Doing good vs.

avoiding bad in prosocial choice: A refined

test and extension of the morality preference hypothesis

Ben M. Tappin 1* and Valerio Capraro 2

1
Royal Holloway, University of London, 2Middlesex University London

*Corresponding author: [email protected]

Journal of Experimental Social Psychology, forthcoming

Abstract
Prosociality is fundamental to human social life, and, accordingly, much research
has attempted to explain human prosocial behavior. Capraro and Rand (Judgment
and Decision Making, 13, 99-111, 2018) recently provided experimental evidence
that prosociality in anonymous, one-shot interactions (such as Prisoner’s Dilemma
and Dictator Game experiments) is not driven by outcome-based social preferences
– as classically assumed – but by a generalized morality preference for “doing the
right thing”. Here we argue that the key experiments reported in Capraro and Rand
(2018) comprise prominent methodological confounds and open questions that
bear on influential psychological theory. Specifically, their design confounds: (i)
preferences for efficiency with self-interest; and (ii) preferences for action with
preferences for morality. Furthermore, their design fails to dissociate the preference
to do “good” from the preference to avoid doing “bad”. We thus designed and
conducted a preregistered, refined and extended test of the morality preference
hypothesis (N=801). Consistent with this hypothesis, our findings indicate that
prosociality in the anonymous, one-shot Dictator Game is driven by preferences for
doing the morally right thing. Inconsistent with influential psychological theory,
however, our results suggest the preference to do “good” was as potent as the
preference to avoid doing “bad” in this case.

Author note: BMT gratefully acknowledges funding from the Economic and Social
Research Council. We thank Patrick Heck for comments on an earlier draft.
 Introduction
People often pay costs to benefit others; they behave prosocially.
Fundamental to human social life (Fehr & Gächter, 2002; Gintis et al, 2003; Nowak,
2006; Tomasello, 2014), prosocial behavior is often explained by appeal to
reciprocity. If I pay a cost to help you today, you – or others who learn about my
behavior – are more likely to help me tomorrow (Nowak & Sigmund, 2005; Rand &
Nowak, 2013; Trivers, 1971). Defying explanations of this kind, however, prosocial
behavior is frequently observed in contexts where opportunities for reciprocity are
absent. For example, in anonymous, one-shot interactions, individuals often forego
some amount of self-interest to the benefit of strangers (Camerer, 2003).
Behavioral economists have classically sought to explain such behavior by
assuming that individuals have preferences for minimizing inequity or maximizing
efficiency (i.e., social welfare) (Bolton & Ockenfels, 2000; Capraro, 2013; Charness &
Rabin, 2002; Engelmann & Strobel, 2004; Fehr & Schmidt, 1999; Levin, 1998).
According to these influential frameworks, prosocial individuals derive utility –
psychological benefit – from particular social outcomes; thus, realizing those
outcomes offsets the cost of behaving prosocially.
A recent alternative perspective is that individuals derive utility from
performing actions they perceive to be morally right (Bicchieri, 2005; DellaVigna et
al., 2012; Huck et al., 2012; Krupka & Weber, 2013). This perspective accords with
evidence from social psychology that individuals derive utility from seeing
themselves in a positive moral light (Aquino & Reed, 2002; Dunning, 2007) and, in
addition, that prosocial individuals in particular view opportunities for prosocial
action in moral terms; for example, by considering what the morally “right” action is
(Liebrand et al., 1986; Weber et al., 2004).
Building on these converging lines of evidence, recent experimental work
advanced the hypothesis that a generalized morality preference – rather than
preferences for minimizing inequity or maximizing efficiency per se – drives
prosocial behavior in anonymous, one-shot interactions (Capraro & Rand, 2018). In
other words, that a simple preference for doing (what is perceived to be) the morally
“right” thing underpins individuals’ prosociality in these contexts.
In their key experiments, Capraro and Rand (2018) used a “Trade-Off Game”
(TOG) to empirically dissociate the hypothesized morality preference from outcome-
based social preferences for equity and efficiency. In the TOG, participants made a
unilateral choice about how to allocate money between themselves and two other
(passive) people. While one choice minimized inequity – all participants earned the
same amount – the other choice maximized efficiency – participants earned
different amounts, but, together, the group earned more. This design effectively
pitted preferences for equity and efficiency against one another; creating a decision
context where the morally “right” choice was ambiguous. The researchers found
that, framing either choice as the morally appropriate one dramatically affected
participants’ choices, such that the majority chose the option framed as morally
appropriate; be that the equitable or efficient choice.
To support the inference that these moral considerations drive prosociality,
however, required additional evidence. To that end, participants also completed, in
addition to the TOG, a canonical prosocial choice task; either the Dictator Game
(DG), or the Prisoner’s Dilemma (PD). In the latter tasks, participants made a
unilateral choice about how much money to donate to a new (passive) person (DG),

