Perceived fairness and public acceptability of carbon pricing: a review of the literature

ABSTRACT While carbon pricing is widely seen as a crucial element of climate policy and has been implemented in many countries, it also has met with strong resistance. We provide a comprehensive overview of public perceptions of the fairness of carbon pricing and how these affect policy acceptability. To this end, we review evidence from empirical studies on how individuals judge personal, distributional and procedural aspects of carbon taxes and cap-and-trade. In addition, we examine preferences for particular redistributive and other uses of revenues generated by carbon pricing and their role in instrument acceptability. Our results indicate a high concern over distributional effects, particularly in relation to policy impacts on poor people, in turn reducing policy acceptability. In addition, people show little trust in the capacities of governments to put the revenues of carbon pricing to good use. Somewhat surprisingly, most studies do not indicate clear public preferences for using revenues to ensure fairer policy outcomes, notably by reducing its regressive effects. Instead, many people prefer using revenues for ‘environmental projects’ of various kinds. We end by providing recommendations for improving public acceptability of carbon pricing. One suggestion to increase policy acceptability is combining the redistribution of revenue to vulnerable groups with the funding for environmental projects, such as on renewable energy. Key policy insights If people perceive carbon pricing instruments as fair, this increases policy acceptability and support. People’s satisfaction with information provided by the government about the policy instrument increases acceptability. While people express high concern over uneven distribution of the policy burden, they often prefer using carbon pricing revenues for environmental projects instead of compensation for inequitable outcomes. Recent studies find that people’s preferences shift to using revenues for making policy fairer if they better understand the functioning of carbon pricing, notably that relatively high prices of CO2-intensive goods and services reduce their consumption. Combining the redistribution of revenue to support both vulnerable groups and environmental projects, such as on renewable energy, seems to most increase policy acceptability.


