1 
  DEPARTMENT OF POLITICAL SCIENCE
 
EN ROUTE TOWARDS FAIRER 
BURDEN SHARING  
Targeting transport-related carbon taxes 
towards the rich and the rest 
Jacob Schönning Landin 
Master’s Thesis:  30 credits 
Programme:   Master’s Programme in Political Science: Environmental Governance and Behavior 
Date:    26 May 2025 
Supervisor:   Niklas Harring 
Words:   15 754 
2
Abstract 
Carbon taxation is an effective policy instrument for reducing GHG emissions, but its adoption and 
development is often hampered by public resistance. Since distributional fairness has emerged as a key 
determinant of carbon tax acceptability, efforts are needed to explore ways of strengthening carbon 
taxation in a fair manner. In this thesis, I applied experimental survey methodology to test how 
fairness perceptions of a carbon tax increase on car fuels were affected by packaging the tax increase 
with the introduction of a luxury carbon tax. Furthermore, it was examined which of two framings of 
the luxury carbon tax most improved fairness perceptions: one relating the policy to a principle of 
equality or needs, respectively. Based on OLS regressions on newly collected data from 178 adult 
Swedish residents, the policy package was found to significantly improve perceived fairness, with a 
small-to-medium effect size. Results regarding motivation framing were more indecisive, but at least a 
needs frame was found to significantly improve perceived fairness.  
Keywords: carbon taxation, luxury carbon tax, fairness perceptions, fairness principles, policy 
package, EVM, Sweden. 
1. Introduction 
1.1 Carbon taxation

To achieve the goals of the Paris agreement, rapid reductions in GHG emissions are needed, especially in 
rich, developed countries. That in turn requires the adoption of effective climate policy instruments. At the 
same time, policy makers should strive for cost-efficiency in their use of public resources. One of the most 
goal-effective and cost-efficient tools for mitigation is the carbon tax, and as a result, it is widely praised and 
demanded by economists, environmental scientists and ecological economists (Sommer et al 2022; Maestre-
Andrés et al 2019; Jagers et al 2020). Furthermore, it has been argued to be a fair and reasonable instrument, 
because it realizes the polluter pays principle (PPP) (Jagers and Hammar 2009; Hammar and Jagers 2007; 
Jagers et al 2020). However, it is less popular among the general public. In some cases, proposals to adopt 
new carbon taxes have been withdrawn or suspended after strong public resistance (Maestre-Andrés et al 
2019, p. 1187).  
This discrepancy has prompted research on what drives attitudes towards carbon taxes. From that literature, 
fairness perceptions has emerged as an important predictor (Maestre-Andrés et al 2019). In addition, in meta-
analyses of predictors for climate policy instruments in general (Bergquist et al 2022) and for sustainable 
transportation policies (Isaacson et al 2024), it has been found to be the very strongest determinant of all 
included variables. A recent experimental study has even provided evidence for direct causality from 
perceived (distributional) unfairness to attitudes towards carbon taxes and other environmental taxes 
(Bergquist 2025). 
3
Furthermore, the same strand of literature has explored what makes people think of a climate policy as fair or 
unfair, and identified three important aspects: personal costs, decision procedure and distributive 
consequences (Maestre-Andrés et al 2019). Of these, distributive consequences holds a primary position 
(Bergquist et al 2022; Bergquist 2025; Lindvall et al 2024). Policies perceived to impose larger relative costs 
on poor and vulnerable groups in society (so called regressive incidence) are often raising fairness concerns 
or perceived as unacceptable. Meanwhile, when policies are perceived to impose costs in proportion to 
people’s income (neutral cost distribution) or larger costs on the wealthy (progressive incidence), that can 
have a positive effect on acceptability (Maestre-Andrés et al 2019, p. 1194-1195). Hence, changing a policy’s 
cost distribution, or people’s beliefs about it, could affect fairness perceptions.  
Now, whether carbon taxes on car fuels, which are in focus in this thesis, have (strongly) regressive 
incidence or not in high-income countries seems to be an unsettled matter among economists. One study of 
seven European countries including Sweden found that the cost incidence is only weakly regressive, or 
roughly proportional (Sterner 2012). Others, looking at carbon pricing more generally, have concluded there 
is a tendency that these instruments are regressive in high-income countries, but (more) progressive in low- 
and medium-income ones (Ohlendorf et al 2021; see also references in Maestre-Andrés et al 2019, p. 1187). 
However, what matters for fairness perceptions is what people believe, and many people in Sweden and 
other countries clearly think these taxes unfairly affect poor and rural households (see f.e Ewald et al 2022; 
Povitkina et al 2021). 
One alternative to make the distribution of costs and benefits from carbon taxes progressive is to earmark 
revenues for redistributive purposes. They can, for exampled, be recycled either as tax cuts or transfers, and 
either targeted to the vulnerable groups, or to all taxpayers or residents in a country. As could be expected, 
several studies on how public attitudes towards carbon taxes are affected by recycling options have found 
these alternatives to be relatively preferred by respondents (Maestre-Andrés et al 2019, p. 1196) and able to 
increase perceived policy support or acceptability (Maestre-Andrés et al 2019; Carattini et al., 2017; Gevrek 
and Uyduranoglu, 2015; Saelen & Kallbekken, 2011), while some have either failed to find a significant 
effect (Baranzini and Carattini, 2017) or found negative effects (Hilmersson 2024). 
However, what happens if fairness concerns are taken into account at an earlier stage, already in the tax 
design? How does that affect fairness perceptions of the suggested policy? This has not been closely 
examined before. One option in line with this is to package an increase in a broad and effective carbon tax 
with the introduction of a new carbon tax which is more progressive from the start, because of which 
products and consumer groups it targets. Testing such a package (instead of simply the perceived fairness of 
the new progressive carbon tax) enables generalizing the results without requiring representative samples.  
This thesis presents a novel and, to my knowledge, unique test of one such package. A raised carbon tax on 
diesel and petrol is suggested together with the introduction of a new luxury carbon tax on transport products 
only affordable to a small wealthy elite. This approach differs from the revenue recycling approach by 
reversing the logic: instead of reacting to the regressive consequences of ordinary carbon taxes and aiming to 
4
ameliorate them, the new tax prevents such effects by design, and instead produces a highly progressive cost 
incidence by targeting the most wealthy. Of course, this does not undo the potentially regressive effects of 
the suggested tax increase on car fuels. But it is an open question how people evaluate a combination of 
these measures. 
The effect on perceived fairness is in this study examined with data from a novel survey with an 
experimental vignette design. All respondents were suggested the same increase in the existing Swedish 
carbon tax on diesel and petrol, but for those receiving a treatment, this policy was packaged with a 
suggestion to introduce a luxury carbon tax on very expensive means of transport. This was expected to 
increase policy fairness perceptions. Results show that the packaging indeed did make the suggested policy 
change more fair in the eyes of respondents. This is noteworthy, because studies like this, with a novel 
treatment and relatively small sample sizes attempting to manipulate policy attitudes risk finding no 
significant effect. Furthermore, a recent previous experimental study on climate policy attitudes (on Swedish 
residents) including a targeting of wealthy people found an opposite effect - that respondents supported such 
policy designs less (Coleman, Harring and Jagers 2023). However, this thesis does confirm earlier findings 
from Sweden and Turkey that more progressive cost distributions are preferred above more regressive ones 
for for climate policies (Brannlund and Persson 2012) and carbon taxes (Gevrek and and Uyduranoglu 2015) 
alike.  
In addition to the primary exploration sketched above, the thesis also covers an examination of how fairness 
perceptions of the policy package depend on how the luxury carbon tax is framed. Since it can be justified 
both on the basis of an equality principle, and on a needs principle, I test them against each other to see 
whether and how their effects on fairness perceptions differ. Results from this are inconclusive, however.  
Now, to situate this thesis more generally, it examines what makes people think of policy changes as (un)fair. 
The results have particular relevance for research fields and policy areas which deal with the distribution of 
resources, whether economic, environmental or of other kinds. One implication is that some policies aiming 
to adress inequalities between rich and poor might be (relatively) well received by the public.  
This speaks into ongoing discussions about how economic resources and carbon emissions alike are 
concentrated to a small elite. For example, researchers have recently called for a stronger focus on the 
climate emissions of the world’s most wealthy (Otto et al 2019). Although estimates vary, it is clear that the 
very richest globally have an enormous and ungeneralizable impact on GHG emissions. In Sweden, the 
carbon inequalities are not as stark as in other places, but nevertheless large. In this situation, policies such as 
luxury carbon taxes could reduce economic and carbon inequalities at the same time. And the results of this 
thesis suggest that such measures can contribute to making climate policy packages more fair in the eyes of 
the public.  
5
1.2 The case of Sweden

Sweden is an interesting case to study in this context for several reasons. Firstly, it was one of the first 
countries to adopt a carbon tax in 1991, and has gradually reduced the amount of exemptions and increased 
its tax rate and, reaching a level that exceeds the carbon taxes in most or all other countries (Sterner 2020). 
Hence, people have long experience of the tax. Over time, its support has been quite stable (perhaps 
increased somewhat despite large rate increases) and although the tax is relatively unpopular, it is more 
popular than other taxes (Jagers et al 2020; Ewald et al 2022). The tax has also been found to be effective at 
mitigating carbon emission (Sterner 2020; Andersson 2019). Furthermore, people in Sweden show high 
concern for climate change, and the country has long been perceived as a green forerunner and an egalitarian 
country with ”relatively generous welfare policies” (Lindvall et al 2024). In addition, the country has quite 
high levels of trust and low corruption (Ewald et al 2022). 
At the same time, the country has recently experienced heightened tensions around fuel prices. Strong protest 
movements (such as ”Bränsleupproret”) have formed against carbon taxation and its effects, perceived to be 
regressive and therefore unfair to low-income and rural inhabitants (Ewald et al 2022). After the price 
increases following the covid-19 pandemic and the Russian invasion of Ukraine in 2022, the political debate 
has often centered around the same perceived unfairness of the taxes on car fuels (Lindvall et al 2024). The 
previous and current governments have adopted policies designed to reduce fuel prices: energy and carbon 
taxes on car fuels have been cut, and the biofuel mandate quota introduced in 2018 has been drastically 
lowered, and then slightly raised again (Öljemark 2025a; Öljemark 2025b; Lundberg 2024). These changes 
are estimated to have contributed to an observed increase in emissions in 2024, and the current government 
has received heavy criticism for its climate policies, with an emphasis on transport policies, from its own 
expert agencies and other relevant actors(Söderholm 2023; Goldmann 2024; Klimatpolitiska rådet 2025; 
OECD 2025). 
For the hypotheses in this thesis, rapidly increasing economic inequality observed in the country (Cervenka 
2022; Finanspolitiska rådet 2024) may also be relevant. The increasing inequalities are often attributed to the 
country’s abolishment or reductions in most taxes on capital, and is recurrently subject to reports and media 
attention. Furthermore, reports have been published on carbon inequalities in the country (Nilsson Lewis and 
Nelsson 2023; Alestig et al 2024) and highlighted by public service media (Elfström 2024). For example, the 
top 1 % in Sweden emits almost ten times more per person each year than the poorest 50 % (Alestig et al 
2024, p. 40). It is unclear, however, to what extent these trends and estimations have become common 
knowledge, or whether the general public thinks of them in terms of (un)fairness. But if they have been 
noticed, they might contribute to a favorable context for increasing fairness perceptions of carbon taxation by 
suggesting the introduction of a luxury carbon tax. 
6
2. Literature review and research gap 
2.1 Previous research on carbon taxes and fairness perceptions