2
or a simultaneous bilateral choice whether to cooperate with a new person (PD),
respectively.
The key finding in Capraro and Rand (2018) was that participants who made
the choice framed as morally appropriate in the TOG – be that the equitable choice
or the efficient choice – were consistently more prosocial in the DG and PD;
donating and cooperating (respectively) more than participants who chose
otherwise in the TOG. Crucially, this result is inconsistent with stable outcome-
based preferences for equity or efficiency as explanations for prosociality, which do
not predict an association between moral framing in the TOG and prosociality in a
different task, such as the DG/PD. The result is instead consistent with the morality
preference hypothesis, which predicts that individuals sensitive to which choice is
morally right in the TOG – as revealed by the moral framing of those choices – are
also revealed to be more prosocial in the DG/PD; where, in contrast to the TOG, the
morally right choice is unambiguous (Krueger & Acevedo, 2007; Krueger &
DiDonato, 2010).
The implication of Capraro and Rand’s (2018) findings is important: They
suggest their data renders the classic approach to understanding prosocial choice
through social preferences insufficient and, in particular, that an account based on a
fluid preference for “doing the morally right thing” is superior. However, their key
evidence derives from an experimental design that contains several prominent
methodological confounds, and leaves open important questions regarding the
mechanism of the hypothesized morality preference. Below we expand on these
issues.

Self-interest
Consider the choice outcomes in the TOG. The equitable choice always
provided the participants – the chooser, and two passive recipients – the same
allocation; 13 Monetary Units (MU) each. The efficient choice, in contrast, always
provided the chooser with 15 MU, and the passive recipients 23 MU and 13 MU,
respectively. Thus, while the efficient choice clearly results in greater overall gains
for the group – at the cost of equity, as intended – it also results in greater gains for
the chooser themselves. In other words, the choice option meant to reveal a
preference for efficiency is confounded with self-interest. A plausible consequence
of this confound is an overestimate of the proportion of individuals with a preference
for efficiency. An overestimation of this kind may have affected the key result – an
association between TOG choice and prosociality in the DG/PD – in two ways.
First, it may have inflated the association between TOG choice under the
equitable-is-moral frame, and prosociality in the DG/PD. Specifically, this
association may not have been driven by participants with a genuine morality
preference – who choose the equitable option under this TOG frame, and the
prosocial option in the DG/PD – but, rather, by self-interested participants – who
choose the efficient option under this TOG frame, and the self-interested option in
the DG/PD. Indeed, in the worst case, the behavior of self-interested participants
could fully account for the observed association between TOG choice under the
equitable-is-moral frame, and prosociality in the DG/PD.
Second, by the opposite logic, the overestimation of individuals with a
preference for efficiency may have deflated the association between TOG choice
under the efficient-is-moral frame, and prosociality in the DG/PD. This is because

3
some participants making the efficient choice under that TOG frame did so not
because of a general morality preference nudged by the framing, but, rather, for
their own self-interest. Crucially, these participants would not have chosen
prosocially in the DG/PD, thereby deflating the observed association between the
two choices.
These issues directly affect the key evidence – an association between TOG
choice and prosociality in the DG/PD – supporting the morality preference
hypothesis. A remedy to these issues is to remove self-interest from the equation by
design.

Action-inaction asymmetry
Not only do the efficient-is-moral and equitable-is-moral frames differ in the
labels used to describe the two choice options, but, in addition, they differ in which
is the active choice and which is the passive choice. Specifically, in the efficient-is-
moral frame, participants start with an equitable allocation (13 MU each), while in
the equitable-is-moral frame they start with an efficient allocation (15, 23, and 13
MU, respectively). In other words, the moral choice is always framed as an active
choice to change these initial allocations. Choice frame is thus confounded with
active/passive frame.
A substantial body of work in social, moral, and decision-making psychology
indicates that humans perceive inaction differently than action (Baron & Ritov, 2004;
Spranca et al., 1991). For example, regret is greater for actions that lead to negative
outcomes than for inactions that lead to the same negative outcomes (Feldman &
Albarracín, 2017; Zeelenberg et al., 2002); individuals are biased towards
maintaining the status quo in decision-making (Samuelson & Zeckhauser, 1988);
and, in moral judgment, harms caused by action are considered worse than the
same harms caused by inaction (Cushman et al., 2006). Finally, most relevant here,
action framing influences engagement in prosocial behavior (Teper & Inzlicht, 2011),
and there is considerable variation in who exhibits action-inaction asymmetries
(Baron & Ritov, 2004).
Given this evidence, it is probable that the confounding of choice frame with
active/passive frame over- or under-estimated the proportion (and types) of
individuals choosing the morally-framed option in the TOG; with unknown
consequences for the key association between TOG choice and prosociality in the
DG/PD. Decoupling these frames is necessary to make clear inferences about the
effect of choice frame in the TOG.

