the participant. It ranges between measure of the questions’ internal validity (Johannesson et al., 1999; 0.01 and 1400 US$/year. Brouwer et al., 2010). We also gave respondents the possibility to 4. Revenue of the tax for the 0 US The value is calculated within the answer “Don’t Know” to the experimental choices. This enables us to household $/year survey and it depends on the tax level calibrate the analysis towards greater accuracy. (Cashback condition only) and the household income declared by the participant. It ranges between Prior to the survey, we provided a simple introduction of the link 0.07 and 773 US$/year. between the use of fossil fuels, CO2 emissions, and global warming as 5. % decrease in national • 14% Depends on the tax level and the well as an explanation of how a carbon tax could address the problem. emissions (2030) temporal horizon considered: We also reminded participants of the expected temperature increase in (compared to 2005 levels) • 25% the US should global greenhouse gas emissions continue on the current (2050) • 20–75% (2030) • 60–130% (2050) path, and we informed them of the national commitment the US gov­ 6. Temporal horizon by 2030, 2030, 2050 ernment has made to decrease its emission trajectory. This introduction which emissions cuts are 2050 reflects the information that would likely be included in a referendum expressed and mirrors similar surveys in the literature. Following the NOAA’s recommendations, we also reminded individuals of their budget con­ scenarios are usually calculated using 2020 as a baseline and they then straints and of expenses they may have to forego to pay for the policy’s gradually increase. The Energy and Dividend Act bill is also set to in­ costs. This helps participants putting the policy’s costs in relation to crease the tax by a minimum of US$10 per year. For the scenarios in their actual income, thereby potentially reducing hypothetical bias which the tax introduction is delayed to 2024 or 2027, the starting tax (Bishop and Boyle, 2019). level is higher to account for this planned yearly increase and is further augmented to make up for the years of missed emissions cuts and, 3.1.2. Policy attributes consequently, a narrower carbon budget. More specifically, we Table 2 summarizes the policy attributes used to describe the tax and increased the planned starting fee by 3.9 p.p. for every year of delay.7 their consequences on the different scenarios. Each attribute has several The resulting tax fees for the delayed scenarios are reported in Table 2. levels, and attributes are interlinked with each other. We considered In order to reduce the cognitive burden on participants, we decided three different Policy Start Dates with three years in between them, 2021 against displaying tax levels on the choice cards, since such information (immediate), 2024, and 2027. By including three-year delays, we is anyway reflected in the personalized costs and revenue estimates. A modeled for a date which is far enough in the future in the constituents’ similar approach was followed by Douenne and Fabre (2022) and minds yet still possibly within a government’s electoral mandate. By Dechezleprêtre et al. (2022). At the start of the survey, participants were including a 6-year delay scenario, we evaluated the possibility for a required to insert their household annual income, and this information government to introduce a tax that will only be effective in the next was used to compute in real time the estimated cost and revenue that the political mandate. participant would likely face in real life as a consequence of the tax We defined Tax Levels expressed in US$/tCO2 that are either below or through a method further detailed below. within the price range that is considered necessary to reach the Paris Cost of the tax to households. Our computations of the costs and rev­ Treaty objectives of staying within 1.5–2◦ C of global warming by the enues to specific households are based on a study (Ummel, 2016) which end of this century. The estimation of the tax levels needed to achieve focused on the distributional implications on American households of these scenarios has been the subject of several modeling exercises. These exercises have estimations that vary widely and depend on a variety of factors—such as different baseline assumptions and a wide range of pessimist/optimist expectations regarding technological improvements 7 This estimate is based on Furman et al. (2015), who conducted a meta- and policy contexts. In defining taxation levels for the experiment, we analysis of 16 integrated climate mitigation models and found that, on relied on the Stiglitz-Stern review of existing carbon price estimates and, average, a decade of delay in the implementation of climate mitigation policy in line with their analysis, we considered the explicit carbon price reflects in an average increase in abatement costs by 39%. An alternative es­ consistent with Paris temperature targets to be between US$40–80/tCO2 timate by the Brookings Institute finds that delaying the introduction of a carbon tax at US$ 15 in the US economy would require on average a total in­ by 2020 and US$50–100/tCO2 by 2030 (High-Level Commission on crease in the tax by nearly 9% per year of delay to achieve the same cumulative Carbon Prices, 2017). We also included a tax equal to US$15/tCO2, emissions were the tax to start immediately (Mckibbin et al., 2014). We which is the taxation level proposed in the Energy and Dividend Act bill. intentionally used the most conservative estimate to detect whether even small More specifically, we considered a carbon tax at US$ 15, 40, 60, and 80 tax adjustments would affect support for the tax. per tCO2 for policies starting in 2021. We considered that the range in the level of taxation so defined should be the most relevant for policy­ makers. Estimates of tax levels compatible with global temperature 5 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 the previously mentioned Carbon Dividend Act.8 Using a simulation of Fig. 1 includes screenshots of how the survey looked like for an in­ household-level effects (from the American Community Survey and the dividual assigned to the Basecase condition (a) and Cashback condition Consumer Expenditure Survey) and input–output data from the Bureau (b). The introductory wording of our survey was partly borrowed from of Economic Analysis, the study calculated the yearly financial impact the National Surveys on Energy and Environment (2016) and from that a US$15/tCO2 tax would have on the average American family in Carattini et al. (2019) and adapted to meet our ends. each income quintile. We used this study as a benchmark to generate within the survey a personalized estimate of cost implications for each 3.2. Post-experiment survey participant, using a) his/her household income, b) the average con­ sumption of people in the same income quintile, and c) the tax level 3.2.1. Individual time preferences considered on the choice card.9 The expected personalized impact of the We used a common approach in the literature to elicit individual policy as shown in the choice card is thus based on the consumption time preferences which consisted of a multi-pricelist choice task as in patterns and characteristics of the average representative household of Andersen et al. (2006) and Coller and Williams (1999). Respondents the participant’s income quintile, adjusted for her/his income. The costs repeatedly chose between a smaller–sooner (option A) and a larger–later assigned to each tax level were internally consistent—higher tax levels payment (option B), as shown in Fig. 2. The more immediate payment corresponded to higher costs to the household.10 The same study also was set to US$500 immediately, whereas the latter reward was US$500 defined the potential revenues that would be redistributed to families + X, available in 1 year. X is a positive number that increases with each based on their income quintile. We thus used these calculations to iteration. The iterated choices are organized in increasing order as compute the cashback revenue to which the participant household could typically done in the time