SFB Pro-environmental behavior as a 
means of self-signaling: Theory 
823 and evidence 
 
Daniela Flörchinger, Manuel Frondel, 
Stephan Sommer, Mark A. Andor 
 
 
 
 
 
 
 
Nr. 31/2021 
 
 
 
 
 
 
Discussion Paper 
 
Pro-environmental Behavior as a Means of
Self-Signaling: Theory and Evidence
Daniela Flörchinger* Manuel Frondel*†
Stephan Sommer‡* Mark A. Andor*
January 12, 2022
Abstract
Recent research indicates that pro-environmental behavior may be
driven by concerns about one’s moral identity. Using identification
with the environmentalist movement Fridays for Future, this paper de-
velops and empirically tests a straightforward model of self-signaling.
We assume that pro-environmental behavior, here taking the train
rather than the plane for a journey, serves as a means of self-signaling.
On the basis of a large-scale survey experiment with revealed pref-
erences, we find evidence that respondents who receive an identity
*RWI – Leibniz Institute for Economic Research
†Ruhr-University Bochum
‡Bochum University of Applied Sciences
prime in the form of a reminder of their previously stated attitude
towards Fridays for Future are more likely to behave in line with the
movement’s moral principles in that they take the train. Our expla-
nation of this outcome is that individuals attempt to avoid cognitive
dissonance by choosing the more environmentally benign alterna-
tive. Our results suggest that pro-environmental behavior may be
enhanced by appealing to an individual’s self-image so that costly
interventions that are designed to convince subjects of new moral
principles may be unnecessary.
JEL classifications: D81, D91.
Keywords: Pro-social behavior, moral identity, cognitive disso-
nance, self-signaling.
Acknowledgements: We are highly grateful for valuable comments and
suggestions by participants of the 3rd PhD-Seminar Experiments in Cli-
mate and Resource Economics in 2021, the RWI Therapy Seminar, as well
as Lukas Tomberg, and, in particular, Colin Vance. We gratefully acknowl-
edge financial support by the Collaborative Research Center “Statistical
Modeling of Nonlinear Dynamic Processes” (SFB 823) of the German Re-
search Foundation (DFG), within Project A3, “Dynamic Technology Mod-
eling” and by the Federal Ministry of Education and Research (BMBF) un-
der grant 03SFK5C0 (Kopernikus Project ARIADNE), grant 01UT1701A
(Project LICENSE), and grant 01LA1823A (Project Eval-Map II).
1 Introduction
Standard economic theory assumes that individual behavior is driven by
preferences over outcomes and the resulting material utility. Recent re-
search, however, has highlighted the role of identity in decision-making,
as well as its utility implications, called affective utility. In a pioneering ar-
ticle, Akerlof and Kranton (2000) were the first to introduce the notion of
identity as an additional ingredient of economic theory. These authors in-
clude the affective utility associated with identity-confirming actions, such
as pro-environmental or, more generally, pro-social behavior, into the util-
ity function (Akerlof and Kranton, 2000).1
In a related article, Bénabou and Tirole (2011) develop a model in which
former actions, such as taking the train or the plane for a journey, serve as
a means of self-signaling, that is, as a signal about one’s moral identity to
oneself. Core features of their model are that (i) individuals are not fully
aware of their true moral identity, which they infer from their past actions,
and (ii) that they derive affective utility from having a positive self-image.
Therefore, a key assumption of the model of Bénabou and Tirole (2011)
is that when deciding upon an action, in addition to material utility, an
individual also takes into account the affective utility that results from the
implications that an action has for the self-image.
1The notion of identity is defined as ”a person’s self-image as well as her assigned
categories” (Akerlof and Kranton, 2000). Self-image is a psychological concept denom-
inating an individual’s “[...] view or concept of oneself. [It] is a crucial aspect of an
individual’s personality that can determine [...] a sense of general well-being” (American
Psychological Association, 2021).
1
Adding to this strand of research, this paper follows Bénabou and Ti-
role (2011) and develops a straightforward model of self-signaling that
we test experimentally using the identification with the environmentalist
movement Fridays for Future. Acting pro-environmentally, which is cap-
tured in the experiment by taking the train rather than the plane for a jour-
ney, is hypothesized to be a means of self-signaling. In a large-scale survey,
in which an experiment with revealed preferences was embedded, nearly
6,000 German household heads had to decide on whether they would pre-
fer to win a voucher for a train ride or for a flight, which was won by one
out of 50 participants.
Half of the sample was randomly assigned to a treatment group, whose
subjects were reminded of their previously stated attitude towards Fridays
for Future immediately before making their trip decision. These subjects
were then asked to actively confirm or revoke this attitude. Following the
experimental literature that has used priming to disentangle the impact
of identity on behavior from preferences over outcomes, this reminder
served as an identity prime that temporarily increases the salience of pro-
environmental identity – for an overview on priming in economics, see
Cohn and Maréchal (2016).
Our empirical findings indicate that subjects of the treatment group,
who were reminded of their previously stated attitude towards Fridays
for Future, are more likely to choose the train than those of the control
group, who did not receive such a reminder. A possible explanation of
this outcome is that subjects of the treatment group attempt to avoid cogni-
2
tive dissonance by choosing the more environmentally benign alternative.
Cognitive dissonance typically arises from inconsistencies between an in-
dividual’s behavior and her self-image (Aronson, 1969). If, for example,
an individual considers herself to be an honest person, she will experience
cognitive dissonance while lying.
We further find that individuals who are less secure about their sup-
port of Fridays for Future are more likely to opt for the environmentally
benign transport mode. Our results suggest that pro-environmental be-
havior may be enhanced by appealing to an individual’s self-image so that
costly interventions that are designed to convince subjects of new moral
principles, such as large-scale information campaigns , are not necessarily
required.
The remainder of this paper is structured as follows: The next sec-
tion provides a literature review, while Section 3 introduces our theoretical
model. The experimental design and hypotheses are explained in Section
4. Section 5 provides a detailed description of our data, while Section 6
presents the empirical results. A suite of robustness checks is reported in
Section 7. The last section summarizes and concludes.
2 Literature Review
Related to the work of Bodner and Prelec (2003) and Bénabou and Tirole
(2004, 2006, 2011), in which individuals infer their identity from past be-
3
havior, our analysis is embedded in the rich behavioral economics litera-
ture on identity, self-signaling, and cognitive dissonance, with the latter
two concepts originating from social psychology.2 The economic implica-
tions of cognitive dissonance were first analyzed in the seminal paper of
Akerlof and Dickens (1982), who introduced the notion that individuals
have preferences over their beliefs about the state of the world, but can
modify these beliefs to a certain extent.