Doing good vs. avoiding bad

An influential hypothesis in social psychology is that immoral, negative, or
otherwise “bad” stimuli weigh more heavily than their “good” counterparts in human
cognition and behavior (Baumeister et al., 2001; Rozin & Royzman, 2001; Vaish et
al., 2008; see Corns, 2018 for a recent critique).
Consistent with this hypothesis, recent evidence suggests that “self-
righteousness” – manifested in, for example, the average person rating themselves
morally superior to the average person (Tappin & McKay, 2017) – are greater for
immoral than moral stimuli (Klein & Epley, 2016, 2017). Relatedly, the correlation
between individuals’ life satisfaction and their self-perception is reportedly stronger
if the latter is computed as the distance between individuals’ “real” and “undesired”

4
selves vs. between their “real” and “desired” selves (Ogilvie, 1987). In other words,
those data suggest the type of person individuals want to avoid being weighs more
heavily (in their life appraisal) than the type of person they would ideally like to be. A
similar asymmetry manifests in the psychology of moral regulation. In particular, in
the distinction between proscriptive morality – what we should do and be – and
prescriptive morality – what we should avoid doing and being. Whereas the former
is considered discretionary and a matter of personal preference, the latter is
considered mandatory and strict (Janoff-Bulman et al., 2009).
To implement the choice framing in the TOG, the two choice options were
jointly framed as moral and immoral, respectively. Individuals choosing the morally-
framed option may thus have been motivated by a preference to do “good” (e.g., a
desire to be moral), or motivated by a preference to avoid “bad” (e.g., an aversion to
being immoral). These distinct preferences are confounded in the TOG design.
Given the preceding evidence, it is plausible that individuals’ choices were
motivated more by a preference to avoid “bad” than to do “good”. Furthermore,
assuming this hypothesis, a further plausible hypothesis is that participants who
were motivated by a preference to do “good” (vs. avoid “bad”) in the TOG were
more likely to behave prosocially in the DG/PD. For example, because the
preference to avoid bad may reflect a general desire to avoid punishment, whereas
a preference to do good may reflect a desire to do good for its own sake. That is,
the latter preference is more diagnostic of true prosocial motivation.

The current study

Here we address the methodological confounds and open theoretical
questions in Capraro and Rand (2018), and, thus, provide a refined and extended
test of the morality preference hypothesis. To do so, we design and implement an
improved Trade-Off Game (TOG), and test for an effect of choice frame on TOG
choice (Hypothesis 1), and for an association between framing of the TOG and
prosociality in a different task, the DG (Hypothesis 3). We also test two novel
hypotheses bearing on existing psychological theory. First, that the effect of choice
frame on TOG choice is greater under an avoid “bad” than do “good” moral frame
(Hypothesis 2) (cf. Baumeister et al., 2001; Rozin & Royzman, 2001; Vaish et al.,
2008). Second, that the association between choosing the morally-framed option in
the TOG, and prosociality in the DG, is greater under a do “good” than avoid “bad”
moral frame (Hypothesis 4).

Methods

The hypotheses, design, sampling and analysis plan were preregistered on

the Open Science Framework (protocol: https://osf.io/9nphs/).

Participants
We sought to collect N=200 participants per treatment, giving a total N=800.
We determined this sample size by multiplying the N-per-treatment in Capraro and
Rand (2018) study 3 by 1.5x; the study most conceptually similar to that which we
reproduce here. Sensitivity power analyses (reported in SI) for our key hypothesis
tests indicated we had sufficient power (>.80) to detect standardized effect sizes
conventionally considered small (r =.10). A total of N=801 participants completed

5
the study. Participants were recruited online via Amazon’s Mechanical Turk (AMT)
(for the validity of AMT, see e.g., Arechar et al., 2018; Horton et al., 2011; Paolacci
et al., 2014; Thomas & Clifford, 2017), and were located in the US at the time of
taking part. All participants provided informed consent. This study was reviewed
and approved by Middlesex University, and Royal Holloway, University of London
ethics procedures.