preference literature. X is calculated using an potentially be entitled in the Cashback condition. The range of values of annual rate of return between 5 and 50% on the principal and the the costs and revenues from the tax are included in Table 2. magnitude of the rewards match those typically used in intertemporal We then described the common benefits each policy alternative choice experiments (Andersen et al., 2006). Interest is compounded supposedly delivers in terms of Percentage Change in the National CO2 monthly based on major US bank’s business practices. The point at Emissions with respect to 2005 levels.11 These emissions reduction ob­ which individuals switch from choosing the more immediate payoff to jectives were estimated based on the tax level and were expressed under preferring the future income identifies the individual’s discount rate. different Temporal Horizons—that is, the emission change that could be Falk et al. (2018) show that this intertemporal choice task is an accurate reached by either 2030 or by 2050. To associate tax levels with their predictor of time discounting, even when the choices are hypothetical. potential to reduce emissions we retrieved data on the absolute values To increase the precision of our estimates, we discarded inconsistent by which US CO2 emissions should decrease compatibly with different responses from our final sample (i.e. respondents repeatedly switching global warming scenarios. These data were taken from the Climate Ac­ between the sooner and the later amounts as the later reward increased). tion Tracker (2021), which also includes estimates of national emission paths under current scenarios.12 For the purpose of our survey design we 3.2.2. Additional data associate their range of projections progressively, by matching In addition, we recorded data on: increasing tax levels with increasing emission cuts,13 while keeping the Stiglitz-Stern estimate as a benchmark (the explicit carbon price iIncome consistent with Paris temperature targets) (High-Level Commission on iiEducation level Carbon Prices, 2017). iii Political identity iv Area classification: Rural, town, metropolitan v Environmental values 8 vi Climate change beliefs We choose this study, since it is, at time writing, the only impact assessment vii Social expectations regarding other’s willingness to pay for the of the distributive implications of the Carbon Dividend Act. Although the study was commissioned by Climate Citizens Lobby (CCL), who advocate for the environment policy, we have no reason to assume there was a conflict of interest, since the viii Self-assessed level of selfishness/altruism on a 1–5 scale report was undertaken by an independent researcher at the University of ix Trust in political institutions: 0–10 index based on expressed trust Pennsylvania and who has no other linkage to CCL. towards the US Congress, the US Federal Government, and 9 More specifically, we multiplied the participant income by the average % politicians. income loss that corresponds to her/his income quintile (as estimated by Ummel, 2016) and divided the amount by US$15 to obtain the cost to that The following sociodemographic information on participants was specific household of each US$1 of carbon tax; we then multiplied such amount already available from Prolific: age, sex, nationality, current country of by the tax level considered on the choice card. These estimates give individuals residence/state of US residence, country of birth, employment status, an idea of the potential impact that such a policy would have on an average and student status. We enriched our dataset by collecting information on representative households in their income range, adjusted for their actual the US states that had a carbon pricing scheme in place at the time of the declared income. An alternative approach, could have been to generate esti­ mates based on participants’ estimated household energy and travelling ex­ survey. penditures (Douenne and Fabre, 2022). 10 This in contrast to other studies that let policy attributes vary independently 3.3. Data and descriptive statistics (e.g. allowing for combinations where lower tax levels have higher cost im­ plications as in Alberini et al., 2018). We fielded our survey between July 29 and August 4, 2021, on 11 Year 2005 was chosen as baseline since it is the year by which US Prolific Academic. Our sampling was restricted to US residents aged 18 nationally-determined pledges are expressed. or older. Remuneration for participation in the survey was an amount 12 The Climate Action Tracker estimates national emission pathways and proportional to the time spent on the survey paid at US minimum wage. whether they align with the Paris Agreement long-term temperature objectives Median completion time was five minutes, with 75% of respondents by using a carbon-cycle/climate model (MAGICC6). The methodology of the completing the survey in three–nine minutes. A total of 1092 re­ model is described in Meinshausen et al. (2009, 2011) and Rogelj et al. (2012). 13 Within this experiment we merely use a broad range of existing projections spondents participated in our survey, and 1013 (92.8%) completed it to as reference points to see how likely these scenarios are to influence partici­ the end. Of these, 13 (1.3%) were excluded from the analysis for failing pants’ preferences. The probability of occurrence of each emission abatement attention checks, thus leaving us with 1000 individuals who formed our scenario, or the assumptions they are derived from is not the focus of this final survey sample. They replied to four experimental choice sets, thus exercise. generating a total of 4000 observations. While random assignment of 6 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 Fig. 1. Introduction to the experiment and examples of choice task. participants across treatments should lead to homogenous groups, we of the ANOVA test of equal variances (for the continuous variables) and calculated descriptive variables of the sample by group to ensure that the p-value of the Kruskal-Wallis test (for the ordinal variables). We the groups were balanced (Table 3). detected slight imbalances in the re-partition of individuals across The distribution of age, educational attainments, and individual gender, ethnicity, income, and political partisanship. These imbalances discount rate were homogenous across groups as shown by the p-values were accounted for in the econometric analysis through control 7 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 Fig. 2. Intertemporal choice task. 8 M. Silvi and E. Padilla Rosa Table 3 Main descriptive statistics of the sample by treatment group. TREATMENTS p-value FINAL US from F SAMPLE POPULATION (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) test Variables 2021/ 2021/ 2024/ 2024/ 2027/ 2027/ 2021/ 2021/ 2024/ 2024/ 2027/ 2027/ 2030 2030 2030 2030 2030 2030 2050 2050 2050 2050 2050 2050 Cashback Basecase Cashback Basecase Cashback Basecase Cashback Basecase Cashback Basecase Cashback Basecase 46.0 43.6 43.8 45.3 46.3 43.3 42.1 43.3 45.2 43.9 45.3 43.2 44.3 46.9a Mean age 0.459 (0.882) (0.882) (0.942) (0.938) (0.995) (0.890) (0.901) (0.921) (0.922) (0.838) (0.853) (0.903) (0.261) (0.905) Female (%) 49% 49% 65% 52% 51% 45% 50% 52% 49% 53% 48% 49% 0.004 51% 51%a Ethnicity (White/ 72% 66% 72% 75% 76% 73% 62% 74% 80% 70% 77% 72% 0.006 72% 72%a non-Hispanic) Median Household $66,338 $67,692 $68,338 $64,648 $64,994 $56,044 $65,939 $63,063 $55,874 $70,645 $66,494 $64,753 0.062 $65,330 $65,712a Income Education (bachelor’s and 61% 60% 55% 53% 61% 64% 58% 69% 60% 52% 62% 55% 0.991 59% 48%b higher) Individual Discount Rate 5.44 5.58 5.62 5.62 5.27 6.21 5.57 4.97 5.59 4.71 5.78 5.76 5.51 9 0.137 NA (mean on a scale (0.198) (0.180) (0.193) (0.195) (0.187) (0.205) (0.206) (0.187) (0.201) (0.187) (0.188) (0.205) (0.056) 1–11) Identifies with the Republican party 16% 34% 16% 25% 26% 22% 21% 29% 14% 18% 23% 23% 0.000 22% 24%c (%) N. participants 80 82 81 85 80 86 84 80 85 83 86 88 NA 1000 NA a National gender, age, ethnicity, and income data came from the US Census Bureau (2021). b Educational attainment data for Americans aged 25–64 came from the OECD (2021). c The percentage of Americans considering themselves Republicans is the June 2021 figure from Gallup Consulting (2021). Energy Economics 125 (2023) 106815 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 variables. To show the comparability of our sample to the US national population, we included statistics in Table 3 on American households using the latest data from the US Census Bureau (2021), the OECD (2021), and Gallup Consulting (2021). Despite minor differences in average age (44.3 vs. 46.9), political partisanship (22% vs. 24% iden­ tifying as Republicans), and educational attainment (59% vs. 48% with at least a bachelor’s degree14), our sample matches well the average US household. 