By virtue of its potential to change incentive structures while generating revenue, the carbon tax has been 
found to be one of the most efficient tools for mitigating climate emissions in terms of goal fulfillment as 
well as costs (Sommer et al 2022; Maestre-Andrés et al 2019; Carattini et al 2018). So why is it that, despite 
its appeal among experts and researchers, and despite a demand for more climate action by large majorities 
in many countries, the carbon tax often meets resistance among the wider public (Ibid)? 
Much research has been undertaken to uncover what drives public attitudes to carbon taxes and other climate 
policy instruments (for reviews, see Drews and van den Bergh 2016; Klenert et al 2018; Carattini et al 2018; 
Maestre-Andrés et al 2019). A wide range of factors have been examined, and among these, fairness 
perceptions has been found to be the strongest determinant of public attitudes to a policy. With a large effect 
size, fairness perceptions outdo not only socioeconomic factors, but also ideology and climate related beliefs 
for climate policies (Bergquist et al 2022) and (sustainable) transportation policies (Ejelöv and Nilsson 2020; 
Isaacson et al 2024). The only other comparable factor behind policy support is perceptions on how effective 
the policy is, sometimes found to have an almost as strong effect as fairness perceptions (Bergquist et al 
2022).  
However, there is some evidence suggesting that these two policy-specific beliefs might be causally 
interrelated (Bolderdijk et al 2017; Isaacson et al 2024; Bergquist 2025). First, two experimental studies 
report that perceived (distributional) fairness affected perceived effectivity of environmental taxes (Bergquist 
2025) or kilometer charging of cars (Bolderdijk et al 2017), in the latter case mediated through policy 
support. Secondly, a meta-analytic study finds that the effect of effectivity perceptions on public opinions - in 
this case concerning ”push” policies for sustainable transportation - is mediated by fairness perceptions, and 
may even be moderated by fairness perceptions (Isaacson et al 2024). These results suggests that the direct 
effect of fairness perceptions on policy attitudes is considerably stronger than the direct effect of effectivity 
perceptions. Isaacson and colleagues (2024) also suggests that the fairer a policy is perceived, the more 
important effectivity perceptions become for policy attitudes, although it should be noted that this interaction 
effect was non-significant. In other words, perceived unfairness is the biggest hurdle to overcome: only once 
it is, effectivity perceptions become more relevant.  
Just like effectivity perceptions have been found to be affected by policy support (Bolderdijk et al 2017), 
researchers have considered the possibility that fairness perceptions would not be a determinant of policy 
attitudes, but rather an effect of them, acting as ad hoc justifications of pre-made choices (for example, it is 
discussed by Isaacson et al 2024, p. 14). However, through four reported experimental studies, Bergquist 
could test this and demonstrate a causal link from (distributional) fairness perceptions to policy support 
(2025). Likewise, while targeting effectivity perceptions failed to affect support in the other experimental 
study, targeting (distributional) fairness perceptions did not (Bolderdijk et al 2017).  
7
Importantly, some researchers argue that fairness is more important for opponents than supporters of carbon 
taxes: while unfairness perceptions is a primary source of opposition, the main motivations behind support is 
rather perceptions about its necessity for the climate or its efficiency than about it being fair (Povitkina et al 
2021, p. 2). Aware of this distinction, some researchers have chosen to focus on unfairness perceptions 
exclusively (Povitkina et al 2021). The focus on unfairness perceptions and policy opposition can be justified 
with the claim that in order for a policy to be politically feasible, it is more important to avoid majority 
opposition than to garner majority support, since neutral people are usually assumed to accept proposals. 
Hence, studies on how manipulating fairness perceptions can affect policy attitudes often relate their results 
to whether or not a majority of respondents can accept a policy or not, rather than whether they actively 
support it (f.e Lindvall et al 2024). The present study follows this focus by reporting shares of unfairness 
ratings for each experimental group. 
On a similar note, some have argued that organized groups of strong opponents to a carbon tax policy can be 
one of the biggest hurdles to implementation, in combination with lack of public support (Ewald et al 2022). 
Just like opponents in general, strong opponents of carbon taxes in Sweden have also been found to be driven 
largely by fairness concerns (Ewald et al 2022; Lindvall et al 2024). Hence, if perceived fairness is increased, 
both groups might shrink and, by extension, the policy can become more feasible. 
Fairness perceptions are commonly examined in one of three specified forms: personal, procedural and 
distributional fairness (Maestre-Andrés et al 2019)1. For the purposes of this study, distributional fairness is 
given center stage, while personal fairness plays a minor role. Procedural fairness is not examined in this 
study.  
Now, distributional fairness relates to how costs and benefits are distributed among groups in society, and 
whether any groups are perceived to bear an unfair burden of costs. Low-income households and rural 
residents are two of the most common groups perceived to bear an unfair burden as a result of carbon 
taxation (Maestre-Andrés et al 2019; Ewald et al 2022, Povitkina et al 2021). When people instead evaluate 
policys according to personal fairness, they look at the perceived costs and consequences for themselves. 
Higher perceived personal costs usually means lower support, and since perceived personal costs usually 
increase with higher tax rates, that can cause support to drop (Maestre-Andrés et al 2019; Carattini et al 
2017). 
Compared with personal and procedural fairness, distributional fairness has been found more important. In a 
meta-study of fifteen determinants of climate policy opinions, Bergquist and colleagues found that 
perceptions of distributional fairness had the strongest effect (r=0.73), while personal fairness had a weaker 
effect (r=0.17) (2022). Subsequent studies have confirmed this picture, and complemented with findings 
suggesting that procedural fairness only plays a marginal role, if any, for acceptability of a carbon tax 
1 Some studies have included some general form of fairness perceptions, see Maestre-Andrés et al 2019, p. 1195. 
Technically, this study asks about fairness in general terms as well, but the treatments are designed to affect 
distributional consequences. 
8
increase (Lindvall et al 2024; Bergquist 2025). Importantly, one of these studies was conducted with Swedish 
participants (Lindvall et al 2024). In other words, many people seem to oppose carbon taxes primarily 
because of distributive concerns. However, it should be noted that exploratory evidence suggests that 
distributional fairness may interact with personal fairness (operationalized as perceived costliness in 
Bergquist 2025).  
A frequent approach among researchers to addressing resistance based on distributional consequences is to 
test how various uses of the carbon tax revenues can mitigate fairness concerns and increase public support 
of the tax. Various suggested revenue recycling options have been tested. Two of these are redistributive ones 
which can make the net effect of a carbon tax progressive: transfers or tax rebates directed to poor and 
vulnerable, and equal lump-sum dividends or tax rebates for a whole population. Both of these recycling 
schemes compensate potential losers of the carbon tax, ideally changing distributional fairness and personal 
cost perceptions. While the progressively of targeted schemes is obvious, lump-sum transfers tend to be 
progressive in the sense that ”low-income households are likely to receive [larger compensations] than the 
cost increase that they suffer” (Carattini et al 2018, pp. 7-8). And indeed, these options have both been found 
to increase support for carbon taxes, and to be preferred above most alternatives (Maestre-Andrés et al 2019; 
Carattini et al., 2017; Gevrek and Uyduranoglu, 2015; Saelen & Kallbekken, 2011). 
However, these options are quite consistently outperformed in terms of popularity by environmental 
earmarking, an option with no obvious redistributive effects. The reason is probably that it addresses 
effectiveness concerns among people who underestimate or fail to understand the Pigouvian mechanism of a 
carbon tax (Maestre-Andrés et al 2019; see also Carattini et al 2017). Now, it should be noted that revenue 
earmarking for environmental purposes may come with inflexibility issues and inefficiency costs in the use 
of public resources (Gevrek and Uyduranoglu, 2015). In addition, this kind of earmarking is in many 
countries, such as Sweden, not a real option because of constitutional limitations (Lindvall et al 2024; 
Sterner 2020).  
Now, although previous literature has found that distributional unfairness is very important for carbon tax 
resistance, it has primarily focused on compensating the perceived biggest losers of the cost distributions, 
rather than on targeting the groups on the other side of the spectrum - the perceived relative winners, 
privileged by their lower carbon dependence for basic needs or by the ease with which they can switch to 
low-carbon consumption alternatives. Since rural residents and low-income households are often perceived 
as some of the most vulnerable to carbon taxes, many people should reasonably think that the least 
vulnerable are urban residents and wealthy people.  
However, few studies have explored how making these groups targets of climate policy affects policy 
attitudes or fairness perceptions. I will here discuss three that have tested such an approach. First, two studies 
applied choice experimental design to examine how varying cost distributions for climate policies or carbon 
taxes specifically affected policy support in Sweden (Brannlund and Persson 2012) and Turkey (Gevrek and 
Uyduranoglu 2015) respectively. In the Swedish study, the progressive alternative explicitly suggested that 
”Higher-income citizens pay a larger share […] of income”. Both studies found that income neutral cost 
9
distributions were preferred somewhat above regressive ones, and that progressive cost distributions were 
preferred even more. 
Secondly, with a conjoint survey experiment design, Coleman, Harring and Jagers recently explored how 
varying several climate policy attributes affected public support in Sweden (2023). Among the varied 
attributes were policy targeting and policy type, and they also tested for interactions between policy type and 
other attributes. Surprisingly, the study found that targeting wealthy individuals was the least popular 
alternative, slightly decreasing policy support compared with the average alternative. Instead, targeting 
consumers made no difference, while targeting companies, especially large companies, increased support 
marginally.  
The preference for targeting large companies rather than individuals aligned with their expectations, but not 
the aversion against targeting wealthy individuals (Ibid, pp. 423-424, 429). The authors refrain from 
speculating in what could explain this latter finding, but their reporting of an interaction test suggest a 
preliminary explanation. In a table displaying how support for each policy type was affected by the other 
attributes, there is an important exemption to the tendency of people to dislike targeting wealthy individuals: 
taxes (Ibid, p. 431). For taxes, targeting wealthy individuals seems to have no significant effect on support 
compared with the average support for that instrument. This runs contrary to the authors’ concluding claim 
that ”we consistently find that people are less supportive of policies that target wealthy individuals”. I find 
this small deviation relevant, because while taxes impose clear costs on their targets, two of the other three 
instruments - information campaigns and subsidies - do not. Subsidies even benefit their targets, and 
interestingly, the aversion against targeting the rich seems to be the strongest for subsidies, coupled with a 
preference for targeting consumers with subsidies. Possibly, even information campaigns and regulations 
could be perceived as benefitting their targets when these are wealthy individuals, although this seems 
unlikely since the same negative effect is not visible for targeting large companies with these instruments.  
In sum, these findings suggest that people dislike climate policies benefitting wealthy individuals, but that  
carbon taxation that directs larger new costs on this group than on others might receive higher support than if 
the costs are shared proportionally or regressively. It seems likely that this is mainly because of differences in 
distributional fairness perceptions. However, the evidence is still rather scarce and somewhat ambiguous, so 
more research is needed here. Below, I discuss how I explore this research gap. 
2.2 Research contribution