By inducing cognitive dissonance as a centerpiece of our analysis, we
build on numerous other experimental studies that attempt to provoke
cognitive dissonance among participants. For example, in a large-scale
field experiment among the customers of the UK’s leading renewable en-
ergy supplier, Gosnell (2018) causes cognitive dissonance by appealing
to the customers’ identity as a conscious consumer. Her results indicate
that, in general, dissonance-inducing messaging increases the willingness
to switch from paper to online billing. This result is in line with our find-
ing that the likelihood of choosing the more environmentally benign al-
ternative in form of the train is increased by the induction of cognitive
dissonance.
Other empirical studies have provided support for pro-social behavior
as a means of self-signaling. For instance, analyzing choices in modified
dictator games, Dana et al. (2007), Grossman (2014), Grossman and Van
Der Weele (2017), and Matthey and Regner (2011) show that subjects who
make selfish choices tend to avoid information on whether their decision
2See Bem (1972) on self-perception theory, which is the basis of self-signaling, and
Festinger (1962) on cognitive dissonance.
4
harms the other player – a result that is perfectly in line with self-signaling
theory: Information ignorance facilitates the maintenance of a positive
self-image while choosing the allocation that yields larger material utility.
In a similar study by Tonin and Vlassopoulos (2013), participants make
three consecutive decisions to allocate an endowment between themselves
and either the experimenter or a charity. One of these decisions is ran-
domly chosen to be implemented. At the end of the experiment, subjects
are given the opportunity to opt out of the donation and keep the entire en-
dowment to themselves. Although this allocation had been feasible from
the very beginning, a substantial fraction of subjects choose to opt out.
This result implies that allocations comprising positive donations were not
purely motivated by preferences over outcomes but served to self-signal
altruism, at least partially.
Gneezy et al. (2012) and Dubé et al. (2017) find evidence for self-signal-
ing in field settings where a consumption good is bundled with a dona-
tion to a charity. By contrast, the experimental results by Grossman (2015)
support a model of social signaling, but provide little evidence for self-
signaling. Our study adds to this literature by investigating self-signaling
in a setting in which subjects have to choose between two travel modes,
where taking the plane is socially more harmful, but yields larger material
utility than an environmentally benign train ride.
5
3 Theoretical Model
To theoretically analyze how self-signaling influences the decision between
a pro-social and a selfish action, we now develop a straightforward two-
period model, thereby following the theory of moral behavior of Bénabou
and Tirole (2011). It is based on a general model of identity in which peo-
ple care about ”who they are” and infer their identity from past actions.
Our model is grounded in the theory of cognitive dissonance, because in-
dividuals have an incentive to choose an action that is in line with their
desired self-view, just to avoid discrepancies between their actions and
their self-image. Otherwise, such a discrepancy would lead to cognitive
dissonance. Another central element of our model is that individuals are
uncertain about their moral identity, i. e. the strength of their moral con-
cerns, and take their prior actions as signals to make inferences about their
true identity.
3.1 Model Setup
In our two-period model (see Figure 1), at t = 0, individuals choose be-
tween a selfish and a pro-social action a ∈ {S, P}, where S designates the
selfish and P the pro-social action for which it is common knowledge that
it is socially preferable. Individuals derive material utility U(a) from ac-
tion a and affective utility V (v̂) from their self-image v̂ on their true iden-
tity v. The affective utility V (v̂) results from the individuals’ intention to
“be true to myself” in their decisions, to “maintain my integrity”, to “not
6
betray my values”, to “be able to look at myself in the mirror”, etc. (Bén-
abou and Tirole, 2011).
It is assumed that U(S) > U(P ) and, hence, rational individuals would
choose a = S if moral concerns and considerations about the self-image
play no role at all, that is, if V = 0. In contrast, given that action P is
socially preferable, a purely altruistic individual would always choose P .
At t = 0, individuals are uncertain about their true moral identity v,
that is, whether they are selfish or altruistic. For the sake of simplicity, it
is assumed that there are only two types of individuals that differ with
respect to the strength of their moral concerns, indicated by v ∈ {vL, vH}
with vH > vL > 0. While not knowing their true type v, at t = 0, individu-
als hold an a-priori belief about their identity v, that is, they either believe
to be of the high-moral type vH or of the low-moral type vL. Due to the un-
certainty about their moral identity v, individuals take their prior actions
as signals to make inferences about their true identity v. There is extensive
empirical evidence that both people judge themselves by their actions and
many decisions are shaped by a concern to achieve or maintain a desirable
self-view (Bénabou and Tirole, 2011).
Self-inference Assumption: At t = 1, individuals are aware of their true
identity v only with probability λ. In contrast, with probability 1 − λ, in-
dividuals are unaware of their true identity v, but instead infer their true
type from past behavior and actions. 1 − λ may therefore be thought of
as the degree of malleability of the self-image through actions (Bénabou and
Tirole, 2011), reflecting the possibility that the motivation for deeds may,
7
for instance, be forgotten or repressed.
Hence, according to the Self-inference Assumption, at t = 1, an indi-
vidual’s self-image v̂ equals the true type v with probability λ, but with
probability 1− λ the true type v has to be inferred from past actions:

v with probability λ,v̂ =  (1)v̂(a) with probability 1− λ,
where v̂(a) denotes the self-image derived from past action a. To rule out
the polar case that the self-image v̂ equals the true type v with probability
100%, we assume that 0 ≤ λ < 1.
If an individual has chosen the pro-social option P , she will consider
herself to be of the high-moral type vH and if she has acted selfishly by
choosing action S, she will think of herself as the low-moral type vL:
v̂(P ) = vH > vL = v̂(S). (2)
In short, the chosen action a serves as a means of self-signaling, that is, as
a signal of one’s true identity v in that it leads to the self-image v̂(a).
When deciding on which action to take, in addition to material utility
U , individuals are supossed to take into account the affective utility V :
V (v̂) ≡ ∂Vsv̂,with s > 0, > 0, (3)
∂v̂
where the affective utility V depends on the self-image v̂ that an individual
8
t = 0 t = 1
A priori belief 𝜌 Choice of action 𝑎 Probability λ < 1 that v
about true 𝑈(𝑎) is revealedidentity Material utility:  „Affective“ utility:     svො
Figure 1: Timing of our two-period model
has at t = 1 and s measures the importance of the self-image v̂ for affective
utility. s depends on the individual’s awareness of identity considerations
and is assumed to be situation-dependent, that is, it can be varied exoge-
nously, for instance, by increasing the salience of social categories through
priming, as it is done in our experiment with the reminder on the attitude
towards the Fridays for Future movement. Note that if self-image were to
be completely irrelevant and, hence, s = 0, affective utility V would van-
ish. To rule out this extreme case, we assume s > 0.