3.4. Analysis To establish the determinants of the support for the tax, we estimated logit regressions of the form: ( Y(i,c) =F PolicyTemporalSetting(p) , PolicyCost&Revenues(i,p,t) , PolicyRevenueUse(r) , IndividualCharacteristic(i) , (1) ) StateLevelControls(s) Fig. 3. Percentage of yes votes by treatment attribute. The response variable Y(i,c) records whether the individual (i) voted Source: own computations based on experimental data. Error bars are calculated in favor of the carbon tax design contained in one of the four choice sets on the basis of 95% confidence intervals. (c) to which he/she was exposed. It is modeled as a function of: As shown in Fig. 3, we find the biggest difference in acceptance rates a) one of the six temporal settings of the policy design (p); across revenue use. Predictably, tax acceptance is higher when revenue b) the costs and the revenues that the policy eventually imposes on the use is channeled back to citizens in the form of cashback payments, household, calculated based on the individual (i)’s household in­ reaching 67% on average (Kruskal-Wallis’ X2(1) = 25.242, p-value = come, the policy temporal setting (p), and the four tax baseline levels 0.00). When revenues are simply conferred to the federal government (t); budget, yes votes reach 59% on average. The tax approval rates appear c) the policy’s use of revenues (r) as either cashback payments or to be highest for the scenarios that introduce the tax in 2024 following a payments to the general federal budget; 3-year delay as compared to the immediate introduction and the 6-year d) personal characteristics of individual (i), including political and delay scenario (Kruskal-Wallis’ X2(2) = 4.919, p-value = 0.08). In our environmental beliefs, personal discount rate, and demographic design, delaying the introduction of the tax implies a higher tax rate and data; and hence higher cost implications for the average household proportionate e) state level controls (s), including whether the state in which indi­ to the length of the delay. This was intentionally done to reflect the fact vidual (i) lives has already initiated a carbon pricing scheme. that delaying the start of a carbon tax would need to be compensated with higher starting fees to still reach the Paris Agreement global Controls d) and e) increase the precision of our estimates and correct warming mitigation objectives. Presumably due to the effect of indi­ for the slight sociodemographic imbalances observed between groups. vidual discounting of future outcomes,16 participants were more likely to commit to a carbon tax if delayed by 3 years, despite the slight cost 4. Results increase it would bring about for their households. However, it appears that the cost increase in the 2027 scenario is big enough to more than 4.1. The temporal distribution of carbon taxes’ costs and rewards that offset the effect of individual discounting and any potential increase in increase their public acceptance yes votes that could have been brought about by a further delayed introduction. Expressing the emission reduction achieved by the tax by We start with a graphical representation of our key results. The different temporal horizons had a statistically significant effect on variable of interest is the percentage of participants voting in favor of the acceptance rates: on average, the 2050 horizon achieved a higher introduction of the carbon tax over the total number of respondents. In consensus of 65% compared to 61% for scenarios in which 2030 emis­ the Annex we report the percentage of “Yes” votes over the total after sions abatement objectives were mentioned (Kruskal-Wallis’ X2(1) = excluding participants who voted “Don’t Know”. Below we re-evaluate 5.992, p-value = 0.01). Constituents saw the tax more favorably when it the results by coding “Don’t know” answers as “No”, which provides a was framed in a longer-term horizon with more ambitious environ­ more conservative estimate of carbon tax support.15 There are clear, mental objectives. In light of the slight socio-demographic unbalances statistically significant variations in the acceptance rates, across policy observed between treatment groups, we computed average treatment attributes and groups. The proportion of yes votes ranged between effects (ATEs) using the propensity score matching method as robustness 55.5% and 70%, depending on treatment group (F: 3.71, Pr > F = check. After including the four variables which showed unbalances and 0.000). These percentages are in line with the latest available opinion additional socio-demographic controls, ATEs calculated through this polls that registered approval rates of 50% (five years earlier in 2016) or method were statistically significant: i) introducing the policy 67% (Puskin and Mills, 2017; Leiserowitz et al., 2021), although we acknowledge that similarly to such polls, survey experiments may overestimate support. 16 The fees and associated household costs were intentionally undiscounted for three reasons. First the acceptable objective rate of discounting to apply to future outcomes is highly controversial; second, the concept of discounting is 14 The percentage of Americans with minimum a bachelor degree comes from too complex to explain to laypeople and third, we wanted to let individuals to the OECD and it is possibly an underestimate since it refers to 2019, while the apply their own discount rate. Since we measure individual discount rates rate of Americans with a university degree keeps growing year on year and is exogenously after the experiment, we are able to use our estimated individual likely to be higher for 2021. discount rates to disentangle the effect of individual discounting, temporal 15 “Don’t Know” answers averaged 7% of the total across treatments. setting and tax level. 10 M. Silvi and E. Padilla Rosa Table 4 Regression output (Marginal effects). Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Logit Logit Logit Logit Logit Logit Logit Logit Logit Logit Logit If vote cert.> 3c If vote If Basecase If Cashback If Net ≥0 If Net < 0 cert.>3c Policy characteristics 0.073*** 0.059* 0.040 0.054 0.052 0.065 0.129** 0.001 0.062 0.056 Policy Start 2024 (dummy) (0.019) (0.071) (0.257) (0.116) (0.146) (0.308) (0.021) (0.986) (0.329) (0.201) 0.092*** 0.075** 0.100*** 0.129*** 0.130*** 0.040 0.223*** 0.089 0.143 0.164*** Policy Start 2027 (dummy) (0.005) (0.028) (0.005) (0.000) (0.000) (0.562) (0.000) (0.117) (0.059) (0.001) − 0.037 − 0.031 − 0.038 − 0.027 − 0.034 − 0.023 − 0.089* − 0.047 − 0.033 − 0.056 − 0.037 Horizon 2030 (dummy) (0.152) (0.239) (0.135) (0.342) (0.236) (0.437) (0.100) (0.287) (0.386) (0.313) (0.307) 0.077*** 0.077*** 0.057** 0.084*** 0.085*** 0.097* Revenue as cashback (dummy) (0.003) (0.004) (0.044) (0.003) (0.004) (0.076) − 0.00005*** Hh Cost (0.013) 0.00006* Hh Revenue (0.077) − 0.002*** − 0.002*** − 0.002*** − 0.002*** − 0.003*** − 0.003*** − 0.004*** − 0.001*** − 0.001** − 0.003** Tax level (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.023) (0.000) Individual characteristics − 0.007* − 0.008* − 0.006 − 0.017** − 0.009 − 0.005 − 0.00002 − 0.010* Intertemporal disc. rate (0–10 scale) (0.062) (0.062) (0.173) (0.053) (0.187) (0.411) (0.998) (0.073) − 0.002 *2024 Start (0.830) 11 0.016* *2027 Start (0.102) 0.010 *2030 Horizon (0.216) − 0.001 *Cashback (0.873) − 0.439*** − 0.300*** − 0.297*** − 0.292*** − 0.299*** − 0.305*** − 0.274*** − 0.449*** − 0.283*** Republican Party (dummy) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) − 7.84 − 7.75 − 7.97 − 1.12 − 1.63 e− 07 − 1.26 e− 06*** − 7.77 e− 07* 3.06e− 07 − 7.35e− 07* Income e− 07** e− 07** e− 07** e− 06** (0.614) (0.001) (0.074) (0.832) (0.102) (0.020) (0.024) (0.022) (0.038) − 0.112* − 0.109* − 0.085 − 0.110* NoRespEnvironment (dummy) (0.074) (0.077) (0.204) (0.070) − 0.192*** − 0.196*** − 0.189*** − 0.194*** − 0.174*** − 0.217*** − 0.285*** − 0.173*** Selfishness (dummy) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) − 0.124 *** − 0.131 *** − 0.155*** − 0.133*** − 0.140 *** − 0.111 *** − 0.093 − 0.152 No social norm (dummy) (0.000) (0.000) (0.000) (0.000) (0.001) (0.009) (0.114) (0.114) 0.031*** 0.030*** 0.031 *** 0.030*** 0.040*** 0.023*** 0.028** 0.040*** Political trust (1–10 scale) (0.000) (0.000) (0.000) (0.000) (0.000) (0.017) (0.047) (0.000) Energy Economics 125 (2023) 106815 − 0.414 *** − 0.423 *** − 0.498 *** − 0.428 *** − 0.619*** − 0.387*** − 0.514 − 0.510*** Clim. change denial (dummy) (0.000) (0.000) (0.000) (0.000) (0.000) (0.006) (0.211) (0.000) Demographic controls − 0.004*** − 0.004*** − 0.005*** − 0.004*** − 0.007*** − 0.003** − 0.004** − 0.006*** Age (0.000) (0.000) (0.000) (0.000) (0.000) (0.026) (0.029) (0.00) 0.047* 0.028 0.028 0.057* 0.027 0.028 0.060 0.023 0.041 Female (dummy) (0.096) (0.358) (0.361) (0.082) (0.383) (0.520) (0.181) (0.706) (0.273) (continued on next page) M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 immediately and expressing emission objective by the 2030 horizon decreased acceptance rates (respectively, ATE = − 0.303, p = 0.000; and If Net < 0 − 0.129** − 0.012 − 0.024 (0.850) (0.538) (0.034) 0.212 0.000 2664 Logit (12) ATE = − 0.034, p = 0.042). Including cashback payments increased Y Y support (ATE = 0.048, p = 0.003). These differences in approval rates Notes. * Significant at <0.10, ** significant at <0.05, *** significant at <0.01. Values reported are Marginal Effects at the Means MEM, p-values in parentheses. Errors are clustered at the individual level. are a sign that the temporal setting of a tax in terms of its start date and the temporal horizon by which its emission reduction targets are If Net ≥0 (0.528) (0.091) (0.654) 0.109* 0.065 0.056 0.303 0.000 1136 Logit expressed have an influence on its political acceptance. In line with the (11) Y Y above results, the treatment group that gathers the highest acceptance corresponds to a carbon tax that redistributes its revenues to constitu­ ents, that starts three years in the future and with emission reduction If Cashback − 0.039 (0.631) (0.995) (0.614) objectives expressed by 2050. Nearly half of all survey participants 0.035 0.002 0.222 0.000 Logit 1888 (10) Y Y (48%) voted in favor of the policy for all tax levels; 18% always voted against, and 2% always responded “Don’t know”. The remaining par­ ticipants changed their support in function of the tax level, usually Fewer observations due to missing data on US state of residence. c Individuals rated the certainty of their votes on a scale from 1 to 7, from very uncertain to very certain. If Basecase switching from “Don’t know” to “No”, from “Yes” to “No” or from “Yes” − 0.149** (0.358) (0.654) (0.035) 0.072 0.022 0.240 0.000 Logit 1936 (9) to “Don’t Know” as the tax level increased. Y Y Table 4 reports our regression estimations of marginal effects. We use logit models with clustered standard errors at the individual level (to account for the fact that repeated responses from the same individuals − 0.072 (0.789) (0.924) (0.155) are not independent). Our dependent variable is a binary 0–1 dummy 0.014 0.003 0.204 0.000 Logit 3844 (8) Y Y capturing whether individuals replied in favor of introducing the tax. The model shows the effect of individual and policy attributes on the likelihood that individuals would vote in favor of a national referendum on the introduction of a carbon tax. Prior to running the regressions we cert.>3c − 0.015 − 0.062 (0.796) (0.596) (0.261) If vote crosschecked for potential multicollinearity among the explanatory 0.018 0.235 0.000 Logit 3420 (7) Y Y variables, by examining bivariate correlations, the Variance Inflation Factors (VIF) and tolerance levels. Across all covariates simultaneously included in the regressions, VIF and tolerance levels fall well below 10 and well above 0.10, respectively, thus ruling out multicollinearity − 0.070 (0.880) (0.971) (0.174) 0.008 0.001 0.201 0.000 3844 concerns. The first four covariates are dummy variables identifying the Logit (6) Y Y treatments. We find that asking individuals to pre-commit to a carbon tax that starts 3 or 6 years in the future increases acceptance by 7 p.p. and 9 p.p., respectively, compared to introducing it immediately, when the effect of the cost of the tax on households is controlled (column 1). − 0.004 − 0.072 (0.937) (0.936) (0.163) 0.002 0.187 0.000 3844 Logit (5) This suggests that individuals would have a preference for postponing Y Y the policy as long as possible, if the policy’s cost were unaffected by this delay, and it confirms our earlier interpretation of more people voting in favor of policy designs starting in 2024 compared to 2027. The promise (0.086) 0.053* 3660b 0.121 0.000 Logit of cashback payments increases acceptance by nearly 7 p.p. These es­ (4) N Y timations are statistically significant at the 99% confidence level. We do not find a statistically significant effect of expressing environmental objectives for 2030 versus 2050 on policy support. This policy attribute seems to be ineffective on its own, once other policy attributes and tax 0.005 0.016 4000 (3) N N levels are accounted for. As an additional robustness check, we repeated the same regression by excluding individuals that expressed some de­ gree of uncertainty regarding their votes (column 2). The estimations If vote cert.> 3c confirm the results from the previous regression both in terms of sta­ tistical significance and signs. As expected, and in line with the literature 0.014 0.000 3568 Logit (2) N N (Sælen and Kallbekken, 2011; Gevrek and Uyduranoglu, 2015), approval decreases with tax rates and the associated additional cost that a tax imposes on households (column 3).17 Predictably, tax acceptance p-value F test nearly statistically significant at 0.11. increases with each additional US$ of revenue from the cashback 0.014 0.000 payments. 4000 Logit (1) N N In column 4, we addressed the slight gender, income, ethnicity, and political imbalances observed across treatment groups. The addition of these control variables left our previous results unchanged. We included a dummy variable which captures whether or not living in a state that has already adopted carbon pricing in some form positively affects in­ Educ.: min. bachelor’s degree dividual acceptance rates; the coefficient is statistically significant at the Ethnicity and nationality Unemployed (dummy) State level carbon tax State level controls Table 4 (continued ) State of residence Rural (dummy) 17 As mentioned in the methods section, cashback payments and household p-value F test Observations (dummy) costs are calculated based on the participant’s declared income. For this reason, Pseudo R2 Variables we substitute household cost and revenues from the tax with tax level due to their high correlation with income. For the same reason we excluded the Policy Start dummies in this regression. b a 12 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 99% confidence level, and it has a positive sign.18 connected to rural living. This is in line with Umit and Schaffer (2020), Column 5 focuses on the individual level characteristics affecting who found that individuals who are more dependent on energy are more choice. When we conjoin individual characteristics and policy attributes averse to carbon taxes. in the same regression (column 6), our aforementioned findings are confirmed in terms of statistical significance and signs. Amongst the 4.2. The effect of intertemporal discounting on tax acceptance individual level characteristics that affect choices the most, we find that denial of the occurrence of climate change reduces the probability to Individual 1-year discount rates correlate negatively with policy