Against the background outlined above, I want to examine a new approach to fairness perceptions. Instead of 
the rather reactionary approach of asking how to direct revenue to address fairness concerns around carbon 
taxes, I ask more preventatively whether carbon taxation can be designed to be perceived as more fair from 
the start. Since fairness perceptions are so tightly connected to distributional consequences, and people seem 
to prefer progressive distributions before regressive ones, this requires a policy which is immediately 
progressive, and ideally which people easily perceive as progressive. Are there any carbon taxes or other 
10
climate policy instruments that not only are placing costs primarily on the most wealthy, but also are easily 
perceived as such? Almost by definition, such a policy would have to primarily target carbon-intensive 
consumption by the rich. As I outline below, one policy alternative fulfilling this description is a luxury 
carbon tax.  
To some extent, testing such a policy might explore the consistency of people’s fairness perceptions. It seems 
rational that if people react to existing carbon taxes primarily on the grounds that they are unfair to low-
income households (and rural inhabitants), then that resistance should wane if carbon taxes are instead 
obviously directed towards the perceived relative winners: the wealthy (and urban dwellers) (as found by 
Brannlund and Person 2012 as well as Gevrek and Uyduranoglu 2015). But are people consistent in how 
they understand fairness in climate politics when the privileged are targets of increased costs as when the 
underprivileged are hit? This question could be extended to a more general context: do people believe 
policies primarily burdening the top of the social ladder to be fair to the same extent as they understand 
policies mainly burdening the bottom to be unfair? 
These latter questions are in this thesis addressed somewhat indirectly. Instead of simply asking about the 
perceived fairness of a new and untested, clearly progressive carbon tax - which would have required a 
representative sample to be generalizable and thereby relevant - the new policy is treated as a complement to 
an increase in an existing, ordinary carbon tax: the Swedish carbon tax on diesel and petrol. Presumably, 
respondents perceive this tax as regressive, or at least far less progressive than the new tax.  
With this design, this thesis also extends previous research on policy-packaging. According to Wicki and 
colleagues (2020), the purpose of policy-packaging is to minimize the trade-off between policy effectiveness 
and feasibility observed in various policy areas. The authors present three definitions of policy-packaging, of 
which the only relevant one for their as well as this paper is horizontal packaging, which refers to ”the 
simultaneous deployment of two or more measures that aim at the same policy target” (Ibid, p. 602). They 
furthermore reiterate a need for more systematic and experimental empirical studies on the effect of 
packaging on feasibility and public acceptability. However, in their own conceptualization as well as 
empirical investigation, they seem to assume that a policy package does not include more than one ”push” 
measure, but may include several other kinds of instruments whose purpose is to increase the acceptability 
and effectivity of the ”primary” push measure. This in turn seems based on an assumption that no push 
measures can work to increase the acceptability and effectivity of other push measures. 
Now, that assumption may be true for taxes and other instruments with regressive effects, or which impose 
costs on broad segments of a population. However, certain taxes have, on the contrary, progressive effects 
and only impose costs on a small share of the population. So the question is, can the introduction of a very 
progressive carbon tax make a simultaneous increase of a more regressive carbon tax possible? I argue that 
such taxes should, like certain other policy instruments, have the potential to work as what Wicki and 
colleagues label as ”ancillary measures”, increasing policy feasibility of the primary measure in a policy 
package (2020). They would most likely do this primarily by affecting fairness perceptions, which is what I 
measure in this thesis. 
11
Now, these two specific policies were chosen purposefully. They were both directed towards transport-
related emissions in order to avoid potential complexities arising from suggesting policies targeting different 
sectors. A focus on one sector only should make the package more internally consistent and so more 
comprehensible. The ordinary carbon tax is expected to provide a certain point of reference against which the 
new tax can be compared, and I believe this comparison to be facilitated by the similarities between the 
measures. However, it should be noted that I cannot substantiate these claims, and it seems like the effect of 
addressing different policy domains in the same package is little explored (Heyen and Wicki 2024, p. 789 
seem to suggest addressing that in future research). 
Furthermore, transport-related emissions account for a large share - more than a third - of emissions globally 
(IEA 2023). Similarly, in Sweden, the share is above 40% if including international transports by sea and air, 
and about a third if excluding them (Trafikverket 2024; Naturvårdsverket 2024). About 90% of the country’s 
emissions from domestic transports come from road transports (Ibid), of which a majority come from from 
(fossil-fuel) driven passenger cars (Naturvårsverket 2024). While a carbon tax on diesel and petrol, targeting 
fossil-fuel driven cars, affect a relatively large share of the country’s emissions (and by extension those of the 
average Swedish resident; Ibid), the extravagant means of transport that would be targeted by the luxury 
carbon tax outlined below cause very large per-capita emissions for those who consume them (Otto et al 
2019; Wallace and Welton 2024). Therefore, there is a certain mitigation potential in targeting each of these 
emission sources, even though the logic on why to focus on them differs. 
From the above discussions, it may seem as though it is obvious that people would perceive the outlined 
policy package as more fair than an isolated increase in the carbon tax on car fuels. To see that it is not, it 
may help to look at the bigger picture. By exploring how the impacts of climate policy changes on the 
distribution of carbon emissions and economic resources affects fairness perceptions, this paper relates to a 
wider discussion on how people want public goods and bads to be distributed in general.  
Now, although economic inequality has been rising in many countries for decades, public outrage about 
those trends is often remarkably absent, and proposals that would exacerbate inequalities further are often 
met with little resistance. In fact, parties running on agendas that increase inequalities through tax cuts and 
welfare retrenchments often receive relatively broad support in many countries (including Sweden), even 
among voters that lose out personally on their policies. On top of this, progressive policies often fail to be 
implemented. An illustrating example of this tendency is how a large majority of Swiss voters recently 
rejected a ballot proposal to tax the super rich to fund a redistributive policy (Emmenegger and Marx 2019). 
In this context, it seems far from obvious that a tax targeted towards the richest could increase the fairness of 
a policy package.  
However, there seems to be something peculiar about how people react to climate policy proposals. Given 
the above findings in previous literature, people seem to be very attentive to how climate policies affect the 
distribution of resources within their country. So far, this attention has mostly been expressed as resistance to 
carbon taxes and other policy measures considered unfair because of how they disproportionately affect the 
12
poor and rural inhabitants. But perhaps, by suggesting certain suitable instruments, this attentiveness to 
fairness in the climate policy domain can be turned into a vehicle for redistribution? 
Lastly, the perceived fairness of a proposal might depend on other factors than policy design. As will be 
evident below, a luxury carbon tax can be motived on the basis of at least two different common fairness 
principles: an (outcome) equality principle and a needs principle. This might have implications for how fair it 
is perceived, as the principles may not resonate equally well with the public. That invites to testing a second 
variable: policy framing around different fairness principles.  
Now, my use of the concept of framing may clash with that of other studies. In this thesis, framing refers to 
the way a luxury carbon tax is described and motivated to respondents in a vignette. Essentially, the policy 
proposal is the same, but the text focuses on one of the two fairness principles. By testing which of the two 
framings that is considered most fair, this paper advances knowledge on which of these principles best 
resonates with the (Swedish) public. On a more practical note, it might also contribute to knowledge on how 
(progressive) climate policies can be communicated to the public in order to increase fairness perceptions. 
3. Theory 
3.1 The case for a luxury carbon tax

In essence, a luxury carbon tax is a tax that would be levied on carbon-intensive luxury products that only 
the wealthiest can afford. Examples of such products would be the ownership and use of private jets, luxury 
yachts and first-class commercial air tickets (Wallace and Welton 2024). Furthermore, the tax rate would be 
set at a rate calculated not just in relation to emission amounts, but also in relation to the wealth and 
purchasing power of the targeted consumer group, in order to potentially have any effect on behaviour. Since 
the target group in question is very wealthy, the tax rate would accordingly need to be substantially higher 
per tonne of emitted CO2-equivalents than any existing carbon tax affecting average consumers (Benoit 
2020; Wallace and Welton 2024). Such a tax should have the potential for direct as well as indirect effects on 
climate mitigation. Direct effects could be achieved by deterring from consumption of targeted behaviours 
by raising their costs. Indirect effects would be achieved if such a tax managed to weaken the so called 
”Veblen effect”: that conspicuous consumption among the very richest trickles down along the income 
distribution, forcing people to buy ”position goods” to keep their place in the social hierarchy. Ideally, it 
could make carbon-intensive positional goods widely morally condemned (Wallace and Welton 2024).  
However, for the purposes of this thesis, what matters is how such a policy would be perceived by the public 
when suggested alongside an ordinary carbon tax increase. Complementing an ordinary carbon tax increase 
with a luxury carbon tax could potentially increase fairness perceptions, as I argue below. 
13
3.2 Fairness principles

