Public acceptance and the German energy transition - An Experimentalstudy on distributional preferences

Ole Kutzschbauch∗, Stefan Traub

Chair of Behavioral Economics,Helmut-Schmidt University Hamburg, Germany

August, 2015

Abstract

The German energy transition, the so called Energiewende, is one of the most ambitiousprojects in Germany’s recent history. Without public support it is most likely that thechallenging goals of the energy transition will not sustain the next decades. As Frondelet al. (2015) point out, a financial support for renewables is crucial to increase their shareto the total energy supply. Since the costs for the transition will increase, the share ofenergy consumption in the expenses of the households will rise. It is clear that this higherburden will lead to a more heated political debate. This paper concentrates on the distribu-tional issues of the German energy transition and its impact on the public acceptance. Weexperimentally investigate how the costs of the energy transition should be divided amongthe society. In particular, we show that the question whether the costs can be determinedor not changes the basis of the decision-making of our respondents.

Keywords: energy transition, distributional preferences, cost uncertainty, justice

preliminary and incomplete:do not quote without permission of the authors

∗Corresponding author. Helmut-Schmidt-University Hamburg, Holstenhofweg 85, D-22043 Hamburg, Germany. Phone: +4940 6541 3194, fax: +49 40 6541 2042, ole.kutzschbauch@hsu-hh.de.

1 Introduction

The success of profound social, economic and technological changes depends heavily on the public perception.

Apart from the debates about the intent and purpose, costs are the focus of interest. Without a social

agreement with respect to the amount and type of financing the costs, those changes are unlikely to reach

their intended goals.

One of the most ambitious projects in Germany’s recent history is the energy transition away from fossil

towards renewable energy sources, the so-called Energiewende. This massive state intervention in the energy

market has far-reaching consequences to society. This paper concentrates on the distributional issues of the

German energy transition and its impact on the public acceptance.1

Today, global warming has been widely accepted as a serious threat for mankind. It has affected energy

and environmental policies throughout the world. At least since the enactment of the Renewable Energy

Act (EEG) in 2000, the German government actively pushed the development towards a more sustainable

energy supply in the German energy market. Highly subsidizing the use of renewable energy sources and

increasing the efficiency of energy consumption in general have since then been the main vehicles of this

turnaround. As Frondel et al. (2015) point out, a financial support for renewables is crucial to increase the

share of green electricity to the total energy supply. Otherwise, green technologies could not compete with

conventional energy sources because of the higher costs in the transition phase. Germany sets the demanded

share of renewables for electricity consumption to at least 50 % in 2030. Although today the total capacities

of renewable energy sources are close to the conventional energy sources, the share of renewables on the

consumption level is still at 25 %. Therefore a challenging path lies ahead. Without a significant increase

of todays capacities of renewable energy sources or a massive technological progress the goals of the energy

transition will not be reached.

The energy costs for the households already increased since the opening of the energy market from 13.94

ct/kWh to 28.81 ct/kWh, where just the EEG surcharge itself increased from 2000 to 2014 from 0.20 ct/kWh

to 6.17 ct/kWh (see Figure 1).

Figure 1: Development of the electricity price and EEG surcharge in Cent per Kilowatthour(kWh) (BDEW, 2015)

If you take a typical household with an energy consumption of around 3000 kWh per year, the costs for these

households increased from 6 Euro for the EEG surcharge and 420 Euro for the electricity in total in the year

2000 to 185 Euro EEG surcharge and 865 Euro for electricity in total per year in 2014. Since the wanted

share of renewables to the energy consumption is by far not reached, it is most likely that the energy costs

will rise further.

1Obviously, the question how the costs should be divided among the society is only one aspect defining public acceptance.

For an overview, see for example Zoellner et al. (2008), Musall and Kuik (2011) or Huijts et al. (2012)

2

Therefore we experimentally investigate how the costs of the energy transition should be divided among

the society. We abstain from modeling efficiency concerns and possible reciprocity to solely focus on the

distributional preferences.

The remainder of this paper is structured as follows: In Section 2 we explain the experimental design. In

Section 3 we present our results. The paper concludes in Section 4 with a short discussion of our findings.

2 The experiment

The experiment was conducted in Bremen, Germany. Our sample consisted of 374 visitors of a well visited

shopping center in the city. By changing our subject pool from standard subjects (regular students) to

common citizens, we conducted a so called artefactual field experiment (Harrison and List, 2004; Charness

et al., 2013)2. We personally addressed and invited the subjects to participate in a scientific study about

justice. In the recruitment phase, we have not addressed the energy transition for two reasons: First, by not

defining the topic, we excluded a potential unwanted preselection for people with a general interest in the

energy transition. Second, we also tested in one scenario if people have different distributional preferences

for a general public project than in case of the energy turnaround. To minimize the risk of bias, we therefore

abstained from the green framing in the recruiting phase.

The participants were told that the study would take around 15 minutes and that they would receive a

guaranteed show-up fee of 5 Euro. In addition, they had the chance to win up to 160 Euro during the course

of the experiment .

After the participants were successfully recruited, our staff members lead them to our mobile lab.3 Before

the subjects started the experiment, they were welcomed and asked to draw a lottery ticket. On each ticket,

there was a number with 4 digits. With the number the respondent was selected to the treatment and

household type during the experiment. The ticket also decided whether or not the subject would win their

gained residual personal income at the end of the experiment. In order to preserve our subjects’ anonymity,

all payoffs were made in cash, right after the experiment was finished.

Before the subjects could start with their experimental task, they were carefully instructed how to use the

notebooks and the computer software4. Staff members ensured comfortable volume levels of the headphones

and enjoyable brightness values for each participant. The experiment was fully computerized. The subjects

were guided through their tasks via audio instructions5 and could always repeat the instructions or ask an

associate for help.

In order to reduce hypothetical bias, we analyzed the individual choice behavior with a strictly neutral cheap

talk script combined with high monetary incentives.6 By ’cheap talk’, we mean that subjects get informed

about the hypothetical character of the experiment, but are asked to behave as realistic as possible. Addi-

tionally, subjects choices have real financial consequences depending on their actual distributional decision

in the experiment.