Introduction
With the entry into force of the 2015 Paris Agreement state leaders have accepted that climate change poses a serious threat to the biosphere and human welfare. However, the majority of nationally determined contributions (NDCs) submitted under the Paris Agreement are not matched by sufficiently effective national policies (Victor et al., 2017). Many economists, environmental scientists and politicians have called for economy-wide carbon pricing, whether in the form of a tax or emissions trading, as a cornerstone of an effective climate policy package to implement the Paris Agreement. The main reason is that carbon pricing increases relative prices of high-versus low-carbon goods and services, in theory resulting in an effective as well as efficient way of reducing CO 2 emissions (Aldy, Krupnick, Newell, Parry, & Pizer, 2010;Nordhaus, 2010). In particular, all prices in the economy should reflect direct and indirect CO 2 emissions generated during the entire production cycle of associated products and services. In other words, carbon pricing represents a systemic policy affecting all carbon-based decisions in the economy, both discouraging highcarbon practices by consumers and producers and stimulating cleaner energy production. Moreover, carbon prices do not draw on government budgets but generate revenues that can be used for funding public investments in climate innovation and adaptation, or for compensating any inequitable effects (High-level Commission on Carbon Prices, 2017).
Despite such arguments in favour, and over 50 carbon pricing schemes implemented or planned around the world (World Bank and Ecofys, 2018), the instrument still meets considerable public and political resistance. Low public acceptability influenced the cancellation of a carbon pricing scheme in 2014 in Australia (Crowley, 2017), the rejection of ballot initiatives in 2016 and 2018 to impose a carbon tax in Washington State (Carattini, Kallbekken, & Orlov, 2019) or the popular protests in France at the end of 2018 which led to the suspension of a carbon tax for six months (Willsher, 2018). Perceived (un)fairness of carbon pricing has been found to be significantly correlated with its (low) public acceptability 1 (Clayton, 2018;Dreyer & Walker, 2013;Hammar & Jagers, 2007). We assume that carbon pricing fairness relates to individuals' perceptions arising from a judgment process regarding the policy instrument (Grasso, 2007). Accounting for people's opinions and public acceptability of climate change policies is important as these affect political choices (Anderson, Böhmelt, & Ward, 2017;Dresner, Dunne, Clinch, & Beuermann, 2006). The lack of acceptability 2 is also one reason why most existing carbon pricing schemes are weak, that is, they exhibit relatively low price levels. Almost half (46%) of emissions covered by carbon pricing are priced at less than US$10/tCO 2 e (World Bank and Ecofys, 2018). This is substantially lower than the price level thought to be consistent with achieving the Paris Agreement temperature target, which according to the High-Level Commission on Carbon Prices, should be at least US$40-80/tCO 2 by 2020 and US$50-100/tCO 2 by 2030 (High-level Commission on Carbon Prices, 2017). Moreover, some research suggests that increasing people's perceived fairness in the context of a CO 2 car tax can diminish scepticism about the effectiveness of the policy (Bolderdijk, Steg, Woerdman, Frieswijk, & De Groot, 2017). Taken together, unravelling fairness perceptions related to carbon pricing can contribute to an improved design of planned and existing carbon pricing schemes so as to make them more consistent with the targets of the Paris Agreement.
Here we examine which aspects of perceived fairness have been addressed in studies and how they affect policy acceptability. This assessment involves collecting concerns raised by the general public and sometimes specific stakeholders, such as policy-makers and business representatives. Our analytical framework, shown in Figure 1, is derived from findings of the fairness literature applied to carbon pricing and the related issue of environmental taxation (Dreyer & Walker, 2013;Kim, Schmöcker, Fujii, & Noland, 2013). It describes that people's judgement of fairness is influenced by perceptions of individual (fairness to me) and collective (fairness to others) consequences of the policy, as well as its design and procedural aspects. One dimension of perceived fairness related to individual consequences can be interpreted as: are the consequences to the self ('me') fair or unfair. Such perceived personal consequences include expected higher economic costs for the individual  and less personal freedom because of regulatory constraints (Baron & Jurney, 1993). Perceived collective consequences of carbon pricing, i.e. fairness to others or everyone, include distributional effects. Examples are disproportional costs for households that are poor or live in remote areas (Clayton, 2018;Jagers & Hammar, 2009), due to them spending a relatively large share of their disposable income on carbon-intensive goods and services like home heating, with the latter being relatively car and thus petrol dependent. Carbon pricing without complementary measures or specified uses of its revenues tends to be regressive (Baranzini, Goldemberg, & Speck, 2000;Grainger & Kolstad, 2010), notwithstanding that some research suggests that it might be progressive or neutral in certain low-and medium-income countries (Dorband, Jakob, Kalkuhl, & Steckel, 2019).
With regard to underlying perceptions of procedural fairness, we considered a wide range of aspects, based on a definition of procedural fairness as: the proper adherence to the rules relevant to a procedure and, by extension, as the correct application of such rules to all cases which are alike within the terms of the rules (formal or informal, explicit or implicit) which are consistently applied to all individuals. (Hay, 1995, p. 501) Hence, for instance, people's low trust in politicians using revenues generated by carbon pricing in a proper way is an aspect included in this category. The reason is that when people express lack of trust that the government will keep its promises about revenue use, it refers to a disregard of a relevant procedural rule, that is, keep your promises. We have also included 'distrust in government' as a relevant procedural aspect. When people express lack of trust in politicians because, for instance, they promote carbon pricing in order to generate additional governmental funds rather than mitigate climate change, it refers to a relevant procedural rule, that is, adhere to the presumed goal of the policy which is to mitigate climate change.
Studies further indicate that certain allocations of the revenues generated by carbon pricing can change people's perceptions of the fairness of the policy because, for instance, they ameliorate regressive impacts. Consequently, the level of acceptability depends on how carbon pricing revenues are used (Carattini, Baranzini, Thalmann, Varone, & Vöhringer, 2017;Jagers, Martinsson, & Matti, 2018;Klenert et al., 2018). In particular, we will review here evidence from empirical studies to get a more systematic understanding of whether using revenues from carbon pricing in order to make the policy fairer contributes to more policy acceptability than other uses.
We have included both carbon taxes and tradable permit (cap and trade) schemes. Carbon taxes apply a duty to goods and services in accordance with their carbon content. Cap and trade schemes set a limit on CO 2 emissions and distribute CO 2 permits among emitters accordingly. These permits can be bought and sold by companies or consumers in a market. To regulate emissions from individuals, a tradable permit scheme could either be implemented upstream, i.e. amongst fuel producers, or downstream, i.e. amongst individual consumers.
There are very few studies dealing with people's perceptions of fairness of upstream tradable permit schemes because their impacts are difficult to perceive. However, there are some studies dealing with people's opinions over downstream schemes known as personal carbon allowances.
Previous literature reviews have been undertaken regarding climate policy support in general (Drews & van den Bergh, 2016) and carbon pricing in particular (Carattini, Carvalho, & Fankhauser, 2018;Klenert et al., 2018). Our paper differs from these because our review focuses more thoroughly on how individuals judge personal, distributional and procedural aspects of carbon taxes and cap-and-trade and their perceived fairness; whereas others covered many more factors influencing public policy support.  Figure 1 illustrates connections between the fairness aspects discussed, the perceived fairness of carbon pricing and acceptability. It accounts for people recognizing carbon pricing to involve personal, distributional and procedural aspects, which jointly affect people's perceived overall fairness of carbon pricing as pointed out in the literature (analysed in Section 3.1). In turn, this perception influences acceptability of the instrument, which is addressed in Section 3.2. An innovative element of this review is the attention given to the use of carbon pricing revenues and how this affects perceived fairness and acceptability (addressed in Section 4). This involves two channels. The first one corresponds mostly to revenue uses (e.g. redistributive) that affect people's perceptions of personal consequences, distributional effects and procedural aspects of carbon pricing and, consequently, influence people's perception of fairness and policy acceptability. The second channel refers to environmental and other uses of revenues that affect policy acceptability without being mediated by fairness. This review should be regarded as an exploratory attempt to address the relevance of fairness in explaining people's acceptability of carbon pricing. Thus, we shed light on people's perceptions of personal, distributional and procedural aspects that influence people's perceived fairness and thereby have a significant role in policy acceptability.
The remainder of the paper is structured as follows. Section 2 explains the procedure used to select primary studies to be reviewed. Section 3 systematically reports findings of the review regarding the impact of carbon pricing on policy acceptability through perceptions as discussed in relation to Figure 1. Section 4 analyses whether people prefer uses of carbon pricing revenues that make the policy fairer. Section 5 discusses potential reasons that explain why people show high concern over distributional effects of carbon pricing but do not rank highly revenue uses aimed at making the policy fairer. Section 6 provides conclusions and recommendations for improving the acceptability of carbon pricing proposals.