Next, it is theoretically interesting which aspect of the luxury carbon tax, if any, that would most strongly 
align with peoples’ sense of fairness2. To examine this, I will apply two of three common fairness principles 
discussed in the literature: equality and need, which are often contrasted with the equity principle (Deutsch 
1975). These concepts have been used empirically to examine how people prefer public goods and bads to be 
distributed (Hammar and Jagers 2007; Povitkina et al 2021). For example, in one study, Swedish respondents 
were asked to what extent they supported each of these three principles applied to the distribution of carbon 
emission reductions. Then they used a regression analysis to examine how this support predicted support for 
a carbon tax increase among the general population, frequent car users and non-frequent car users 
respectively (Hammar and Jagers 2007). Another study employed a more qualitative approach with open-
ended questions to examine why American respondents thought carbon taxes were unfair. They grouped 
responses into seven categories, and then concluded by connecting these to the three fairness principles 
common in the theoretical literature (Povitkina et al 2021). In this study, I will employ the equality and needs 
principles by emphasizing different aspects of the luxury carbon tax to the respondent groups that receive the 
policy package. 
Now, previous studies have usually applied these principles on the distribution of one resource at a time 
(such as Hammar and Jagers 2007). However, here they are applied to two resources simultaneously: a) 
carbon emissions and b) money. This is because a luxury carbon tax affects the distribution of both in similar 
ways, and because a sizable share of respondents are likely to strongly doubt that a carbon tax affects carbon 
emissions at all (Ewald et al 2022). Under such circumstances, it makes little sense not to mention anything 
related to money in a text designed to promote a luxury carbon tax. Likewise, if effects on emissions were 
not mentioned, people might confuse the luxury carbon tax with a plain luxury tax.  
Regarding the specific principles, the most straightforward of the three is the need principle. It posits that 
distribution should be aimed at need fulfillment (Hammar and Jagers 2007). This principle aligns well with 
the first argument below (section 3.4) for why the policy package should be perceived as more fair: because 
a) it targets unnecessary luxury products that no-one fundamentally needs which, as a result, those who do 
buy should relatively easily be able to do without, and b) because the wealthy consumers that can afford it 
also have the greatest capacity to pay more tax. Both of these aspects are framed in a need treatment vignette. 
Next, equality can be defined either in terms of procedure (everyone receives equal shares of what is to be 
distributed) or in outcome (the good or bad ends up more equally distributed than before) (Hammar and 
Jagers 2007). Outcome equality aligns best with the purposes of this study, since it is more easily applied to 
justify a luxury carbon tax. Outcome equality is emphasized in the second argument below (section 3.4): that 
2 In choosing to relate the luxury carbon tax to fairness perceptions this way, as well as in the formulation of H2 and H3 
in section 3.4 below, I have been much inspired by and followed largely the structure and discussion of Hilmersson, 
2024. 
14
a luxury carbon tax promotes a more equal distribution of both a) carbon emissions and b) economic 
resources. This is framed in an equality treatment vignette.  
Lastly, the principle of equity is a merit-based principle, which posits that goods and bads should be 
distributed according to how much people have contributed to them (Hammar and Jagers 2007). Equity is 
loosely connected to the Polluter Pays Principle (PPP) (Ibid), which implies that the social costs of emissions 
should be payed by the consumer or producer of the product causing the emissions. According to Hammar 
and Jagers, the existing Swedish carbon tax is designed according to equity and PPP, since its consumer costs 
are directly associated with emissions (2007, p. 380). Therefore, one could assume in a study like this that 
the chosen control group proposal to simply raise the carbon tax on car fuels best realizes equity/PPP. 
However, the purpose of the control vignette is to provide a point of comparison for the treatments. Hence, it 
will not include a special equity framing and I will not infer that respondents perceive it as a realization of 
equity, both because it would confuse interpretation of the treatment effects, and because it goes beyond the 
scope of this thesis.  
3.3 Research questions

In short, I want to test whether and how the policy package described above changes fairness perceptions 
compared with a control proposal to only raise the existing Swedish carbon tax on diesel and petrol.  
My first research question is therefore:  
How are fairness perceptions of a carbon tax increase on car fuels affected by packaging the policy proposal 
with the introduction of a luxury carbon tax on expensive, emission-intensive travelling? 
I am also interesting in how framing the luxury carbon tax around an equality or a needs principle affects 
fairness perceptions. Hence, my second research question is: 
How are fairness perceptions of such a policy package affected by which fairness principle is framed in the 
proposal? 
3.4 Hypotheses

In this section, I present hypotheses on how fairness perceptions are affected by the experimental treatments. 
First, turning a single policy carbon tax increase on diesel and petrol into a policy package by adding a 
luxury carbon tax introduction should have a positive impact on fairness perceptions. Below I will elaborate 
on why and let my first hypothesis mirror this prediction. Secondly, I will test two different frames, each 
focusing on one fairness principle: equality (in outcome) and need. Since it seems debatable which of these 
15
frames could be more successful at increasing fairness perceptions, I will discuss and put forth two 
contradicting hypotheses concerning this.  
To start with, the policy package should be perceived as more fair than an ordinary carbon tax in isolation, 
for several reasons. First, as Wallace and Welton argue (2024), a luxury carbon tax targets some of the most 
unnecessary consumption-based emissions. A quite common criticism against ordinary carbon taxes is that 
they make it harder and costlier for certain people, such as rural inhabitants, to fulfill their basic needs (see 
for example Povitkina et al 2021). However, by definition, no-one is in need of (extremely) luxurious means 
of travel, or forced to consume luxuries (Wallace and Welton 2024, p 1174). On the contrary, as Wallace and 
Welton argue, the emissions they cause undermine the needs fulfillment of the very poorest people globally 
(2024, p. 1175). If people perceive climate change as an important problem, which a majority in Sweden 
even among protesters of ordinary carbon taxes seem to do (Ewald et al 2022), then one reason for that is 
likely to be that it threatens people’s basic needs. And by that logic, taxes targeting the most unnecessary 
emissions should increase perceived fairness of carbon taxation.  
In addition, since the products covered by the suggested luxury carbon tax are already very costly (I follow 
Wallace and Welton by exemplifying with luxury yachts, private jets and first and business class commercial 
aviation travel; 2024), the only ones who can afford them are the most wealthy groups in society. Therefore, 
that part of the policy package would only hit people who have a great - arguably the greatest - capacity to 
pay higher taxes or to avoid them by choosing low-carbon alternatives (Benoit 2020). Meanwhile, since the 
suggested ordinary carbon tax increase is equally large in all experimental conditions, its effect on needs-
fulfillment would be the same. Hence, to the extent that people perceive carbon taxation as (un)fair because 
of how it affects the fulfillment of basic needs, the policy package should be perceived as more fair than an 
ordinary carbon tax increase on its own.  
Second, a luxury carbon tax should promote a more equal distribution of both carbon emissions and 
economic resources compared with any scenario without that tax. This is because it not only avoids raising 
costs for low- and middle- income people, but also does raise costs for high-income people, and the new tax 
revenues would benefit everyone by going to the general budget. As a result, the policy package should 
produce a more equal outcome than the standalone ordinary carbon tax increase. This should positively affect 
fairness perceptions of the policy package to the extent that carbon taxation is perceived as (un)fair because 
it (increases) decreases economic inequality.  
Third, taxes that target the wealthy have been found to be among the most popular taxes, and it seems likely 
that this is largely because of how fair they are perceived. In a recent study on public attitudes to a wealth tax 
in the U.K., a wealth tax received high support, people overwhelmingly preferred taxes targeting the 
wealthiest above those with a much broader scope, and when asked about design features of a wealth tax, 
many respondents were again keen to only burden the wealthiest (Rowlingson et al 2021). The authors 
concluded that fairness beliefs played an important role for views on the tax base, and that a guiding fairness 
principle seems to have been ’capacity to pay’.  
16
Furthermore, when one study surveyed Swedes about various tax increases to finance climate investments, a 
wealth tax was among the most popular alternatives, roughly as popular as a carbon tax increase in a 
nationally representative sample, and clearly more popular in a protest group sample (Ewald et al 2022). 
Another study, framing a wealth tax as one of several welfare policy proposals, similarly found a wealth tax 
to be among the most popular of the included instruments (Khan et al 2023), so people in Sweden seem quite 
supportive of taxes targeting the rich for climate investments and other purposes. Even though these specific 
studies did not measure fairness perceptions, it seems reasonable that taxes which are strongly supported are 
also perceived as quite fair, given that other studies have found such a strong relation between fairness 
perceptions and attitudes to certain taxes. Compared with a wealth tax, the luxury carbon tax could be 
perceived as even more fair for targeting not just wealth in general, but morally condemnable uses of wealth 
(Wallace and Welton 2024). To conclude, empirical evidence indicates that taxes targeting the wealthy are 
very popular, and to the extent that this is because they are perceived as fair, this should imply that the policy 
package will raise fairness perceptions compared with the ordinary carbon tax increase.  
Fourth, the very mentioning of a ’luxury carbon tax’ might make distributional consequences even more 
salient in people’s minds than it already is with ordinary carbon taxes, by highlighting some of the most 
superfluous emission sources and the people that most easily can absorb increased costs or lower their 
emissions. However, it is debatable to what extent such an effect could be of importance in this study, since 
all respondents were explicitly asked to reflect about the general fairness of their proposal, which to most 
people should mean to primarily think about its distributional consequences. 
Lastly, of course, one factor might somewhat negatively pull down general fairness perceptions of the policy 
package: increased costs and negative personal fairness perceptions among the wealthy people who are 
affected by the luxury carbon tax. However, this should not affect general fairness perceptions much, both 
because personal fairness is much less important than distributional consequences as predictor of support, 
and so should also be less important as predictor of general fairness (Bergquist et al 2022), and because those 
(very) wealthy people only make up a tiny portion of the population. Because of my sampling methods, I 
might not reach a single person from that group. 
As a result, my first hypothesis is:  
H1: A carbon tax increase on car fuels is perceived as more fair if it is packaged with a luxury carbon tax, 
than if suggested in isolation. 
To answer my next hypotheses, a framing on how the luxury carbon tax realizes the need principle will be 
compared to one on how it realized outcome equality. Since no previous study has been conducted 
connecting fairness principles with a clearly progressive tax, it is hard to say which frame would have a 
stronger effect on fairness perceptions. Similar studies, such as Hammar and Jagers (2007), Sommer and 
colleagues (2022) and Reeskens and van Oorschot (2013), have usually applied a procedure equality 
principle in their comparisons with an equity and need principle. Their conclusions on people’s relative 
17
preferences for these principles are hard to extend to this context, in which an outcome equality principle as 
well as a needs principles can be used to justify one and the same policy.  
The closest study is from Hilmersson (2024), which similarly framed equal tax rebates with an outcome 
equality-based and needs-based treatment, respectively, to examine the effect on policy support. She found 
that the needs-based treatment, emphasizing the redistributive effects of equal tax rebates, led to higher to 
support than the equality treatment. However, both treatments caused a reduction in support compared with 
the control group.  
Against this background, I present my two competing hypotheses together: 
H2: Framing the luxury carbon tax around (outcome) equality makes fairness perceptions increase more 
than framing that the luxury carbon tax targets the most unnecessary emissions. 
H3: Framing that the luxury carbon tax targets the most unnecessary emissions makes fairness perceptions 
increase more than framing that the luxury carbon tax promotes equality. 
4. Method 
In this chapter, I first describe and motivate the choice of research method and design. After that, I describe 
how I operationalize my variables, before briefly discussing ethical considerations. I also describe the 
process of creating a pilot survey before making survey revisions and distributing the actual survey. Finally, I 
describe the use of statistical techniques.  
4.1 Research method and design: between-person EVM