The potential payoff was divided in two separate parts, the individual payoff and the donation for a charitable

organization. The residual income after deducting the subjects contribution to the energy transition defined

the potential payment for the participant. Every amount that was contributed to the energy transition

reduced the participants payoff and was donated to an charitable organization of the subjects choice. The

odds to win were 1 to 10. All winner tickets were determined by a random mechanism before the session

started. Winners received their individual payoff based on their decisions in the first and second part of the

experiment, which will be described later. The experiment consisted overall of three parts. The subjects had

2(Charness et al., 2013) name them ”extra-laboratory experiments”.3The lab was based on a less frequented place in the mall, so that anonymity could be ensured. The workplace of each

subject consisted of a notebook and one headphone.4The experiment is programmed in Visual Studio 2013.5A professional voice-over actor from the company Voice Archive in Denmark recorded the instructions in German.6see for discussions about the goods and bads about cheap talk scripts, (Cummings and Taylor, 1999; List, 2001) and

(Aadland and Caplan, 2006).

3

to approach the sections in the following order: (i) a test for social preferences, (ii) a voting task for their

preferred distribution of the costs of the energy transition, (iii) a post-experimental questionnaire concerning

the participants personal characteristics and their attitudes towards the energy transition and the climate

change in general.

2.1 Social preferences test

As the distributional choices for the costs of the energy transition by the respondents can heavenly rely on

the type of social preferences, we used the double-price-list technique applied by Balafoutas et al. (2012).7

Here the subjects have to choose between egalitarian and unequal allocations. The participants make ten

binary choices in total. The respondent have to divide a given amount of money units8 between herself, as

the decision maker, and another randomly matched participant, the so called passive player. The 10 choices

are placed in two allocation blocks (see Table 1 for the payoff mechanism).

Table 1: Social preferences test - payoff mechanism

Advantageous(Disadvantageous) Block

Asymmetric distribution: Left Egalitarian distribution: Right

Your Payoff Passive Players’ Payoff Your Payoff Passive Players’ Payoff

e+ x e− g(e+ g) e+ e e+ e

Notes: e = 50, g = 30 and x ∈ {−20,−10, 0, 10, 20}.

In the first, the subject has to make five choices between an egalitarian distribution (e:e) and an asym-

metric distribution (e + x:e − g) ,where x ∈ {−20,−10, 0, 10, 20}. In the second block the egalitarian

distribution remains the same, while the unequal distribution is modeled as follows: e + x:e + g , where

x ∈ {−20,−10, 0, 10, 20}. In both blocks e = 50 and g = 30 remain unchanged. By the switching point of

the decision maker from the equal to the asymmetric distribution, we can attach the specific type of social

preferences to the subject.9

In the first (advantageous inequality) block, a subject who only maximizes their own payoff, switches to the

unequal distribution not before x = 0. If she changes her choice before that, the respondent is willing to

sacrifice her own payoff in order to minimize the income of the passive player and thereby will be defined as

spiteful. If the decision maker is not switching after x = 10 from the egalitarian to the unequal allocation,

she is willing to reduce her own payoff to acquire an equal distribution, this type can be called inequality

averse (IAV).10

In the second (disadvantageous inequality) block, the decision maker has to decide between an efficient or

an egalitarian distribution. The efficient allocation involves a disadvantageous distribution for herself. The

earlier the subject switches from the equal allocation to the asymmetric distribution, the more she is willing

to sacrifice her own payoff to increase efficiency. Therefore we can use her choices to measure the efficiency

preferences of the respondents. As we are only interested in the distributional preferences and exclude in

the later voting choice efficiency concerns, we focus in the second block on the own payoff maximizer. To

maximize her own payoff, the subject will not switch to the equal distribution before x = 0. We define

subjects as egoistic(EGO), if they do not change their choice before x = 0 in the advantageous block to the

unequal distribution and choose the egalitarian allocation in the disadvantageous block if x ≤ 10.

The subjects receive a combined payoff of one of the ten choices as a decision maker and one of the ten

choices as a passive player. To rule out reciprocity, it is ensured that no decision maker will be matched with

7This test is based on the Equality Equivalence Test developed by Kerschbamer (2015).8Throughout our experiment we talk about ”money units” (MU) instead of Euros. The MU/Euros exchange rate of 5 to 1

is well-known from the beginning and communicated repeatedly.9A rational subject will only change once from the equal distribution to the asymmetric allocation and should not redirect

his binary choice.10A more detailed specification to determine the specific type of social preferences can be found in (Balafoutas et al., 2012)

4

passive players in reverse order.

2.2 Distributional choice

The main part of the experiment is designed to obtain subjects’ distributional preferences regarding the costs

of the German energy transition within a heterogeneous society. Subjects choose their favorite distribution

for the society out of the position of one income group. As we point out that the public acceptance of the

energy transition heavily relies on the individual burden, we are interested in the self-centered fairness view

of the participants and exclude any ”impartial observer” framing in our experiment . In our experiment we

introduce the subjects to a fictional society and inform the participants about the different household types

with their heterogeneous incomes. The household type is denoted by i ∈ {A,B,C}. Each household type

receives a fixed endowment xi ∈ {A = 200, B = 375, C = 625}, which remains the same over all treatments.

It is important to point out, that only the individual endowment defines the heterogeneity of our society.

Due to time restrictions in an artefactual field experiment, we abstain from addressing further parameters,

e.g., effort. To induce effort, a performance test would have been needed to introduce a second reasonable

parameter for the distributional choice. Another defining parameter for heterogeneity would have lead to

more treatments in order control for the specific effects of each parameter. Therefore we postponed those

questions to future experiments.

In the scenarios with the energy transition framing, the subjects are briefly informed about the German

energy transition and that our society wants to follow suit.11 In the neutral project scenario we speak

only about a general public project that has to be financed. There is no further description. After the

introduction the subjects have to ”vote” for their favorite distribution regarding the costs of the energy

transition, respectively the public project.

Figure 2: Sample screen of voting on the distribution of the costs of the energy transiton.