Literature search and method of analysis
An extensive literature search of studies dealing with people's opinion on carbon pricing schemes with a focus on perceived fairness was conducted using the academic database 'Scopus'. We aimed at covering a wide range of fields, from economics to psychology and political science. To this end, a variety of search terms were used: (accepta* OR view OR attitude OR opinion OR preference) AND ('carbon pric*' OR 'carbon trad*' OR 'carbon tax' OR 'ecological tax reform' OR 'cap and trade' OR 'energy tax' OR 'environmental tax' OR 'climate tax' OR 'permit trad*' OR 'emission trad*' OR fee OR dividend) AND (equity OR fairness OR 'recycling revenue' OR 'distribut*' OR equality OR social OR poverty OR rebate OR earmark OR justice). We further included in the sample relevant studies cited in the reference list of papers found through Scopus. Studies were included if they: (i) involved analysis of primary empirical data and (ii) made explicit reference to carbon pricing or (iii) addressed other pricing instruments (e.g. fuel taxes) in the context of climate change. Note that the third criterion means that we did not include more general studies of environmental taxes or road pricing. While possibly not exhaustive, this procedure ensures a comprehensive coverage of the literature.
A total of 43 empirical studies were identified, among which 37 deal with carbon taxes and similar schemes (such as environmental tax reform), three studies with cap and trade schemes and three studies with both schemes together. We extracted the following information from each study: authors; journal and year of publication; location of the study (country); research objective; method; year of data collection; sample size and representativeness; type of stakeholder questioned; discipline or research area to which the study belongs; characteristics of the carbon pricing instrument; whether or not carbon pricing was implemented in the considered country at the time of the study; people's perceptions of personal consequences, distributional effects and procedural aspects; and effects of these on people's acceptability of carbon pricing; preferences for uses of revenues and their effects on fairness and policy acceptability.
Most studies were published after 2006 with only three studies as exceptions. In terms of representativeness of the studies, 18 of the 29 quantitative studies reported that they were representative of the scale at which the study was conducted: three of them at the city level and the others at the whole country or state level. Our results show that English-speaking countries are overrepresented in the sample of studies analysed. Most studies were undertaken in the USA (10 papers; 18% of the whole sample of studies), followed by the UK (eight papers, 14%), and Australia and Sweden with six papers (11%) undertaken in each country (see Figure 2). Moreover, almost all countries except Australia, India and South Africa are North American or European.
In order to analyse personal, distributional and procedural aspects associated with carbon pricing as perceived by respondents in primary studies, we identify excerpts from the articles reviewed related to concerns expressed by respondents about carbon pricing. We cluster the excerpts that expressed identical or similar aspects of carbon pricing and systematically coded them in order to establish different types of personal, distributional or procedural concerns. We counted the number of papers that reported each concern. Note that all opinions reported belong to the general public, except when a concrete group of stakeholders is specified, which mainly corresponds to business representatives and decision-makers. The broad spectrum of studies we wanted to include, and the diversity of methodological approaches they used, generated challenges in terms of data analysis and comparability. Most studies did not ask people to list their perceived personal, distributional and procedural concerns but rather the researchers evaluated people's perceptions of policy impacts presented to them. Thus, our results should be understood in terms of which perceived personal, distributional and procedural concerns are more or less frequently addressed in the literature and people's opinion regarding them. With the aim of assessing the effects of perceptions on acceptability of the policy instrument, we counted the number of papers that found a relationship between respondents' perceptions and acceptability of the policy instrument. We distinguish between results from qualitative (interviews and focus groups) and quantitative (questionnaire survey) studies while also adding nuances for each as they cover different types of data. For example, the quantitative studies report the statistical significance of the relationship between perceived effects and policy acceptability. To assess people's preferences for different uses of revenues, we used the available information on how people rank them. Next, we assessed whether those uses that can potentially make the policy fairer are also the ones most preferred. Finally, we examined the influence of the use of revenues on instrument acceptability by counting the number of papers that found a relationship between these two variables.

Perceived personal consequences, distributional effects and procedural aspects
The articles reviewed identified a variety of perceived personal consequences, distributional effects and procedural aspects of carbon pricing. Table 1 provides a list of these effects and aspects that appeared in more than one study along with how many studies mentioned each instance.

Perceived personal effects
The most mentioned personal consequence of carbon pricing in the reviewed studies is that it may imply higher energy prices and less purchasing power. 3 This effect was reported in 13 studies. For instance, Carattini et al. (2017) find that an argument widely used by the opponents of carbon taxes in Switzerland was that higher carbon tax rates would increase the consumer price of energy. Respondents in Kaplowitz and McCright (2015) referred to structural constraints that citizens face, such as the lack of alternatives to driving as one argument against a gasoline tax which would generate unacceptable additional expenditures. Among these studies, six of them addressed the reduction of purchasing power. For instance, many Swiss respondents (67% of a sample of 338 individuals) in  were concerned about a reduction in consumption and purchasing power due to the implementation of a carbon tax. The loss of jobs as a potential impact of carbon pricing was mentioned in five studies. Irish and UK citizens participating in focus groups undertaken by Clinch et al. (2006) stated that the implementation of an environmental tax reform could lead to job losses in Table 1. Perceived aspects of carbon pricing, ordered by frequency in primary studies.