approve it by 42 p.p. People who identify with the Republican Party are approval, the estimated effect is statistically significant with negative 20 p.p. less likely to vote in its favor (these estimates are statistically sign (columns 5,6,8). Ceteris paribus, impatient individuals are less significant with negative signs). Income negatively correlates with tax likely to vote in favor of the tax, presumably because by placing less acceptance, potentially reflecting the fact that individuals with a higher importance on the future, they are less inclined to accept additional income incur higher cost increases in absolute terms and receive smaller short-term costs for the long-term preservation of the climate. The cashback payments. However, this finding may also reflect other relationship between individual support of environmental preservation income-related characteristics, such as preferences for redistribution. policies and temporal preferences is well documented in the literature We also find that individual beliefs concerning the environment and (Fehr and Leibbrandt, 2011; Bruderer Enzler, 2015). Once we control for society in general affect choice. Individuals who decline any personal individual discount rates in our regressions, the effect of dealying the responsibility to care for the environment, or those who think that policy’ start to 2024 loses statistical significance, corroborating our others would not pay a price to protect the environment are less likely to earlier intuition that individual’s preference to postpone the policy by vote for the tax. As a measure of pro-social orientation, we asked par­ three years is entirely explained by individual discounting. In Column 8 ticipants whether they thought that one should focus on his/her own life we include interaction effects between individual discount rates and the and not bother too much with helping others. People who agreed with four dummy variables identifying other policy attributes. We find a this statement were 30 p.p. less likely to approve of the tax. In line with statistically significant interaction between individual discount rates previous findings in the literature (Umit and Schaffer, 2020), we find and policy start date. The positive sign indicates that the effect of indi­ that our index measuring trust in institutions on a 0–10 scale positively vidual discounting on policy acceptance is attenuated when the tax correlates with tax approval. The more an individual trusts the federal starts 6-years later. This result could be interpreted as an indication of government, the US Congress, and politicians, the more likely she/he is hyperbolic discounting, with participants discounting farther outcomes to avail of carbon pricing at the federal level. By looking at demographic at declining rates. It supports the idea that impatient individuals, while characteristics, we find that carbon tax support decreases with age. We generally more opposed to the introduction of a carbon tax, may lower did not find a statistically significant effect of unemployment or of ed­ their resistance, if they had a chance to pre-commit to a policy that starts ucation.19 We find some evidence that females are more likely to be in later in time. This is in line with the literature highlighting that people favor of a carbon tax, however this result was not always statistically with hyperbolic discounting have a desire for pre-commitment mecha­ significant across regressions. Repeating the regression on the subset of nisms (Ashraf et al., 2006). As mentioned earlier, the effect of dis­ participants who were highly certain of their choices does not change counting on policy choice seems to depend on the expected economic these results (column 7). We compute interaction terms between the effect of the policy: only households that expect to be negatively policy’s temporal attributes and revenue use, and between all policy impacted from the policy seem to discount the future benefits that the attributes and the net economic impact of the tax on participants. The carbon tax may bring. Negative time preferences seemingly do not estimated effects are not statistically significant and are thus not re­ activate when individuals expect a neutral or positive impact of the ported here. However, when we run split regressions separating par­ policy on their income. This is possibly a sign that when the carbon tax is ticipants in the Basecase and Cashback condition (columns 9 and 10), we attached to a private economic compensation which is delivered at the find that the temporal attributes of the policy lose relevance. In addition, same time, individuals do not perceive the choice situation as one separating individuals who expect a net loss from the tax, compared to involving an intertemporal tradeoff. those who do not, we find that the effect of temporal frame and indi­ vidual discounting is only statistically significant for the former (col­ 4.3. The political feasibility of a Paris Agreement’s objective-compatible umns 11 and 12). In addition, we find that expectations about the tax economic effects of the tax are highly relevant for individuals living in rural areas: when they do not expect to receive a dividend or when they Fig. 4 depicts the percentage of yes votes by taxation level. Each expect a net loss, they are less likely to vote in its favor by 15 and 13 p.p. individual voted the same carbon tax design four times (as defined by respectively. Conversely, including cash-back transfers effectively start of the policy, temporal emission horizon, and revenue use), with counters opposition from rural areas. It is possible that people living in only variations in the taxation level (and the associated costs, revenues, rural areas anticipate that the cost imposed on them by the tax may and emission cuts). The four baseline tax levels voted upon were 15, 40, actually be higher than the ad-hoc figures included in the choice cards, 60, and 80 US$. These rates were adjusted proportionally to the delay owing to longer commuting distances and fossil fuel consumption with which the policy was introduced for individuals in the 2024 and 2027 delay conditions. As the Figure shows, acceptance is highest in the first choice set when the tax rate and the cost implications for house­ 18 Additional regressions not reported here included controls for the emission holds are lowest and it gradually decreases as the costs increase. Qual­ abatement objectives of each design. The coefficients are statistically significant itative analysis of the sentences written by participants to explain their and have a negative sign, while the coefficients for household costs lose sig­ choices in the pilot phase confirms that for some there is a switch point nificance. This is clearly due to multicollinearity between the cost of the tax and at which the cost of the tax simply becomes too high to the individual the amount of emissions it promises to abate. As per our experimental design either in relation to his/her income, the policy’s environmental objec­ the percentage of emissions that a carbon tax design can achieve depends on the tives, or both. It should be noted that in the Basecase condition, in­ tax level and consequently, a tax that pursues more ambitious environmental dividuals are predictably more sensitive to tax increases, and the objectives also translates into a higher cost for households. 19 Beyond the dummy variable for people with at least a bachelor degree percentage of yes votes decreases more markedly in this condition, which is included in the regressions we also tested as alternative measure an whereas it appears to be more stable for participants that receive divi­ educational attainment scale from 1 to 10 and a dummy variable for people dends. This shows that in a carbon tax policy with a cashback program, who currently are students. They also did not give any statistically significant individuals would potentially accept a level of taxation that would be result. politically unfeasible in the absence of this revenue redistribution. 