Denoting total intertemporal utility by W , W is given by
W (a) ≡ E[U(a) + V (v̂)], (4)
where the expectation is taken with respect to the distribution (λ, 1− λ) of
self-image v̂ ∈ {v, v̂(a)} and hence
W (a) = U(a) + λV (v) + (1− λ)V (v̂(a)). (5)
9
3.2 Behavior
At t = 0, a rational individual solves the intertemporal utility maximiza-
tion problem
maxaW (a) (6)
by comparing W (S) and W (P ). The difference in intertemporal utility W
between choosing action P and S yields the incentive I to behave pro-
socially:
I ≡ W (P )−W (S)
= U(P ) + λV (v) + (1− λ)V (v̂(P ))− [U(S) + λV (v) + (1− λ)V (v̂(S))]
= U(P )− U(S) + (1− λ)s[vH − vL]. (7)
It becomes evident from this expression that individuals are more likely
to behave pro-socially the larger the material utility U(P ) from the pro-
social action P is. Furthermore, from expression (7), the following two
propositions can be derived.
Proposition 1: An individual is more likely to behave pro-socially
(a) the more important her self-image is for her affective utility V , that
is, the larger s is, and
(b) the more malleable her self-image is, that is, the larger 1− λ.
Proof: From expression (7) follows that (a) ∂I = (1 − λ)[v − v
∂s H L
] > 0 and
(b) ∂I(1− ) = s[vH − vL] > 0, as vH > vL, s > 0, and λ < 1.∂ λ
10
Proposition 2: An increase in an individual’s importance s of her self-
image implies a particularly strong incentive to behave pro-socially for
those individuals with a highly malleable self-image, that is, when 1−λ is
large.
2
Proof: ∂ I(1− ) = vH − vL > 0, as as vH > v∂s∂ λ L.
Both propositions point to the role of former actions as a signal for indi-
viduals to infer their moral identity: First, the more important self-image
is for affective utility, that is, the larger s, the more important it is to signal
a high-moral identity vH by behaving pro-socially (Proposition 1 (a)). Sec-
ond, the greater the malleability of one’s self-image, that is, the larger 1−λ,
the more important it is to send a positive signal by behaving pro-socially
(Proposition 1 (b)). Third, according to Proposition 2, a high malleability
of one’s self-image is reinforced by an increase in its importance, leading
to particularly strong incentives to behave pro-socially.
4 Experimental Design and Hypotheses
Drawing on an incentivized survey experiment conducted in Germany in
late 2019, we apply the theoretical model presented in the previous section
to the context of pro-environmental behavior and individuals’ identifica-
tion with the environmentalist movement Fridays for Future. In a discrete
choice task, respondents had to decide on whether they would prefer to
win a voucher worth 40 euros either for a flight or for a railway trip, the
11
latter of which is more environmentally benign and, hence, socially pre-
ferred.3 One out of 50 survey participants was randomly selected to win
such a voucher.
For many trips, the plane is frequently the faster and more conve-
nient option than the train and, hence, taking the plane corresponds to
activity S of our theoretical model, which is associated with the larger
material utility. Recently, though, because of its high emissions of green-
house gases, flying has been discredited by an anti-flying movement that
is spearheaded by Fridays for Future. The anti-flying agenda is particularly
pushed by Greta Thunberg, the central figure of Fridays for Future, who re-
ceived enormous attention from the media for avoiding a flight to the UN
climate summit in New York in 2019. Instead, for environmental reasons,
even for the very long trip from Europe to the USA, she took the boat. At
other instances, she preferred taking the train (CNN, 2019; Spiegel, 2019).
Accordingly, traveling by train is the option that is conceived in our exper-
iment to correspond to the socially preferable activity P of our theoretical
model.
At the beginning of the survey, we elicited the respondents’ general
attitude towards the Fridays for Future protests by asking the following
question: “Currently, in many cities students are protesting for more cli-
mate protection every Friday both during and outside school hours. What
is your attitude towards these so-called Fridays for Future protests?”. Re-
spondents answered on a 5-point Likert scale, indicating the degree of
3The voucher was valid either for the German Railways (Deutsche Bahn) or the website
of Flightgift, where flights from a large variety of airlines can be booked.
12
support for the protests – for the response options, see Question A13 in
the appendix.
Later in the survey, half of the sample was randomly assigned to the
treatment group, for which the salience of the self-image was increased
through an identity prime: Immediately before deciding upon the type of
voucher, either for a train or plane trip, subjects of the treatment group
were reminded of their previously stated general attitude towards Fridays
for Future and were asked to confirm or revoke this attitude. Conditional
on their previously stated attitude, subjects of the treatment group were
asked: ”Would you agree that, overall, you don’t support/have a neutral
attitude towards/support the Fridays for Future movement?”.
While the control group did not receive any such questions, nor any
other additional information, the treatment was designed to arouse cog-
nitive dissonance, that is, to raise the respondents’ awareness about the
inner conflict between the morally appropriate behavior, i.e. taking the
train, and the more convenient choice of taking the plane. Given this ex-
perimental design, from Proposition 1 (a), we now derive the following
hypothesis:
Hypothesis 1: Subjects of the treatment group, for which the treatment
of reminding respondents about their attitude towards Fridays for Future
protests should increase the importance s of the self-image, are more likely
to choose the train than subjects of the control group.
Our theoretical model further predicts that respondents who are less
13
sure about their true identity, in formal terms, for which 1 − λ is larger
than for others, are more likely to behave pro-socially – see Proposition
1(b). Following from Proposition 2, these respondents will also react more
strongly to the treatment, which increases the importance s of the self-
image.
To construct a measure of insecurity about the own moral identity,
we exploit the answers to a series of questions with respect to the re-
spondents’ opinion on various aspects of the Fridays for Future movement,
such as skipping classes for protests, climate policy as a question of inter-
generational justice, and the importance of protests relative to students’
own environmental behavior. Respondents who are very sure about their
support for Fridays for Future should consistently express positive attitudes
towards the movement in all these aspects.
Hence, to measure respondents’ insecurity about these aspects, we took
the standard deviation of a respondent’s answer to Question A13, which
elicited the overall attitude towards Fridays for Future, as well as to Ques-
tion A14, and then standardized this insecurity measure – for details, see
Question A13 and A14 and the description of the construction of this mea-
sure in appendix A.2, as well as Table 3 in the subsequent section. Due to
this standardization, we are able to interpret the corresponding regression
coefficient in terms of standard deviations of insecurity.
Based on this insecurity measure, we are now able to formulate two ad-
ditional hypotheses that correspond to Proposition 1 (b) and Proposition
2, respectively:
14
Hypothesis 2: Corresponding to a larger malleability, 1 − λ in our theo-
retical model, respondents who are more insecure about their true moral
identity, and whose answers to questions about various aspects of Fridays
for Future display more variance, are more likely to choose the train.