Procedure
Participants began by playing a Dictator Game (DG). In the DG, they were
given $0.10 and they had to decide how much, if any, to give to another anonymous
participant who received no starting money allocation. Participants could donate in
increments of $0.01; from $0.00 to $0.10. The participant was informed that the
other person had no active choice and would only receive what they decide to give.
We asked two comprehension questions to ensure that participants understood the
payoff structure of the DG prior to their decision. Specifically, we asked which
choice (1) maximized their own payoff, and which choice (2) equalized their payoff
with that of the other person. Participants who failed either or both comprehension
questions were prevented from completing the survey (this condition was made
explicit in the consent form). Those who passed the comprehension questions were
then asked to make their DG decision.
Following the DG, participants played an improved Trade-Off Game (TOG). In
this TOG, participants (“choosers”) had to decide between two choice options that
affected their own payoff and the payoff of two other people; the latter being
passive recipients who did not make any choices. One option was “equitable”, in
the sense that it minimized payoff differences among the three participants;
specifically, they each earned $0.13. The other option was “efficient”, in the sense
that it maximized the sum of the payoffs of the three participants; specifically, the
chooser earned $0.13, while the other two people earned $0.23 and $0.13,
respectively. Importantly, in this improved TOG design, because the chooser earns
$0.13 by making either choice, the confounding of self-interest with preferences for
efficiency is eliminated. Furthermore, because participants are not told that one or
the other state of money distribution ([13, 13, 13] or [13, 23, 13]) initially holds, both
choice options are rendered equal in terms of active/passive frame.
Before reading the TOG instructions, participants were randomly assigned to
one of four versions of the TOG, each corresponding to a particular framing
combination in a 2x2 between-subjects design:

•   TOG frame: Give – Do Good

•   TOG frame: Give – Avoid Bad
•   TOG frame: Equalize – Do Good
•   TOG frame: Equalize – Avoid Bad

We experimentally manipulate whether the efficient (“Give”) or equitable

(“Equalize”) choice is framed as morally appropriate (choice frame1: Give, Equalize).

1
Following Capraro & Rand (2018), in our preregistered protocol and analysis script we labelled the
efficient-is-moral frame the “Give” frame, and the equitable-is-moral frame the “Equalize” frame. For
consistency, we follow that convention here.

6
Furthermore, we also manipulate whether the moral framing emphasizes doing
“good” or avoiding “bad” (moral frame: Do Good, Avoid Bad):

•   Under the Give – Do Good frame, the efficient option is labelled “be
generous, Option 1”, and the equitable option is “Option 2”
•   Under the Give – Avoid Bad frame, the efficient option is labelled “Option 2”,
and the equitable option is “be ungenerous, Option 1”
•   Under the Equalize – Do Good frame, the efficient option is labelled “Option
2”, and the equitable option is “be fair, Option 1”
•   Under the Equalize – Avoid Bad frame, the efficient option is labelled “be
unfair, Option 1”, and the equitable option is “Option 2”

Importantly, notice that the experimental manipulation of moral frame

decouples the preference to do “good” from the preference to avoid “bad”. After
making their decision in the TOG, participants provided standard demographic
information, at the end of which they were given the completion code needed to
submit the survey on AMT. After the end of the survey, we downloaded the data file
and computed the bonuses, which were paid on top of the base participation fee
received by all participants ($0.50). No deception was used. We refer to the SI for
verbatim experimental instructions (available here: https://osf.io/m7w2s/). We report
all measures, manipulations and exclusions.

Results

Data analysis was conducted in R (v.3.4.0, R Core Team, 2017) using

RStudio (v.1.1.423, RStudio Team, 2016). R packages used in analysis and figures:
ggplot2 (v.2.2.1, Wickham, 2009), plyr (v.1.8.4, Wickham, 2011), dplyr (v.0.7.4,
Wickham et al., 2017), reshape (v.0.8.7, Wickham, 2007), gridExtra (v.2.3, Auguie &
Antonov, 2017), effsize (v.0.7.1, Torchiano, 2017), data.table (v.1.10.4-3, Dowle et
al., 2017). The raw data and code to reproduce all results and figures in this paper
are available at https://osf.io/x5stj/.