The ”voting” task is constructed as follows (see Figure 2):12 On the left box of the screen, the different

household types i ∈ {A,B,C} with their individual endowment (budget) xi are displayed. The household of

the subject i is indicated by the color green. In the first box right from the budget, subjects can see the con-

sequences of their distributional choice ti. The contribution to the energy transition (labeled Energiewende)

of each household and the resulting payoff ci for each income group are displayed in the payoff box. On

the right-hand side the effect of the current individual distributional choice is illustrated in a histogram.

10The Question if ”fairness lies in the eye of the beholder?” is discussed for example in (Konow, 2009) or (Babcock and

Loewenstein, 1997).11The participants are briefed about the general goal to shift from fossil to renewable energy sources and the more efficient

use of energy. We want to know subjects’ distributional preferences which apply to the overall picture of the energy transition

and not certain aspects of it. Therefore a detailed description of the political strategies to achieve those targets is not given.12In the further course, we describe the voting mechanism on the basis of the energy transition scenarios. The setting of the

neutral project scenario is the same as in the former scenarios with the exception that a public project substitutes the energy

transition as the project to be financed.

5

The endowment of the other households xj 6=i are represented by the gray bars, the budget of the subject is

indicated in green. The particular share of the other households tj 6=i to the costs of the energy transition

is displayed by the color red, while the participants contribution ti is marked by a lighter coloring. Above

the histogram the exogenous given costs of the energy transition G are shown. Below the histogram a slide

bar is displayed. The position of the slide bar decides on the respondents favorite distribution regarding the

costs of the energy transition.

We investigate the distributional choices of our subjects in five different scenarios. Our treatment structure

is constructed as follows (see Table 2):

Table 2: Treatment overview

Scenario

energy

transition

frame

social pro-

tection

mechanism

public costs

certain

income

group

certain

Baseline(BAS) X - X X

Minimum Requirement(MR) X X X X

Costs Uncertainty(CU) X - - X

Veil of Ignorance(VOI) X - X -

Neutral Project(NP) - - X X

T1) The Baseline scenario is already described above. Each subject votes out of the position of one

household group for his favorite distribution. The voting task has no further institutional restrictions.

T2) In the Minimum Requirement scenario we introduce an instrument to ensure a certain endowment

after the voting process for the household with the lowest income xmin. In our setting, we call this

institutional instrument a social protection mechanism. In this case, the xmin-household has a secure

income of 75 MU. Therefore the xmin-household can only by requested up to a amount of 125 MU.13

The difference has to be carried by the other households. All subjects are informed if their vote activates

the social protection mechanism.

T3) In the Cost Uncertainty scenario the value of the costs of the energy transition cannot be specified.

G lies between 400 and 600MU . The decision changes in the following way: Instead of knowing the

absolute consequences of the individual vote, the subjects only know the percentage share of each

household of the costs of the energy transition. The changed decision-making situation is displayed in

an altered screen. Instead of a histogram, the consequences of the subjects’ vote is pictured by a pie

chart.

T4) In the Veil Of Ignorance scenario the subjects are not assigned to one household group before the voting

process is over. The participant should now reveal her ”true” distributional preferences, without the

self-interested bias caused by the knowledge of her position in the society.14

T5) In our last treatment, we test in the neutral project scenario if the energy transition needs to be seen

different than other public projects. We analyze this question in relation to the distribution of the costs

for those projects. Therefore we simply talk about a common public project with no further description

and exclude any energy transition framing.

The consequences of the individual voting is subject to restrictions of the following distribution key ti:

ti = (1− τ)× GN + τ × (xi−xmin)∑N

j=1(xj−xmin)×G

1362,5 % of the low income households endowment.14see Rawls (2009); Harsanyi et al. (1953).

6

where ti is the individual contribution to the energy transition, τ ∈ [0, 1] is the distributional parameter. τ

defines if the resulting distribution will be regressive, proportional or progressive. N is the total number of

household types in our society. xj is the individual endowment of each household type including the endow-

ment of the participant. xmin is defined as the lowest endowment within the society xmin = min {xi, ..., xn}.G are the total costs for the energy transition.15

The selection of τ leads to the following consequences for the individual contribution of each household type:

Table 3: Consequences of selected τ

household type i ∈ A,B,CA B C

τ ∈ [0, 1] 0 1 0 1 0 1

ti ↗ ↘ ↗ ↘ ↘ ↗Notes: ↗ = higher individual contribution ti,

↘ = lower individual contribution ti.

You can see that the household types A and B have to pay a higher amount the more τ moves towards 0.

In contrast, the individual contribution ti of household type C is higher the more τ moves towards 1. The

voting process stays the same in all five scenarios. After the individual vote on their favored distribution

scheme is finished, subjects are informed about the public decision for the chosen distribution of the costs of

the energy transition.

The public decision arises as follows: One vote per household type is selected for the poll. The desired

distribution of the voter in the median position then determines the public division of the costs.16 The

individual payoff is then calculated as the respective income after the deducting the individual share of the

costs of the energy transition. The contribution to the energy transition thus defines the amount that could

be spent to a charitable organization of the subjects’ choice.

2.3 Postexperimental questionnaire

After voting on their favorite distribution, the participants are asked to express their attitudes and opinions

towards the energy turnaround and the climate change in general. The questionnaire starts after the voting

task in order to prevent potential motivational bias in the energy transition framed scenarios and not to

distort the neutral project scenario.

Before we start with the preferences for the energy transition, e.g., climate change, we ask the participants

about the current share of renewables to the total energy supply in Germany.17 The general attitude towards

the climate change is asked in four cases with possible answers shown in the brackets: (i) On what basis

should global emission rights be distributed? (equal distribution, compensation of historical emissions or

grandfathering).(ii) What are we supposed to do in the case of the climate change? (prevent or adapt to the

consequences).(iii) Who is regarded responsible financing the climate change? (government, the economy or

both). (iv) Finally, we ask the respondent if she tries to reduce her energy consumption and if so for what

reason (no; yes, because of financial issues; yes, because of environmental reasons).