Hypothesis 3: In line with Proposition 2, the treatment effect of the re-
minder with respect to the attitude towards Fridays for Future, which raises
the importance s of the self-image, should be stronger for respondents
who are more insecure about their true moral identity and, hence, whose
answers to the questions about various aspects of Fridays for Future display
more variance.
5 Data
Our analysis relies on data collected by the survey institute forsa. forsa
maintains a panel of about 80,000 members, being representative of the
German-speaking population – for more information on forsa’s household
panel, see http://www.forsa.com. The survey was addressed to house-
hold heads, defined as those individuals who are responsible for the house-
hold’s financial decisions. Data collection is based on a state-of-the-art tool
that allows panelists to fill out the questionnaire using either a television
or the internet. The survey could be interrupted and resumed at any time.
Respondents in our survey tend to be older, wealthier, and more educated
than the average of the German population (see Table A1 in the appendix).
This implies that the results of our study are only valid for this particular
15
sample and cannot be extrapolated to the entire German population.
The survey period spanned from October 16 to November 6, 2019.
6,549 household heads were recruited to fill in the questionnaire. Out of
these, 553 dropped out prior to the experiment or refused to participate
in it and 48 did not provide their attitude towards the Fridays for Future
protests, such that the sample size for the experiment amounts to 5,948
respondents. Dropout rates hardly differ between treatment and control
group, implying that selection bias is not an issue.
The summary statistics, reported in Tables 2 and 3, indicate that ran-
domization was successful. Socio-economic characteristics, distance to the
nearest airport, pro-environmentalist attitude, and attitudes towards Fri-
days for Future are very similar for the treatment and control group. Only
minor differences emerge for the gender composition and the frequency
of traveling by plane.
Figure 2 illustrates the relationship between the respondents’ insecu-
rity about their attitude towards Fridays for Future and their past travel
behavior by presenting the density of the insecurity measure for both re-
spondents who indicated to never take the plane and those who fly at least
sometimes. The figure shows that the group with subjects who never fly is
less sure about their attitude towards Fridays for Future than the opposite
group.
Furthermore, a t test reveals that the mean values of insecurity between
the two groups are statistically different at the 1% significance level. This
16
Table 2: Means of Socio-demographic Characteristics across Treatment and
Control Groups, as well as the Whole Sample
Whole Treatment Control
sample group group t statistics
Female 0.418 0.407 0.428 1.587
Age 56.3 56.3 56.3 0.119
At least technical college 0.327 0.330 0.325 -0.407
Employed 0.516 0.507 0.525 1.337
Children < 14 years in household 0.151 0.154 0.148 -0.588
Distance to nearest airport in km 34.2 34.1 34.3 0.331
Flyers 0.696 0.685 0.706 1.736
Environmentalist attitude 3.647 3.653 3.641 -0.497
Inclination towards Green party 0.192 0.187 0.198 1.038
Area of residence:
Urban area 0.359 0.365 0.353 -0.927
Peri-urban area 0.432 0.427 0.436 0.690
Rural area 0.209 0.208 0.210 0.253
Type of nearest airport:
Large airport 0.710 0.705 0.715 0.861
Medium airport 0.290 0.295 0.285 -0.861
Household size:
1 person 0.269 0.274 0.264 -0.851
2 persons 0.473 0.465 0.481 1.236
3 persons 0.136 0.137 0.135 -0.188
4 persons 0.090 0.090 0.091 0.242
≥ 5 persons 0.031 0.035 0.028 -1.404
Net household income:
Income < 1,200 Euro 0.076 0.079 0.073 -0.881
Income 1,200 - 2,700 Euro 0.361 0.371 0.350 -1.512
Income 2,700 - 4,200 Euro 0.337 0.331 0.343 0.898
Income ≥ 4,200 0.226 0.219 0.234 1.280
Note: t statistics for testing the equality of means across treatment and control
groups are reported in the last column.
17
Table 3: Means of various Attitudes towards the Fridays for Future Move-
ment across Treatment and Control Group
Whole Treatment Control
Sample group group t statistics
Insecurity about attitude towards Fridays for Future 0.004 0.008 -0.000 -0.322
Attitude towards Fridays for Future:
Negative attitude 0.309 0.306 0.311 0.381
Neutral attitude 0.233 0.238 0.229 -0.820
Positive attitude 0.458 0.456 0.460 0.342
Pupils should protest only in free time:
disagree 0.274 0.269 0.279 0.892
neutral 0.083 0.082 0.084 0.227
agree 0.643 0.649 0.637 -0.961
OK to skip classes:
disagree 0.490 0.484 0.497 0.962
neutral 0.097 0.097 0.098 0.095
agree 0.412 0.419 0.406 -1.034
Question of intergenerational justice:
disagree 0.144 0.140 0.147 0.791
neutral 0.111 0.112 0.110 -0.212
agree 0.745 0.748 0.743 -0.484
Pupils should rather change their own behavior:
disagree 0.286 0.288 0.285 -0.296
neutral 0.146 0.146 0.147 0.073
agree 0.567 0.566 0.569 0.218
Note: t statistics for testing the equality of means across treatment and control groups are
reported in last column.
finding is in line with Proposition 1 (b): Subjects who are less secure about
their true identity are more likely to exhibit identity-confirming behavior,
that is, they are more likely to choose the more environmentally friendly
means of transport.
Not least, our data reveals that supporting the Fridays for Future move-
ment is strongly correlated with other indicators of subjects’ environmen-
18
N = 2066
N = 4419
-4 -2 0 2 4
Insecurity about FfF
Never-flyers Flyers
Figure 2: Insecurity about the attitude towards Fridays for
Future by flying behavior
tal attitudes. For instance, Figure A1 presented in the appendix illus-
trates that respondents who have a positive attitude towards Fridays for
Future are more likely to score higher on a scale measuring general pro-
environmentalist attitude than other respondents. It is likewise not sur-
prising that Fridays for Future supporters are more inclined towards the
Green party (Figure A2) than non-supporters.
6 Empirical Results
Claiming that subjects of the treatment group are more likely to choose the
train than subjects of the control group, Hypothesis 1 can be easily tested
on the basis of a t test on the difference between the shares of train users
across treatment and control group, as the assignment to these experimen-
19
Density
0 .2 .4 .6
tal groups was random.
