Opposing Attitudes Towards Environmental
Taxes: A Price Dilemma?
A Case of the Swedish Attitudes on Gasoline Emission Taxes
Martin Bergström Arturo Ignacio López Casanueva
Supervisor: Katarina Nordblom
Master’s Thesis in Economics, 30 hec
Spring 2023
Graduate School, School of Business, Economics and Law, University of Gothenburg,
Sweden
Abstract
This thesis investigates how people’s attitudes toward environmental taxes are shaped
and whether changes in end-consumer prices of gasoline have a direct effect on carbon tax
perception. A theoretical model is constructed to represent the individual’s choice problem
using a utility function. An empirical model is then set up based on the theoretical model
and uses cross-sectional survey data between 2011-2020 to analyze people’s attitudes towards
environmental taxes using gasoline pump prices and personal characteristics. The results in
this thesis suggest a negative relationship between pump prices and public opinion on a
carbon tax. We also investigate whether the magnitude of this relationship could differ
among population groups, however we find no significant effects. These findings contribute
to the field of research on Pigouvian taxes and how to address greenhouse-gas emissions by
putting emphasis on prices as an addition to prior determinants, as well as what it means
for policy makers aiming to implement such taxes.
1
Acknowledgements
We want to express our utmost gratitude to our supervisor Katarina Nordblom for ex-
cellent guidance and discussions throughout the whole process. Moreover, we want to thank
Jens Ewald for his help. We also thank the SOM Institute and SND for providing quality
data that made our empirical analysis possible. A big thanks to our partners Mercedes and
Julia for putting up with us and standing besides us along the way. Finally we want to thank
our families, friends and the amazing Hyllan crew for their support throughout our studies.
Not to forget HN for giving us motivation to carry on.
2
There and back again. A hobbit’s tale by Bilbo Baggins
— J.R.R. Tolkien, The Hobbit
3
Contents
1 Introduction 5
2 Literature Review 8
3 Theoretical Framework 10
3.1 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4 Data and Methodology 17
4.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5 Results & Analysis 26
6 Conclusion 31
Appendices 38
A Theoretical Model 38
A.1 Utility Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
A.2 Consumer Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
A.3 Assumptions and Implications . . . . . . . . . . . . . . . . . . . . . . . . . . 38
A.4 Agreement on tax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
A.5 Differentiation w.r.t. Gasoline Pump Prices . . . . . . . . . . . . . . . . . . 40
A.5.1 First term . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
A.5.2 Second Term . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
A.6 Comparative Statics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
B Additional Results 42
4
1 Introduction
The climate debate continues, and with growing evidence of environmental damages coming
from our consumption and pollution, more eyes are fixed on the health of the planet. The
consumption of energy and fossil fuels has a negative impact on the environment by releasing
greenhouse gases such as carbon dioxide (CO2) into the atmosphere, and in the last decades,
consumption of such products has increased in society (Ritchie, 2022). Therefore, the debate
on how to regulate consumption to reduce greenhouse gas emissions has become relevant.
Setting a tax on consumption that leads to negative externalities is one of the methods used
to regulate markets that emit significantly and is increasingly applied all over the world
(OECD, 2023).
Pigouvian taxes, as they are mostly referred to, are set to correct for negative effects
caused by market activities (Pigou, 1920). From a very basic economic perspective, the
tax increases the price which in turn distorts consumer demand, consequently causing a fall
in consumption and thus a reduction of the negative externality. From this mechanism,
the idea of Pigouvian taxes is to reach an optimal level of consumption from a sustainable
point of view. However, individuals might not only care about economic efficiency, and
although the theory seems very straightforward, issues hindering the full use of the tax are
recurringly appearing. Ewald, Sterner, and Sterner (2022) highlight public support and
economic interests as obstacles, where companies within carbon intense sectors continuously
protect their economic interests.
Public support, which will be the focus area of this thesis, is emphasized by the authors
as crucial to implementing a Pigouvian tax mechanism. Increasing or applying a tax to a
good that is causing harm in some way might not only be perceived as a means to deal with
that externality but rather as conflicting with individual preferences. Since many people
depend on fossil fuels to transport themselves, people have been shown to perceive the price
high enough without the tax (Povitkina et al., 2021). Recent examples are the Yellow Vest
5
movement in France and Bensinupproret 2.0 in Sweden where increases in gasoline taxes, or
current taxes being perceived as too high, led to public resistance (Douenne and Fabre, 2022;
Ewald, Sterner, and Sterner, 2022).
The purpose of this study is to analyze the impact that prices of fossil fuels have on
attitudes towards environmental taxes and what the economic and social determinants of
opinion towards these taxes are. Specifically, we focus on gasoline prices and their effect
on carbon tax opinion. This is done using a national survey sample in Sweden between the
years 2011-2020. A meaningful question is if environmental attitudes are inherent or can
be shaped by the given context. In times of uncertain and unfavorable economic outlooks
where global pandemics, bottlenecks in supply chains, and wars cause major macroeconomic
effects, households are more likely to be responsive to price changes. This was shown during
the Swedish election in 2022, where political parties proposed tax reductions for gasoline in
order to gain public support.
By understanding whether pump prices of gasoline affect public opinion on the carbon
tax associated with it, policymakers can gain important insight into how to address tax
opposition. If drastic price changes do cause opposition towards a carbon tax, politicians
in favor of the tax face potential challenges in keeping the tax, as opposing parties can
take advantage of and gain public support from the emerged dissatisfaction caused by price
changes. However, this study alone could not be used to draw such causal inferences but
would help guide research toward whether prices could play a crucial part in shaping attitudes.
In a small country such as Sweden, the production price of gasoline is difficult to control inside
the country, which makes taxes a flexible tool to alter the consumer cost of gasoline from
within the country. Even though a price shock might have derived from an exogenous shock, it
could on an individual level be reasonable to focus on the components that one’s government
can affect, i.e., the taxes. From this arises the possibility of an interesting phenomenon,
where people could use the tax as a scapegoat for high gasoline prices.
6
The final price of gasoline in Sweden includes an energy tax and a carbon emission tax
(both of them are taxed per liter), with the VAT tax of 25% added after all these are
summed up (Skatteverket, 2023). The prices used for the analysis in the study are the
weekly weighted average gasoline prices in Sweden, obtained from the European Commission
(2023). Although gasoline prices vary a bit from region to region, the prices are still quite
uniform across Sweden, therefore the weighted average price is a good measure of the prices
across Sweden.
Prior studies discuss the perception of unfairness as a major driver for opposition to-
wards carbon taxes, where the public perceives the tax to be regressive or to be unequally
implemented across countries and regions (Ewald, Sterner, and Sterner, 2022). The case of
fairness has also been linked to pump prices being perceived as too high, making a carbon
tax difficult to motivate (Povitkina et al., 2021). While recent studies have gathered factors
of what affects public opinion towards a carbon tax, this study contributes to this collection
by adding the aspect of end-consumer prices on gasoline. In other words, this study helps
determine whether people’s opinion towards the tax is given not only by formerly found
determinants but also by exogenous factors such as the gasoline price. By finding whether
gross prices of gasoline act as a determinant for attitudes on carbon taxes, policy makers
can bear in mind how to shape future tax schemes. For example, if uncontrolled exogenous
shocks cause tax opposition through prices, the Pigouvian mechanism might need to be con-
structed to respond to such shocks. Moreover, our study provides a theoretical model fit to
observe the consumer’s utility when the price is introduced to the utility model, and how
price changes affect the utility gained from taxes. Since most studies within the field solely
focus on conducting empirical studies, ours bring value in terms of a combination of empirics
and theory.
With the help of our theoretical findings, together with previous literature on the topic, we
present hypotheses on how carbon tax attitudes are shaped by pump prices. Not only do we
7
believe pump prices of gasoline to affect carbon tax attitudes, but we also aim to understand
whether this effect differs in size among various groups in society. More specifically, we
compare this effect between people that are more or less concerned about the climate, as
well as between those dependent on a car or not1. Using cross-sectional data from the SOM
Institute between 2011-2020 (University of Gothenburg, SOM Institute, 2022), we analyze
Swedish attitudes with an ordinal logit regression.
The rest of this study contains the following; chapter 2 includes a literature review on
carbon taxation and public opinion. Moreover, a theoretical model is presented in chapter
3 to explain how theory can support our hypotheses, which will be introduced thereafter.
In chapter 4, data and methodology are described. The results of the empirical analysis are
displayed and discussed in chapter 5. Chapter 6 concludes.
2 Literature Review
The reasons why people support or oppose an environmental tax have not completely reached
consensus among researchers. However, there are recurrent studies of some factors playing a
larger role. While earlier findings point towards self-interest, area of living, and education to
affect one’s opinion of an environmental tax (Löfgren and Nordblom, 2006; Hammar, Jagers,
and Nordblom, 2008), other drivers might be easier to affect by policy makers. For people
to believe that the tax is efficient, trust in politicians could be a key component (Hammar
and Jagers, 2006). From a global perspective, the European countries had relatively higher
environmental taxes than the rest of the world, which could have been a signal of higher
trust in their respective governments (Sterner and Köhlin, 2003). Löfgren and Nordblom
(2006) and Löfgren and Nordblom (2009) emphasize awareness towards the problem as well
1Note that during the period studied the share of electric vehicles in Sweden was very small; accounting
for approx. 6% in 2020 including hybrid cars (Trafikanalys, 2021). Therefore we do not take this into account
when conducting our study.