The general position towards the energy transition is analyzed on the basis of three cases: (i) Subjects have

15During the experiment, G is hold constant at the level of 600 MU , except for the Veil Of Ignorance scenario.16The subjects were randomly assigned to one of the five workstations. The median voters decision is separately computed

for each treatment from previous subjects’ τ that is already saved on the workstation. If the subject is the first participant on

one of the workstations, the poll is conducted by computer-simulated additional votes of the other household types.17Up to the time of the experiment, the current figures for 2013 were 24.1 percent of the total energy supply in Germany.

The latest (estimated) figures for 2014 are 26.1 percent share of renewables on the total energy supply (AG Energiebilanzen

e.V., 2014)

7

to choose the topic they associate the most with the term ”Energiewende”.18(ii) The participants are asked

how much Germany should do against the consequences of the climate change in global context (more than

the other countries, the same as the others, less than the others).(iii) The respondents have to state their

opinion if higher costs are acceptable to achieve the goals of the German energy transition (totally agree,

somewhat agree, rather disagree, totally disagree). The last field of our interest focuses on the cost aspect of

the energy transition: (i) The participants have to state their position towards the current financial burden

of the energy transition (too high, at the right level, too low). (ii) Then we ask: which principle should define

the distribution of the costs? (costs-by-cause principle, profit-taker principle, equal distribution among all).

(iii) The last question is one specific towards the EEG surcharge in Germany. We ask the subjects about

their position towards the fact that energy intensive companies are freed from the additional costs of the

energy transition (totally agree, somewhat agree, rather disagree, totally disagree).

Furthermore, we collect the participants socio-demographic data. Besides age, education, gender, domestic

circumstances, income, profession, religion and a self-evaluation for risk-preferences, we also ask if the subjects

already have profited by supportive measures in case of energy saving (such as thermal protection or energy

consulting). All answers are entered into the computer by selected items (based on likert scales) from

predefined lists. If subjects do not or can not answer one question they always have the possibility to choose

”no respond” or ”i don’t know”. After the questionnaire is finished, the lottery starts and the participants

are informed whether they have won or not.

3 Results

Figure 3 pictures the median value for the selected τ on the collective and the individual level.19 Underneath

the figure, table 4 provides the data basis for the graph. Next to the median and its range of deviation, the

table also reports the mean values and standard deviation.

Figure 3: Selected τ on collective (left figure) and individual (right figure) level by treatment.

18The possible answers are: Reduction of greenhouse gas emissions; increase the percentage of renewable energy; achieve

greater energy efficiency; nuclear phase-out.19Because of the independence of the every observation for each household type, we can create i3 observations on the collective

level. The Veil Of Ignorance Scenario is the same on both levels as the respondents have not chosen their τ out of the position

of household type and therefore a permutation would not be reasonable.

8

Table 4: Selected τ by treatment

selected τ

collective level individual level

treatment NQuantiles

NQuantiles

0.25 Median 0.75 0.25 Median 0.75

Baseline 15000 0.48 0.5 0.5 74 0.40 0.5 0.62

Minimum

Requirement15625 0.45 0.52 0.58 75 0.40 0.52 0.62

Cost Uncertainty 15625 0.25 0.37 0.44 75 0.19 0.36 0.45

Veil Of Ignorance 75 0.49 0.6 0.75 75 0.49 0.60 0.75

Neutral Project 16875 0.48 0.50 0.59 76 0.35 0.50 0.61

Notes: N = number of observations, collective level = data from permutation,

indivudual level = actual collected data

As can be seen in the baseline scenario subjects prefer a more proportional share of the costs of the energy

transition (τ = 0.5). There is also almost no effect in the introduction of a social protection mechanism in the

MR treatment for the xmin-households. The median τ is only slightly higher than in the BAS scenario(0.52

to 0.5). Comparing the favored financing mechanism of the energy transition (BAS) with the financing

mechanism of other public projects(NP), we find that the median τ is exactly the same in both scenarios

(τ=0.5). The Cost Uncertainty scenario leads to a more regressive distribution (τ = 0.37/0.36). Finally in

the Veil Of Ignorance scenario the votes lead to a more progressive distribution(τ=0.6).20

In the next subsections we will first analyze how institutional changes shape the distribution scheme of the

society and afterwards test these findings against the individual characteristics, namely the social preferences

and the attitude towards the energy transition. We want to ensure that we have found real treatment effects

and not differences in the personal features between the scenarios.

3.1 Institutional level

From the outcomes displayed in figure 3 and table 4, we see that most institutional changes lead only to

small deviation from the proportional distribution we have seen in the BAS scenario. The energy transition

(public projects) seems to be seen as a task for the whole society, but with the limitation that no specific

group in the population is overburdened. The respondents thereby follow the ability-to-pay principle.

The introduction of a social protection mechanism for low income household in the MR scenario leads to no

significant change in the voting behavior. The range of the deviation from the median τ in the BAS and MR

scenarios only falls below the critical level (τ <= 0.37) in the choices of the household type A (low income).

This result can be interpreted in the way that a minimum requirement for low-income households is already

embedded in the general distributional preferences of the society.

Furthermore, the question if the energy transition is seen different than other public projects is important.

In the public discussion and media reporting it often seems that the energy transition has an exceptional

position due to its importance for the whole society. Our data can not support this picture. At least from a

distributional point of view, the energy transition has no outstanding position to other public projects in the

public perception. In both scenarios (BAS and NP) the society votes for a proportional tax scheme (τ = 0.5

in both cases).

So far, the context in the distributional decision making and state interventions lead to no significant differ-

ences. The distributional preferences of our respondents seem to be robust in these ways. But if we change

the position and the actual costs for the society, we find significant changes in the distributional choices of

the respondents. The observation of a fair and just distribution without personal bias in the (VOI ) scenario

and the role of ”cost uncertainty” in the (CU ) scenario, lead to significant treatment effects (see table 5).

20Table 9 in Appendix A gives a more detailed overview of the selected τ , including the choices of each household type. Table

10 in the Appendix pictures the financial impact of the distributional votes. It shows the share on the costs of the energy

transition (public project) and the budgetary burden as consequences of the selected τ .