Number of studies
Perceived personal effects Higher energy prices and less purchasing power 13 Beuermann & Santarius, 2006;Carattini et al., 2017;Dresner, Jackson, et al., 2006;Harwatt, Tight, Bristow, & Gühnemann, 2011 α ;Kaplowitz & McCright, 2015;Klok et al., 2006;Lo, 2013  their countries. Four studies reported a concern for less comfort and well-being as a consequence of carbon pricing instruments. Beuermann and Santarius (2006) reported that German respondents were worried about the implications of an environmental tax reform on their daily lives, particularly regarding the maintenance of their living standards, due to an increase in fuel prices. Four studies indicate that carbon pricing limits freedom of choice. For instance, Owen, Edgar, Prince, and Doble (2008) found that some UK participants in focus groups expressed considerable reluctance to the idea of imposing limits on individual carbon emissions. People were concerned about the government exerting influence over how they live and what they consume. Three studies mention voluntary behaviour punished rather than appreciated. This effect relates to 'motivational crowding-out' which denotes that carbon pricing weakens people's intrinsic motivations and decreases their voluntary efforts to reduce carbon emissions (Frey & Jegen, 2001). This effect links to fairness perceptions in the sense that some people feel they are unfairly 'punished' by a carbon tax because they already undertook voluntary 'climate action' (e.g. Baranzini, Caliskan, & Carattini, 2014). That is, they would have to incur an extra cost due to the tax, no matter whether they have already undertaken voluntary climate action.

Perceived distributional effects
The distributional impacts of carbon pricing that received most attention, namely in 14 studies, were impacts on poor households. Respondents in Beuermann and Santarius (2006), Clinch and Dunne (2006) and Kallbekken and Aasen (2010) expressed the concern that when people have few low-carbon options available, a carbon tax affects poor households harder than wealthier ones. Unlike the latter, the former suffer a relatively high cost or are even forced to refrain from consuming the taxed good or service. Clinch and Dunne (2006) reported that, at the time of their study, Ireland had one of the highest fuel poverty rates in Europe. This may have contributed to respondents worrying that implementing an environmental tax reform might hinder access to heating fuel for many households, particularly for those outside the tax system and thus unable to benefit from a reduction in labour taxes. Among business representatives interviewed, only those in Ireland expressed concern about effects on poor households whereas those from Denmark, France, Germany and the UK did not pay much attention to poverty worsening as a result of environmental tax reform .
Eleven studies referred to a policy burden unevenly distributed between firms and households or among the latter. Some of these identified concerns about the regressive nature of the instrument in a general way. For instance, in Baron and Jurney (1993), most individuals thought that a gasoline tax to combat global warming would unfairly distribute the costs of change. Other studies considered it unfair that the burden of the policy is ultimately passed onto consumers through the increase in prices of CO 2 -intensive products . Some people alleged that companies would be able to write off most of the policy burden in a way that the ordinary tax payer would not. Respondents in other studies, such as Owen et al. (2008), focused on the implications for high-income people and argued that they will be able to afford whatever level of taxation is imposed on them, generating an unfair situation with respect to low-income households.
Respondents in five studies expressed concerns over the implications of the policy burden unevenly distributed among provinces, states or countries. There are two different scales, between and within countries. In the first case, Brannlund and Persson (2012) reported that Swedish respondents preferred a reduction of CO 2 emissions taking place at the European level (EU) rather than only at the scale of Sweden in order to share the burden of reducing emissions. They also found that Swedish people appear to be equally concerned about the social and geographical distribution of the policy burden. With regards to uneven distribution within the same country, Lachapelle, Borick, and Rabe (2012) show that a majority of Americans (59%) and Canadians (62%) do not appear to worry much about fairness implications due to an uneven distribution of the policy burden among the state and the provincial levels because they reject the idea that action taken by their respective state/province should be conditional upon the actions of others. Four studies showed that people are concerned about impacts on rural households. Respondents in Baranzini et al. (2014) suggested this, as there are fewer possibilities of substitution between private and public transportation in the countryside than in urban areas.
Next, participants in four studies mentioned impacts on the elderly and people in need of care because they probably need to spend more time at home and have higher expenditure of fuel for heating their homes. For instance, Dresner, Jackson, and Gilbert (2006) reported this concern among UK respondents. Four studies identify concern about exemptions and privileges given to some sectors. Some participants in the study in Dresner, Jackson, et al. (2006) criticized the Climate Change Levy in the UK for avoiding taxation of the coal industry. Carson, Louviere, and Wei (2010) found that a small majority (53.7%) of Australian respondents was in favour of not giving special treatment to energy intensive sectors. Not surprisingly perhaps, business representatives expressed a completely different opinion, highlighting the positive effect of exemptions on firms' competitiveness and criticized carbon taxes Klok, Larsen, Dahl, & Hansen, 2006).
The initial distribution of permits has been identified as one of the main determinants of distributional impacts of tradable permits. For instance, Bristow, Wardman, Zanni, and Chintakayala (2010) analysed the preferred distribution of permit allocation from a potential personal carbon trading scheme among UK residents. The preferred option was the one based on an equal allocation to all with extra permits for those with greater needs, for example, living in rural areas, with poor housing or having any disability. On the contrary, the most unacceptable allocation was the one whereby permits were allocated to adults based on a Government assessment of needs.
A study by Hammar and Jagers (2007) analysed people's perception of the distribution of the burden of reducing CO 2 from private car transportation taking into account three distributional principles (equity, needs-based and equality). It found that most respondents (75%) perceived as fair that people who pollute the most decrease their emissions more than others, in line with what they call an 'equity principle'. Fewer respondents (54%) judged as fair that people who need their cars the least, e.g. who have good access to public transportation, should decrease their emissions more than others, in line with a so-called 'needs-based principle'. Less than half of the respondents (48%) judged as fair that everybody should decrease their emissions at the same rate, e.g. by 10%, in line with what they call an 'equality principle'.