6.1 Shares of Respondents who Choose to Travel by Train
Relative to the control group, the figures reported in Table 4 indicate that
the share of respondents who opt for the train is significantly higher in
the treatment group, whose subjects were asked to confirm or revoke their
previously stated attitude towards Fridays for Future immediately before
making their traveling decision. Overall, with a share of 79.1% of train
users in the treatment group (see last row of Table 4), the difference to the
control group amounts to 3.9 percentage points and is statistically signifi-
cant at the 1% significance level, thereby confirming Hypothesis 1.
Table 4: Shares of Respondents who Choose to Travel by Train
Attitude towards Treatment group Control group Difference between
Fridays for Future # Obs. Share # Obs. Share treatment and control group
Negative attitude 918 0.743 918 0.664 0.078**
Neutral attitude 712 0.754 675 0.751 0.003
Positive attitude 1,366 0.843 1,359 0.813 0.030*
Total 2,996 0.791 2,952 0.753 0.039**
Note: * and ** denote statistical significance at the 5% and the 1% level, respectively.
Focusing on those participants who indicated a positive attitude to-
wards Fridays for Future, in line with our expectation, we find that the share
of train users in the treatment group is 3.0 percentage points higher than
for the control group, whereas for those with a neutral attitude towards
20
Fridays for Future, the respective difference is virtually vanishing. In con-
trast, as can be seen from the last column of Table 4, the treatment effect
doubles for those participants who had stated a negative attitude towards
the Fridays for Future movement.
At first glance, the strong treatment effect for participants with a neg-
ative a-priori attitude towards Fridays for Future may appear puzzling.
Nonetheless, this result seems plausible, as it is in line with the theories
of moral balancing (Nisan and Horenczyk, 1990) and conscience account-
ing (Gneezy et al., 2014). According to these theories, individuals balance
moral and immoral actions against each other, trying to maintain a sat-
isfactory moral self-image (Ploner and Regner, 2013).4 In other words,
individuals are more likely to behave morally right after having violated
a social norm and vice versa.
As the public debate in Germany was dominated in the survey year
2019 by climate politics, and Fridays for Future received a lot of media cov-
erage, expressing a negative attitude towards the movement may be in-
terpreted as a violation of a social norm. When subjects of the treatment
group are reminded of their violation of the social norm by previously
stating a negative attitude towards Fridays for Future, respondents may
feel guilty and may try to compensate the feeling of guilt by choosing the
morally superior alternative for traveling, that is, the train.
4The theory of moral balancing has been confirmed in numerous empirical studies –
see e. g. Nisan and Horenczyk (1990), Monin and Miller (2001), Ploner and Regner (2013),
Gneezy et al. (2014), as well as Blanken et al. (2015) for a meta-analysis.
21
6.2 Linear Probability Model Estimation Results
To analyze the relationship between the insecurity about one’s attitude
towards Fridays for Future and the likelihood of opting for the train, we es-
timate a linear probability model (LPM), the results of which are reported
in Table 5. For starters, it bears noting that the results of the first spec-
ification, which merely includes the attitude and treatment dummies, as
well as the corresponding interaction terms, exactly mimic the shares of
those participants who choose the train – see Table 4. For instance, adding
the estimates on the coefficients of the treatment dummy and its interac-
tion term with positive attitude, 0.003 and 0.027, respectively, yields pre-
cisely the difference of 0.03 between the treatment and the control group
for those who indicated a positive attitude towards Fridays for Future (Ta-
ble 4). While such results reconfirm the validity of Hypothesis 1, it is also
of note that the coefficient estimates for the treatment dummy and its inter-
action terms with attitude towards Fridays for Future are jointly significant
in all specifications.
With respect to Hypothesis 2, which presumes that respondents who
are less secure about their attitudes towards the movement are more likely
to opt for the train, we find that our insecurity measure is uncorrelated
with travel mode choice. Yet, when interacted with an indicator of having
stated an overall positive attitude, the interaction term turns out to be pos-
itive and statistically significant. This outcome is not surprising, as only
individuals who view themselves as supporters of the Fridays for Future
movement tend to confirm this identity by choosing an action that is in
22
Table 5: Results of a Linear Probability Model (LPM) on the Likelihood to
Choose the Train
(1) (2) (3)
Coeff. Std. Err. Coeff. Std. Err. Coeff. Std. Err.
Positive attitude 0.062** (0.020) 0.070** (0.024) 0.092** (0.028)
Negative attitude -0.087** (0.023) -0.056* (0.028) -0.053 (0.031)
Treatment 0.003 (0.023) 0.017 (0.026) 0.045 (0.029)
Insecurity - - -0.016 (0.015) -0.014 (0.017)
Treatment * Positive attitude 0.027 (0.027) 0.008 (0.031) -0.030 (0.035)
Treatment * Negative attitude 0.075* (0.031) 0.055 (0.036) 0.042 (0.040)
Insecurity * Positive attitude - - 0.038* (0.016) 0.037* (0.018)
Insecurity * Negative attitude - - -0.030 (0.019) -0.025 (0.021)
Insecurity * Treatment - - 0.019 (0.013) 0.012 (0.015)
Constant 0.751** (0.017) 0.739** (0.021) 0.679** (0.049)
Gender No No Yes
Age No No Yes
College degree No No Yes
Employed No No Yes
Income No No Yes
Household size No No Yes
Children No No Yes
Urban/rural No No Yes
Distance to airport No No Yes
Airport type No No Yes
Observations 5,948 5,936 4,728
Adjusted R2 0.019 0.022 0.059
Note: * and ** denote significance at the 5%, and 1% level, respectively. The reference cate-
gory is neutral attitude towards Friday for Future.
line with its demands. In detail, for respondents who have expressed a
positive attitude towards Fridays for Future, a one-standard-deviation in-
crease in the insecurity about this attitude is associated with a 3.7 to 3.8
23
percentage points higher probability of choosing the train than respon-
dents with a neutral attitude (Table 5), thereby supporting Hypothesis 2.
With respect to Hypothesis 3, claiming that respondents who are more
insecure about their attitude towards Fridays for Future react more strongly
to the treatment, we find an interaction term of the treatment dummy and
our insecurity measure that is positive, yet not different from zero in sta-
tistical terms. Thus, Hypothesis 3 is not supported.
Taken together, our results provide support for the interpretation of
pro-environmental behavior as a means of self-signaling: First, the more
an individual cares about her self-image, here because its salience was in-
creased through the identity prime, the more likely she is to act in line
with the values corresponding to her desired self-image, thus reducing
cognitive dissonance. Second, the more uncertain an individual is about
her support for the values in question, the more likely she is to reassure
herself of this identity by choosing a pro-environmental action.
7 Robustness Checks
The fact that students skip classes to attend Fridays for Future protests is
highly controversial and might play an important role for both the respon-
dents’ attitude towards the movement and for the security with respect
to their attitude. In a robustness check, we therefore use items (a) and
(b) of Question A14 to exclude respondents from the analysis who gener-
24
ally support the Fridays for Future movement, but are opposed to skipping
classes.