13 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 additional unit increase in the climate belief scale, political scale, trust index, and environmental responsibility scale (calculated based on the equation in column 6, Table 4). The highlighted areas in the graphs correspond to individuals that match the four characteristics just described. Across the four groups, inclusion in the Cashback group and in the treatment group in which the tax starts in 2027, followed by a tax design that starts in 2024, have the biggest impact on the probability of approving the tax, and these effects are statistically significant. These effects are positive for any value of the four scales under consideration; they are strongest for individuals identifying as Republicans, people with low trust in public institutions, and people who deny any personal environmental responsibility. On the contrary, these treatments are not as effective on individuals who do not recognize that climate change is Fig. 4. Percentage of yes votes by taxation level, Cashback versus Basecase. occurring. These graphs indicate that one of these policy features or a Source: own computations based on experimental data. Taxation levels shown combination thereof may help to persuade a potentially politically in labels are baseline levels, that are adjusted upward for the delayed scenarios. relevant section of the American population to approve a carbon tax. In particular, the fourth choice set which corresponds to the tax level 5. Conclusions that has the highest probability of reaching the Paris global warming objectives falls short of the majority approval in the Basecase condition The literature on intertemporal choice and cooperation tells us that with 49% of yes votes, while it appears to be politically feasible in the cooperation is less likely to arise in a scenario in which costs are Cashback condition. Fig. 5 shows the temporal treatment that maximizes immediately imposed for rewards that are shared and will occur further the percentage of yes votes by focusing only on the fourth choice. The in the future. This is the standard temporal context of a carbon tax which figure shows that if we focus on taxation levels consistent with Paris imposes from its beginning costs on a constituency for environmental global temperature objectives, but we do not want to earmark revenues, benefits that accumulate slowly and will only be tangible in the distant then the ideal policy horizon that maximizes acceptability is the 2024/ future. The intertemporal dimension of carbon taxes and its effect on 2050 horizon. public acceptability has not been previously studied. This is the first paper to explore whether the temporal contextualization of carbon taxes can be designed in a way that increases their popularity. In particular, 4.4. Political relevance analysis we analyzed whether asking people to pre-commit to a carbon tax that starts a few years from now increases its public support and whether There are pools of participants that are significantly more likely to be individuals prefer that its target emission cuts are expressed by a more against a carbon tax. Even though they may constitute a minority, they proximal or distant date. This is also the first study to analyze whether may still represent a politically relevant group. It is thus of interest to individual temporal preferences affect the acceptability of carbon taxes. understand whether or not there is a temporal framing of a carbon tax We find that it is possible to design carbon taxes that optimize the that can persuade them. As noted earlier, among the analyzed individual temporality of its costs and benefits to the public’s preferences. Ceteris characteristics, the primary ones that negatively affect tax acceptance paribus, individuals are more likely to commit to a carbon tax if the are: a) the belief that climate change is not currently happening; b) economic costs it imposes on households can be slightly shifted to the political identification with the Republican Party; c) distrust in in­ future (by postponing the start of the policy by a few years). This is stitutions; and d) refusal to ascribe to personal responsibility for pro­ possibly due to individual discounting as individuals typically discount tecting the environment. future costs. A key caveat to this finding is that temporal preferences are Fig. 6.a–d shows the marginal effects of policy attributes (policy start dynamic and potentially inconsistent, as people who claims to be in in 2024 or 2027, 2030 as emission horizon, and Cashback) for each 80% 70% AVERAGE 64% 60% AVERAGE 49% 50% 40% 30% 20% 10% 0% 2021 2024 2027 2021 2024 2027 Cashback Basecase 2030 2050 Fig. 5. Percentage of yes votes by treatment group, taxation baseline 80 USD$. Source: own computations based on experimental data. 14 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 Fig. 6. Marginal effects of policy attributes on nay-sayers. Source: own computations based on experimental data. Notes. Highlighted areas in the graphs corresponds to: a.“Climate change is not happening”; b. “Identifies as Republican”; c. “Mostly distrust public institutions”; d. “Denies personal environmental responsibility”. favor of a policy that starts in three years, may actually change their is intentionally an underestimate. We show that when individuals are mind three years from now. Clearly, the increase in popularity of a confronted with the short-term opportunity cost that delaying the tax delayed policy start can be advantageous only if it comes with a pre- would potentially imply, they are not necessarily in favor of indefinitely commitment mechanism, such as a binding popular vote through postponing it. When the price of carbon is adjusted to take the delay into which constituents can irrevocably pre-commit to the preference they account, we see that individuals start to trade off tax postponement stated a few years earlier. against the containment of costs. The exact switch point would ulti­ However, delaying the introduction of a carbon tax has clear bio­ mately depend upon the adjusted tax rate, individual discounting, and physical consequences as the carbon keeps accumulating in the atmo­ how the cost on households is derived, among other things. As our sphere and the carbon budget that remains in order to stay within the descriptive findings in Fig. 3 show, our conservative estimate of a tax Paris global warming objectives narrows with each year of delay. A increase which is proposed to meet the Paris objectives with a six-year delayed introduction would, therefore, require increasing the ambition delay is high enough to decrease tax approval rates as compared to of the tax by either adjusting its starting level or by scheduling steeper 2021 and 2024. In the longer-term, such a tradeoff is likely to resolve year-on-year increases. In this paper, we tested the former hypothesis, itself, as the increased cost of mitigation for longer delays is enough to and we adjusted starting tax levels in proportion to the delay. A precise erode any additional support which might come from the appeal of estimate of this adjustment is even more complex than estimating the shifting costs to the future. taxation level required to meet the Paris objectives if the tax were to The political message is clear: while precommitment to costs that are start immediately because it relies even more heavily on predictions and slightly delayed may increase the immediate political acceptance of a assumptions regarding the evolution of global emissions, the availability tax and increase the likelihood that a national referendum would pass, of low carbon technology and climate feedback loops. Attempting an that increased support would be merely temporary and artificial unless exact estimation is beyond the scope of our paper, but we propose an the short-term opportunity costs of waiting were disclosed. Procrasti­ adjusted tax level which falls within the range of plausibility, and which nating on the approval of urgently needed carbon taxes for fear of voter 15 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 opposition may not payoff in the long term, as the public opposition that politicians try to avoid today will likely increase proportionally to the length of the delay and the