Perceived procedural aspects
The procedural aspect of carbon pricing that created most concern among respondents was distrust in the government, mentioned in 12 studies by a significant amount of respondents. Such distrust appears to take various forms. For instance, in a study on Sweden, Hammar and Jagers (2006) showed that 64% of a sample of 1270 respondents expressed a low level of trust in politics in general terms. A study on Ireland by Clinch and Dunne (2006) reported distrust in the capacity of the government to implement an environmental tax reform. In addition, Beuermann and Santarius (2006) found that distrust of German respondents in the government was motivated by the insufficient information about carbon pricing provided by the administration. A related procedural aspect, identified by nine studies, relates to distrust in the government for the use of carbon pricing revenues. These indicated a lack of trust that the government would keep its promises about revenue use (e.g. lowering labour taxes) or guaranteeing revenue neutrality (see e.g. Beuermann & Santarius, 2006;Hsu, Walters, & Purgas, 2008).
Seven studies reported dissatisfaction with governmental information provision about the policy. For instance, Kallbekken and Aasen (2010) found for Norway that one-third of respondents thought providing information about policy instruments such as taxes is necessary to make them understandable. Moreover, respondents in Klok et al. (2006) declared that the government should provide information about whether CO 2 reduction objectives had been met. Similarly, both business representatives and the general public from Germany expressed their concerns about making environmental tax reform more transparent (Beuermann & Santarius, 2006). Respondents from three studies referred to the insufficient consultation of social partners in designing and implementing carbon pricing. Deroubaix and Lévèque (2006) showed that controversies about environmental tax reform in France emerged due to a very small number of protagonists taking part in key decisions concerning its design, a process characterized by confidential negotiations with energy-intensive industries' representatives and exclusion of unions and NGOs. Respondents in two studies expressed distrust in market mechanisms to deal with CO 2 reduction. For instance, Lo, Alexander, Proctor, and Ryan (2013) found that the possibility of cheating and market manipulation was deemed to be the biggest concern surrounding emission trading schemes by respondents.
Only one study, for the UK, compared people's perceptions of the three policy instrumentscarbon tax, personal carbon trading and upstream carbon trading (Owen et al., 2008). Some participants in the focus groups thought that (downstream) personal carbon trading is a more equitable approach than either (upstream) emissions trading or a carbon tax. However, the latter two instruments were considered as being easier to implement.

Influence of perceptions on policy acceptability
Now we move to the next step in our analysis, which is guided by the question whether perceived fairness of carbon pricing affects public acceptability of it. Table 2 presents a summary of our findings.

Influence of perceived personal effects on acceptability
Several perceived personal consequences were found to have an effect on policy acceptability (see Table 2). Here we explain the effects found in more than one study. Seven quantitative studies and one qualitative study reported that higher energy prices and less purchasing power decreased acceptability of carbon pricing. For instance,  found that affirming that carbon taxes are an issue for purchasing power is linked with about 15% lower probability of accepting the instrument. In the case of the qualitative study, respondents giving importance to individual economic interests were less inclined to vote positively for carbon pricing (Beuermann & Santarius, 2006). Three studies found that perceptions of potential loss of jobs due to carbon pricing negatively influenced acceptability, although two of them were not significant Carattini et al., 2017). Another relevant effect positively influencing people's acceptability of carbon pricing found in two studies was the perception that the policy does not limit freedom of choice.

Influence of perceived distributional effects on acceptability
With respect to distributional effects, four studies found that a neutral cost distribution, that is, all citizens pay the same share of income, increases policy acceptability. This is illustrated by two studies in which a change from a regressive cost distribution (all citizens pay the same amount) to a neutral one affected positively the acceptance of the policy. Similarly, Baranzini and Carattini (2014) found that some people would accept a tax if it was proportional to income and consumption, therefore based on the situation of each household. Meanwhile, Hammar and Jagers (2007) assessed perceived fairness based on people's preference for one of the three distributional principles for distributing the burden of carbon taxation on private car transportation. They found that what they call the 'equality principle' seems most important in increasing acceptability of a higher carbon tax, that is, everybody should decrease their emissions the same, e.g. by 10%. The perception that the instrument could be regressive was found to influence its rejection in two quantitative studies and another qualitative study, although it was also found to not significantly affect acceptability in . The progressive distribution of the costs of the instrument determined people's acceptability of carbon pricing in three studies. As an example, Hammar and Jagers (2007) revealed that policy acceptability increased if those who should decrease their emissions the most are those who needed their cars the least, e.g. who had access to public transportation. This distribution of the policy costs followed what the authors called a 'needs-based principle'. People who accepted the carbon tax in Dreyer and Walker (2013) and Dreyer, Walker, McCoy, and Teisl (2015) had a higher belief in ensuring that large industries are subject to a carbon price than those who did not.

Influence of perceived procedural aspects on acceptability
Concerning procedural aspects, four quantitative studies and four qualitative ones found that policy acceptability increases when people are satisfied with governmental information provision about the policy. Two of them (Klok et al., 2006;Lo et al., 2013) showed that this is particularly the case when the information emphasizes the effectiveness of the policy instrument. Nevertheless, another study found that having information about the instrument did not affect acceptability significantly (Rhodes, Axsen, & Jaccard, 2014). Trust in the government was reported in four quantitative studies and three qualitative studies as increasing the probability of acceptability of carbon pricing. Hammar and Jagers (2006) further reported a strong role of trust showing that motorists who trust their politicians were no more likely to resist CO 2 tax increases than high-trusting persons with no access to a car. On the contrary, low-trusting people with access to a car were the group found to be most strongly against an increased CO 2 tax.