This study applies a form of experimental vignette methodology (EVM). Atzmüller and Steiner defines a 
vignette as a ”short, carefully constructed description of a person, object, or situation, representing a 
systematic combination of characteristics” (2010, p. 128). 
Following that, this study randomly assigns all respondents to one of tree experimental groups through a 
survey design in the used software Qualtrics. As long as the groups are large enough, they should be 
comparable with each other on all relevant attributes, which enables causal inference (Angrist and Pischke 
2014, pp. 12-16). Each groups receives a vignette, describing a carbon tax proposal. The first part of the text 
is the same for all, motivating and proposing and ordinary carbon tax. The two treatment groups then 
receives a longer text, adding a luxury carbon tax to the proposal. However, the framing of the described 
purpose and effects behind the luxury carbon tax are different between the two treatment groups, reflecting 
either an equality or a need principle. 
18
Now, EVM studies aim to strike a good balance between internal and external validity. By so doing they 
present a viable alternative both to representative surveys, which have high external and low internal validity, 
and to conventional experiments that instead are strong on internal validity (establishing causal relationships) 
but weak on external validity (Atzmüller and Steiner 2010; Aguinis and Bradley 2014). EVM studies 
overcomes some of the weaknesses of these other methodologies by both enabling larger samples than 
laboratory experiments and, through the vignette experiment, stronger causal identification than traditional 
surveys (Ibid).  
Aguinis and Bradley argue that EVM is very useful to determine the nature and direction of causal 
relationships, because it gives researchers the ability to include relevant variables and exclude confounding 
ones. Likewise, they argue that it is appropriate when it is ethically hard or impossible to conduct traditional 
experiments, because it allows researchers to ”create hypothetical scenarios that adress sensitive 
topics” (2014, p. 357). Indeed, this is the case for this study: a conventional experiment seems practically 
impossible, but thanks to the EVM, the hypotheses can still be tested.  
Next, the literature suggests several different research designs for EVM studies: between-person, within-
person and mixed research design. While between-person designs only let each respondent judge one single 
vignette, and different vignettes are randomly assigned to respondents, all respondents in within-subject 
studies judge the same set of vignettes. Few studies apply a between-person design (Atzmüller and Steiner 
2010), and scholars advise against it because ”[w]ithout other vignettes to serve as referent points for their 
own judgments, responses may not accurately reflect the true judgments of each respondent” (Aguinis and 
Bradley 2014, pp. 360-361). However, they also recommend that researchers choosing this option provide 
participants sufficient information to understand the context as good as possible (Ibid).  
This study applies between-person design, for several reasons. First, this reduces the amount of information 
respondents are presented with. The chosen information texts are relatively long, and giving people even 
more text to analyze could have increased cognitive load and affected response quality or attrition rates. 
Secondly, and more importantly, if respondents were shown vignettes both with and without the luxury 
carbon tax introduction - which would have been a case if including a within-person element - I estimated a 
risk for a demand effect which could have confounded the results. In short, to subtly test the effect of a 
policy design change, the point of comparison should not be revealed. Thirdly, since the two treatment 
vignettes essentially suggest the same policies, presenting both side-by-side would have induced respondents 
into comparing their substantive differences, and likely finding none. This would have undermined the 
purpose of varying the framing around different fairness principles.  
As recommended by Aguinis and Bradley (Ibid), a baseline information is also given to sufficiently 
contextualize the proposals for the respondents.  
Now, the choice in this thesis to compare two framings with each other, and essentially two treatment groups 
with each other, is rather unconventional. Treatment groups are usually not compared with each other, since 
both of them have been manipulated and cannot be assumed to provide some neutral point of comparison. I 
19
readily acknowledge this, but have chosen to follow Hilmersson (2024) in this regard, who also compared 
the treatment groups in a similar manner. 
4.2 Operationalization of variables

Main variables: 
Respondents were presented with the policy proposal vignettes (one for each group). In the formulation of 
these vignettes, I closely followed Hilmersson (2024). Table 1 shows the three vignettes in English and 
Swedish. 
Table 1.  
Group Treatment Vignette Vignette (Swedish)
1 Control Emissions from the transport sector Transportsektorn står för en stor del av 
accounts for a significant part of Sweden’s Sveriges växthusgasutsläpp. 
greenhouse gas emissions. A carbon tax on Koldioxidskatt på fossila bränslen 
fossil fuels is used to reduce such används för att minska sådana utsläpp. 
emissions. Sweden’s carbon tax on diesel Sveriges koldioxidskatt på diesel och 
and petrol is currently 2.81 and 3.27 SEK/ bensin är idag 2,81 respektive 3,27 kr/
liter respectively. liter.  
One alternative to further reduce Swedish Ett alternativ för att framöver minska de 
emissions would be to increase this carbon svenska utsläppen mer vore att höja denna 
tax by 1 SEK/liter. koldioxidskatt med 1 kr/liter.
2 Package with (Insert control text) (Sätt in kontrollgruppens text) 
equality frame
In addition, a luxury carbon tax would be Dessutom skulle man införa en 
introduced in order to promote a more lyxkoldioxidskatt för att främja en mer 
equal distribution of emissions and jämlik fördelning av utsläpp och 
resources. The tax rate would be tillgångar. Skattesatsen skulle vara 
considerably higher per tonne of emitted betydligt högre per ton utsläppt koldioxid 
CO2 than the one on diesel and petrol. This än den på diesel och bensin. Denna 
carbon tax would be levied on means of koldioxidskatt skulle läggas på 
transport exclusively available to the transportmedel som endast är tillgängliga 
wealthiest people in society, such as luxury för de rikaste i samhället, såsom lyxjakter, 
yachts, private jets and flight travel in privatjetplan och flygresor i business 
business class. class.
3 Package with (Insert control text) (Sätt in kontrollgruppens text)  
needs frame
In addition, a luxury carbon tax would be Dessutom skulle man införa en 
introduced in order to strongly raise the lyxkoldioxidskatt för att kraftigt höja 
cost of the most unnecessary transport kostnaden för de mest onödiga 
emissions. This would be levied on the most transportbaserade utsläppen. Den skulle 
luxurious means of transport, such as läggas på de mest lyxiga transportmedlen, 
luxury yachts, private jets and commercial såsom lyxjakter, privatjetplan och 
flight travel in business class. This carbon flygresor i business class. Denna 
tax would hit the consumers that can most koldioxidskatt skulle slå mot de 
easily lower their emissions or pay higher konsumenter som lättast kan minska sina 
taxes without compromising their basic utsläpp eller betala mer skatt utan att 
needs. kompromissa med sina grundläggande 
behov.
Then, perceived fairness was measured with the question ”How fair do you think this proposal is (both 
policies together)?”, with the note in parentesis included only for the treatment groups. The question was 
20
answered by choosing an alternative on a five-point Likert scale, from ”very unfair” via ”rather unfair”, 
”neither fair nor unfair” and ”rather fair” to ”very fair”. 
The variation of vignettes were treated as the independent variable, while fairness perceptions were treated 
as the dependent variable. 
Background variables: 
Before receiving the vignettes, respondents answered questions on which country they live in, how old they 
are, what education level they have achieved, which gender they have, and to what extent the reside in an 
urban or rural area. These questions served two purposes. First, country of residence and age were used to 
filter out responses from people not living in Sweden or under the age of 18, since the survey only focused 
on adult Swedish residents. Secondly, age, education, gender and area of residence were used to check the 
balance between the three experimental groups. A chi-squared test was used for each of the variables to test 
whether the groups despite the random assignment had ended up with any significant differences between 
each other on any of the background variables. If they did, that might suggest a need to control for that 
variable in order to be able to draw valid conclusions from the results.  
4.3 Ethical considerations

Esaiasson and colleagues underscore that studies should be conducted in an ethical manner (2017). They 
especially discuss two areas relevant for experimental studies: gathering informed consent from participants 
and offering the possibility to quit participation at any time and demand the deletion of their personal data. 
Regarding informed consent, researchers should provide correct and satisfactory information on the study, 
but not necessarily complete information. Since informing respondents that they are part of an experiment 
might induce unwanted disturbances, such as demand effects, full disclosure is not to be recommended from 
the start. Instead, researchers should provide correct but incomplete information before gathering consent. 
However, after the experiment is completed, participants should be informed that they were part of an 
experiment and the (real) purpose of the study (Ibid, pp. 348-349, 354). 
Now, it could be added that this thesis does not use any manipulations or question of a very serious nature. 
The most serious is the main question, asking about fairness perceptions of a policy proposal that is designed 
and framed differently between different respondent groups. Overall, this is not very ethically problematic.  
Nevertheless, I gathered informed consent by, presenting respondents at the start of the survey with 
information on how personal data from the survey would be processed in accordance with GDPR. I informed 
participants that they belonged to one of three randomized subsamples, what kind of information would be 
gathered from them, who would have access to it, that it would be anonymous and that it would, after 
completion of the thesis, be deleted. The text also included some information about me, and about the 
purpose of the study. It further informed participants that they would give their consent by answering the 
21
survey, that their participation is voluntary and that their consent could be withdrawn before submission by 
cancelling the survey. Following this information, they were further prompted to actively consent by 
checking a box reading ”I consent”. Without doing so, they could not proceed.  
After answering the last questions in the survey, participants were informed about the experimental character 
of the survey. They were briefly told about all the experimental conditions, and that the purpose of the 
experiment was to examine any differences in how fair the proposals were perceived.  
It would have been too technically demanding to delete specific responses upon request, so that was never 
offered. However, after completing this thesis, I will delete the information gathered from the survey as 
promised. Furthermore, in the analysis, I excluded all participants with missing responses, even only on the 
very last questions. I did this because due to how the Qualtrics survey was set up and closed, respondents 
who actively left the survey before submitting by closing the browser tab could later re-enter at the point 
where they left, and their answers are likely to have been gathered a week after they entered the survey. By 
using their data, I would withdraw on the promise that their consent could be withdrawn at any time by 
cancelling the survey.  
4.4 Data collection procedure

I used the survey software Qualtrics as tool for data collection. To reach Swedish as well as non-Swedish 
speakers, the survey was available in Swedish and English.  
Before creating and distributing the actual survey, I created a pilot and distributed it to a small group close 
me, receiving 17 responses. The purpose of the pilot was to test the quality of the whole survey: whether 
questions were clear and comprehensible, and whether the manipulation worked as intended. After the pilot, 
I decided to exclude potential control or moderating variables, such as ideology or car ownership, both 
because testing interactions went beyond the scope of this thesis, and because sensitive and unnecessary 
control questions might increase survey dropouts. I also redesigned the operationalization of fairness 
principles. Lastly, I adjusted the manipulation checks, removing one question only focusing on how the 
proposal would affect the wealthy, which seemed to fail to display a successful manipulation. For a similar 
reason, I changed the wording in the one kept somewhat, from ”the wealthiest/poorest” to ”the wealthy/poor” 
(see below), and increased its response alternatives from three to five to capture more nuanced changes. 
Instead, two new manipulation checks were added to test the varying fairness principle framing.  
I then gathered the data by distributing the final survey via various channels. For EVM studies, Esaiasson 
and colleagues recommend group sizes of at least 30-40 respondents for each experimental group (2017, p. 
347). My aim was to get a bit above those figures, to be able to capture effects even if they were not very 
large, which I particularly expected when comparing the two treatments with each other. 
22
To maximize sample size as well as participant diversity, I employed three channels for recruiting 
participants: posters on public boards, social media, and face-to-face recruitment. For each of these channels, 
I created a unique url-link and a qr-code based on that link. These links contained embedded information 
about which channel each respondent had been recruited through. The purpose of this was to be able to 
examine whether results were affected by the recruitment channel. However, such an examination was never 
carried out, because a large majority of answers were recruited through Facebook and text messages.  
During survey distribution, some responses were flagged by Qualtrics as bots. After discovering this, I 
decided to flag duplicate responses as well.  
Commitment check: 
Based on a recommendation from Qualtrics (Geisen 2022), I used a commitment check instead of an 
attention check. By comparing the use of a commitment check and various forms of attention checks, they 
concluded that a commitment check was the most effective tool to raise the quality of the data. Closely 
following their formulations and method, I informed respondents that it is important they provide thoughtful 
answers to each survey question in order for me to receive accurate measures of their opinions. Then I asked 
whether they commit to that, providing three alternatives - essentially ’maybe’, ’yes, ’no’ - and only treated 
’yes’-responses as passing the commitment check. All other answers were filtered out.  
Manipulation check: 
It was important to know whether the manipulation worked as expected. One expectation was that the 
manipulation changed how people thought their proposal affected the cost distribution between the wealthy 
and the poor in society. The policy package should be perceived as shifting burdens from the poor on to the 
wealthy visávi the ordinary carbon tax increase.  
In line with this, the first manipulation check asked respondents: 
If you compare wealthy and poor people in society, which of these groups do you think would be most 
affected by the proposal?  
Answers were on a five-point Likert-scale, ranging from ”The wealthy would be much more affected” to 
saying the same about the poor. Moderate alternatives replaced ”much” with ”somewhat”, and a neutral 
alternative suggested that they would be close to equally affected. 
The other two manipulation check questions were designed to capture to what extent each treatment 
successfully made people perceive the framing they were intended to receive. The second check asked to 
what extent the proposal was promoting equality. The third check asked to what extent the proposal focused 
on people’s needs. 
23
4.5 Statistical analysis