associated increase in mitigation costs that The authors have nothing to disclose. the policy would need to include. The higher the opportunity cost of waiting, the less likely is a delayed introduction to win any additional Acknowledgements support. Politicians could leverage this and include the additional costs that an increased risk of waiting could pose for the climate in their We thank two anonymous reviewers and colleagues at the Depart­ carbon tax proposition and communications. ment of Applied Economics of Autonomous University of Barcelona and In addition, there are clear unknowns about tipping points and Ghent University for their valuable comments and suggestions. This feedback mechanisms that imply a greater risk of catastrophic global research was supported by the Spanish Ministry of Science and Inno­ warming events if climate mitigation policies are postponed and the vation, under grants PID2021-126295OB-I00 and BES-2016-078156. concentration of CO2 in the atmosphere becomes higher (Furman et al., 2015). Shifting a larger portion of the abatement effort to the future is Appendix A. Supplementary data also inefficient as it leads to avoidable welfare losses compared to a more balanced intertemporal distribution of efforts (Gollier, 2022). The most Supplementary data to this article can be found online at https://doi. efficient policy is thus the policy that starts immediately as the mitiga­ org/10.1016/j.eneco.2023.106815. tion costs increase with each year of delay. In the absence of cashback payments, or in the presence of individual economic losses, people’s References preference to delay the policy comes at a future welfare loss. Thus, any carbon tax proposal or postponement thereof that is passed on by the Andersen, S., Harrison, G.W., Lau, M.I., Rutström, E.E., 2006. Elicitation using multiple price list formats. Exp. Econ. 9, 383–405. https://doi.org/10.1007/s10683-006- population should not only include the economic opportunity cost of 7055-6. waiting, but it should also mention the resulting welfare losses and the Anderson, S., Marinescu, I.E., Shor, B., 2023. Can Pigou at the polls stop US melting the increase in long-term environmental risks. poles? J. Assoc. Environ. Resour. Econ. 10 (4), 903–945. https://doi.org/10.1086/ 722970. In the short-term there is thus also a policy tradeoff, in which pre- Arora, P., Peterson, N.D., Krantz, D.H., Hardisty, D.J., Reddy, K.S., 2012. To cooperate or committing to a policy that is postponed by a few years may increase not to cooperate: using new methodologies and frameworks to understand how its public support but worsen its efficiency and its effectiveness in affiliation influences cooperation in the present and future. J. Econ. Psychol. 33 (4), 842–853. https://doi.org/10.1016/j.joep.2012.02.006. averting climate catastrophe. This finding applies only to the Baseline Arrow, K., Solow, R., Portney, P., Leamer, E., Radner, R., Schuman, H., 1993. Report of scenario, and for individuals who expect a net loss from the tax. Beyond the NOAA Panel on Contingent Valuation. disclosing the opportunity costs of waiting, another effective way to Ashraf, N., Karlan, D., Yin, W., 2006. Tying Odysseus to the mast: evidence from a commitment savings product in the Philippines. Q. J. Econ. 121 (2), 635–672. counteract this preference for postponing the policy is to redistribute the https://doi.org/10.1162/qjec.2006.121.2.635. tax revenues to citizens. Adding a personal gain, such as a private eco­ Bachus, K., Van Ootegem, L., Verhofstadt, E., 2019. ‘No taxation without hypothecation’: nomic compensation, that is delivered at the same time as the individual towards an improved understanding of the acceptability of an environmental tax costs of the tax leaves the temporal dimension of choice in the back­ reform. J. Environ. Policy Plan. 21 (4), 321–332. Baranzini, A., van den Bergh, J.C.J.M., Carattini, S., Howarth, R.B., Padilla, E., Roca, J., ground; neutralizing the effect of individual discounting and of the 2017. Carbon pricing in climate policy: seven reasons, complementary instruments, policy’s temporal context on tax support. When choice is no longer just and political economy considerations. Wiley Interdiscip. Rev. Clim. Chang. https:// between incurring in a private cost for the sake of a long-term shared doi.org/10.1002/wcc.462. Bateman, I.J., Cole, M., Cooper, P., Georgiou, S., Hadley, D., Poe, G.L., 2004. On visible benefit, but there is also short-term private benefit from the policy, in­ choice sets and scope sensitivity. J. Environ. Econ. Manag. 47 (1), 71–93. dividuals appear to no longer perceive the choice situation as one Bechtel, M.M., Scheve, K.F., van Lieshout, E., 2020. Constant carbon pricing increases involving an intertemporal tradeoff. To maximize support for the most support for climate action compared to ramping up costs over time. Nat. Clim. Chang. 10, 1004–1009. https://doi.org/10.1038/s41558-020-00914-6. efficient policy, the ideal carbon tax formulation should thus contain a Benartzi, S., Thaler, R.H., 2013. Behavioral economics and the retirement savings crisis. dividend redistribution component, at least at its introduction. This also Science 339 (6124), 1152–1153. https://doi.org/10.1126/science.1231320. appeals to groups of constituents that would typically oppose the tax. Bergquist, M., Nilsson, A., Harring, N., et al., 2022. Meta-analyses of fifteen determinants of public opinion about climate change taxes and laws. Nat. Clim. Chang. 12, Furthermore, people are more likely to favor a carbon tax if justified 235–240. https://doi.org/10.1038/s41558-022-01297-6. by more ambitious albeit more temporally distant environmental ob­ Bishop, R.C., Boyle, K.J., 2019. Reliability and validity in nonmarket valuation. Environ. jectives; for example, climate neutrality by 2050 is a more evocative Resour. Econ. 72, 559–582. Brouwer, R., Dekker, T., Rolfe, J., Windle, J., 2010. Choice certainty and consistency in goal than merely halving carbon emissions by 2030. A limitation of our repeated choice experiments. Environ. Resour. Econ. 46, 93–109. experimental design is that we included just two uses of revenues out of Bruderer Enzler, H., 2015. Consideration of future consequences as a predictor of all available alternatives. This was done to contain the already high environmentally responsible behavior: evidence from a general population study. number of treatments. Future research could expand to other alternative Environ. Behav. 47 (6), 618–643. https://doi.org/10.1177/0013916513512204. Carattini, S., Kallbekken, S., Orlov, A., 2019. How to win public support for a global revenue uses not included in the paper. It could also build on this work carbon tax. Nature 565, 289–291. https://doi.org/10.1038/d41586-019-00124-x. by looking at the effect of the timing of the cashback payments versus Carson, R.T., Groves, T., 2007. Incentive and informational properties of preference the timing of the increased costs that households would face on a daily questions. Environ. Resour. Econ. 37, 181–210. Center for Climate and Energy Solutions, 2021. U.S. State Carbon Pricing Policies. https basis. ://www.c2es.org/document/us-state-carbon-pricing-policies/. Cherry, T.L., Kallbekken, S., Kroll, S., 2014. The impact of trial runs on the acceptability Inclusion and diversity statement of environmental taxes: experimental evidence. Resour. Energy Econ. 38, 84–95. https://doi.org/10.1016/j.reseneeco.2014.06.005. Climate Action Tracker, 2021. USA Country Summary. https://climateactiontracker.org/ We worked to ensure gender and ethnic balance in the recruitment of countries/usa/. human subjects. Coller, M., Williams, M.B., 1999. Eliciting individual discount rates. Exp. Econ. 2, 107–127. https://doi.org/10.1023/A:1009986005690. Curry, O., Price, M., Price, J.M., 2008. Patience is a virtue: Cooperative people have CRediT authorship contribution statement lower discount rates. Person. Indiv. Diff. 44 (3), 780–785. https://doi.org/10.1016/ j.paid.2007.09.023. Dechezleprêtre, A., et al., 2022. Fighting Climate Change: International Attitudes toward Mariateresa Silvi: Conceptualization, Methodology, Software, Climate Policies. OECD Economics Department Working Papers: No. 1714. OECD Formal analysis, Investigation, Writing – original draft, Writing – review Publishing, Paris. https://doi.org/10.1787/3406f29a-en. & editing, Funding acquisition. Emilio Padilla Rosa: Writing – review Douenne, T., Fabre, A., 2022. Yellow vests, pessimistic beliefs, and carbon tax aversion. & editing, Supervision, Project administration, Funding acquisition. Am. Econ. J. Econ. Policy 14 (1), 81–110. 