Influence of perceived overall fairness on acceptability
Five quantitative studies and one qualitative study found that perceived overall fairness positively affects public acceptability of carbon pricing. For instance, Dreyer et al. (2015) and Dreyer and Walker (2013) measured perceived overall fairness of a carbon tax that held industries responsible for emissions they release and included a  Beuermann & Santarius, 2006*;Carattini et al., 2017;Carson et al., 2010;Dresner, Dunne, et al., 2006*;Hsu et al., 2008;Klok et al., 2006*;Lo et al., 2013  compensation package for households indirectly impacted. The latter study found not only that individuals who reported higher levels of perceived fairness were more likely to find the policy acceptable but also that for a oneunit increase in the levels of perceived fairness, an individual was about two and a half times more likely to support the policy. Only one study from the sample showed no significant relationship between perceived fairness and acceptability or support for the instrument (Baron & Jurney, 1993).

People's preferences for use of carbon pricing revenues
A unique feature of carbon pricing is that it generates revenues that can be used for various purposes, with potential consequences for equity, the environment and political support. Information regarding how people feel about different revenue uses can arguably improve political decisions about these by knowing which uses can count on more public support and can therefore increase overall political acceptability and feasibility. For this reason, we shed light on the preferred uses of revenues by the general public that received attention in the studies reviewed. We focussed on whether people prefer uses of revenues that ameliorate the regressive effects of carbon pricing. We therefore distinguished between three main types of uses of the revenues: environmental, redistributive and other type of uses. Environmental uses refers to funding projects that reduce CO 2 emissions or fund low-carbon energy sources; redistributive uses denote achieving a less regressive outcome; and other uses cover the reduction of existing taxes or inclusion in the general government budget. The most preferred use of revenues was found to be funding environmental projects, as reported in fifteen studies ( Table 3). The preference for using the revenues for environmental purposes depended on the kind of project proposed, where funding research and development of renewable energies was the 'environmental project' most mentioned, whereas purchasing foreign carbon credits received the lowest preference in Carattini et al. (2017). Admittedly, most studies did not properly specify the concrete environmental allocation of the revenues to have an accurate assessment of people's preferences. For instance, they used general terms such as 'environmental earmarking' or 'environmental projects' without it always being clear what this entailed. Even studies that referred to renewable energy projects were unclear as governments do not tend to fund directly such public projects, but at best subsidize innovation or adoption of private environmental projects such as on renewable energy. Another revenue-use option within the environmental uses preferred in two studies was funding technological innovation and energy efficiency, such as grants for residential insulation or subsidies for energy-efficient innovations. This option was the most popular among business representatives Dresner, Dunne et al., 2006) who claimed that they needed technology grants and financial support to ensure improvements in energy efficiency. Some business representatives, though, regarded using revenues for investment subsidies as unfair for punishing companies that already undertook energy efficiency efforts voluntarily and rewarding ones that did not (Klok et al., 2006). Several revenue-use options dealing with redistribution in order to obtain a less regressive outcome were identified in the studies. Redistributing revenue to vulnerable groups such as low income households and the elderly is purposely designed to be progressive by providing lower-income households with a higher amount of the tax revenue. The elderly are separately mentioned as they are especially vulnerable to low temperatures and therefore to higher energy prices. It was found to be the most preferred option in three studies and the second one in three other studies. Equal share of revenue given to each taxpayer was the second preferred option in two studies and the third option in one study. It is considered progressive because fixed amounts of compensation account for a greater proportion of income in low-income households. Moreover, since low-income households tend to spend less, in absolute terms, on energy consumption than high-income households, then the former will receive more through the monetary transfer than the cost increase they suffer (Chiroleu-Assouline & Fodha, 2014; Klenert & Mattauch, 2016).
Within other type of uses, recycling revenue through reducing existing taxes, which is known as environmental tax reform, is considered the preferred option and the second best option by two studies each. However, reducing existing taxes did not receive much support from the general public, and was found to be the least favoured option in seven studies. There are mainly three reasons that can explain this opposition.
Firstly, people do not trust that the government will actually do as promised. Dresner, Jackson, et al. (2006) found that people 'did not believe that other taxes would be reduced in reality or that the money would be spent on what was promised.' Secondly, people are sceptical about the idea of a double dividend of environmental tax reform (Deroubaix & Lévèque, 2006;Klok et al., 2006). Double dividend refers to recycling carbon pricing revenues by reducing distortionary taxes (e.g. labour taxes, income taxes), which may have positive impacts on economic growth, employment, or technological development (Bovenberg & van der Ploeg, 1994). Some studies recognize the difficulty of achieving a double dividend due to the complex interactions between carbon pricing policies and the fiscal system (see Goulder (2013) for a detailed identification of the necessary conditions). Thirdly, people question the environmental effectiveness of shifting the tax because if people have the same amount of money at the end, there is no incentive to reduce consumption of taxed-products . In this sense, people appear to have little appreciation that the relative cost of goods and services would change as a consequence of a tax shift. Tax rebates received support as the second most preferred option in five studies. Baranzini et al. (2014) report that people preferred it because it could be a way to reward those companies and households reducing their energy demand. Financing public transport was the most preferred option in one study but the least in another. Finally, the uses that received least support were government deficit reduction and revenues going into the general budget.  Klok et al., 2006) 1 (Dresner, Jackson, et al., 2006) 2 Hsu et al., 2008) Redistributive Redistributing revenue to vulnerable groups (low-income households, elderly) 3 Carson et al., 2010; 3 Gevrek & Uyduranoglu, 2015) 3 (Baranzini et al., 2014;Klok et al., 2006;Rotaris, 2017 Only one study showed explicitly the extent to which different uses of revenues affect people's perception of fairness of the instrument. Jagers et al. (2018) analysed the effects of two different uses of CO 2 tax revenues in a proposal to increase this tax in Sweden, a general cut in income tax for all Swedes and an individual tax rebate only to those paying the CO 2 tax. The study reported that the fairness rating increases from the option without any compensation to the option for equal compensation through a general cut in income tax and further increases with an individual tax rebate. However, neither equal nor individual compensation have any noticeable effect on policy support.
Our results show that options that use the revenues generated by carbon pricing to make the policy fairerredistributing revenue to vulnerable groups or equal share of revenue given to each taxpayerwere preferred in certain studies whereas environmental uses of revenues tended to be more preferred overall.