9

Table 5: Treatment effects

comparison of

compared

treatments

Mann-

Whitney

U-Test

household A household B household C

All

household

types

BAS vs. MRz -0.643 -0.126 -0.274 -0.917

p 0.5201 0.8994 0.7839 0.3589

BAS vs. CUz 0.842 2.360** 4.109** 4.112***

p 0.3996 0.0183 0.0000 0.0000

BAS vs. VOIz - - - -2.720***

p - - - 0.0065

BAS vs. NPz -1.462 0.253 0459 -0.553

p 0.1437 0.8002 0.6463 0.5802

In order to test the distributional choices without any bias from self-interest, we introduced the Veil Of

Ignorance scenario. With τ = 0.6 we have a weak significant change in the tax scheme towards a more

progressive cost distribution. Figure 4 displays these findings.

Figure 4: Baseline scenario vs. Veil Of Ignorance scenario: selected τ

If we assume that subjects only can make a fair and just decision from an ”original position” without knowing

their later position in the society (Rawls, 2009), it is interesting that respondents who know their position

(at least in the low and medium income group) choose a tax scheme that leads to a higher share on the costs

of the energy transition than in the ”fair and just” allocation under the Veil.21 It seems that despite a fair

distribution individuals want to contribute a noticeable share.

We have seen until now that the participants favor a proportional or progressive distribution of the costs.

There seems to be a social agreement among the population that within their capabilities each household

type should contribute something to the public project. If we now introduce cost uncertainty (risk), the

distributional choices are changing as the whole society moves towards a more regressive tax scheme (see

figure 5). The selected τ is moving away from the median value of 0.5 towards zero among all household

types (but only significant within the household types B and C). The strongest effect can be seen in the high

income group C, with a change from ∆− τ = 0.21 in comparison to the BAS scenario. As the median τ also

diverges in the middle income group B towards a more regressive distribution (median tau = 0.41), we can

not simple argue with the lack of social preferences in risky situations.

21Remember that τ → 1 leads to a higher share on the costs for the low and medium income groups (see table 3)

10

Figure 5: Baseline scenario vs. Cost Uncertainty scenario: selected τ by household types

It rather seems that through cost uncertainty, we have a change in the basis of decision-making. Without cost

uncertainty, a majority of the votes is based on the ability-to-pay-principle. With the introduction of risk,

we see that the polluter-pays-principle also plays an important role. This result follows more a consumption

tax, where each individual take his share of the costs according to her energy use in the context of the

energy transition. Our finding has a strong policy implication. As the goals of the energy transition are

very ambiguous and the success cannot be foreseen, the effect of cost uncertainty strongly indicates that the

polluter-pays principle in the financial mechanism for the energy transition should be included.

The next subsection controls our observed treatment effects for individual characteristics and attitudes.

3.2 Individual level

The classical economic theory says that involuntary transfers for the provision of a public good will completely

crowd-out voluntary transfers. As (Andreoni, 1990) introduced the ”warm-glow of giving” - hypothesis, he

proofed that crowding-out will be incomplete because individuals care about giving. Our design makes sure

that all subjects receive the same utility from the public good independent of their individual contribution.

As we keep the costs of the energy transition (e.g. public project) constant, it would be only be rational to

contribute something if the respondent benefit from her own contribution or has social preferences.

In the case of the energy transition the former is only reasonable, if the respondents have a positive attitude

towards the intent and purpose of the project.

Table 6: Attitude towards the energy transition

households

A B C All

∆-WTP in % θ=1-0.45 %

(0.5148)

+ 1,21 %

(0.2610)

+5,04 %

(0.3427)

+0,95 %

(0.5361)

Notes: ∆-WTP = Deviation between the willingness-to-pay of supporters and

opponents of the energy transition, θ[0, 1] = Attitude towards the energy transition,

significance tested with the Mann-Whitney U-Test, p-value in parentheses.

Table 6 shows the changes in willingness-to-pay (WTP) for subjects with a positive attitude towards

11

the energy transition.22 It is obvious that the attitude towards the energy transition does not lead not to

a higher WTP and by that does not influence the distributional choice of the respondents. This finding is

accentuated by the comparison of the treatments BAS and NP. Here we also find no significant difference

in the respondents choices between those scenarios (see table 4 above). Our data indicates that in the soci-

ety the energy transition is not perceived differently than any other public project, at least on the issue of

financing those projects.23

If the context does not matter, the selection of the distributional key should depend on the respondents

social preferences. Subjects with inequality aversion(IAV) likely vote for a tax key that leads to a more

equal income distribution. This effect should be seen over all income groups and all given scenarios. The

respondents should therefore choose τ → 1. In addition, respondents with egoistic(EGO) preferences always

maximize their own payoff. The effect of these preferences depends on to which household type these subjects

are assigned. If we would have only egoistic subjects, this should also lead to a more progressive tax key

as it is rational for the households with low and medium incomes to choose τ → 1 against the high income

household with τ → 0.

Table 7: Selected τ by social preferences

social preferences scenariohouseholds

A B C All

∆-selected τIAV

BAS, MR, CU, NP+ 0.00

(0.7952)

-0.06

(0.9007)

+ 0.04

(0.9221

+ 0.05

(0.7782)

VOI - - -+0.06

(0.1287)

EGO BAS, MR, CU, NP+0.21***

(0.0013)

+0.10*

(0.0610)

+0.07

(0.7368)

+0.07***

(0.0039)

Notes: ∆-selected τ = Deviation from median τ if the respondents have certain social preferences,

IAV = Inequality Aversion, EGO = egoistic preferences, significance tested with the Mann-Whitney U-Test,

p-value in parentheses.

As can be seen in Table 7, we find no concrete evidence that IAV players choose a significant different

distribution than other subjects. But what it is obvious is that EGO respondents who are assigned to

the low and medium income group vote for more progressive tax scheme. As IAV and EGO types should

have the same tendencies in the low and medium income groups and could share also the same personal

characteristics, we divided these groups and controlled our results in Appendix A table 12. We can see from

this table that our results persist.24

In order to test our treatment effects, we conduct a linear regression analysis with the logit transformed

τ as the dependent variable.25 Table 5 list the regressions for seven models. The benchmark category is

the BAS scenario with household type B (medium income group). Model I contains only the treatment

variables. Model II includes the household types A (low income) and C (high income), Model III controls

for the selected social preferences and the risk self perception. Model IV merges the variables from Model

I to III. The attitude towards the energy transition and climate change and the personal characteristics are

tested in Model V and VI. Finally, all variables are combined in Model VII.