With respect to Hypothesis 1, claiming that subjects of the treatment
group are more likely to choose the train than subjects of the control group,
the results are robust to this reduction in the sample size, which decreases
by about 1000 observations. The treatment effect of the identity prime
increases to 4.6 percentage points in total (last column of Table 6), rather
than 3.9 points as reported in Table 4, and to 4.8 percentage points for
respondents with a positive attitude towards Fridays for Future .
Table 6: Shares of Respondents who Choose to Travel by Train when excluding
Respondents who Support Fridays for Future, but Oppose Skipping
Classes for Protests
Attitude towards Treatment group Control group Difference between
Fridays for Future # Obs. Share # Obs. Share treatment and control group
Negative attitude 918 0.743 918 0.664 0.078**
Neutral attitude 712 0.754 675 0.751 0.003
Positive attitude 846 0.863 871 1 0.048**
Total 2,476 0.787 2,464 0.741 0.046**
Note: * and ** denote significance at the 5% and 1% level, respectively.
For this subset of respondents, we have also re-estimated the linear-
probability model the results of which are presented in Table 7. These
results are quite similar to those presented in Table 5. A notable exception
is that the interaction term between insecurity and the indicator of hav-
ing stated an overall positive attitude towards the movement decreases in
magnitude and becomes statistically insignificant (Table 7). This suggests
that the positive correlation between our insecurity measure and the pref-
25
erence for the train is mainly driven by respondents who support Fridays
for Future, but do not like the aspect of skipping classes for protests.
Table 7: Results of a LPM on the Likelihood to Choose the Train when ex-
cluding Respondents who Support Fridays for Future, but Oppose
Skipping Classes for Protests
(1) (2) (3)
Coeff. Std. Err. Coeff. Std. Err. Coeff. Std. Err.
Positive attitude 0.064** (0.021) 0.073** (0.026) 0.109** (0.030)
Negative attitude -0.087** (0.023) -0.059* (0.028) -0.052 (0.032)
Treatment 0.003 (0.023) 0.014 (0.027) 0.043 (0.030)
Insecurity - - -0.014 (0.016) -0.013 (0.018)
Treatment * Positive attitude 0.045 (0.029) 0.029 (0.035) -0.017 (0.039)
Treatment * Negative attitude 0.075* (0.031) 0.060 (0.037) 0.047 (0.041)
Insecurity * Positive attitude - - 0.020 (0.018) 0.017 (0.020)
Insecurity * Negative attitude - - -0.030 (0.019) -0.024 (0.021)
Insecurity * Treatment - - 0.015 (0.015) 0.008 (0.017)
Constant 0.751** (0.017) 0.741** (0.022) 0.630** (0.054)
Gender No No Yes
Age No No Yes
College degree No No Yes
Employed No No Yes
Income No No Yes
Household size No No Yes
Children No No Yes
Urban/rural No No Yes
Distance to airport No No Yes
Airport type No No Yes
Observations 4940 4928 3922
Adjusted R2 0.022 0.023 0.063
Note: * and ** denote significance at the 5% and 1% level, respectively. The reference cate-
gory is a neutral attitude towards Fridays for Future. The treatment dummy and its interac-
tion with attitude towards Fridays for Future are jointly significant in all specifications.
26
.5
.4
.3
.2
.1
0
-4 -2 0 2
Insecurity about attitude
Confirmed attitude
Did not confirm attitude
Figure 3: Insecurity about the attitude towards Fridays for
Future by answer to the treatment question with
which subjects either confirmed their attitude to-
wards Fridays for Future or not.
Finally, we take a closer look at the relationship between whether sub-
jects of the treatment group confirm their previously stated attitude to-
wards Fridays for Future and their insecurity about their attitude. Figure 3
illustrates that respondents who are less sure about their attitude towards
the Fridays for Future movement are more likely to confirm their attitude
when answering the treatment question, being consistent with our Hy-
pothesis 2: Respondents who are less sure about their attitude are more
likely to confirm it – either verbally or through identity-confirming behav-
ior.
27
Density
8 Summary and Conclusion
Ever since the notion of identity has been introduced into economic theory
in a seminal paper by Akerlof and Kranton (2000), numerous experimen-
tal studies have demonstrated its impact on behavior (Cohn and Maréchal,
2016). Assuming that individuals are not necessarily aware of their own
moral identity, but may infer it from their past behavior, the literature
on the economics of identity has been complemented by models of self-
signaling (Bénabou and Tirole, 2004, 2006, 2011; Bodner and Prelec, 2003).
In this article, we have developed a straightforward theoretical model
of self-signaling from which we have derived three hypotheses that have
been empirically tested on the basis of an incentivized experiment. It was
embedded in a large-scale survey, in which nearly 6,000 German house-
hold heads had to decide on whether they would prefer to win a voucher
for a train ride or for a flight. Acting pro-environmentally, in the exper-
iment by taking the train rather than the plane for a journey, is hypothe-
sized to be a means of self-signaling.
Exploiting the information about the identification with the environ-
mentalist movement Fridays for Future, half of the sample was randomly
assigned to a treatment group, whose subjects were reminded of their
previously stated attitude towards Fridays for Future immediately before
making their trip decision and were asked to actively confirm or revoke
this attitude. Following the experimental literature that has used priming
to disentangle the impact of identity on behavior from preferences over
28
outcomes, this reminder served as an identity prime that temporarily in-
creases the salience of pro-environmental identity.
We find that, first, respondents who care more about their self-image
because they received a reminder of their previously stated attitude to-
wards the Fridays for Future movement are significantly more likely to
choose the environmentally benign transport mode, here the train rather
than the plane, than respondents who do not receive a reminder. Second,
respondents who are less secure about their true moral identity, evidenced
by a larger variance in their answers to questions about various aspects
of Fridays for Future, are more likely to reassure themselves of their pro-
environmental identity by choosing the train.
Taken together, our results provide support for the interpretation of
pro-environmental behavior as a means of self-signaling: First, the more
an individual cares about her self-image, here because its salience was in-
creased through an identity prime, the more likely she is to act in line with
the values corresponding to her desired self-image, thus reducing cogni-
tive dissonance. Second, the more uncertain an individual is about her
support for the values in question, the more likely she is to reassure her-
self of this identity by choosing a pro-environmental action. In the end,
our empirical results suggest that to increase pro-environmental or, more
generally, pro-social behavior, it may be sufficient to appeal to an indi-
vidual’s self-image so that costly interventions, such as information cam-
paigns, which are designed to convince individuals of new moral princi-
ples, may be unnecessary.