8
as labeling of the tax to be potential elements creating higher support for implementing
environmental taxes. In terms of consumption behavior, Ghalwash (2007) studies consumer
sensitivity towards consumption prices where an increase in the environmental tax connected
to the good causes a greater reduction in consumption compared to the same monetary
change in the producer price, signaling that people might react stronger in terms of behavior
towards policy changes compared to exogenous shocks or changes made by firms. Focusing
on attitudes instead, the present study will focus on the effect that general price fluctuations
of gasoline, including the carbon tax, have on the perception of these taxes.
More recent studies on attitudes towards environmental taxes have continued to point
towards various determinants, indicating many components play a role in the matter. Flood,
Islam, and Sterner (2010) show through demand elasticity that initiating the tax might be
difficult in a country while increasing it where the tax is already high might prove easier.
The reason could be due to public habit formations, where new consumption patterns such
as driving electric cars cause people to be less affected by changes in carbon taxes. Impor-
tantly, studies in different countries prove that factors could also be of dissimilar significance.
A survey conducted in Norway showed that one’s worry about the environment is greatly
affecting the view on an environmental tax, while self-interest was of much less importance
(Kallbekken and Sælen, 2011). Fairbrother, Sevä, and Kulin (2019) found similar evidence,
but only among countries with high political trust. In cases where political trust is low, deter-
minants such as environmental concern seem to not affect the attitude toward carbon taxes,
which speaks for trust being an important factor to gain public support. While trust has
been recurringly important to gain support for the tax, newer studies have also found other
aspects to affect public opinion. Douenne and Fabre (2022) finds people to overestimate the
negative aspects of the tax such as considering it regressive, and where compensatory mea-
sures are induced, they believe themselves to be worse off than they are. Criticism towards
the efficiency of the tax is also brought up as a hindrance. Moreover, opposition could also
mistakenly perceive the tax as higher than it is, showing the need for clear communication
9
from policy makers (Umit and Schaffer, 2020).
More recent studies show that the perception of fairness plays an important role in shaping
public attitudes toward carbon taxes. From a wide study by Bergquist et al. (2022), the
most important determinant of climate tax policy attitudes was fairness. Equal results were
found by Ewald, Sterner, and Sterner (2022). Furthermore, a specific study on why people
consider carbon taxes unfair found that people consider the consumer price to be high enough
already (Povitkina et al., 2021). There was also a consensual opinion that these taxes will
economically harm people with low-income and those living in rural areas the most, causing
further perception of unfairness. The present study contributes by further investigating
whether this feeling of unfairness could be directly influenced by changes in the pump price
of gasoline, and whether or not its effect could differ among groups in the society.
3 Theoretical Framework
3.1 The Model
The theoretical framework used for analyzing the opinion and environmental attitude will
follow a similar setup as used by Hammar, Jagers, and Nordblom (2008). The individuals
have the utility function (1). The utility function is assumed to have additive separability
for the simplicity of the analysis. The consumption is composed of two consumption goods.
The individual’s consumption of good F is cF , which represents the individual’s consumption
of gasoline fuel and its consumption causes a negative externality by emitting CO2 into the
environment. The consumption of the alternative good A is cA and does not cause a negative
externality. The goods are assumed to be substitutes, yet not perfect substitutes, which
allows for capturing different effects for different groups across the population.
X is the total negative externality on the environment caused by the consumption of the
10
good cF across the population. In plain words, it is the total impact on the environment
caused by the greenhouse gas emissions of gasoline consumption. The marginal environmental
damage of the consumption of good cF is measured by the damage parameter γF , where
γF > 0. The total impact of the population’s consumption is calculated by the summation
of consumption across the total population H. This linear damage form of the externality X
follows Hammar, Jagers, and Nordblom (2008) structure. The damage caused has a negative
impact on the individual, perceived and/or health related.
Similar to Putler (1992), the utility function is in terms of L which is the loss (or gain)
on utility experienced by the individual from the difference between ticket price p̂ of gasoline
and a reference price RP that the individual forms arbitrarily before he or she solves the
consumption problem. The reference price is, therefore, exogenous and does not intervene in
the budget constraint, simplifying the problem. As one can see L is structured as a bargain
price, although we want to focus on the negative effect, where the observed prices are higher
than the reference price. In that case, the individual will experience disutility when observing
prices exceeding their reference price from the perception he or she gets from paying a higher
price. On the opposite way, when experiencing ticket prices below the reference price, the
consumer experiences a utility gain from perceiving a “bargain” price. The main difference
between this utility function and the one used in Putler (1992) is that the perceived loss
(gain) from prices is not marginal in this study, thus L does not affect the consumer choices.
U(cF , cA, X, L) (1)
Where:
p̂cF + cA∑= Y
X = γ HF h=1 cF,h
L = RP − p̂
p̂ = pF + τF
11
The ticket price p̂ of the consumer good cF includes the respective CO2 emission tax
τF and is the price the individual faces when purchasing gasoline at the gas station. The
individual does not separate the different price structures included in the gasoline prices when
filling the tank and thinks only of the final ticket price p̂ when purchasing it. The individual
faces a budget constraint tied to their total income Y. For simplicity the individual’s income
is assumed exogenous. The price of the alternative good cA is normalized and the alternative
good has no consumption tax.
3.1.1 Assumptions
The model is supported by the assumptions shown below in (2):
∂U ∂2U ∂U ∂2U
> 0; < 0; > 0; < 0
∂c 2 2F ∂cF ∂cA ∂cA (2)
∂U ≤ ∂
2U ∂U ∂2U
0; ≤ 0; ≥ 0; ≥ 0
∂X ∂X2 ∂L ∂L2
The individual has a positive and decreasing marginal effect on utility with both con-
sumption goods. The marginal effect on the utility of negative impact on the environment
is negative. The second derivative of utility in terms of the environmental impact is nega-
tive, meaning that the disutility from pollution increases at a growing rate. The “bargain
price” structure of L in the utility function has a positive and increasing effect on the utility
function; although the effect of prices on L is negative, thus prices affect the utility function
through L negatively.
Since this study’s objective is not actually to find the efficiency of the environmental tax
in question, the effects the tax has on consumption are not necessary to analyze. The study
is focused on the perception of the environmental tax, and more specifically, on the effects
that prices have on the perception of the tax. Therefore, the problem is solved as follows.
First, the individual will solve the consumption problem following the utility function. When
maximizing its consumption, we assume the individual does not take into account the negative
externality produced from his consumption of gasoline, thus taking X as exogenous. Note
12
that L does not affect the individual’s consumption problem. The individual’s attitude
towards the current tax level τF can be evaluated by (3):
∣
∂U
∂τ ∣∣∣ (3)F τF
If (3) is negative, that means that the current tax rate τF is higher than the individual
would like and vice-versa. If (3) is equal to 0 then the tax rate is currently at the optimal
level for the individual. A negative (3) means the individual does not agree with a tax
increment, while a positive result means he or she agrees with a tax increment. Theoretically,
if the current tax is at the individual’s optimal level, it would result in disagreement with an
increment in tax levels at the current moment, ceteris paribus. Equation (4)2 is the derivative
of the utility function in terms of the tax, therefore when evaluated at τF will give us the
sign of (3), which tells whether the individual wants the tax to be higher, lower, or remain
as it is.
[ ]
H
∂U ∂U ∂c ∑F ∂U ∂cA ∂U ∂cF,h
= + + γF −
∂U
(4)
∂τF ∂cF ∂p̂ ∂cA ∂p̂ ∂X ∂p̂ ∂L
h=1
By substituting the results obtained from the individual’s consumer problem (equation
(10) in Appendix A), into equation (4) we can rewrite it as equation (5).
[ ]
H
∂U − ∂U ∂U
∑ ∂cF,h ∂U
= cF + γF − (5)
∂τF ∂cA ∂X ∂p̂ ∂L
h=1
Equation (5) shows the different things affecting the individual’s perception of the tax.
The first term of the equation, − ∂U c , is important as it allows to see the effects of the indi-
∂c FA
vidual’s consumption characteristics on his/her tax opinion; note that the effect is negative.
This first term is the marginal utility of the alternative good times how much gasoline the
individual consumes. For example, individuals who drive a lot will observe a high cF nat-
urally, having a high negative impact on their tax perception. Individuals with low income
will tend to observe a high marginal utility of cA and thus will also have a large negative
2Note that ∂p̂∂τ = 1. For the detailed steps to reach equations (4), (5) and (6) see Appendix A.F
13
impact on the tax perception. Such groups of the population should be more reluctant to an
environmental tax increase.
The environmental attitudes of the individual are characterized by ∂U , where the larger
∂X
the absolute value of this, the higher the environmental concern about the externality the
individual has. Therefore, this element of equation (5) captures the effect of the individual’s
environmental concern on the perception of taxes, where one can see that the effect is positive.
The expression ∂U shows the direct effect that the change in the ticket price from the
∂L
change in the tax has on the tax perception through the bargain price structure; in simpler
terms, how the individual perceives and reacts at the new price at the moment of purchase
given a tax increase. When ∂U = 0 the problem is reduced to the traditional economic
∂L
theory, where the consumption and externality are the only things affecting the individual’s
optimal tax rate. In any other case, the individual experiences an additional disutility through
gasoline prices, negatively affecting their perception of the environmental tax.