The dummy variables that enter the regressions are: (i) Minimum Requirement Scenario (MR): To test for

22θ = 1 is determined if respondents taking the following positions: (i) The society should counteract the consequences of the

climate change, (ii) higher costs are acceptable for the goals of the energy turnaround and (iii) the current amount of the costs

for the energy transition is acceptable or could be higher.23Table 11 in the Appendix A displays the influences of the attitude towards the energy transition and the climate change in

general.24We also tested for the personal characteristics of the subjects. You find the results of the regression in table 13 in the

Appendix A.25Hereby we can solve the problem of the limitation of the [0, 1]-interval for τ and can use the data in a regular linear

regression.

12

the effects of institutional interventions. (ii) Cost Uncertainty Scenario (CU): To test for the consequences for

the introduction of risk. (iii) Veil Of Ignorance Scenario (VOI): To control for self interested bias. (iv) Neutral

Project Scenario (NP): To check for framing effects in the context of the energy transition. (v) household

type A (low income) & C (high income): To Test for the effects of different income positions in comparison

to the benchmark household. (vi) egoistic preferences (EGO) and risk self perception (risk SP): Influence

of having egoistic preferences and subjects describing themselves as risk averse. In addition, we test for the

attitudes and personal characteristics described in section 2.3.

Table 8: Effects on institutional and individual level

Model I Model II Model III Model IV Model V Model VI Model VII

constant-0.045 0.082 0.007 0.400 0.424 0.333 0.806

(0.189) (0.262) (0.264) (0.363) (0.981) (0.837) -1.670

MR0.056 0.055 0.037 -0.023 -0.537 -0.238 -1.066**

(0.275) (0.276) (0.325) (0.328) (0.444) (0.345) (0.507)

CU-0.785*** -0.787*** -0.954*** -1.018*** -1.407*** -1.020*** -1.660***

(0.233) (0.231) (0.318) (0.315) (0.422) (0.329) (0.503)

VOI0.520* 0.518* 0.624* 0.623* -0.198 0.484 -0.574

(0.275) (0.276) (0.351) (0.347) (0.484) (0.409) (0.573)

NP0.178 0.175 0.129 0.102 -0.359 -0.105 -0.777

(0.285) (0.286) (0.409) (0.407) (0.558) (0.491) (0.755)

household type A

(low income)

- -0.021 - -0.316 -0.079 -0.221 -0.307

(0.215) (0.269) (0.379) (0.312) (0.475)

household type C

(high income)

- -0.357 - -0.773** -0.542 -0.668* -0.561

(0.221) (0.313) (0.430) (0.362) (0.527)

EGO- - 0.383* 0.393* 0.595* 0.351 0.731*

(0.229) (0.218) (0.346) (0.231) (0.408)

risk SP- - -0.351*** -0.379*** -0.359** -0.310** -0.246

(0.115) (0.115) (0.152) (0.132) (0.188)

Attitudes No No No No Yes No Yes

Personal

CharacteristicsNo No No No No Yes Yes

r2 0.066 0.075 0.127 0.159 0.220 0.188 0.268

N 375 375 234 234 146 202 132

Notes: Dependent Variable: selected τ (logit transformed). Standard Errors in parentheses. Linear

regression with White’s heteroscedasticity-robust covariance matrix. Significant coefficients are

marked with one asterisk if p ≤ 0.10, two asterisks if p ≤ 0.05, and three asterisks if p ≤ 0.01.

The regression strongly supports our findings on the institutional level that cost uncertainty leads to a more

regressive distribution. Furthermore the effect of the veil of ignorance can be seen. It is slightly significant on

the 10 % level and vanishes, when we test for personal characteristics and attitudes. The impact of egoistic

preferences shown in table 7 is supported by our regression. The fit of the regressions is in the range of the

usual noise in experimental data.

The regressions support our findings that the proportional distribution prevails over nearly all scenarios and

the respondents follow the ability-to-pay principle. From our data the energy transition and public projects

in general are seen as a task for the whole society, with the limitation that no social group is overburdened.

This result is quite robust until cost uncertainty is introduced. With the introduction of risk, the voting

behavior seems to follow no longer just the ability-to-pay principle but rather includes also the polluter-pays

principle.

13

4 Conclusion

In this paper, we have experimentally studied the distributional preferences of the respondents in the context

of financing the energy transition in Germany. By that, we gave new insights to the existing literature on

willingness-to-pay for green energy.26As far as we know this paper is the first of its kind, where distributional

preferences in the context of the climate change are examined in experiments.

We collected 375 observations from non-standard subjects in the field. The individuals have made their deci-

sions in a strictly neutral cheap talk design with strong monetary incentives so that the data is as meaningful

as possible. Respondents voted for their favorite distribution in five scenarios to test different institutional

changes. In addition, we elicited subjects’ social preferences and asked for their sociodemographic data as

well as their attitude towards the energy transition and the climate change in general.

We find that subjects generally follow the ability-to-pay principle (see figure 3 and table 4 in section 3). The

participants have voted in most of the treatments for a proportional distribution, where all social groups

contribute their share to the total costs without overburden a single group. This basis of decision-making also

explains why the introduction of a social protection mechanism in the Minimum Requirement scenario has

no significant effects. A mechanism to avoid the danger of a financial overloading of low income households

seems to be embedded in the distributional preferences of the society.

Another outcome is that the energy transition is not seen different than other public projects at least in the

matter of financing those political projects. These findings are supported both by the comparison of the

baseline and the neutral project scenario and the insignificant influence of the attitude towards the energy

transition of the subjects in the distributional choice.

Furthermore, we find that the respondents are willing to take their share of the costs despite a ”fair and

just” allocation of the costs. If we compare the BAS and VOI scenarios, we see that the low and medium

income households, which are informed about their position in the society, are choosing a distribution that

leads to a higher share for themselves compared to the choices of the social planner.