Influence of people's preferences for revenue use on carbon tax acceptability
We now analyse whether particular uses of revenues from carbon pricing as discussed in Section 4.1 affect its public acceptability. An overview of the results is given in Table 4. In general, the level of acceptability increases when revenues are recycled in some way, rather than going to the general government budget. For instance,  found that acceptability increased to 64% of a sample in Switzerland when revenues were recycled according to people's preferred recycling option, which was 15% more than without earmarking. However, 36% of the sample considered that earmarking revenues were not sufficient for them to accept the carbon tax.
Earmarking for environmental projects, in particular renewable energy development, affected positively people's acceptability of carbon pricing in four quantitative studies and in four qualitative papers. Nevertheless, it decreased acceptability in one study . One explanation for the discordant result found by Table 4. Effects of people's preferences for revenue use on policy acceptability.

Use of revenues
Significant effect Non-significant γ
With regards to revenue-use options that make carbon pricing less regressive, we found that redistributing revenue to vulnerable groups and an equal share of revenue given to each taxpayer increased people's acceptability in five studies (one non-significant) in case of the former and two studies in case of the latter (one non-significant). Within other types of revenue uses, an environmental tax reform (reducing existing taxes) was found to affect negatively the acceptability of the instrument in six studies. However, two of them had a non-significant effect (Bristow et al., 2010;Carattini et al., 2017). Significant effects were found for a proposal for an environmental tax reform to reduce labour costs in general terms (Beuermann & Santarius, 2006;Dresner, Dunne, et al., 2006) and a general cut in income tax for all Swedes (Jagers et al., 2018). However, another study found that reducing taxes on income as well as goods and services positively affected people's acceptability of carbon pricing (Hsu et al., 2008). If we look closer at preferences among different types of stakeholders, studies found that implementing an environmental tax reform and using the revenues for giving grants to industry increased carbon pricing acceptability among business representatives (Klok et al., 2006). On the contrary, Deroubaix and Lévèque (2006) found that using carbon pricing revenues to reduce existing taxes decreased business representatives' acceptability of carbon pricing.
Using the revenues for refunding citizens through a tax rebate showed contradictory results. One study (Kaplowitz & McCright, 2015) found that policy acceptability increased among Americans when the extra revenue was refunded equally to American families via a tax rebate. In contrast, a study by Jagers et al. (2018) showed that an individual compensatory scheme directing tax rebates only to those paying the CO 2 tax decreased policy support among Swedish citizens. Funding public transportation positively influenced people's acceptability of carbon pricing in three studies.

Discussion
Our review has shed light on perceived personal consequences, distributional effects and procedural aspects that co-determine perceived fairness, which in turn affects policy acceptability. It further shows that people's perceived fairness of carbon pricing positively affects its acceptability.
We have found that people expressed high concern over the uneven distribution of the policy burden, with a particular focus on poor people. Furthermore, the perception that the instrument can be regressive reduced people's acceptability of the policy whereas a neutral and progressive distribution of the policy costs increased instrument acceptability. Somewhat surprisingly, people did not show a clear preference for using the revenues generated by the carbon pricing scheme in ways that make the policy fairer by reducing its regressive effects. Instead, funding renewable energy projects was most often the preferred option. Several factors might explain this.
Firstly, results showed that many people are sceptical that taxation will change behaviour much on its own. This scepticism possibly underlies the finding that people think revenues should be used to reinforce emissions reduction by allocating them to relevant environmental projects. Baranzini et al. (2014) found that people considered that if the revenues are used for something different than environmental purposes, then the government would lose its credibility regarding how urgent addressing climate change is.
Secondly, people seemed to confuse regulatory taxes aimed at changing behaviour (from high-to lowcarbon consumption or production) with taxes aimed at generating public revenues. Consequently, they seemed to incorrectly confuse the main purpose of environmental taxes with the second type, i.e. raising revenues and thus considered it logical to use these revenues for environmental purposes. Some hope for public understanding of this difference comes from Kallbekken and Aasen (2010), who found that more than half of their participants understood the principal purpose of environmental taxes as regulating and reducing emissions. This suggests people are able to understand regulatory taxation and that information explaining the purpose of carbon taxes can create broader support for them. However, Kallbekken, Kroll, and Cherry (2011) found no evidence that providing information about the functioning of Pigouvian (regulatory) taxes reduces tax aversion.
Thirdly, a generally low preference for revenue uses to decrease regressive effects of carbon pricing might be due to a lack of either awareness or trust that carbon pricing with appropriate revenue recycling is progressive or fair. According to Carattini et al. (2017), making distributional effects salient increased people's preference for progressive designs of carbon taxesthrough payments for low-income households or by an equal share of revenue given to everyone. A fourth explanation, illustrated by Baranzini et al. (2014), is that, despite being concerned about the regressive impacts of carbon pricing, some people did not consider redistributing revenue to low-income households a viable option due to its large administrative burden. Having said all this, it should be noted that redistributing revenue to low-income households was still the second most favourite revenue use option, and significantly increased people's acceptability of carbon pricing.
The preference for redistributive uses of the revenue is linked to the socioeconomic characteristics of people surveyed. For instance, Carattini et al. (2017) found that low-income households showed a marked preference for progressive designs. Gevrek and Uyduranoglu (2015) provided similar evidence in this regard. This suggests that individual socioeconomic characteristics influence people's concerns about distributional and procedural effects. Future research could look more systematically into the relationship between differences in socioeconomic characteristics and fairness perceptions, as well as the links between perceived personal and distributional effects, and how all this in turn affects policy acceptability.
Procedural aspects seem to play a greater role in determining the preference for revenue uses than distributional effects, especially regarding environmental earmarking. Some studies linked people favouring environmental earmarking with distrust in governments, as earmarking constrains policymaking and thus functions as a credible commitment device (Deroubaix & Lévèque, 2006). Procedural aspects can explain the low preference for environmental tax reform (lowering other taxes) mainly because people do not trust that the government will actually keep its promise.
It is widely accepted that all countries, in particular large CO 2 emitters, need to reinforce their climate policy in order to fulfil the Paris Agreement. Carbon pricing is a recommended policy for that goal. However, a lack of public acceptability can jeopardize its effective implementation and stability. Paying attention to policy acceptability and designing carbon pricing schemes that can count on people's support is thus necessary to ensure policy roll out. The effects of carbon pricing on personal circumstances, distributional effects and procedural aspects as well as its perceived fairness should be carefully analysed in each specific case as they will depend on the precise design of the carbon pricing scheme. Recent French protests against the implementation of a carbon tax illustrate the urgency and need for better understanding of the reasons why people accept carbon pricing schemes or not. The French case demonstrates that people's perceptions of fairness of the policy instrument are complex and depend on a broad set of contextual factors. For instance, lack of trust in government and the implementation of other regressive policies by French President Macron -such as the simultaneous cancellation of a wealth tax -likely had considerable influence on the popular rejection of the tax.