Since there is an ambiguous effect on the tax perception the individual’s characteristics
will determine whether (5) is positive, negative, or zero; in other words, whether the individ-
ual supports the proposal to increase taxes or not. Whenever the effect of climate concern
has a larger effect than the income effect and the direct effect of prices, the individual will
agree with a carbon tax increase, making expression (5) positive. Whenever the other effects
dominate the environmental concern, the individual will disagree with a tax increase. Exam-
ples of such situations could be when individuals weigh a bargain price heavily L, or when
they drive a lot or have a high marginal utility on the alternative consumption good.
To observe how a change in pump prices affects the individual’s tax attitude, we take the
total differential of equation (5) with respect to the pump prices, which yields equation (6):
14
∂U ∣∂
∂τ ∣∣ [ ]H 2 [ ]∣ H−∂2U ∂c 2 ∑ ∑ 2 2F A= cF − ∂U ∂cF ∂ U ∂cF,h ∂U ∂ cF,h ∂ U+ γF + γF − (6)∂p̂ ∂c2 ∂p̂ ∂c ∂p̂ ∂X2 ∂p̂ ∂X ∂p̂2 ∂L2τF A A h=1 h=1
Assuming that the demand function for gasoline is convex across the population, equation
(6) will be negative for people who have a price inelastic demand for gasoline3. Several
studies show that gasoline demand is price inelastic in Sweden and across countries in general
(Coglianese et al., 2017; Dahl, 2012; Flood, Islam, and Sterner, 2010).
3.2 Hypotheses
Supported by the theoretical model we formulate the following three hypotheses about how
prices affect the perception of emission taxes on fuel consumption.
Hypothesis 1. Increases in pump prices of gasoline will lead to a lower acceptance of carbon
taxes on fuel.
Given that gasoline is a quite price inelastic good and recent studies about Sweden report
a very price inelastic demand, an increase in gasoline prices will generally lead to a rejection
of increasing carbon emission taxes. Since the consumer pays the ticket price that includes
the taxes, one does not think about whether the portion of the tax is large or not. There
could be a tendency to blame something whereas taxes are potentially used as the scapegoat.
Pigouvian taxes have as purpose to distort consumer preferences through taxation in order to
regulate the market in question. Hypothesis 1, therefore, suggests that ∂U > 0, which means
∂L̂
that individuals observe an additional direct negative effect from price increases, independent
of the income and substitution effects and their environmental attitudes.
Hypothesis 2. Population groups who are highly dependent on consuming gasoline will
observe a larger negative impact in carbon tax attitudes from an increase in pump prices of
gasoline.
3The detailed steps to this result are available in Appendix A.
15
Such consumers will most likely be affected more by an increase in price, as equation
(5) illustrates the direct effects of a price increase on the tax opinion. Equation (6) shows
2
the indirect effects involved, where the equation will not only be negative, but −∂ U ∂cA
∂c2
c −
A ∂p̂
F
∂U ∂cF < 0, this will be the case given that these groups tend to have a very inelastic demand
∂cA ∂p̂
of gasoline4. Highly dependent consumers might feel unfairness in increasing environmental
taxes given their high dependence on fuel, such as people driving daily to work. For example,
people living in rural areas that need to drive for commuting could lack alternatives such
as public transport. These people will have a very high-income effect from a tax increase
or price increase, thus making the previous expression of equation (6) negative and large in
magnitude. Since consumers do not separate taxes at the moment of consumption, they will
tend to seek an answer for the price increase and possibly blame the tax.
Hypothesis 3. Population groups who are more concerned about the climate will observe a
smaller negative impact in carbon tax attitudes from an increase in pump prices of gasoline.
People with high concerns about the environment will tend to support fuel emission taxes
as they have stronger inherent environmental attitudes independent of whether fuel prices
are high or not. Therefore, fuel prices will have a lower or null impact on attitude towards
fuel emission taxes for this group. Such people that are aware and concerned about the
emissions released by gasoline consumption will have a large and negative ∂U . Even though
∂X
the consumption problem is solved independently of the negative externality and the bargain
price, individuals with high environmental concerns will possibly reflect it by consuming a
larger portion of their budget in the alternative good rather than consuming gasoline. The
expression −∂2U ∂cA2 cF − ∂U ∂cF < 0 in equation (6) will be smaller in magnitude, by having a∂cA ∂p̂ ∂cA ∂p̂
smaller consumption of gasoline. This expression could potentially be positive for people with
a more price elastic demand for gasoline, who can for example choose to commute by using
a bicycle or public transport instead of driving, thus easily substituting fuel consumption for
the alternative good.
4See Appendix A for details.
16
4 Data and Methodology
4.1 Data
To answer the hypotheses, we merge survey data from the SOM surveys between the years
2011 and 2020 with weekly average gasoline pump prices in Sweden, available at the European
Commission (European Commission, 2023). The weekly prices are linked to when each
respondent hands in their survey. From 2011 onwards the survey question on whether the
individual agrees with an increase in emission taxes was included. Thus, the investigated
period is between 2011 and 2020, which is the latest publication from the SOM Institute at
the time of this study. More specifically, the study covers responses from the years 2011,
2012, 2013, 2014, 2015, 2019, and 2020 as they are the years when the relevant questions
to the study were included in the survey. The list of variables and their respective sources
are presented in Table 1 as well as their corresponding descriptive statistics in Table 2. The
surveys conducted by the SOM Institute are annual, where a randomized sub-sample of the
Swedish population is selected each year, making it cross-sectional data.
Table 3 shows how well the sample represents the Swedish population in terms of gender,
age, education, and political orientation. Noteworthy is that younger people (age 16-35) are
under-represented by over 4%, whereas people between 56-75 years are over-represented by a
similar proportion. All in all, older people respond to the survey to a greater extent except for
people between 76-86 years. It is also clear that people with higher education respond to the
survey more frequently. Regarding ideology, our sample represents the country’s population
quite well. In general, the sample used for the study resembles well to Sweden’s population.
Our dependent variable is based on a five-point Likert scale, where “Increase the carbon
tax on gasoline” can be answered on a scale from “Very bad proposal” to “Very good pro-
posal”. Initially, one would be interested to see whether there exists opposition towards the
carbon tax. There is evidence of opposition to be found in political debates, such as the elec-
17
Table 1: Variable List and Sources
Source
Proposal: Increase the CO2 tax on gasoline SOM-institute (2011-2020)
Gasoline pump price European Commission 2023
Drives car frequently (dummy) SOM-institute (2011-2020)
Rural residential area (dummy) SOM-institute (2011-2020)
Concern about: Changes in the Earth´s climate SOM-institute (2011-2020)
Subjective placement on ideological left-right scale SOM-institute (2011-2020)
Female (dummy) SOM-institute (2011-2020)
Age SOM-institute (2011-2020)
Gross household income SOM-institute (2011-2020)
Educational attainment SOM-institute (2011-2020)
Table 2: Summary Statistics
Obs. Mean Std. Dev. min max
Proposal: Increase the CO2 tax on gasoline 13,822 2.66 1.25 1 5
Gasoline Pump Price 13,822 14.28 0.82 12.07 16.02
Drives car frequently (dummy) 13,822 0.66 0.47 0 1
Rural residential area (dummy) 13,822 0.40 0.49 0 1
Concern about: Changes in the Earth´s climate 13,822 2.26 0.73 1 3
Subjective placement on ideological scale 13,822 3.05 1.17 1 5
Female (dummy) 13,822 0.51 0.50 0 1
Age 13,822 51.17 17.44 16 85
Gross household income 13,822 3.34 1.35 1 5
Educational attainment 13,822 2.27 0.71 1 3
18
Table 3: Representation of sample
Sample (%) Sweden (%) Diff. (%)
Gender
Female 50.63 50.68 0.06
Male 49.37 49.32 -0.06
Age
16-25 9.63 14.12 -4.49
26-35 12.38 17.19 -4.81
36-45 15.72 15.95 -0.23
46-55 17.77 15.60 2.17
56-65 19.85 14.89 4.96
66-75 17.44 12.68 4.76
76-85 7.21 9.57 -2.36
Education
Low (Elementary school or less) 15.33 19.01 -3.68
Medium (Highschool but no university) 42.18 44.86 -2.68
High (University studies/degree +) 42.49 36.13 6.36
Political Orientation
Clearly to the left 10.59 10.40 0.19
Somewhat to the left 22.40 22.60 -0.20
Neither left nor right 29.34 31.00 -1.66
Somewhat to the right 26.49 25.60 0.89
Clearly to the right 11.19 10.50 0.69
*Source: Statistic Sweden and SOM Institute
19
tion campaign run by the right-winged parties in Sweden, where a reduction of the energy
tax by 80 cents per liter was implemented on the 1st of January 2023 (Regeringskansliet,
2022. Moreover, prior studies show that not only do taxes rank as the least popular means
to tackle environmental issues (Fairbrother, 2022), but that opposition groups have formed
in recent years (Ewald, Sterner, and Sterner, 2022; Douenne and Fabre, 2022). Using our
sample from the SOM surveys, we can also observe a large share of the respondents opposing
the proposal of an increase in a carbon tax (Table 4).