The most important finding is seen when we introduce cost uncertainty. The society moves from a clear

proportional distribution towards a more regressive tax scheme (see figure 5 in section 3).This behavior can

not be explained solely by the effect of risk on social preferences. Therefore it needs another explanation. It

seems that the basis of the decision making moves from the ability-to-pay principle to a compromise, which

also includes the polluter-pays principle. As the subjects cannot foresee the certain value of their share, our

data indicates that consumption of the energy seems to be a fair basis for the distribution of the costs.

Our results have important policy implications for the acceptance of the energy transition. If the society

is certain that the goals of the energy transition will be achieved as planned and the costs are relatively

certain a proportional distribution on the basis of the ability-pay-principle is preferred. This contradicts

the current financial mechanism: the EEG surcharge with his regressive characteristics (see Frondel et al.

(2015)). However, if the costs for the ambitious objectives of the government are perceived unstable, we see a

movement towards a capitation of the costs. This scenario is quite reasonable as we already have debates in

the implementation on the important grid development from north Germany to Bavaria. If uncertainty about

the costs takes hold, the regressive character of the EEG surcharge could support the public acceptance of

the energy transition.

This result leads to a dilemma for the politicians. On the one side, they like to ensure the society that the

goals of the energy transition will be reached and therefore the costs are certain. Under these circumstances

the current finance mechanism (the EEG surcharge) would be the wrong instrument to ensure the public

acceptance. On the other side, a commitment by the government that the cost of the energy transition are

uncertain and that there is a risk in the energy transition would support the surcharge as an instrument but

certainly would reduce the trust in the success of the energy transition and in the government themselves.

Therefore, future research should explore the underlying principles for the distributional choices in more

detail and especially focus on the role of cost uncertainty (risk) to the decision making process.

26Prominent examples are Menges et al. (2005); Menges and Traub (2009); Grosche and Schroder (2011)

14

References

Aadland, D. and A. J. Caplan (2006). Cheap talk reconsidered: New evidence from cvm. Journal of Economic

Behavior & Organization 60(4), 562–578.

AG Energiebilanzen e.V. (2014). Bruttostromerzeugung in deutschland ab 1990 nach energietragern. Website.

available online at http://www.ag-energiebilanzen.de/index.php?article_id=29&fileName=ausw_

05112014ov_engl.pdf; downloaded on 24.06.2015.

Andreoni, J. (1990). Impure altruism and donations to public goods: a theory of warm-glow giving. The

economic journal, 464–477.

Babcock, L. and G. Loewenstein (1997). Explaining bargaining impasse: The role of self-serving biases. The

Journal of Economic Perspectives, 109–126.

Balafoutas, L., R. Kerschbamer, and M. Sutter (2012). Distributional preferences and competitive behavior.

Journal of economic behavior & organization 83(1), 125–135.

BDEW (2015). Bdew zum strompreis der haushalte. Website. available online at https:

//www.bdew.de/internet.nsf/id/9D1CF269C1282487C1257E22002BC8DD/$file/150409%20BDEW%

20zum%20Strompreis%20der%20Haushalte%20Anhang.pdf;downloadedon24.06.2015.

Charness, G., U. Gneezy, and M. A. Kuhn (2013). Experimental methods: Extra-laboratory experiments-

extending the reach of experimental economics. Journal of Economic Behavior & Organization 91, 93–100.

Cummings, R. G. and L. O. Taylor (1999). Unbiased value estimates for environmental goods: a cheap talk

design for the contingent valuation method. American Economic Review, 649–665.

Frondel, M., S. Sommer, and C. Vance (2015). The burden of germanys energy transition: An empirical

analysis of distributional effects. Economic Analysis and Policy 45, 89–99.

Grosche, P. and C. Schroder (2011). Eliciting public support for greening the electricity mix using random

parameter techniques. Energy Economics 33(2), 363–370.

Harrison, G. W. and J. A. List (2004). Field experiments. Journal of Economic Literature, 1009–1055.

Harsanyi, J. C. et al. (1953). Cardinal utility in welfare economics and in the theory of risk-taking. The

Journal of Political Economy 61(5), 434.

Huijts, N., E. Molin, and L. Steg (2012). Psychological factors influencing sustainable energy technology

acceptance: A review-based comprehensive framework. Renewable and Sustainable Energy Reviews 16(1),

525–531.

Kerschbamer, R. (2015). The geometry of distributional preferences and a non-parametric identification

approach: The equality equivalence test. European Economic Review 76, 85–103.

Konow, J. (2009). Is fairness in the eye of the beholder? an impartial spectator analysis of justice. Social

Choice and Welfare 33(1), 101–127.

List, J. A. (2001). Do explicit warnings eliminate the hypothetical bias in elicitation procedures? evidence

from field auctions for sportscards. American economic review, 1498–1507.

Menges, R., C. Schroeder, and S. Traub (2005). Altruism, warm glow and the willingness-to-donate for green

electricity: an artefactual field experiment. Environmental and Resource Economics 31(4), 431–458.

Menges, R. and S. Traub (2009). Who should pay the bill for promoting green electricity? an experimental

study on consumer preferences. International Journal of Environment and Pollution 39(1-2), 44–60.

15

Musall, F. D. and O. Kuik (2011). Local acceptance of renewable energya case study from southeast germany.

Energy Policy 39(6), 3252–3260.

Rawls, J. (2009). A theory of justice. Harvard university press.

Zoellner, J., P. Schweizer-Ries, and C. Wemheuer (2008). Public acceptance of renewable energies: Results

from case studies in germany. Energy Policy 36(11), 4136–4141.