29
A Appendix
A.1 Relevant Survey Questions (Translated from German)
A. General
Question A12: How often do you travel by plane?
– More than 5 times per year
– 3 to 5 times per year
– Once or twice per year
– Less than once per year
– Never
– Don’t know/No answer
Question A13: Currently, students in many cities are protesting for more climate
protection every Friday during and outside school hours. What is your attitude
towards these so-called ”Fridays for Future” protests?
– (1) I don’t like them at all
– (2) I don’t like them
– (3) neither in favor nor I don’t like them
– (4) I’m in favor
– (5) I’m totally in favor
– Don’t know/No answer
Question A14: Please indicate the extent to which you agree with the following
statements about the "Fridays for Future" protests.
Items (randomized):
a. Students have the right to protest for climate protection. However, this
should not be done during school hours, but during their free time.
i
b. Students have the right to protest for climate protection during school hours
as well. Otherwise, they would not get the necessary attention from politi-
cians and the public.
c. The protests have an important political function. In this way, the students
show that the children’s future also depends on the climate policy decisions
of today’s adults and is thus a question of intergenerational justice.
d. The importance of the protests is overrated. Behavioral changes of the stu-
dents for climate protection would be more important and effective.
Scale:
– fully agree
– rather agree
– neither agree nor disagree
– rather disagree
– totally disagree
– don’t know / no answer
KOMP. Experiment
Question KOMP1: Do you plan to take a private trip within the next year using
either the train or the plane? (Note: Please do not consider a trip where you
would take a car (either as a driver or passenger) or a bus).
– Yes
– No
– Don’t know/ No answer
If respondent is in the treatment group FfF:
Question KOMP2: Filtering into different groups depending on answer to Question
A13.
For participants who answered Question A13 with 1 or 2.
ii
Would you agree that, overall, you do not support the "Fridays for Future" move-
ment?
– Yes
– No
– Don’t know/ No answer
For participants who answered Question A13 with 3.
Would you agree that, overall, you have a neutral attitude towards the "Fridays
for Future" movement?
– Yes
– No
– Don’t know/ No answer
For participants who answered Question A13 with 4 or 5.
Would you agree that, overall, you support the "Fridays for Future" movement?
– Yes
– No
– Don’t know/ No answer
If KOMP1 = Yes
In the following, you can win a travel voucher worth 40 euros for your upcoming
trip. The winners will be selected randomly. One out of 50 respondents will
receive a voucher. You can choose whether you would prefer to receive a flight
voucher or a rail voucher for your trip, should you be one of the winners. Flight
vouchers can be redeemed via the Flightgift website with more than 300 different
airlines, while rail vouchers can be redeemed with Deutsche Bahn AG.
If KOMP1 = No
Imagine that you are planning to take a private trip using either the train or the
iii
plane. In the following, you can win a travel voucher worth 40 euros for your
upcoming trip. The winners will be selected randomly. One out of 50 respondents
will receive a voucher. You can choose whether you would prefer to receive a
flight voucher or a rail voucher for your trip, should you be one of the winners.
Flight vouchers can be redeemed via the Flightgift website with more than 300
different airlines, while rail vouchers can be redeemed with Deutsche Bahn AG.
For everyone:
Question KOMP3: If you are among the winners, which voucher do you choose?
– Flight voucher worth 40 euros
– Train voucher worth 40 euros
PV: Psychological Control Variables
The order of questions in this section was randomized
Now we would like to ask you some more questions about the environment in
general.
Question PV3:
To what extent do you personally agree with the following statements?
Items (randomized):
a. I am worried when I think about the environmental conditions our children
and grandchildren will probably have to live in.
b. There are natural limits to growth that our industrialized world has long
since reached or exceeded.
c. Environmental protection should be a priority for Germany, even if it inter-
feres with economic growth.
d. In order to preserve our natural livelihoods, we must all be willing to limit
our standard of living.
SOE. Socioeconomic Characteristics
iv
Finally, please answer a few questions about yourself. Your data will be
treated absolutely confidentially in accordance with the data protection regula-
tions.
Question SO2: What is your highest vocational training or (technical) college
degree?
– No degree
– Apprenticeship or vocational internship of at least 12 months
– Vocational preparation year
– Apprenticeship, vocational training in the dual system
– Preparatory service for the intermediate civil service in public administra-
tion
– Vocational qualification from a vocational school/college, completion of a
1-year school in the healthcare sector
– 2- or 3-year school of health care
– Technical college degree (master craftsman, technician or equivalent de-
gree)
– Vocational academy, technical academy
– Degree from a university of applied sciences
– University of applied sciences degree, also engineering degree
– Degree from a university, scientific college, art college
– Doctorate
– No answer
Question SOE3: Which of the following applies to you? Please select only one
answer option.
– I am employed or working (incl. trainees, persons on parental leave or
partial retirement)
– I am a pupil
– I am a student
– I am a pensioner or retiree
v
– I live from income from capital assets, renting or leasing
– I receive maintenance/allowances from my spouse, partner, parents, rela-
tives or other persons - including persons from outside the household.
– I am a housewife/ husband or take care of children and/or persons in need
of care.
– I receive unemployment benefit I
– I receive unemployment benefit II or social benefits (Hartz IV benefits)
– I receive social welfare or basic income support in old age or in case of
reduced earning capacity
– None of the above options applies to me
– No answer
Question SOE6: What is the total monthly net income of your household? This
is the sum of wages, salaries, income from self-employment, or pensions after
deduction of taxes and social security contributions. Please also include income
from public assistance, income from renting, leasing, housing allowance, child
benefit and other income.
– less than 700 Euro
– 700 to less than 1,200 Euro
– 1,200 to less than 1,700 Euro
– 1,700 to less than 2,200 Euro
– 2,200 to less than 2,700 Euro
– 2,700 to less than 3,200 Euro
– 3,200 to less than 3,700 Euro
– 3,700 to less than 4,200 Euro
– 4,200 to less than 4,700 Euro
– 4,700 to less than 5,200 Euro
– 5,200 to less than 5,700 Euro
– 5,700 Euro and more
– No answer
vi
Question SOE8: In Germany, many people tend to vote for a particular political
party for a long time, although they also vote for different parties from time to
time. How about you: Are you - in general - inclined toward a particular party?
And if so, which one?
– CDU / CSU
– SPD
– AfD
– FDP
– The Left Party
– Bündnis 90 / The Greens
– Another party
– No party
– Don’t know / no answer
A.2 Insecurity Measure
Question A13 elicited respondents’ general attitude towards Fridays for Fu-
ture and Question A14 asked for their agreement with four different state-
ments about the movement. We coded the answers to all these items on a
5-point Likert scale, such that higher values on the scale imply stronger ap-
proval of the movement. The measure on the insecurity of a respondent’s
attitude towards Fridays for Future was then calculated as follows: First,
we took the standard deviation of an individual’s answers to all items of
Questions A13 and A14 to obtain a variable measuring the heterogeneity
of the respondent’s attitude towards different aspects of the movement.