Table 4: Frequency Table of Proposal to Increase the CO2 tax on gasoline
Frequency Percent Cumulative
Very bad proposal 3,118 22.56 22.56
Rather bad proposal 3,333 24.11 46.67
Neither good nor bad proposal 3,678 26.61 73.28
Rather good proposal 2,491 18.02 91.30
Very good proposal 1,202 8.70 100.00
Total 13,822 100.00
Our main explanatory variable is the gross pump price of gasoline throughout the period
from 2011 to 2021. As can be seen in figure 1, our time frame does not include any major
price shocks. In past years, however, prices in Sweden have surged, suggesting a revisit of
the subject in the near future. What is also observed is that the carbon tax has remained
almost constant during the period, ranging between 2,44-2,66 SEK. In order to investigate
how prices could affect attitudes for each individual when responding to the survey, we use
the reported date when someone’s form was received by the SOM Institute, and then assign
the most recent weekly price prior to that date5. We assume most people are quite up to date
5Since some people respond to the survey a few months later than it was given to them, data from pump
prices for some months of 2021 is used.
20
Figure 1: Gasoline Pump Prices and Carbon Tax on Gasoline
when it comes to gasoline prices at the pump, and from our hypothesis, have this in mind
when responding to the idea of increasing carbon taxes for petrol. It is worth mentioning
that since the surveys are handed out randomly and not to the same individuals each year,
we lack the opportunity to conduct a panel data study. However, since our main independent
variable consists of weekly prices, we can link these to the dates the respondents handed in
their forms.
The other independent variables of main interest are environmental concern, car usage,
area of living, and ideology. From earlier studies, one can see that people concerned for
the environment generally support a Pigouvian tax (e.g. Bergquist et al., 2022; Kaplowitz
and McCright, 2015). While Kallbekken and Sælen (2011) found little to no predictive
power from accessing or using a car towards supporting a carbon tax, other studies find a
negative relationship between car usage and supporting a carbon tax (Hammar, Jagers, and
Nordblom, 2008; Umit and Schaffer, 2020). Moreover, people living in cities are found to
perceive a carbon tax more positively compared to people living in the countryside (Douenne
21
and Fabre, 2022; Ewald, Sterner, and Sterner, 2022). Both using a car frequently and living
in rural areas could imply a higher car dependency, which is included in one of our hypotheses
of interest. All these variables are ordinal whereas environmental concern is based on a four-
point Likert scale from ”Of no concern at all” to ”Of major concern”, ideology goes from
far-left to far-right on a 5-point scale, and area of living is based on a 7-point scale ranging
from “Pure countryside” to “Big city: central”. As for environmental concern, we make the
variable into a 3-level scale, grouping the least environmentally concerned people6. First of
all, the groups include very few observations, with 311 and 2,064 respectively. Moreover,
people could exaggerate their concerns which would explain the low amount of observations
in the group expressing the least concern. Just like Hammar and Jagers (2006), we create a
dummy for car usage, whereas the most frequent drivers are given the value 1. A dummy is
also made for the variable area of living, whereas we group people living in the countryside
and in small towns in our dummy Rural.
Ideology could also explain differences in attitudes regarding an environmental tax, where
left-wing people seem to be more positive towards taxes in general. Jagers, Martinsson, and
Matti (2019) examine people’s attitude when compensatory measures are introduced to find
that right-wing people tend to consider it a good idea, while left-wing people generally prefer
a plain carbon tax, irrespective of income.
The rest of the covariates control for individual traits and characteristics. These are
age, sex, household income level, and education level. As a control variable age squared is
also included because of the quadratic relationship between age and our dependent variable.
Our variable controlling for sex is constructed as a female dummy where being female = 17.
The household income level is divided into 5 levels created by the SOM Institute8. In our
6In our regressions the least environmentally concerned group, as well as the ideology group being politi-
cally centered, are omitted, as they are the reference groups.
7Our sample included 19 observations where respondents selected “others” as sex. These are excluded
from the regression since we lack observations to draw any conclusions for that group.
8The levels are divided in the following order of annual household income in SEK: 100,000-200,000; 200,001-
22
regressions, people in the middle income level (Medium) are omitted. Educational levels are
divided into 3 different levels, where the first is elementary school or less, the second level is
high-school education and the top level is university level education. Lower education level
people are omitted in the regression.
A limit to our study arises due to the narrow range of years, because the main question of
our analysis has been included in the SOM survey since 2011. One clear limitation of the data
is the missing years of 2016, 2017, and 2018, as well as data from years prior to 2011. For the
three years excluded in this time period, there were too many missing variables of interest for
us to be able to perform the analysis. Continuing to pose this question in the coming years of
the survey would be interesting for further research, since a longer time period could capture
more global events causing shifts in gasoline prices, such as the most recent changes observed
in figure 1. Another variable relevant to control for is government trust, which seems to be
positively related to accepting a carbon tax (Umit and Schaffer, 2020; Ewald et al., 2022;
Fairbrother, 2019). However, this variable is only available for the years 2011-2013 and 2019
in the SOM survey, making our regression lose too many observations and years. Thus,
we exclude this variable in our main regression9. Furthermore, the survey lacked questions
regarding perceived tax efficiency as well as how fair the tax is considered to be, which puts
some additional limitations to our study since these factors could be important determinants
for the attitudes (Rotaris and Danielis, 2019; Ewald, Sterner, and Sterner, 2022).
The variables chosen for the analysis have relatively low correlations between each other.
The correlations matrix heatmap is presented in figure 2. A relatively large correlation is
found between whether the individual frequently drives a car and lives in a rural area, given
that people in rural areas are often more dependent on cars for transport as public transport
300,000; 300,001-500 000; 500,001-700,000; 700,001+.
9We run a separate regression for only these years including government trust. This does not change our
main regression results, and government trust is negatively related to carbon tax attitude and is statistically
significant.
23
Figure 2: Correlation Matrix Heatmap
is not very efficient in these areas. The correlation is still not alarmingly high between these
two variables, with a value of approx. 0.23. On the control variables, one can find a naturally
high correlation between household income level and education level. Gasoline prices are not
correlated with any variable. There are no correlations above .4 in absolute values.
4.2 Methodology
We estimate a model with Proposal: To increase the CO2 tax on gasoline as the dependent
variable, and weekly pump prices of gasoline as the main regressor. In the model we control
for demographic and psychological socioeconomic variables. We link the survey responses to
the relevant weekly gross pump price before answering the survey. Since the questionnaire is
mostly answered analogously, we use information on the date of handing in the survey and
tie that to the relevant prices prior to that date. We use weekly price data, not adjusted for
inflation. The reason is that when people purchase and consume the good, they might not
24
consider the level of the price in relation to potential inflation.
This study uses an ordered logit regression to empirically analyze the hypotheses. Since
dealing with an ordinal outcome variable, using the ordered logit helps to estimate the ex-
pected probabilities of each different category outcome from the dependent variable. Al-
ternative to the ordered logit, the ordered probit, and LPM estimations are also suitable
methods for the present analysis. The ordered logit and ordered probit are very similar, and
the main difference between these two methods is the distribution used for estimating the
probabilities; it is logistic for the logit and standard normal for the probit. As for the LPM,
the estimation results are much easier to interpret. However, to use an LPM, we would need
to make the outcome variable binary. By doing so, we would lose a lot of information given
that the variable initially has five different possible outcomes. Another issue of the LPM
model is that it can potentially have results with probabilities over 1 and under 0, given that
it is a linear estimation, making it a less accurate method for this type of categorical data.
Since we cannot measure the difference between the rank levels of our dependent variable,
OLS and binary models would not be suitable. The ordered logit, however, does not assume
the ranks to be of equal distance between each other, which makes it a good fit for this study.
Notwithstanding, the estimations from the ordered probit and the LPM have very similar
results to the ones obtained with the ordered logit (See Table 11 in Appendix B).
As mentioned before, the interpretation of the results is not as straightforward as if using
an LPM or a regular OLS estimation. Therefore, to analyze the magnitude of the results,
one needs to check the marginal effects since the coefficients obtained from the ordered logit
do not explain the size of the effect on the explanatory variable but only indicate whether
the effect is positive or negative.
25
The main regression is estimated as follows in equation (7)10:
Prop.to.increase.CO2.tax = gasoline.prices ∗ β1 + driving.car ∗ β2 + ∗env.concernβ3
+ area.of.living ∗ β4 + ideology ∗ β5 + age ∗ β6 + age2 ∗ β7 + income ∗ β8
+ education ∗ β9 + female ∗ β10 + year ∗ β11 (7)
Since we conduct a regression on data over time, a time trend of carbon tax attitudes
or gasoline prices might cause omitted variable bias. This omitted variable bias could arise
if there is a change in the trend of attitudes, which would not be captured by the other
variables. Therefore we include the variable year. Estimation without time-trend has the
same results.
5 Results & Analysis
The results of the ordered logit estimations are presented in Table 511. Model (1) is the base
estimation of the model. In order to gain even more detailed regression results, one could
include yearly dummies to control for all the variation between the years, which means that
we only measure effects from price differences within the year. This is referred to as time
fixed-effects. The results for the estimation including time fixed-effects with year dummies
were not included in the analysis because of the small price variations within each year.