Data exclusions
N=288 (26.45%) participants answered one or more of the comprehension
questions incorrectly, or did not answer these questions, and were thus prevented
from completing the study (following our preregistered protocol). Of the remaining
N=801 participants who completed the study, there were N=15 (1.87%) duplicate
responses according to participants’ IP address/unique Mechanical Turk ID. In line
with our preregistered protocol, we excluded these duplicates, retaining the earliest
responses only—defined by the date/time they began the study. Finally, after these
exclusions, we identified N=2 (0.25%) participants that dropped out of the study
prior to making their decision in the TOG, and are thus unable to be included in the
analysis (leaving N=784 for analysis).

Hypotheses 1 & 2

Preregistered analyses

7
 We first test whether participants were more likely to choose the efficient
option in the TOG under the “give” choice frame than under the “equalize” choice
frame (Hypothesis 1). We then test whether this framing effect was stronger under
the “avoid bad” moral frame than under the “do good” moral frame (Hypothesis 2).
To that end, we fit a binomial logistic regression model with two dummy-coded
treatment variables as predictors: choice frame [0=equalize frame, 1=give frame]
and moral frame [0=avoid bad, 1=do good], and choice in the TOG as the
dependent variable [TOG choice: 0=equitable choice, 1=efficient choice].
Consistent with Hypothesis 1, choice frame predicted TOG choice in the
expected direction, Odds Ratio (OR) = 3.55, 95% CI [2.34, 5.40], Z = 5.94 p <.001.
A majority of participants chose the efficient option under the give choice frame
(69.21%), whereas only a minority of participants chose this option under the
equalize choice frame (39.90%).
Inconsistent with Hypothesis 2, there is no statistically significant interaction
between choice frame and moral frame, OR = 0.91 [0.50, 1.64], Z = -0.32, p =.752.
In other words, the effect of choice frame appeared largely independent of whether
the choice was framed as “doing good” or “avoiding bad”. Both H1 and H2 results
remain similar after adjusting for age, gender [not female=0, female=1] and
education [0=less than college, 1=college or above] in the model: Main effect of
choice frame OR = 3.74 [2.45, 5.72], Z = 6.11, p <.001; interaction between choice
frame and moral frame OR = 0.89 [0.49, 1.61], Z = -0.39, p =.697. The proportion of
choices in each of the four treatments is displayed in Figure 1.

8
Figure 1. Proportion of efficient and equitable choices as a function of
treatment. Error bars are ± 1 SEM.

Hypothesis 3

Preregistered analyses
Next, we test the hypothesis that participants who make the ‘moral’ choice in
the TOG – that is, choose the efficient option under the “give” frame, or choose the
equitable option under the “equalize” frame – donate more to their partner in the DG
(Hypothesis 3). We fit a linear regression model with two dummy-coded variables as
predictors: TOG choice frame [0=equalize frame, 1=give frame] and the choice the
participant made in the TOG [0=equitable choice, 1=efficient choice], respectively.
The DV is amount donated in the DG [from 0 to 10].
Consistent with Hypothesis 3, there is an interaction in the predicted
direction, b = 1.68, SE = 0.40, t = 4.17, p <.001. Under the equalize frame,
participants who made the equitable choice donated more in the DG (M = 3.40, SD

9
= 2.60) than participants who made the efficient choice (M = 2.38, SD = 2.57), t
(389) = 3.81, p <.001, hedges’ g = 0.39, 95% CI [0.19, 0.60]. This pattern was
reversed under the give frame. There, participants who made the equitable choice
donated less in the DG (M = 2.53, SD = 2.75) than participants who made the
efficient choice (M = 3.19, SD = 2.80), t (391) = -2.19, p =.029, hedges’ g = -0.24 [-
0.45, -0.02]. (Note: All t-tests are post-hoc.) This interaction between choice frame
and TOG choice remains similar after adjusting for age, gender, and education, b =
1.67, SE = 0.40, t = 4.14, p <.001. The interaction pattern is displayed in Figure 2
(panel A). As seen in Figure 2, DG donations follow an approximately bimodal
distribution peaking over donations of 0 and 5. We thus conducted exploratory
analyses to test the robustness of the preceding linear regression results (reported
in SI). These were consistent with the linear regression models.

10
Figure 2. Violin plots (A) and boxplots (B) of DG donations as a function of
choice frame, and the choice the participants made, in the TOG. A, points
denote the mean values and error bars are ± 1 SEM. B, bolded centre lines denote
the median values.