16 M. Silvi and E. Padilla Rosa Energy Economics 125 (2023) 106815 Drews, S., van den Bergh, J.C.J.M., 2016. What explains public support for climate Meinshausen, M., Raper, S.C.B., Wigley, T.M.L., 2011. Emulating coupled atmosphere- policies? A review of empirical and experimental studies. Clim. Pol. 16 (7), 855–876. ocean and carbon cycle models with a simpler model, MAGICC6: Part I – model https://doi.org/10.1080/14693062.2015.1058240. description and calibration. Atmos. Chem. Phys. 11, 1417–1456. Falk, A., Becker, A., Dohmen, T., Enke, B., Huffman, D., Sundeand, U., 2018. Global National Surveys on Energy and Environment, 2016. Data Tables 2016. University of Evidence on Economic Preferences. Quart. Jour. Econ. 133 (4), 1645–1692. https:// Michigan. https://closup.umich.edu/research-projects/national-surveys-energy-a doi.org/10.1093/qje/qjy013. nd-environment/data-tables. Fehr, E., Leibbrandt, A., 2011. A field study on cooperativeness and impatience in the OECD, 2021. Adult Education Level. Retrieved August 9, 2021, from. https://data.oecd. Tragedy of the Commons. J. Public Econ. 95 (9–10), 1144–1155. https://doi.org/ org/eduatt/adult-education-level.htm#indicator-chart. 10.1016/j.jpubeco.2011.05.013. Ostrom, E., 1990. Governing the Commons: The Evolution of Institutions for Collective Furman, J., Shadbegian, R., Stock, J., 2015. The Cost of Delaying Action to Stem Climate Action. Cambridge University Press, Cambridge. Change: A Meta-Analysis. VoxEU.org. 25 February. Palan, S., Schitter, C., 2018. Prolific.ac—a subject pool for online experiments. J. Behav. Gallup Consulting, 2021. Party Affiliation. Retrieved August 9, 2021, from. https://news. Exp. Financ. 17, 22–27. https://doi.org/10.1016/j.jbef.2017.12.004. gallup.com/poll/15370/party-affiliation.aspx. Peer, E., Brandimarte, L., Samat, S., Acquisti, A., 2017. Beyond the Turk: alternative Gevrek, Z.E., Uyduranoglu, A., 2015. Public preferences for carbon tax attributes. Ecol. platforms for crowdsourcing behavioral research. J. Exp. Soc. Psychol. 70, 153–163. Econ. 118, 186–197. Available at: https://doi.org/10.1016/j.ecolecon.2015.07.020. https://doi.org/10.1016/j.jesp.2017.01.006. Gibbard, A., 1973. Manipulation of voting schemes: a general result. Econometrica 41 Puskin, D., Mills, S.B., 2017. Moving the Needle of Americans Support for a Carbon Tax. (4), 587–601. https://doi.org/10.2307/1914083. National Survey on Energy and Environment, Issues in Energy and Environmental Gollier, C., 2020. Climate economics from theory to policy. EAERE Mag. 8, 6–8. Winter Policy. Number 30. 2020. Rogelj, J., Meinshausen, M., Knutti, R., 2012. Global warming under old and new Gollier, C., 2022. The Cost-Efficiency Carbon Pricing Puzzle. TSE Working Paper. scenarios using IPCC climate sensitivity range estimates. Nat. Clim. Chang. 2, Hardin, G., 1968. The tragedy of the commons. Science (New York, NY) 162 (3859), 248–253. https://doi.org/10.1038/nclimate1385. 1243–1248. https://doi.org/10.1126/science.162.3859.1243. Sælen, H., Kallbekken, S., 2011. A choice experiment on fuel taxation and earmarking in Hendrickx, L., van der Kooij, R., Poortinga, W., 2001. Temporal factors in resource Norway. Ecol. Econ. 70 (11), 2181–2190. dilemmas. Acta Psychol. 108 (2), 137–154. Satterthwaite, M.A., 1975. Strategy-proofness and Arrow’s conditions: existence and High-Level Commission on Carbon Prices, 2017. Report of the High-Level Commission on correspondence theorems for voting procedures and social welfare functions. Carbon Prices. World Bank. J. Econ. Theory 10 (2), 187–217. Jacquet, J., Hagel, K., Hauert, C., Marotzke, J., Röhl, T., Milinski, M., 2013. Intra- and Savin, I., Drews, S., Maestre-Andrés, S., van den Bergh, J.C.J.M., 2020. Public views on intergenerational discounting in the climate game. Nat. Clim. Chang. 3 (12), carbon taxation and its fairness: a computational-linguistics analysis. Clim. Chang. 1025–1028. https://doi.org/10.1038/nclimate2024. 162, 2107–2138. https://doi.org/10.1007/s10584-020-02842-y. Johannesson, M., Blomquist, G.C., Blumenschein, K., Johansson, P., Liljas, B., Thaler, R.H., Benartzi, S., 2004. Save more TomorrowTM: using behavioral economics to O’Conor, R.M., 1999. Calibrating hypothetical willingness to pay responses. J. Risk increase employee saving. J. Polit. Econ. 112, S1. https://doi.org/10.1086/380085. Uncertain. 18, 21–32. The White House, 2021. FACT SHEET: President Biden Sets 2030 Greenhouse Gas Kotchen, M.J., Turk, Z.M., Leiserowitz, A.A., 2017. Public willingness to pay for a US Pollution Reduction Target Aimed at Creating Good-Paying Union Jobs and Securing carbon tax and preferences for spending the revenue. Environ. Res. Lett. 12 (9) U.S. Leadership on Clean Energy Technologies. https://www.whitehouse.gov/briefi https://doi.org/10.1088/1748-9326/aa822a. ng-room/statements-releases/2021/04/22/fact-sheet-president-biden-sets-2030-g Leiserowitz, A., Maibach, E., Rosenthal, S., Kotcher, J., Carman, J., Wang, X., reenhouse-gas-pollution-reduction-target-aimed-at-creating-good-paying-union- Goldberg, M., Lacroix, K., Marlon, J., 2021. Politics & Global Warming. Yale jobs-and-securing-u-s-leadership-on-clean-energy-technologies/. Program on Climate Change and Communication Report. https://climatecommuni Tiezzi, S., Xiao, E., 2016. Time delay, complexity and support for taxation. J. Environ. cation.yale.edu/publications/politics-global-warming-december-2020/5/. Econ. Manag. 77, 117–141. https://doi.org/10.1016/j.jeem.2016.03.002. Libecap, G.D., 2008. State regulation of open-access, common-pool resources. In: Umit, R., Schaffer, L.M., 2020. Attitudes towards carbon taxes across Europe: the role of Ménard, C., Shirley, M.M. (Eds.), Handbook of New Institutional Economics. perceived uncertainty and self-interest. Energy Policy 140, 111385. https://doi.org/ Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-69305-5_22. 10.1016/j.enpol.2020.111385. Louviere, J., Hensher, D., Swait, J., Adamowicz, W., 2000. Stated Choice Methods: Ummel, K., 2016. Impact of CCL’s Proposed Carbon Fee and Dividend Policy: A High- Analysis and Applications. Cambridge University Press, Cambridge. Resolution Analysis of the Financial Effect on US Households (Prepared for Citizens’ Mannix, E.A., 1991. Resource dilemmas and discount rates in decision making groups. Climate Lobby). J. Exper. Soc. Psychol. 27 (4), 379–391. https://doi.org/10.1016/0022-1031(91) US Census Bureau, 2021. United States Data 2019. Retrieved August 9, 2021, from http 90032-2. s://data.census.gov/cedsci/profile?q=United%20States&g=0100000US. consulted Mannix, E.A., Loewenstein, G.F., 1993. Managerial Time Horizons and Interfirm on 09/08/2021. Mobility: An Experimental Investigation. Organiz. Behav. Human Decis. Proces. 56 US Congress, 2021. 117th Congress. . Current legislative activities database, Retrieved (2), 266–284. https://doi.org/10.1006/obhd.1993.1055. August 9, 2021, from: https://www.congress.gov/bill/117th-congress/house McKibbin, W.J., Morris, A.C., Wilcoxen, P.J., 2014. The Economic Consequences of Delay -bill/2307/cosponsors. in U.S. Climate Policy. The Brookings Institute, Washington D.C.. https://www.broo Wang, M., Rieger, M.O., Hens, T., 2016. How time preferences differ: evidence from 53 kings.edu/wp-content/uploads/2016/06/03_economic_consequences_delay_us_c countries. J. Econ. Psychol. 52, 115–135. limate_policy.pdf Weng, W., Morrison, M.D., Boyle, K.J., Boxall, P.C., Rose, J., 2021. Effects of the number Meinshausen, M., Meinshausen, N., Hare, W., Raper, S.C.B., Frieler, K., Knutti, R., of alternatives in public good discrete choice experiments. Ecol. Econ. 182, 106904. Frame, D.J., Allen, M.R., 2009. Greenhouse-gas emission targets for limiting global warming to 2 ◦ C. Nature 458, 1158–1162. 17