16

A Appendix

Table 9: Detailed overview of selected τ

treatments household type NQuantiles

Mean S.D.0.25 Median 0.75

Baseline

Scenario

A 24 0.35 0.5 0.59 0.49 0.22

B 25 0.44 0.5 0.66 0.52 0.29

C 25 0.48 0.5 0.5 0.47 0.17

All 74 0.40 0.5 0.62 0.49 0.23

Minimum Requirement

Scenario

A 25 0.34 0.55 0.65 0.5 0.28

B 25 0.45 0.52 0.58 0.54 0.20

C 25 0.4 0.5 0.6 0.49 0.23

All 75 0.40 0.52 0.62 0.51 0.24

Cost Uncertainty

Scenario

A 25 0.39 0.42 0.55 0.44 0.18

B 25 0.21 0.41 0.48 0.36 0.18

C 25 0.17 0.29 0.32 0.25 0.15

All 75 0.19 0.36 0.45 0.35 0.19

Veil Of Ignorance

ScenarioAll 75 0.49 0.60 0.75 0.58 0.24

Neutral Project

Scenario

A 26 0.50 0.50 0.74 0.58 0.21

B 25 0.25 0.50 0.62 0.5 0.29

C 25 0.23 0.48 0.58 0.44 0.28

All 76 0.35 0.50 0.61 0.51 0.27

Notes: N = number of observations, S.D. = standard deviation

17

Table 10: Consequences of selected τ : Share on the costs of the energy transition ti/G and budgetary burden ti/xi.

Base

line

Min

imum

Req

uirement

Cost

Unce

rta

inty

Veil

ofIgnora

nce

NeutralPro

ject

AB

CAll

AB

CAll

AB

CAll

All

AB

CAll

(m

edia

n)τ

0.5

0.5

0.5

0.5

0.5

50.5

20.5

0.5

20.4

20.4

10.2

90.3

60.6

0.5

0.5

0.4

80.5

ti/G

household

A

(lo

win

com

e)

0.1

70.1

70.1

70.1

70.1

50.1

60.1

70.1

60.1

90.1

90.2

40.2

10.1

30.1

70.1

70.1

70.1

7

household

B

(m

ediu

min

com

e)

0.3

10.3

10.3

10.3

10.3

10.3

10.3

10.3

10.3

20.3

20.3

20.3

20.3

10.3

10.3

10.3

10.3

1

household

C

(hig

hin

com

e)

0.5

20.5

20.5

20.5

20.5

40.5

30.5

20.5

30.4

90.4

90.4

40.4

70.5

60.5

20.5

20.5

10.5

2

ti/xi

household

A

(lo

win

com

e)

0.5

0.5

0.5

0.5

0.4

50.4

80.5

0.4

80.5

80.5

90.7

10.6

40.4

0.5

0.5

0.5

20.5

household

B

(m

ediu

min

com

e)

0.5

0.5

0.5

0.5

0.5

0.5

0.5

0.5

0.5

10.5

10.5

10.5

10.4

90.5

0.5

0.5

0.5

household

C

(hig

hin

com

e)

0.5

0.5

0.5

0.5

0.5

20.5

10.5

0.5

10.4

70.4

70.4

20.4

50.5

40.5

0.5

0.4

90.5

Notes:A

=D

ecisio

nsby

household

type

A,B

=D

ecisio

nsby

household

type

B,C

=D

ecisio

nsby

household

type

C,All

=D

ecisio

nsofall

household

types,ti/G

=Tax

/Costsofthe

energy

transitio

n(public

proje

ct),ti/xi

=Tax/personalin

com

e,despiteτ

all

results

are

presented

inpercent.

18

Table 11: Influence of attitudes towards the climate change and energy transition

Attitudes

counteract the consequences

of the climate change.

-0.026

(0.028)

Acceptance of higher costs

for the energy transition.

-0.025

(0.022)

Agreement to the level

of the current financial burden.

-0.002

(0.012)

Germany should at least do as much

as the other countries.

0.013

(0.042)

Emission rights should be

distributed by capitation.

0.007

(0.021)

The whole society is responsible

for the energy transition.

-0.052*

(0.028)

The financial burden of the energy transition

should be distributed among the whole society.

0.009

(0.026)

Disagreement that energy intensive companies

should be freed from the eeg surcharge.

-0.018

(0.023)

Subjects, who reduce their energy use

because of ecological reasons.

0.071***

(0.022)

Subjects, who estimated the correct share

of renewables on the total energy supply.

-0.035

(0.024)

Subjects, who associate the energy transition

with the increase of the share of renewables

to the total energy supply.

-0.014

(0.019)

r2 0.148

N 144Notes: Dependent Variable: wtpin%. Standard Errors in paren-

theses. Linear regression with White’s heteroscedasticity-robust

covariance matrix. Significant coefficients are marked with one

asterisk if p ≤ 0.10, two asterisks if p ≤ 0.05, and three asterisks

if p ≤ 0.01.

Table 12: Selected τ by social preferences

social preferences scenariohouseholds

A B C All

∆-selected τ

IAVBAS, MR, CU, NP

0.00

(0.7952)

-0.06

(0.9007)

+ 0.04

(0.9221

+ 0.05

(0.7782)

VOI - - -+0.06

(0.1287)

EGOBAS, MR, CU, NP

+0.21***

(0.0013)

+0.10*

(0.0610)

+0.07

(0.7368)

+0.07***

(0.0039)

IAV & EGO-0.01

(0.9661)

-0.065

(0.1045)

-0.085

(0.5264)

0.00

(0.2487)

Notes: ∆-selected τ = Deviation from median τ if the respondents have certain social preferences,

IAV = Inequality Aversion, EGO = egoistic preferences, EGO = no egoistic tendencies,

significance tested with the Mann-Whitney U-Test, p-value in parentheses.

19

Table 13: Influence of personal characteristics

Personal Characteristics

Age-0.046

(0.158)

Gender-0.582***

(0.215)

Income-0.007

(0.079)

Level of education0.068

(0.117)

Religious0.262

(0.209)

Householdsize0.068

(0.094)

Ownership-0.097

(0.245)

Vote for the green party0.271

(0.297)

Vote for the SPD party-0.428*

(0.242)

profiteer from subsidies0.103

(0.270)

r2 0.042

N 318Notes: selected τ (logit transformed). Standard Errors in paren-

theses. Linear regression with White’s heteroscedasticity-robust

covariance matrix. Significant coefficients are marked with one

asterisk if p ≤ 0.10, two asterisks if p ≤ 0.05, and three asterisks

if p ≤ 0.01.

20