The hypotheses were tested by running dummy-variable regressions in Stata. This technique is useful when 
the independent variable is categorical and the dependent variable is continuous, and the researcher is 
interested in how mean values in the DV change between the IV categories (Mehmetoglu & Jakobsen, 2022. 
pp. 98-101). In my case, I had a categorical IV (the experimental conditions), and a DV which is not strictly 
continuous (fairness perceptions of the policy proposals). However, the DV can be assumed to approximate a 
continuous scale, with equal distances between the five values. People might interpret the words ”very” and 
”rather” differently, but when put on a Likert-scale, with a neutral middle-point, I think it is fair to assume 
that, on average, people will assume the four intervals to be equally large.  
For between-person EVM studies, Aguinis and Bradley recommend using data analytic techniques such as 
MANOVA, ANOVA and ANCOVA (214, p. 364). I considered dummy-variable regressions appropriate as 
well, since regressions with several dummy variables correspond directly to an ANOVA (just like regressions 
with one dummy variable correspond directly to an independent t-test; Mehmetoglu & Jakobsen, 2022, pp. 
101, 106). They both assess differences in mean scores on an outcome variable between different groups.  
To conduct the regressions and the rest of the analysis, I exported the data from Qualtrics into Stata. In Stata, 
I created dummy variables for each of the experimental conditions as well as one for the two treatments 
pooled together, and ran the regressions. The regression results presented below are from regressions with 
only one dummy variable, comparing two groups each. However, I also ran a regression estimating the 
means of all three groups, the results of which differed only marginally from the ones presented, with p-
values displaying (in)significance on the same level of confidence.  
5. Results and analysis 
5.1 Data

The survey was open from 17 April to 29 april 2025 (when I closed it, I left it open for respondents who 
already started the survey to complete it). The goal was to gather at least 150-200 answers. In total, by 5 
May, 224 responses had been collected. Among these, 7 were flagged by Qualtrics as potential bots, and 
were therefore excluded from the main analysis. Another 4 were flagged as duplicates, one of which I 
included because I could see that it was gathered through personal contact, and I knew that I had lent my 
own phone twice to respondents I recruited that way. In addition, 17 people failed the commitment check, 14 
did not answer that they live in Sweden, and two were younger than 18. Lastly, 15 people either actively or 
passively cancelled the survey after the initial questions but before submitting on the last ones, so I had to 
assume they withdrew consent. In sum, 52 responses were excluded. As a result, 172 responses were kept 
and used for the analysis. 
24
5.2 Descriptive statistics, randomization analysis and frequency 
distribution

In table 2, I present how the sample and each experimental group was divided according to age, education, 
gender and area of residence. In short, the total sample was skewed towards young adults (aged 25-34), 
highly educated, males and people living in urban areas. This is not representative of the Swedish population 
on the whole, but that was also not a goal of the study.  
More relevant, however, is how the experimental groups differ from each other by these characteristics. Here 
we see that for the binary variable gender, the share of female respondents is very similar in all three groups. 
As could be expected, differences become larger for the other three variables, with more categories. For 
example, the control group has more from the youngest age group, from the group who completed vocational 
education or similar, and people living in a town or smaller city, but fewer in the age group 55-64, in the 
group with highest education and of those living in a city than the other two experimental groups.  
Table 2. Respondent characteristics by experimental group, displaying frequencies and shares. 
Control Equality treatment Need treatment Total
N 58 (33.7%) 58 (33.7%) 56 (31.5%) 172 (100.0%)
Age
18-24 10 (17.2%) 5 (8.6%) 3 (5.4%) 18 (10.5%)
25-34 26 (44.8%) 28 (48.3%) 23 (41.1%) 77 (44.8%)
35-44 9 (15.5%) 6 (10.3%) 9 (16.1%) 24 (14.0%)
45-54 6 (10.3%) 5 (8.6%) 12 (21.4%) 23 (13.4%)
55-64 2 (3.4%) 6 (10.3%) 8 (14.3%) 16 (9.3%)
65-74 5 (8.6%) 7 (12.1%) 0 (0.0%) 12 (7.0%)
75-84 0 (0.0%) 1 (1.7%) 1 (1.8%) 2 (1.2%)
Education
Primary 1 (1.7%) 0 (0.0%) 0 (0.0%) 1 (0.6%)
Some secondary 1 (1.7%) 0 (0.0%) 1 (1.8%) 2 (1.2%)
Secondary 8 (13.8%) 8 (13.8%) 2 (3.6%) 18 (10.5%)
Vocational or similar 7 (12.1%) 3 (5.2%) 0 (0.0%) 10 (5.8%)
Some University but 
no degree 8 (13.8%) 7 (12.1%) 6 (10.7%) 21 (12.2%)
University - Bachelors 
Degree 18 (31.0%) 15 (25.9%) 25 (44.6%) 58 (33.7%)
25
Graduate or 
professional degree 15 (25.9%) 25 (43.1%) 22 (39.3%) 62 (36.0%)
Gender
Female 28 (48.3%) 25 (43.9%) 25 (45.5%) 78 (45.9%)
Area of residence
In a city 12 (20.7%) 23 (39.7%) 24 (43.6%) 59 (34.5%)
In a town or smaller 
city 39 (67.2%) 25 (43.1%) 23 (41.8%) 87 (50.9%)
In a village 4 (6.9%) 5 (8.6%) 3 (5.5%) 12 (7.0%)
On the countryside 3 (5.2%) 5 (8.6%) 5 (9.1%) 13 (7.6%)
To test whether the randomization of the three groups had been successful, I performed chi-square tests for 
each of the background variables. The chi-square test is designed to test the relation between two categorical 
variables (Mehmetoglu and Jakobsen 2022, pp. 49-50), in this case between each background variable and 
the assignment of treatments. By conducting the test on the experimental groups, one can induce whether the 
randomization worked as intended, or if the groups became so different by any variable that it risks affecting 
the results. The null hypothesis is that there are no significant differences between the groups for the variable 
that is tested, and if the p-value reaches below 0.05, the null hypothesis can be considered refuted.  
I ran the chi-square test on age, education, gender and area of residence, respectively (table 3). For gender 
and education, the p-values indicate that the groups are not significantly different. For age and education, the 
p-value is close to 0.05, and for area just below 0.1. This is not significant at 95% level, but at 90%. In other 
words, there is a small risk that these differences might have affected results somewhat. However, since the 
p-values were above 0.05, it was not reasonable to control for any of these variables when testing the main 
effects, as it would likely introduce more noise than it would make the estimation more accurate. 
Table 3. Chi-squared test results of randomization for each background variable.
Variable P-value
Age 0.060
Education 0.068
Gender 0.891
Area of residence 0.098
Lastly, I examined how answers on the DV were distributed for each of the three experimental groups. 
Skewed distributions might result in floor or ceiling effects, which could lead to an underestimation of the 
actual effect. Table 4 shows how the frequency of answers for each available fairness perception rating, as 
well as the mean scores. As the mean scores are close to the middle-point value (3), and especially that of the 
control group, there is no reason to suspect a floor or ceiling effect.  
26
Table 4. Frequency distribution and mean values of fairness perceptions for each experimental group. Shares 
and standard deviations in parenthesis. 
Control Equality treatment Need treatment Total
N 58 58 56 172
Fairness 
perceptions:
Very unfair 12 (20.7%) 10 (17.2%) 7 (12.5%) 29 (16.9%)
Rather unfair 11 (19.0%) 10 (17.2%) 7 (12.5%) 28 (16.3%)
Neither fair nor 8 (13.8%) 3 (5.2%) 5 (8.9%) 16 (9.3%)
unfair
Rather fair 21 (36.2%) 18 (31.0%) 19 (33.9%) 58 (33.7%)
Very fair 6 (10.3%) 17 (29.3%) 18 (32.1%) 41 (23.8%)
Mean score 2.966 (1.350) 3.379 (1.497) 3.607 (1.384) 3.314 (1.429)
5.3 Results

To test my hypotheses, I ran four bivariate OLS regressions, all reported in table 5. Model 1 estimated the 
effect on perceived fairness of receiving any of the treatment vignettes compared with receiving the control 
vignette. In model 2, the effect of the equality treatment against the control was estimated by excluding the 
need treatment group. Likewise, in model 3, the equality group was excluded to estimate the effect of the 
need treatment. Lastly, model 4 estimated the effect of receiving the need treatment instead of the equality, 
excluding the control. 
Table 5. OLS regression of the effects of treatments on fairness perceptions. 
Models (1) (2) (3) (4)
Treatments (pooled) 0.526** (0.228)
Equality treatment 0.414 (0.265)
Need 0.642** (0.256) 0.228 (0.270)
Constant 2.966*** (0.185) 2.966*** (0.187) 2.966*** (0.179) 3.379*** (0.189)
N 172 116 114 114
Adjusted R2 0.0247 0.0124 0.0446 -0.0026
Note: Standardized coefficients are presented. Standard errors in parenthesis. * p < 0.10, ** p < 0.05, *** p < 
0.01 
The first and the third model showed positive significant effects at the 95% confidence level. Receiving any 
treatment increased fairness perceptions by almost half a point on the five-point Likert-scale, which means 
Graph 1 27
Note: From left to right: model 1, 2, 3 and 4. Dotted line indicates baseline for each regression.
that respondents in the treatment groups on average rated their proposal as somewhat more fair than the 
control group (p-value=0.022). This result confirms H1.  
Model 2 and 3 then examines which of the treatments most strongly drive this positive effect. On the face of 
it, both treatments produce positive effects of a similar size. However, only the need treatment effect was 
significant (p-value=0.014), while the equality treatment effect was only close to significant at the 90% 
confidence level (p-value=0.121). The results of these two models indicate some support to H3, and lack of 
support for H2. However, to test this more clearly, I also ran model 4. The equality condition was here 
chosen as baseline in order to produce a positive coefficient in light of the results in model 2 and 3. Once 
again, the effect was positive, but insignificant. 
Graph 1 illustrates these findings with bar charts and confidence intervals. Judging by the CI:s, it is hard to 
tell whether model 2 fails to find a significant effect of the equality treatment because there is none, or 
because the sample size is too small. Since the effect in model 4 is clearly insignificant, however, it is not 
possible to confidently confirm H3 and refute H2.  
In order to enable comparisons of these results with other studies, Cohen's d effect size was calculated for the 
two significant effects from model 1 and 3, but also for model 2 in order to provide an figure guiding the 
sufficient sample size estimates of potential future studies re-testing the equality treatment (Sullivan and 
Feinn 2012). These calculations were in turn based on mean and standard deviation estimates from t-test 
comparisons of each respective pair of groups. The results of these calculations can be seen in table 6.  
28
Table 6. Effect size estimations.
Pooled treatments vs Equality vs control Need vs control
control
Cohen's d 0.372 0.290 0.469
N 172 116 114
According to Sullivan and Feinn (2012), a Cohen's d of 0.2 indicates a small effect size, while one of 0.5 
indicates a medium effect size. All the captured effects here fall in between small and medium size, but while 
the (insignificant) equality treatment effect approaches small size, the need treatment effect approaches 
medium.  
Lastly, I present differences in the shares of unfairness ratings in table 7. This is because they might have 
particular importance for political feasibility, as discussed in 2.1.  
Table 7. 
Control Equality treatment Need treatment Treatments (pooled)
N 58 58 56 114
Unfair total 23 (39.7%) 20 (34.5%) 14 (25.0%) 34 (29.8%)
(very+rather)
5.4 Manipulation checks