Conclusions
We reviewed the role of perceived fairness in studies addressing people's acceptability of carbon pricing. A main finding is that peoplés perception of carbon pricing as fair strengthens its policy acceptability. The review indicated that fairness perceptions cover personal consequences, distributional effects and procedural aspects.
Our results reveal further that higher energy prices and reduction of purchasing power are the personal consequences mentioned most often, whereas impacts on poor households and uneven distribution of the policy burden between firms and households or among the latter are the most perceived distributional effects. Distrust in government and over the good use of revenues generated by the carbon pricing instrument were frequently mentioned concerns related to procedural aspects of the policy. Our results showed that expected higher energy prices, less purchasing power and a regressive incidence of the policy decrease policy acceptability. Several procedural aspects increase carbon pricing acceptability, such as people's satisfaction with governmental information provision about the policy, and people trusting their government.
Our review indicates that although people express high concern over the uneven distribution of the policy burden, they often prefer to use carbon pricing revenues for environmental projects rather than inequity compensation. In addition, it should be noted that redistributing revenue to vulnerable groups still comes out as the second most favourite revenue use option, which significantly increases people's acceptability of carbon pricing.
Although our results are mostly from North American and European countries, several general policy insights can be provided in order to increase public acceptability through making carbon pricing fairer. Carbon pricing needs to be implemented as a complementary instrument to other policies, in particular ones that ensure the development of products and services that are less CO 2 -intensive. A common concern expressed regarding procedural aspects was that insufficient information is provided by governments about carbon taxation and the use of revenues. Hence, providing more information could be a relatively cheap and effective way to increase public acceptability. Moreover, it would likely improve people's trust in government, which was the main concern expressed about procedural aspects.
Preferences for using the revenues to make the policy progressive may strengthen in the future, particularly in countries that suffer considerable (or rising) socioeconomic inequality. In this sense, our results suggest that a combination of redistributing revenue to vulnerable groups and funding environmental projects, such as on renewable energy, seems the best option to explore in the future in order to significantly increase policy acceptability. In fact, an analysis of the various countries that had implemented carbon pricing schemes by 2013 revealed that almost none of them combine these two revenue uses (Carl & Fedor, 2016). New carbon pricing schemes have been implemented recently, which allows for further research to compare our findings regarding public acceptability of carbon pricing recycling options against how revenue is recycled in more recently implemented carbon pricing schemes.
Future research is also needed regarding public policy acceptability in low-and middle-income countries, given that carbon pricing policies are spreading there as well. Moreover, further exploring public preferences for different types of environmental projects in which to invest revenues is necessary, because the nature of, and realistic prospects for, such projects are not very clear. For example, renewable energy projects tend to garner much support but are often undertaken by private partners rather than the public sector. Public support in this case could take the form of innovation or adoption subsidies, which might, however, trigger different public responses in terms of fairness perceptions and policy acceptance than the more general notion of 'funding environmental projects'. We encourage future work to test and refine proposed frameworks for carbon pricing, taking systematically into account all the relationships described.

Notes
1. The literature we review here uses different terms to refer to similar concepts, such as policy acceptance, acceptability and support. These are not always clearly defined or distinguished. For this reason, we abstain from making a sharp distinction between them as well. For an in-depth discussion of the concepts, see Kyselá, Scasný, and Zverinová (2019). 2. Factors other than perceived fairness affect public acceptability of climate policy, such as negative effects on international competitiveness and exports, or more generally on the economy and employment. 3. Terms in italics denote impacts identified in Table 1.

Disclosure statement
No potential conflict of interest was reported by the authors.