Because of this, the time-fixed effects did not produce significant results and are not included
in the results tables given that those results do not bring any additional information to the
discussion. Models (2-3) include interaction terms to capture the effects of the interaction
between gasoline prices and car dependency and climate concern respectively. Model (4)
includes both interaction terms simultaneously. For the relevance of this thesis, we only
present a plot of the average marginal effects of the main independent variable Gasoline
10Models 2 and 3 have the same regression but each with an interaction term (e.g., gaso-
line.prices*driving.car). Model 4 includes both interaction terms.
11Note that these coefficients can only tell the direction and significance of the effect. To interpret magni-
tude average marginal effects need to be used.
26
Pump Price for model (1) in this section12. The tables showing the marginal effects of the
estimation model (1) are presented in Appendix B (Tables 6-10).
The estimations for our models (1-4) yield results in favor of Hypothesis 1, where an
increase in gasoline pump prices has a significant negative effect on carbon tax attitudes. In
other words, the results indicate that people experiencing higher gasoline prices tend to view
a proposal to increase carbon taxes more negatively. In the same way, all models displayed
suggest that people that are more dependent on gasoline will tend to support the proposal
to increase the environmental tax less, which can be seen from the negative and significant
coefficients for driving a car frequently (with 1% and 5% significance levels in models (1),
(3) and (2), (4) respectively). This can also be observed among people living in more rural
areas, where we find a negative association on a 1% significance level. Models (1) and (2)
present the expected sign and significant results for climate concern, where people with more
concern towards the environment have positive and larger support for a tax increase. Model
(3) and (4) loses significance in climate concern as well as the interaction term between
climate concern and gasoline prices. The coefficients for climate concern in estimations (3)
and (4) are both insignificant. These coefficients represent the respective categories of climate
concern given that the price is 0. Therefore, these results lack economic meaning and their
insignificant results do not come as a surprise. The interaction is also insignificant. The
remaining results are similar to the other models and still indicate a negative and significant
effect from driving a car frequently.
The average marginal effects of gasoline prices in model (1) are plotted in figure 313. The
impact on the probability of having each of the outcomes given a 1 SEK increase in the prices
of gasoline is visualized. An increase in prices will increase the probability of having outcome
12The average marginal effects of Gasoline Pump Price for models (2)-(4) are not included since the results
are similar and the interactions are insignificant.
13Average marginal effects are interpreted by how an increase in the given variable affects the probability
of each outcome of the dependant variable happening
27
Table 5: Ordered Logit Estimation Results for Proposal: Increase the CO2 tax on petrol
(1) (2) (3) (4)
Gasoline Price -0.11∗∗∗ (0.02) -0.13∗∗∗ (0.03) -0.13∗∗∗ (0.05) -0.15∗∗ (0.06)
Drives Car Frequently -0.76∗∗∗ (0.04) -1.15∗∗ (0.57) -0.76∗∗∗ (0.04) -1.17∗∗ (0.57)
Climate Concern
Some concern 0.76∗∗∗ (0.05) 0.76∗∗∗ (0.05) 0.46 (0.80) 0.43 (0.80)
Major concern 1.53∗∗∗ (0.05) 1.53∗∗∗ (0.05) 1.31 (0.81) 1.25 (0.82)
Drives Car Freq.×Gasoline Price 0.03 (0.04) 0.03 (0.04)
Climate Concern Interactions
Some concern×Gasoline Price 0.02 (0.06) 0.02 (0.06)
Major concern×Gasoline Price 0.02 (0.06) 0.02 (0.06)
Rural Area Dummy -0.42∗∗∗ (0.03) -0.42∗∗∗ (0.03) -0.42∗∗∗ (0.03) -0.42∗∗∗ (0.03)
Ideology
Clearly to the left 0.74∗∗∗ (0.06) 0.74∗∗∗ (0.06) 0.74∗∗∗ (0.06) 0.74∗∗∗ (0.06)
Somewhat to the left 0.40∗∗∗ (0.04) 0.40∗∗∗ (0.04) 0.40∗∗∗ (0.04) 0.40∗∗∗ (0.04)
Somewhat to the right -0.16∗∗∗ (0.04) -0.16∗∗∗ (0.04) -0.16∗∗∗ (0.04) -0.16∗∗∗ (0.04)
Clearly to the right -0.79∗∗∗ (0.06) -0.79∗∗∗ (0.06) -0.79∗∗∗ (0.06) -0.79∗∗∗ (0.06)
Female Dummy 0.10∗∗∗ (0.03) 0.10∗∗∗ (0.03) 0.10∗∗∗ (0.03) 0.10∗∗∗ (0.03)
Age -0.01∗∗ (0.01) -0.01∗∗ (0.01) -0.01∗∗ (0.01) -0.01∗∗ (0.01)
Age Squared 0.00∗ (0.00) 0.00∗ (0.00) 0.00∗ (0.00) 0.00∗ (0.00)
Household income level
Very low 0.03 (0.05) 0.03 (0.05) 0.03 (0.05) 0.03 (0.05)
Low -0.05 (0.06) -0.05 (0.06) -0.05 (0.06) -0.05 (0.06)
High 0.03 (0.05) 0.03 (0.05) 0.03 (0.05) 0.03 (0.05)
Very high 0.28∗∗∗ (0.05) 0.28∗∗∗ (0.05) 0.28∗∗∗ (0.05) 0.28∗∗∗ (0.05)
Education Level
Medium 0.21∗∗∗ (0.05) 0.21∗∗∗ (0.05) 0.21∗∗∗ (0.05) 0.21∗∗∗ (0.05)
High 0.80∗∗∗ (0.05) 0.80∗∗∗ (0.05) 0.80∗∗∗ (0.05) 0.80∗∗∗ (0.05)
Year of survey -0.01 (0.01) -0.01 (0.01) -0.01 (0.01) -0.01 (0.01)
Observations 13,822 13,822 13,822 13,822
Pseudo R2 0.092 0.092 0.092 0.092
Standard errors in parentheses; ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
28
Figure 3: Average Marginal Effects of Gasoline Pump Prices
1, ”Very bad proposal”, by approximately 1.6%. An increase in price will increase the
probability by approx. 0.6% of selecting outcome 2 (bad proposal). So an increase in pump
prices will yield an increase in the probability of people rejecting the proposal to increase
the taxes. On the same line, we observe that the increase in prices reduces the probability
of people supporting or being neutral to the proposal to increase the tax. As seen in figure
3, the trend suggests that an increase in price will yield a more negative attitude towards
the proposal. One can suppose that people will either reduce their support for a carbon tax
increase, stop supporting it or deepen their disapproval of the proposal. One interesting thing
to notice is that the largest effect of all from an increase in gasoline prices will be observed in
outcome 1, which is to strongly disapprove the proposal. Another interesting result is that
the largest negative impact is observed in outcome 4. In other words, making it the outcome
with the largest reduction of probability of occurring.
To summarize, the results suggest that pump prices of gasoline indeed have a negative im-
pact on people’s perception of taxes independent of their characteristics, such as consumption
29
patterns like driving a car frequently and attitude towards the environment. As proposed in
hypothesis 1, there is empirical evidence obtained from this analysis that points towards the
direct presence of prices in the individual’s utility functions, as formulated in equation (1).
The other regressions shown in table 11 in Appendix B yield similar results, which supports
our chosen methodology. The results obtained for the interaction terms were insignificant
and therefore we have no evidence to support hypotheses 2 and 3, suggesting that gasoline
price sensitivity towards carbon tax attitudes may not differ between people being more or
less climate concerned, as well as among those being more or less car dependent..
The results of the analysis follow the findings from prior studies (Ewald, Sterner, and
Sterner, 2022; Fairbrother, Sevä, and Kulin, 2019; Bergquist et al., 2022; Umit and Schaffer,
2020). People using a car more frequently are more likely to disagree with an increase in
the tax according to the obtained results signaling self-interest, in contrast to the findings
in Kallbekken and Sælen, 2011. Self-interest effects can also be found when looking at
area of living, as people in larger cities tend to support the tax increase to a higher degree
than those living in rural areas. This could reflect on the level of substitutability of the
individual mentioned in the Theoretical Framework, where people in population-dense areas
have more alternative transportation methods available to choose from when prices of gasoline
are high. Ideology is another characteristic that seems to have a clear trend where people
considering themselves on a left-wing scale are more likely to support the tax increase than
right-wing individuals. A higher tendency towards acceptance of the tax proposal can also
be found among people with higher environmental concern, higher education, and females.
No significant results regarding income groups can be found except for the top-level income
group.
30
6 Conclusion
The issue of carbon emissions is growing, while tools made to mitigate emissions such as the
Pigouvian taxes face continuous resistance. This study contributes to understanding what
factors influence support for a carbon tax and thus, helping better understand aspects that
could cause opposition. Most importantly, this study adds to the basket of empirical findings
that changes in the pump price of gasoline could influence attitudes towards the tax, even
though the tax itself has remained constant. From earlier studies, fairness was highlighted as
a key influencing determinant of people’s environmental tax attitudes, whereas high pump
prices could potentially explain part of it. Although our findings indicate that gasoline prices
affect carbon tax attitudes, a more detailed analysis, or perhaps the use of other survey
questions, would be needed to study whether people in Sweden believe the unfairness could
also derive from already high gasoline prices. For example, further studies could analyze how
prices shape attitudes among different income groups. However, better data on income would
be necessary as the current survey offered income data based on levels and uses household
income rather than individual income, failing to accurately capture the individual’s income.
Nonetheless, the results obtained were insignificant for 4 out of 5 income groups, where only
the highest income level had significant results.