To close this chapter, I here report the results of my manipulation checks.  
First, both of the treatments, separately and jointly, had a rather strong and significant effect (β-coefficient>1 
and p<0.01 in all cases) on the first manipulation check, asking which of the two groups ”the poor” and ”the 
wealthy” would be most affected by the proposal they had received. The direction of all these effects were 
from the poor towards the wealthy. This meant that, as expected, the policy package was perceived as 
shifting the relative impacts towards the wealthy in society. 
The results of the second and third checks were less in line with expectation. In the test of equality 
promotion, both of the treatments had a strong and significant positive effect compared with the control (β-
coefficient>1 and p<0.01), and when the treatment groups were run against each other, no significant 
difference was found (although if anything, the equality treatment seemed to have a somewhat larger effect, 
as expected). 
Results from the check on needs focus defy expectations even more. Now, the equality treatment produced a 
significant positive effect (β-coefficient>0.5 and p<0.05), but the need treatment did not (!). However, the β-
coefficient for the effect of the need treatment was positive, and there was no significant difference found 
when the two treatments were instead run against each other.  
29
In sum, this suggests that the treatments only partially worked as intended. For some reason, people in both 
treatment groups perceived that the proposal promoted equality to a similar extent. Furthermore, the third 
manipulation check failed to confirm that people in the need group perceived that their proposal focused on 
needs to a greater extent than the control groups. At the same time, however, results indicated that people 
from the equality group more unexpectedly did perceive a larger focus on needs than the control.  
6. Discussion 
In this thesis, I have tested whether including a new luxury carbon tax in a climate and transport policy 
package can increase perceived fairness among the public. The primary policy to which the luxury carbon 
tax was added was a raised carbon tax on diesel and petrol, and the results confirmed hypothesis 1: people 
indeed believed the package to be more fair than the singular policy proposal. Confirming expectations 
further, the first manipulation check showed that the adding of a luxury carbon tax caused an expected shift 
in people’s estimations of which group would be most affected by the policy change, from the poor towards 
the wealthy. This suggests that the observed change in fairness perceptions was related to a change in 
perceived distributional consequences.  
To reconnect to discussions in 2.2, this implies that people are quite consistent in their (un)fairness 
evaluations of climate policies: just like the burdening of underprivileged groups can make a tax more unfair, 
a progressive targeting of society’s elites can make it more fair in the public’s eyes. Furthermore, this means 
that the perceived fairness and distributional consequences of carbon taxation can be affected at the 
beginning of policy design - before revenue options are even on the table - through the choice of target 
products or consumer groups.  
Against the robust findings in previous literature of the strength of the relation between perceived 
distributional fairness and policy attitudes (Bergquist et al 2022; Isaacson et al 2024), and the established 
direction of causality (Bergquist 2025), the observed effect on fairness perceptions is likely to mediate an 
effect on policy attitudes, albeit a smaller one. This also opens up new avenues for research on policy 
packages. It seems combinations of several push measures is not necessarily a bad idea. Certain taxes do 
have the potential to increase perceived fairness and, by extension, feasibility of others. 
However, to firmly conclude a mediated effect on policy attitudes from this kind of policy package, more 
research is needed. Importantly, one might need a larger sample to capture such a mediated effect, as it is 
likely to be smaller than the direct effect on fairness perceptions. The latter’s size was estimated small-to-
medium, suggesting a mediated effect size approaching small. 
Regarding any potential effects on policy acceptability, it was also interesting that the share of respondents 
who rated their proposal as unfair was lower for both treatment groups than the control group (especially for 
30
the need group). Since these differences were not analyzed with any statistical test they should be interpreted 
with particular caution. Nevertheless, if the differences were significant, it suggests that the treatments 
pushed some people over the important threshold from rating the policy as unfair to rating it as neutral or 
even fair. This could have particular importance for feasibility, given the short discussion on the importance 
of majority acceptability in 2.1. 
In addition, I have tested whether framing the luxury carbon tax based on an (outcome) equality principle or 
on a need principle is most effective at making the package seem more fair. Results from this were more 
inconclusive. Neither H2 or H3 could be confirmed or refuted, although if anything, results were in favor of 
H3. This tendency confirms results from Hilmersson (2024), comparing the effects on the popularity of 
carbon taxes of similar need and equality framings around complementary equal tax rebates, which suggests 
that Swedes favor the needs principle above the equality principle. However, further studies are needed to 
establish or refute a relation between fairness principle framing and fairness ratings of a luxury carbon tax. 
Based on the manipulation checks, it is possible that the differences between the two treatments were too 
small to impact perceptions. 
Future studies are needed to further validate the findings on H1 as well. It is presumably the first test of 
perceptions of a package including a luxury carbon tax, and of that policy whatsoever. This warrants some 
caution as to generalizability of the findings. Likewise, the context could be quite favorable for the 
demonstrated effect. Therefore, similar studies should be conducted in other countries. Another reason for 
caution is that the luxury carbon tax is a very novel policy which has not been publicly debated in the tested 
context, at least not widely, but if it was, it is likely to be subject to political messaging. It is unclear how that 
could alter the effect reported here. 
For future studies on perceptions and attitudes to similar policy instruments, I see three pathways. First, the 
luxury carbon tax could be studied on its own, with representative samples. It is an interesting instrument in 
itself, and should belong to a group of ”eco-social policies”, with implications for both ecological and social 
outcomes, for which public support in Sweden have recently been studied (Khan et al 2023). Besides 
comparing support or fairness perceptions for this policy with that of similar ones, such studies could also 
explore potential interaction effects. For example, Lindvall and colleagues’ (2024) found that a progressive 
compensation scheme was more successful at increasing policy support for carbon taxes among left-leaning 
individuals in Sweden than among right-leaning ones. Based on this, it seems reasonable to expect a luxury 
carbon tax to be more popular among people ideologically to the left than to the right.  
The second pathway is to study other progressive climate taxes, either in isolation or in a similar package as 
here. One alternative would be to explore ”progressive individual carbon taxes” (Dietsch 2024). This 
proposal takes some core concepts of the luxury carbon tax one step further. It suggests the creation of a tax 
schedule covering most major emission sources from individual lifestyle choices, with exponentially 
increasing marginal carbon tax rates depending on the amount of total personal emissions. Since personal 
emissions largely depend on personal wealth, the scheme is argued to put a larger burden on the wealthiest, 
31
just like the luxury carbon tax. By virtue of the complexity and wide coverage of such a scheme, it should 
primarily be suggested to respondents in isolation. 
Lastly, researchers should examine how the effects on perceived fairness and policy attitudes of these kinds 
of instruments or packages can interact not only with ideology, but also with redistributive revenue recycling 
or compensatory policies. Can effects of progressive targeting and progressive redistribution on fairness 
perceptions stack on each other, or would they rather cancel each other out? 
Regarding practical implications, this thesis primarily demonstrates how complementing ordinary carbon 
taxes with more targeted progressive ones can be a constructive tool to make more stringent climate policies 
(more) politically feasible, by increasing the perceived fairness of the policy proposal. This is especially 
relevant for countries in which revenue earmarking is not an alternative, such as Sweden (Sterner 2020; 
Lindvall et al 2024). By facilitating the achievement of climate goals, this contributes to environmental 
sustainability.  
Furthermore, there are implications for social and economic policy domains as well. The fact that the adding 
of a certain progressive tax increased fairness perceptions of a policy package suggests political 
opportunities for policy makers to find new revenue streams for public welfare expenses, and to adopt new 
redistributive policies. In an age of recurring tax cuts, austerity policies, and large financial expectations on 
rich countries to finance international climate-related costs and compensations (Chancel et al 2024), new 
sources of revenue could provide important contributions to economic sustainability.  
Regarding redistribution, the results of this thesis suggest that the wider public might agree with various 
theorists (Dietsch 2024, p. 1587) that the wealthy lack just entitlement to at least a portion of the resources 
they currently own and consume. Instruments such as luxury carbon taxes could be effective tools for 
redistribution of two key resources: money, and carbon emissions. Many other policy tools are available to 
redistribute one or both of these resources. But if climate policies have become a focal point for fairness 
concerns and demands in general, perhaps progressive carbon taxes can provide a new tool for redistribution. 