Second, we standardized the obtained variable by subtracting its mean
value across all respondents and dividing by its standard deviation. Thus,
we obtain a standardized insecurity measure, enabling us to interpret the
corresponding regression coefficients in terms of standard deviations of
vii
insecurity.
viii
A.3 Figures
.6
.4
.2
0
1 2 3 4 5
Environmentalist attitude
Positive attitude
Neutral attitude
Negative attitude
Figure A1: General Pro-environmentalist Attitude by Atti-
tude towards Fridays for Future
.4
.3
.2
.1
0
Negative Neutral Positive
Mean value
95% confidence interval
Figure A2: Inclination towards Green Party by Attitude
towards Fridays for Future
ix
Density
Share with inclination towards
green party
A.4 Tables
Table A1: Comparison of the Sample with the German Population
Sample Germany (2019)
Female 0.418 0.505
At least technical college 0.327 0.281
Employed 0.516 0.519
Age:
< 25 years 0.021 0.240
25 - 35 years 0.093 0.127
35 - 45 years 0.130 0.126
45 - 55 years 0.182 0.149
55 - 65 years 0.226 0.150
≥ 65 years 0.348 0.209
Household size:
1 person 0.269 0.423
2 persons 0.473 0.332
3 persons 0.136 0.119
4 persons 0.090 0.091
≥ 5 persons 0.031 0.035
Net household income:
Income < 1700 euros 0.182 0.305
Income 1700 - 3200 euros 0.375 0.361
Income ≥ 3200 euros 0.443 0.334
Data for the German population is drawn from (Destatis, 2020).
x
References
Akerlof, G. A. and Dickens, W. T. (1982). The economic consequences of
cognitive dissonance. American Economic Review, 72(3):307–319.
Akerlof, G. A. and Kranton, R. E. (2000). Economics and identity. Quarterly
Journal of Economics, 115(3):715–753.
American Psychological Association (2021). APA Dictionary of Psychol-
ogy. https://dictionary.apa.org/. Accessed on October 21, 2021.
Aronson, E. (1969). The Theory of Cognitive Dissonance: A Current Per-
spective. In: Advances in Experimental Social Psychology. 4:1–34.
Bem, D. J. (1972). Self-perception theory. In Advances in Experimental Social
Psychology, volume 6, pages 1–62. Elsevier.
Bénabou, R. and Tirole, J. (2004). Willpower and personal rules. Journal of
Political Economy, 112(4):848–886.
Bénabou, R. and Tirole, J. (2006). Incentives and prosocial behavior. Amer-
ican Economic Review, 96(5):1652–1678.
Bénabou, R. and Tirole, J. (2011). Identity, morals, and taboos: Beliefs as
assets. Quarterly Journal of Economics, 126(2):805–855.
Blanken, I., van de Ven, N., and Zeelenberg, M. (2015). A meta-analytic
review of moral licensing. Personality and Social Psychology Bulletin,
41(4):540–558.
Bodner, R. and Prelec, D. (2003). Self-signaling and diagnostic utility in ev-
eryday decision making. The Psychology of Economic Decisions, 1(105):26.
CNN (25.01.2019). Teen activist tells Davos elite they’re to blame
for climate crisis. https://edition.cnn.com/2019/01/25/europe/greta-
xi
thunberg-davos-world-economic-forum-intl/index.html. Accessed on
July 15, 2021.
Cohn, A. and Maréchal, M. A. (2016). Priming in economics. Current
Opinion in Psychology, 12:17–21.
Dana, J., Weber, R. A., and Kuang, J. X. (2007). Exploiting moral wiggle
room: experiments demonstrating an illusory preference for fairness.
Economic Theory, 33(1):67–80.
Destatis (2020). Bevölkerung und Erwerbstätigkeit. Haushalte und Fam-
ilien. Ergebnisse des Mikrozensus. Statistisches Bundesamt. Fach-
serie 1 Reihe 3, https://www.statistischebibliothek.de/ mir/receive/DE-
Heft_mods_00131398.
Dubé, J.-P., Luo, X., and Fang, Z. (2017). Self-signaling and prosocial be-
havior: A cause marketing experiment. Marketing Science, 36(2):161–186.
Festinger, L. (1962). A theory of cognitive dissonance, volume 2. Stanford
University Press.
Gneezy, A., Gneezy, U., Riener, G., and Nelson, L. D. (2012). Pay-what-
you-want, identity, and self-signaling in markets. Proceedings of the Na-
tional Academy of Sciences, 109(19):7236–7240.
Gneezy, U., Imas, A., and Madarász, K. (2014). Conscience account-
ing: Emotion dynamics and social behavior. Management Science,
60(11):2645–2658.
Gosnell, G. K. (2018). Communicating resourcefully: a natural field ex-
periment on environmental framing and cognitive dissonance in going
paperless. Ecological Economics, 154:128–144.
Grossman, Z. (2014). Strategic ignorance and the robustness of social pref-
erences. Management Science, 60(11):2659–2665.
xii
Grossman, Z. (2015). Self-signaling and social-signaling in giving. Journal
of Economic Behavior & Organization, 117:26–39.
Grossman, Z. and Van Der Weele, J. J. (2017). Self-image and willful ig-
norance in social decisions. Journal of the European Economic Association,
15(1):173–217.
Matthey, A. and Regner, T. (2011). Do I really want to know? A cogni-
tive dissonance-based explanation of other-regarding behavior. Games,
2(1):114–135.
Monin, B. and Miller, D. T. (2001). Moral credentials and the expression of
prejudice. Journal of Personality and Social Psychology, 81(1):33.
Nisan, M. and Horenczyk, G. (1990). Moral balance: The effect of prior be-
haviour on decision in moral monflict. British Journal of Social Psychology,
29(1):29–42.
Ploner, M. and Regner, T. (2013). Self-image and moral balancing: An ex-
perimental analysis. Journal of Economic Behavior & Organization, 93:374–
383.
Spiegel (29.07.2019). Greta Thunberg segelt in die USA.
https://www.spiegel.de/lebenundlernen/schule/greta-thunberg-
segelt-ueber-den-atlantik-in-die-usa-a-1279508.html. Accessed on July
15, 2021.
Tonin, M. and Vlassopoulos, M. (2013). Experimental evidence of self-
image concerns as motivation for giving. Journal of Economic Behavior &
Organization, 90:19–27.
xiii