From our theoretical model, we stated that an increase in prices would cause a negative
individual utility derived from the tax, leading to our hypothesis that a price increase of
gasoline would decrease support for a carbon tax increase. Our empirical findings point
towards gasoline pump price increases having a negative effect on public acceptance of carbon
taxes, and with the recent surges in prices, creating a suitable environment for political parties
to gain votes by proposing environmental tax reductions. Not only does our empirical study
provide an interesting perspective to this field of research, but we also contribute with a
theoretical model aiming to explain the relevance of observing end-consumer prices when
investigating what affects carbon tax attitudes.
31
Sweden is said to have succeeded in implementing environmental taxes on petrol by doing
so over a long period, allowing individuals to adapt to the new levels by experiencing small
price changes. However, our findings pose an interesting question: have people become
accustomed to the tax levels or the gross price of gasoline itself? During our period of
interest, no major price fluctuations occurred. However, the carbon tax also changed very
little during this period, implying the tax could be used as a scapegoat for other factors
increasing the price. Perhaps imposing a more flexible tax scheme would allow adapting to
exogenous price shocks, thus creating a more even price level of gasoline, which could possibly
help obtain higher support for the tax.
This study also aimed to observe whether price changes affect carbon tax attitudes dif-
ferently across different population groups. When testing this between people with different
car habits, as well as between people expressing different levels of climate concern, we find
no significant results. From this, we cannot distinguish people’s carbon tax attitudes to be
more or less sensitive toward a price change. On the contrary, our results suggest a unison
defiance of the tax once prices go up. Even so, there is a clear relationship between car-
dependent people expressing less support for the tax, indicating further proof of self-interest.
Also in line with prior studies, people with a high climate concern have a higher support
for the carbon tax on gasoline. This relates to the findings of Löfgren and Nordblom, 2009
where awareness of the problem could lead to higher support. People in larger cities often
have multiple transportation options, whereas people in the countryside rely on fewer, and
sometimes only their car.
Nonetheless, the period under study is not very large and has small price fluctuations
of gasoline, thus not being able to capture larger volatility in prices (as shown in figure
1). As mentioned earlier, there is a potential for further study, making a similar analysis
including the large spike in fuel prices of recent years, since the survey data available was
until 2020. It is even more interesting to study this event, given that the price shock of fuel
32
is derived from a completely external shock, given the invasion of Ukraine. With these large
fluctuations in price, the results might change, and could potentially support hypotheses 2 and
3, differentiating the effects of the price of gasoline among different population groups. Not
only that, it would be interesting to analyze how people’s attitudes toward the environment
and environmental taxes are affected by such price shocks.
While our study produces interesting findings, our data allows only to investigate to what
extent people support an increase in carbon taxes. By adding the questions about how people
view the carbon tax on gasoline more generally, one could investigate whether price changes
could cause opposition, which this study cannot accurately explain.
Another interesting gap remaining is to include the effects of governmental trust in the
analysis. Unfortunately, the survey data available only included questions on trust for very
few years of the analyzed data, thus sacrificing a significant amount of information. Given
the small price fluctuations in this period, we decided to keep a larger amount of observations
than to focus on the effects of trust. Nevertheless, the results obtained for the ordered logit
estimations including trust in the government produced the same results as the ones obtained
without it, and similarly to prior studies trust in the government had a positive effect on
carbon tax attitudes. On this note, checking whether price sensitivity toward carbon tax
attitudes among people with different levels of government trust could be an interesting
continuation of this study.
This study focuses on environmental taxes on gasoline which as of today is an extremely
important field of research to better understand how efficient policy tools can be implemented
to reduce carbon emissions. Looking ahead toward collective goals such as the Net Zero,
Green Deal 2030, etc., this field of research, more specifically related to environmental taxes,
could potentially lose its importance. This is, of course, if alternative fuels can be successfully
implemented on a large scale.
33
References
Bergquist, Magnus et al. (Mar. 2022). “Meta-analyses of fifteen determinants of public opinion
about climate change taxes and laws”. In: Nature Climate Change 12 (3), pp. 235–240.
issn: 17586798. doi: 10.1038/s41558-022-01297-6.
Coglianese, John et al. (2017). “Anticipation, Tax Avoidance, and the Price Elasticity of
Gasoline Demand”. In: Journal of Applied Econometrics 32.1, pp. 1–15. doi: https:
//doi.org/10.1002/jae.2500. eprint: https://onlinelibrary.wiley.com/doi/pdf/
10.1002/jae.2500. url: https://onlinelibrary.wiley.com/doi/abs/10.1002/jae.
2500.
Dahl, Carol A. (2012). “Measuring global gasoline and diesel price and income elasticities”.
In: Energy Policy 41. Modeling Transport (Energy) Demand and Policies, pp. 2–13. issn:
0301-4215. doi: https://doi.org/10.1016/j.enpol.2010.11.055. url: https:
//www.sciencedirect.com/science/article/pii/S0301421510008797.
Douenne, Thomas and Adrien Fabre (Feb. 2022). “Yellow Vests, Pessimistic Beliefs, and
Carbon Tax Aversion”. In: American Economic Journal: Economic Policy 14 (1), pp. 81–
110. issn: 1945774X. doi: 10.1257/pol.20200092.
European Commission (2023).Weekly Oil Bulletin - Price Developments. Directorate-General
for Energy. https://energy.ec.europa.eu/data- and- analysis/weekly- oil-
bulletin_en. Accessed on 2023/02/13.
Ewald, Jens, Thomas Sterner, and Erik Sterner (Nov. 2022). “Understanding the resistance
to carbon taxes: Drivers and barriers among the general public and fuel-tax protesters”.
In: Resource and Energy Economics 70 (1). issn: 09287655. doi: 10.1016/j.reseneeco.
2022.101331.
Fairbrother, Malcolm (May 2022). “Public opinion about climate policies: A review and call
for more studies of what people want”. In: PLOS Climate 1 (5), e0000030. issn: 2767-3200.
doi: 10.1371/journal.pclm.0000030.
Fairbrother, Malcolm, Ingemar Johansson Sevä, and Joakim Kulin (Nov. 2019). “Political
trust and the relationship between climate change beliefs and support for fossil fuel taxes:
34
Evidence from a survey of 23 European countries”. In: Global Environmental Change 59.
issn: 09593780. doi: 10.1016/j.gloenvcha.2019.102003.
Flood, Lennart, Nizamul Islam, and Thomas Sterner (Mar. 2010). “Are demand elasticities
affected by politically determined tax levels? Simultaneous estimates of gasoline demand
and price”. In: Applied Economics Letters 17 (4), pp. 325–328. issn: 13504851. doi:
10.1080/13504850701735864.
Ghalwash, Tarek (Jan. 2007). “Energy taxes as a signaling device: An empirical analysis of
consumer preferences”. In: Energy Policy 35 (1), pp. 29–38. issn: 03014215. doi: 10.
1016/j.enpol.2005.09.011.
Hammar, Henrik and Sverker C. Jagers (2006). “Can trust in politicians explain individuals’
support for climate policy? The case of CO2 tax”. In: Climate Policy 5 (1), pp. 613–625.
issn: 1469-3062. doi: 10.1080/14693062.2006.9685582.
Hammar, Henrik, Sverker C Jagers, and Katarina Nordblom (2008). “Attitudes towards Tax
Levels: A Multi-Tax Comparison”. In: Fiscal Studies 29 (4), pp. 523–543. url: https:
//about.jstor.org/terms.
Jagers, Sverker C., Johan Martinsson, and Simon Matti (Feb. 2019). “The impact of compen-
satory measures on public support for carbon taxation: an experimental study in Sweden”.
In: Climate Policy 19 (2), pp. 147–160. issn: 17527457. doi: 10.1080/14693062.2018.
1470963.
Kallbekken, Steffen and H̊akon Sælen (May 2011). “Public acceptance for environmental
taxes: Self-interest, environmental and distributional concerns”. In: Energy Policy 39 (5),
pp. 2966–2973. issn: 03014215. doi: 10.1016/j.enpol.2011.03.006.
Kaplowitz, Stan A. and Aaron M. McCright (Dec. 2015). “Effects of policy characteristics
and justifications on acceptance of a gasoline tax increase”. In: Energy Policy 87, pp. 370–
381. issn: 03014215. doi: 10.1016/j.enpol.2015.08.037.
Löfgren, Åsa and Katarina Nordblom (2006). The Importance of Habit Formation for Envi-
ronmental Taxation. Working papers in Economics no.204. Gothenburg, Sweden: Depart-
ment of Economics, University of Gothenburg.
35
Löfgren, Åsa and Katarina Nordblom (Dec. 2009). “Puzzling tax attitudes and labels”. In:
Applied Economics Letters 16 (18), pp. 1809–1812. issn: 1350-4851. doi: 10 . 1080 /
13504850701719660.
OECD (2023). Environmentally related tax revenue. OECD. https://stats.oecd.org/
Index.aspx?DataSetCode=ERTR. Accessed on 2023/03/08.
Pigou, Arthur (1920). The Economics of Welfare. Mac Millan.
Povitkina, Marina et al. (Sept. 2021). “Why are carbon taxes unfair? Disentangling public
perceptions of fairness”. In: Global Environmental Change 70. issn: 09593780. doi: 10.
1016/j.gloenvcha.2021.102356.