In a time with rapidly rising inequalities, which come with a range of social costs (Wilkinson et al 2017), 
climate policies might in that case also contribute to social sustainability.  
References 
Aguinis, H., & Bradley, K. J. (2014). Best Practice Recommendations for Designing and Implementing 
Experimental Vignette Methodology Studies. Organizational Research Methods, 17(4), 351–371.
Alestig, M., Nilsson Lewis, A., & Marbinah, R. (2024). Sveriges ansvar i klimatkrisen. Oxfam Sverige. 
Retrieved February 17, 2025, from https://oxfam.se/wp-content/uploads/2025/01/Rapport-Sveriges-ansvar-i-
klimatkrisen_Mediumres.pdf
Andersson, J. J. (2019). Carbon taxes and CO2 emissions: Sweden as a case study. American Economic 
Journal Economic Policy, 11(4), 1–30. 
32
Atzmüller, C., & Steiner, P. M. (2010). Experimental Vignette Studies in Survey Research. Methodology, 
6(3), 128–138.
Angrist, J. D., & Pischke, J. (2014). Mastering ’Metrics: The Path from Cause to Effect. Princeton University 
Press.
Baranzini, A., & Carattini, S. (2017). Effectiveness, earmarking and labeling: Testing the acceptability of 
carbon taxes with survey data. Environmental Economics and Policy Studies, 19, 197–227.
Benoit, P. (2020). A luxury carbon tax to address climate change and inequality: not all
carbon is created equal. Ethics Int. Aff. Retrieved from: https://www.ethicsandinternationalaffairs.org/online-
exclusives/a-luxury-carbon-tax-to-address-climate-change-and-inequality-not-all-carbon-is-created-equal 
Bergquist, M. (2025). Limiting support for environmental policies: Unfairness is a more critical barrier than 
cost and ineffectiveness. Ambio, 54(2), 350–363.
Bergquist, M., Nilsson, A., Harring, N., & Jagers, S. C. (2022). Meta-analyses of fifteen determinants of 
public opinion about climate change taxes and laws. Nature Climate Change, 12(3), 235–240. 
Bolderdijk, J. W., Steg, L., Woerdman, E., Frieswijk, R., & De Groot, J. I. M. (2017). Understanding 
effectiveness skepticism. Journal of Public Policy and Marketing, 36(2), 348–361. 
Brannlund, R., & Persson, L. (2012). To tax, or not to tax: preferences for climate policy attributes. Climate 
Policy, 12(6), 704–721.
Carattini, S., Baranzini, A., Thalmann, P., Varone, F., & Vöhringer, F. (2017). Green taxes in a post-Paris 
world: Are millions of nays inevitable? Environmental Resource Economics, 68, 97–128.
Carattini, S., Carvalho, M., & Fankhauser, S. (2018). Overcoming public resistance to carbon taxes. Wiley 
Interdisciplinary Reviews. Climate Change, 9(5), e531-n/a. 
Cervenka, A. (2022). Girig-Sverige: Så blev folkhemmet ett paradis för de superrika. ScandBook.
Chancel, L., Bothe, P., & Voituriez, T. (2024). The potential of wealth taxation to address the triple climate 
inequality crisis. Nature Climate Change, 14(1), 5–7.
Coleman, E. A., Harring, N., & Jagers, S. C. (2023). Policy attributes shape climate policy support. Policy 
Studies Journal, 51(2), 419–437. 
Deutsch, M. (1975), Equity, Equality, and Need: What Determines Which Value Will Be Used as the Basis of 
Distributive Justice?. Journal of Social Issues, 31: 137-149. 
Dietsch, P. (2024). A fairer and more effective carbon tax. Nature Sustainability, 7(12), 1584–1591.
Drews, Stefan & van den Bergh, Jeroen C.J.M. (2016). What explains public support for climate policies? A 
review of empirical and experimental studies. Climate Policy 16(7): 855-876. 
Ejelöv, E., & Nilsson, A. (2020). Individual factors influencing acceptability for environmental policies: A 
review and research agenda. Sustainability, 12(6), 2404.
Elfström, C. (2024, March 14). Klimatet: Så mycket mer släpper de rikaste ut. SVT Nyheter. https://
www.svt.se/nyheter/inrikes/klimatet-sa-mycket-mer-slapper-de-rikaste-ut--0s8y2y
Emmenegger, P., & Marx, P. (2019). The Politics of Inequality as Organised Spectacle: Why the Swiss Do 
Not Want to Tax the Rich. New Political Economy, 24(1), 103–124.
33
Esaiasson, P., Gilljam, M., Oscarsson, H., Towns, A. E., & Wängnerud, L. (2017). Metodpraktikan: konsten 
att studera samhälle, individ och marknad.
Ewald, J., Sterner, T., & Sterner, E. (2022). Understanding the resistance to carbon taxes: Drivers and 
barriers among the general public and fuel-tax protesters. Resource and Energy Economics, 70, 101331.
Finanspolitiska rådet. (2024). Economic inequality in Sweden An overview of facts and future challenges. 
Särskilda studier från Finanspolitiska rådet, 2024/1.
Geisen, E. (2022). Improve data quality by using a commitment request instead of attention checks. 
Qualtrics. Retrieved from: https://www.qualtrics.com/blog/attention-checks-and-data-quality/ 
Gevrek, Z. E., & Uyduranoglu, A. (2015). Public preferences for carbon tax attributes. Ecological 
Economics, 118, 186–197. 
Goldmann, M. (2024). Reduktionsplikten höjs: kritik från remissinstanser. Dagens Industri. Retrieved from: 
https://www.di.se/hallbart-naringsliv/remissinstanserna-om-hojda-reduktionsplikten-racker-inte-for-att-
klara-eu-s-krav/ 
Hammar, H., & Jagers, S. C. (2007). What is a fair CO2 tax increase? On fair emission reductions in the 
transport sector. Ecological Economics, 61(2–3), 377–387.
Heyen, D. A., & Wicki, M. (2024). Increasing public support for climate policy proposals: a research agenda 
on governable acceptability factors. Climate Policy, 24(6), 785–794.
Hilmersson, A. (2024). THE DEBATE ABOUT CARBON TAX REBATES: The impact of tax rebates and its 
fairness arguments on policy support in Sweden. MA thesis approved at the Department of political science, 
Gothenburg university.
IEA (International Energy Agency). (2023). Energy System: Transport. Retrieved from: 
https://www.iea.org/energy-system/transport 
Isaacson, S., Jagers, S. C., Helferich, M., & Bergquist, M. (2024). The Role of Perceived Fairness in Public 
Opinion on Sustainable Transportation Policy: A Meta-Analytic Structural Equation Model. The 
International Journal of Climate Change. Impacts and Responses, 17(1), 1–24. 
Jagers, Sverker C., Lokrantz, Elin, Persson, Sara & Sterner, Thomas (2020). Bör utsläpparna betala? in 
Ulrika Andersson, Anders Carlander & Patrik Öhberg (ed.) Regntunga skyar. Göteborgs universitet: SOM-
institutet.  
Klimatpolitiska rådet. (2025). Klimatpolitiska rådets rapport 2025. Retrieved from: https://
www.klimatpolitiskaradet.se/rapport-2025/ 
Khan, J., Emilsson, K., Fritz, M., Koch, M., Hildingsson, R., & Johansson, H. (2023). Ecological ceiling and 
social floor: public support for eco-social policies in Sweden. Sustainability Science, 18(3), 1519–1532. 
Klenert, D., Mattauch, L., Combet, E., Edenhofer, O., Hepburn, C., Rafaty, R., & Stern, N. (2018). Making 
carbon pricing work for citizens. Nature Climate Change, 8(8), 669–677.
Lundberg, L. (2024, September 4). Varför pratar alla om reduktionsplikten? RISE. Retrieved from: https://
www.ri.se/sv/nyheter/blogg/varfor-pratar-alla-om-reduktionsplikten 
Lindvall, D., Sorqvist, P., Jagers, S. C., Karlsson, M., Sjoberg, S., & Barthel, S. (2024). The Role of Fairness 
for Accepting Stricter Carbon Taxes in Sweden. CLIMATE, 2024, Vol. 12, Iss. 11, 12(11).
34
Maestre-Andrés, Sara, Drews, Stefan & van den Bergh, Jeroen. (2019). Perceived fairness and public 
acceptability of carbon pricing: a review of the literature. Climate Policy 19(9): 1186-1204. 
Mehmetoglu, M., & Jakobsen, T. G. (2022). Applied Statistics using STATA: A Guide for the Social Sciences. 
Sage Publications Limited.
Naturvårdsverket. (2024). Klimatet och transporterna. Retrieved from: https://www.naturvardsverket.se/
amnesomraden/klimatomstallningen/omraden/klimatet-och-transporterna/ 
Nilsson Lewis, A. and Nelson, H.,. (2023). Sveriges väg till en jämlik och rättvis klimatomställning. Oxfam 
Sverige. Retrieved from: https://oxfam.se/wp-content/uploads/2023/11/Oxfam.Sveriges-vag-till-en-jamlik-
och-rattvis-klimatomstallning.2023.pdf 
OECD (Organisation for Economic Co-operation and Development). (2025). OECD Environmental 
Performance Reviews: Sweden 2025. Retrieved from: https://www.oecd.org/en/publications/oecd-
environmental-performance-reviews-sweden-2025_91dcc109-en.html 
Otto, I. M., Kim, K. M., Dubrovsky, N., & Lucht, W. (2019). Shift the focus from the super-poor to the 
super-rich. Nature Climate Change, 9(2), 82–84. 
Povitkina, M., Jagers, S. C., Matti, S., & Martinsson, J. (2021). Why are carbon taxes unfair? Disentangling 
public perceptions of fairness. Global Environmental Change, 70, 102356. 
Reeskens, T., & van Oorschot, W. (2013). Equity, equality, or need? A study of popular preferences for 
welfare redistribution principles across 24 European countries. Journal of European Public Policy, 20(8), 
1174–1195.
Saelen, H., & Kallbekken, S. (2011). A choice experiment on fuel taxation and earmarking in Norway. 
Ecological Economics, 70, 2181– 2190. 
Sommer, S., Mattauch, L., & Pahle, M. (2022). Supporting carbon taxes: The role of fairness. Ecological 
Economics, 195, 107359. 
Sterner, T. (2012). Distributional effects of taxing transport fuel. Energy Policy, 41(1), 75–83.  
Sterner, T. The Carbon Tax in Sweden. In Standing up for a Sustainable World; Edward Elgar Publishing: 
Cheltenham, UK, 2020; pp. 59–67.
Sullivan, G. M., & Feinn, R. (2012). Using effect size—or why the P value is not enough. Journal of 
Graduate Medical Education, 4(3), 279–282. 
Söderholm, E. (2023). Ny kritik mot sänkt reduktionsplikt: ”Uppfyller inte kraven”. Vi Bilägare. Retrieved 
from: https://www.vibilagare.se/nyheter/ny-kritik-mot-sankt-reduktionsplikt-uppfyller-inte-kraven 
Trafikverket. (2024). Transporternas klimatpåverkan. Retrived from: https://bransch.trafikverket.se/for-dig-i-
branschen/miljo---for-dig-i-branschen/minskad-klimatpaverkan/transporternas-klimatpaverkan/ 
Wallace, C. G., & Welton, S. (2024). Taxing luxury emissions. Cornell Law Review, 109(5), 1153–1231.
Wicki, M., Huber, R. A., & Bernauer, T. (2020). Can policy-packaging increase public support for costly 
policies? Insights from a choice experiment on policies against vehicle emissions. Journal of Public Policy, 
40(4), 599–625.
Wilkinson, R. G., & Pickett, K. E. (2017). The enemy between us: The psychological and social costs of 
inequality. European Journal of Social Psychology, 47(1), 11–24.
35
Ohlendorf, N., Jacob, M., Minx, J. C., Schröder, C., & Steckel, J. C. (2021). Distributional Impacts of 
Carbon Pricing: A Meta-Analysis. Environmental & Resource Economics, 78(1), 1–42. 
Öljemark, J. (2025a). Dieselskatt. Ekonomifakta. Retrieved from: https://www.ekonomifakta.se/sakomraden/
elfakta/styrmedel/dieselskatt_1208801.html 
Öljemark, J. (2025b). Bensinskatt. Ekonomifakta. Retrieved from: https://www.ekonomifakta.se/sakomraden/
elfakta/styrmedel/bensinskatt_1208679.html