Putler, Daniel S. (1992). “Incorporating Reference Price Effects into a Theory of Consumer
Choice”. In: Marketing Science 11.3, pp. 287–309. issn: 07322399, 1526548X. url: http:
//www.jstor.org/stable/183891 (visited on 03/30/2023).
Regeringskansliet (2022). Tillfälligt sänkt skatt p̊a drivmedel och sänkt skatt p̊a bränslen i
viss värmeproduktion. Regeringskansliet. https://www.regeringen.se/rattsliga-
dokument/proposition/2022/11/prop.-20222317. Accessed on 2023/03/21.
Ritchie, Hannah (2022). Annual CO2 emissions. Our World in Data. https : / /
ourworldindata . org / grapher / annual - co2 - emissions - per - country ? country =
~OWID_WRL. Accessed on 2023/04/26.
Rotaris, Lucia and Romeo Danielis (Feb. 2019). “The willingness to pay for a carbon tax in
Italy”. In: Transportation Research Part D: Transport and Environment 67, pp. 659–673.
issn: 13619209. doi: 10.1016/j.trd.2019.01.001.
Skatteverket (2023). Historik Skattesatser. Skatteverket. https : / / skatteverket . se /
foretag / skatterochavdrag / punktskatter / energiskatter / skattpabransle . 4 .
15532c7b1442f256bae5e56.html. Accessed on 2023/02/20.
Sterner, Thomas and Gunnar Köhlin (2003). “Environmental Taxes in Europe”. In: Public
Finance and Management 3 (1), pp. 117–142.
36
Trafikanalys (2021). Fordon 2020. Sveriges Officiella Statistik. https://www.trafa.se/
globalassets/statistik/vagtrafik/fordon/2021/fordon_2020.pdf. Accessed on
2023/05/29.
Umit, Resul and Lena Maria Schaffer (May 2020). “Attitudes towards carbon taxes across
Europe: The role of perceived uncertainty and self-interest”. In: Energy Policy 140. issn:
03014215. doi: 10.1016/j.enpol.2020.111385.
University of Gothenburg, SOM Institute (2022). The National SOM Survey Cumulative
Dataset 1986-2020. University of Gothenburg.
37
Appendices
A Theoretical Model
A.1 Utility Function
U(cF , cA, X, L) (8)
With: ∑H
X = γ cF,h
h=1
Y = p̂cF + cA
p̂ = pF + τ
L = RP − p̂
A.2 Consumer Problem
The consumer maximization problem the individual faces is the following
max U(cF , cA)
cF ,cA (9)
s.t. Y = p̂cF + cA
From the FOC we obtain:
∂U ∂U
= p̂ (10)
∂cF ∂cA
A.3 Assumptions and Implications
∂U ∂2U ∂U ∂2U
> 0; < 0; > 0; < 0
∂c ∂c2F F ∂cA ∂c
2
A (11)
∂U 2≤ ∂ U ≤ ∂U ≥ ∂
2U
0; 0; 0; ≥ 0
∂X ∂X2 ∂L ∂L2
38
From the previous assumptions we can obtain the following conditions given our budget
constraint we can solve for cA:
cA = Y − p̂cF+
Taking the derivative of cA in terms of p̂ yields the following equation:
[ ]
∂cA − ∂cF= cF + p̂ (12)
∂p̂ ∂p̂
We can see that there is an ambiguous effect in the previous equation, thus, to check what
part dominates we divide everything by cF and obtain:
−1− ηc (13)F
Where ηc is the gasoline price elasticity of demand. The first thing we notice is thatF
this expression is negative if and only if gasoline is inelastic, meaning that −1 < ηc ≤ 0.F
Therefore:
∂cA
< 0 ⇔ −1 < ηc ≤ 0 (14)
∂p̂ F
A.4 Agreement on tax
When the individual solves for his optimal tax rate, she/he solves:
∂U
(15)
∂τF
The utility function U can be formulated from the original as follows:
U(cF , cA, X, L) = U(cF , Y − p̂cF , X, L) (16)
From the previous equation we obtain the following equation:
[ ]
H
∂U ∂U ∂cF ∂p̂ ∂U ∂cA ∂p̂ ∂U ∑ ∂cF,h ∂p̂ ∂U ∂L ∂p̂
= + + γF + (17)
∂τF ∂cF ∂p̂ ∂τF ∂cA ∂p̂ ∂τF ∂X ∂p̂ ∂τF ∂L ∂p̂ ∂τF
h=1
39
Note that ∂p̂ = 1 & that ∂L = −1, thus equation (17) can be written as:
∂τF ∂p̂ [ ]
H
∂U ∂U ∂cF ∂U ∂cA ∂U ∑ ∂cF,h ∂U
= + + γF − (18)
∂τF ∂cF ∂p̂ ∂cA ∂p̂ ∂X ∂p̂ ∂L
h=1
Which can be written as follows by substituting the results from the FOC (equation (10)):
[ ]
H
∂U − ∂U ∂U
∑ ∂cF,h ∂U
= cF + γF − (19)
∂τF ∂cA ∂X ∂p̂ ∂L
h=1
A.5 ∣ Differentiation w.r.t. Gas[oline Pum]p Pric∣ 2 [
es ]
∂ ∂U 2 H H
∂τ ∣∣ −∂ U ∂cA − ∂U ∂cF ∂2U ∑ ∂cF,h ∂U ∑ ∂2c 2F F,h ∂ U= cF + γ2 2 F + γF − (20)∂p̂ τ ∂cA ∂p̂ ∂cA ∂p̂ ∂X ∂p̂ ∂X ∂p̂2 ∂L2F h=1 h=1
A.5.1 First term
The first term of equation (20) is expression (21):
∂2− U ∂cA ∂U ∂cFcF − (21)
∂c2A ∂p̂ ∂cA ∂p̂
There is an ambiguous effect if we assume the market for gasoline is inelastic, where the
first part is negative and the second part positive.
2
−∂ U ∂cA − ∂U ∂c
2
F ⇔ −∂ U ∂cA ∂U ∂cFcF < 0 cF < (22)
∂c2 ∂p̂ ∂c ∂p̂ ∂c2A A A ∂p̂ ∂cA ∂p̂
Dividing by cF and multiplying by 1 (
p̂):
p̂
2
−∂ U ∂cA ∂Up̂ < ηc (23)
∂c2A ∂p̂ ∂c
F
A
From the previous equation (23) we can see that whenever the price elasticity of gasoline
is sufficiently inelastic, that meaning that it is tending to 0, we will find that expression
(21) will be negative. In the case of a perfectly inelastic demand, it will always be negative.
We can also see that when the price is large enough, ceteris paribus, expression (21) will be
negative.
40
Individuals that have a price elastic demand for gasoline will observe that expression (21)
is positive.
A.5.2 Second Term
The second term from equation (20) is expression (24):
[ ∑ ]2 [ ]2 H H∂ U ∂c ∂U ∑ 2F,h ∂ c 2F,h ∂ U
γF + γ2 F − (24)∂X ∂p̂ ∂X ∂p̂2 ∂L2
h=1 h=1
The first and last elements of expression (24) have a negative effect on equation (20). This
can be easi[ly ch∑ecked from]the model assumptions (11). The second element from expression
(24) is ∂U ∑ ∂2cγ H F,hF h=1 2 and its sign is not so clear. The sign of this terms depends on∂X ∂p̂∂2c
the sign of H ∑ F,hh=1 2 . If the demand function for gasoline is convex across the population∂p̂ ∂2c
will imply that H F,hh=1 2 > 0 thus making expression (24) negative.∂p̂
A.6 Comparative Statics
Recalling equation (20):
[ ]
∂U ∣∣∣ 2 [ ]∂ ∣ ∂2 2 ∑H ∑H 2 2∂τ − U ∂cA − ∂U ∂cF ∂ U ∂cF,h ∂U ∂ cF,h ∂ UF = cF + γF + γF −∂p̂ τ ∂c2A ∂p̂ ∂cA ∂p̂ ∂X2 ∂p̂ ∂X ∂p̂2 ∂L2F h=1 h=1
Individuals who are highly dependent on consumption of gasoline, whom have a very
price inelastic demand for gasoline, will observe a negative result for equation (20) assuming
that the demand function for gasoline is convex across the population. In other words,
these individuals will observe a higher disagreement on an increase in environmental taxes
whenever gasoline prices rise. In general, gasoline is a rather price inelastic market, given
the high dependence regular car users are to driving.
41
B Additional Results
Table 6: O. Logit Avg. Marginal Effects for Proposal: Increase the CO2 tax on petrol
Estimation (1)
Gasoline Pump-Price
1 Very bad proposal 0.0162∗∗∗ (0.00)
2 Rather bad proposal 0.00588∗∗∗ (0.00)
3 Neither good nor bad proposal -0.00393∗∗∗ (0.00)
4 Rather good proposal -0.0102∗∗∗ (0.00)
5 Very good proposal -0.00803∗∗∗ (0.00)
Drives Car Frequently
1 Very bad proposal 0.112∗∗∗ (0.01)
2 Rather bad proposal 0.0405∗∗∗ (0.00)
3 Neither good nor bad proposal -0.0271∗∗∗ (0.00)
4 Rather good proposal -0.0699∗∗∗ (0.00)
5 Very good proposal -0.0553∗∗∗ (0.00)
Climate Concern
Some concern
1 Very bad proposal -0.145∗∗∗ (0.01)
2 Rather bad proposal -0.0171∗∗∗ (0.00)
3 Neither good nor bad proposal 0.0638∗∗∗ (0.00)
4 Rather good proposal 0.0665∗∗∗ (0.00)
5 Very good proposal 0.0318∗∗∗ (0.00)
Major concern
1 Very bad proposal -0.249∗∗∗ (0.01)
2 Rather bad proposal -0.0755∗∗∗ (0.00)
3 Neither good nor bad proposal 0.0880∗∗∗ (0.00)
4 Rather good proposal 0.147∗∗∗ (0.00)
5 Very good proposal 0.0901∗∗∗ (0.00)
Marginal effects; Standard errors in parentheses; ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
42
Table 7: O. Logit Avg. Marginal Effects for Proposal: Increase the CO2 tax on petrol cont.
Estimation (1)
Ideology
Clearly to the left
1 Very bad proposal -0.0933∗∗∗ (0.01)
2 Rather bad proposal -0.0544∗∗∗ (0.01)
3 Neither good nor bad proposal 0.0100∗∗∗ (0.00)
4 Rather good proposal 0.0724∗∗∗ (0.01)
5 Very good proposal 0.0653∗∗∗ (0.01)
Somewhat to the left
1 Very bad proposal -0.0547∗∗∗ (0.01)
2 Rather bad proposal -0.0269∗∗∗ (0.00)
3 Neither good nor bad proposal 0.0109∗∗∗ (0.00)
4 Rather good proposal 0.0397∗∗∗ (0.00)
5 Very good proposal 0.0310∗∗∗ (0.00)
Somewhat to the right
1 Very bad proposal 0.0249∗∗∗ (0.01)
2 Rather bad proposal 0.00852∗∗∗ (0.00)
3 Neither good nor bad proposal -0.00770∗∗∗ (0.00)
4 Rather good proposal -0.0156∗∗∗ (0.00)
5 Very good proposal -0.0101∗∗∗ (0.00)
Clearly to the right
1 Very bad proposal 0.138∗∗∗ (0.01)
2 Rather bad proposal 0.0252∗∗∗ (0.00)
3 Neither good nor bad proposal -0.0526∗∗∗ (0.00)
4 Rather good proposal -0.0714∗∗∗ (0.01)
5 Very good proposal -0.0395∗∗∗ (0.00)
Marginal effects; Standard errors in parentheses; ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
43
Table 8: O. Logit Avg. Marginal Effects for Proposal: Increase the CO2 tax on petrol cont.
Estimation (1)
Rural Residential Area Dummy
1 Very bad proposal 0.0614∗∗∗ (0.00)
2 Rather bad proposal 0.0222∗∗∗ (0.00)
3 Neither good nor bad proposal -0.0149∗∗∗ (0.00)
4 Rather good proposal -0.0384∗∗∗ (0.00)
5 Very good proposal -0.0304∗∗∗ (0.00)
Female Dummy
1 Very bad proposal -0.0151∗∗∗ (0.00)
2 Rather bad proposal -0.00545∗∗∗ (0.00)
3 Neither good nor bad proposal 0.00365∗∗∗ (0.00)
4 Rather good proposal 0.00942∗∗∗ (0.00)
5 Very good proposal 0.00745∗∗∗ (0.00)
Age
1 Very bad proposal 0.00167∗∗ (0.00)
2 Rather bad proposal 0.000606∗∗ (0.00)
3 Neither good nor bad proposal -0.000405∗∗ (0.00)
4 Rather good proposal -0.00105∗∗ (0.00)
5 Very good proposal -0.000828∗∗ (0.00)
Age Squared
1 Very bad proposal -0.0000157∗ (0.00)
2 Rather bad proposal -0.00000569∗ (0.00)
3 Neither good nor bad proposal 0.00000381∗ (0.00)
4 Rather good proposal 0.00000984∗ (0.00)
5 Very good proposal 0.00000778∗ (0.00)
Marginal effects; Standard errors in parentheses; ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
44
Table 9: O. Logit Avg. Marginal Effects for Proposal: Increase the CO2 tax on petrol cont.
Estimation (1)
Household Income Level
Very Low
1 Very bad proposal -0.00522 (0.01)
2 Rather bad proposal -0.00175 (0.00)
3 Neither good nor bad proposal 0.00138 (0.00)
4 Rather good proposal 0.00318 (0.01)
5 Very good proposal 0.00240 (0.00)
Low
1 Very bad proposal 0.00793 (0.01)
2 Rather bad proposal 0.00249 (0.00)
3 Neither good nor bad proposal -0.00221 (0.00)
4 Rather good proposal -0.00474 (0.01)
5 Very good proposal -0.00348 (0.00)
High
1 Very bad proposal -0.00424 (0.01)
2 Rather bad proposal -0.00141 (0.00)
3 Neither good nor bad proposal 0.00113 (0.00)
4 Rather good proposal 0.00258 (0.00)
5 Very good proposal 0.00194 (0.00)
Very High
1 Very bad proposal -0.0408∗∗∗ (0.01)
2 Rather bad proposal -0.0161∗∗∗ (0.00)
3 Neither good nor bad proposal 0.00904∗∗∗ (0.00)
4 Rather good proposal 0.0263∗∗∗ (0.00)
5 Very good proposal 0.0215∗∗∗ (0.00)
Marginal effects; Standard errors in parentheses; ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
45
Table 10: O. Logit Avg. Marginal Effects for Proposal: Increase the CO2 tax on petrol cont.
Estimation (1)
Education Level
Medium
1 Very bad proposal -0.0361∗∗∗ (0.01)
2 Rather bad proposal -0.00847∗∗∗ (0.00)
3 Neither good nor bad proposal 0.0129∗∗∗ (0.00)
4 Rather good proposal 0.0199∗∗∗ (0.00)
5 Very good proposal 0.0118∗∗∗ (0.00)
High
1 Very bad proposal -0.119∗∗∗ (0.01)
2 Rather bad proposal -0.0447∗∗∗ (0.00)
3 Neither good nor bad proposal 0.0321∗∗∗ (0.00)
4 Rather good proposal 0.0766∗∗∗ (0.01)
5 Very good proposal 0.0553∗∗∗ (0.00)
Year of survey
1 Very bad proposal 0.00120 (0.00)
2 Rather bad proposal 0.000434 (0.00)
3 Neither good nor bad proposal -0.000290 (0.00)
4 Rather good proposal -0.000750 (0.00)
5 Very good proposal -0.000593 (0.00)
Marginal effects; Standard errors in parentheses; ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
46
Table 11: Additional Estimations for Proposal: Increase the CO2 tax on petrol
O. Probit Logit LPM
Gasoline Pump-Price -0.06∗∗∗ (0.01) -0.12∗∗∗ (0.03) -0.02∗∗∗ (0.00)
Daily Car Use Dummy -0.45∗∗∗ (0.02) -0.74∗∗∗ (0.05) -0.13∗∗∗ (0.01)
Climate Concern
Some concern 0.45∗∗∗ (0.03) 0.73∗∗∗ (0.08) 0.07∗∗∗ (0.01)
Major concern 0.90∗∗∗ (0.03) 1.75∗∗∗ (0.08) 0.25∗∗∗ (0.01)
Rural Area Dummy -0.25∗∗∗ (0.02) -0.40∗∗∗ (0.05) -0.06∗∗∗ (0.01)
Ideology
Clearly to the left 0.42∗∗∗ (0.03) 0.98∗∗∗ (0.07) 0.19∗∗∗ (0.01)
Somewhat to the left 0.23∗∗∗ (0.03) 0.56∗∗∗ (0.06) 0.10∗∗∗ (0.01)
Somewhat to the right -0.10∗∗∗ (0.02) -0.08 (0.06) -0.01 (0.01)
Clearly to the right -0.45∗∗∗ (0.03) -0.58∗∗∗ (0.09) -0.07∗∗∗ (0.01)
Female Dummy 0.05∗∗∗ (0.02) -0.03 (0.04) -0.01 (0.01)
Age -0.01∗ (0.00) -0.01 (0.01) -0.00 (0.00)
Age Squared 0.00∗ (0.00) 0.00 (0.00) 0.00 (0.00)
Household Income Level
Very low 0.02 (0.03) 0.01 (0.07) -0.00 (0.01)
Low -0.03 (0.03) 0.01 (0.08) -0.00 (0.01)
High 0.01 (0.03) -0.00 (0.07) -0.00 (0.01)
Very high 0.16∗∗∗ (0.03) 0.31∗∗∗ (0.06) 0.05∗∗∗ (0.01)
Education Level
Medium 0.13∗∗∗ (0.03) 0.21∗∗∗ (0.08) 0.03∗∗∗ (0.01)
High 0.47∗∗∗ (0.03) 0.90∗∗∗ (0.08) 0.14∗∗∗ (0.01)
Year of survey -0.00 (0.00) 0.01 (0.01) 0.00 (0.00)
Constant -13.07 (14.12) -3.26 (2.27)
Observations 13,822 13,822 13,822
Adjusted R2 0.187
Pseudo R2 0.092 0.173
Standard errors in parentheses; ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
Binary outcome used for Logit and LPM estimations is created using original outcome variable
where 1 is Very good proposal & Rather good proposal, otherwise 0
Note that O. Probit and logit estimation results are not marginal effects
47