Hypothesis 4

Preregistered analyses
In our final preregistered analysis, we test whether the difference in DG
donations between participants who made the ‘moral’ vs. ‘non-moral’ choice in the
TOG is larger under the “do good” frame than under the “avoid bad” frame
(Hypothesis 4). In line with our preregistered protocol, to simplify this analysis, we
collapse across two variables: TOG choice frame [equalize, give], and the TOG
choice the participant made [efficient, equitable]. This provides a new binary
variable denoting whether the participant made the moral choice in the TOG [0=no,
1=yes], where the moral choice is simply defined as either the efficient option under
the “give” frame, or the equitable option under the “equalize” frame. We then fit a
linear regression model with two variables as predictors: moral choice [0=no,
1=yes], and moral frame [0=avoid bad, 1=do good]. As before, the DV is amount
donated in the DG [0-10].
Inconsistent with Hypothesis 4, there is no statistically significant interaction
between moral choice and moral frame on DG donations, b = -0.32, SE = 0.40, t = -
0.79, p =.429. In other words, while participants who made the moral choice in the
TOG tended to donate more in the DG than participants who made the non-moral
choice (i.e., Hypothesis 3), this effect appeared largely independent of whether the
participants made their TOG choice under the moral frame of “doing good” or
“avoiding bad”. Adjusting for age, gender, and education in the model did not
meaningfully change this result, b = -0.33, SE = 0.40, t = -0.81, p =.418 (see SI for
robustness checks).

Discussion
Converging evidence suggests that prosociality in anonymous, one-shot
interactions is not solely motivated by outcome-based social preferences, but that it
is also motivated by what individuals perceive to be the morally right action
(Bicchieri, 2005; DellaVigna et al., 2012; Eriksson et al, 2017; Kimbrough &
Vostroktunov, 2016; Krupka & Weber, 2013); perhaps serving to maintain a positive
moral self-image (Aquino & Reed, 2002; Dunning, 2007). Building on this work,
recent experimental evidence advanced the hypothesis that a generalized morality
preference drives prosocial behavior in anonymous, one-shot interactions like that
in Dictator and Prisoner’s Dilemma games (Capraro & Rand, 2018). This hypothesis
rejects the classic view in behavioral economics that prosociality in these situations
is driven by social preferences for equity and efficiency.
Here we identified prominent methodological confounds and open theoretical
questions in the key experiments reported in Capraro and Rand (2018). In particular,
their Trade-Off Game (TOG) design (i) confounds preferences for efficiency with

11
self-interest, and (ii) preferences for action with preferences for morality. Moreover,
the design fails to dissociate preferences to do “good” from preferences to avoid
doing “bad”. It is highly likely these issues affected the observed association
between choice in the TOG, and prosociality in the DG/PD; the key evidence for the
morality preference hypothesis. Likewise, the failure to decouple the preference to
do “good” from that to avoid doing “bad” leaves the mechanism of the proposed
morality preference unclear, and misses a key prediction from influential
psychological theory; that “bad is stronger than good” (Baumeister et al., 2001;
Rozin & Royzman, 2001; Vaish et al., 2008).
To address these issues, we designed and implemented an improved
TOG/DG experiment – eliminating the confounds identified in the original
experiments reported in Capraro and Rand (2018). In doing so, we replicated the
key results in support of the morality preference hypothesis. We found that framing
one or the other TOG choice as morally appropriate – by labelling the focal choice
“fair” or “generous”, or the counterpart choice “unfair” or “ungenerous” – strongly
affected individuals’ choices. Specifically, approximately 70% of individuals chose
the efficient option when that choice was framed as morally appropriate, dropping
to 40% when the equitable choice was framed as morally appropriate; a swing of
30%, and a reverse in the majority decision. More importantly, we found that
individuals who chose the morally appropriate option in the TOG – be that the
efficient option or the equitable option – were more prosocial in the preceding DG;
donating more money to a stranger. This result was robust to various analytic
specifications, and provides evidence that prosociality (in the DG) is driven by a
preference for doing what is perceived to be morally right.
Our results lend experimental support to various alternatives to outcome-
based preference models (e.g., Alger & Weibull, 2013; Brekke et al., 2003;
Kimbrough & Vostroktunov, 2016; Krupka & Weber, 2013; Lazear et al., 2012; Levitt
& List, 2017). In particular, these models assume that individuals have moral
preferences that guide their prosocial decision-making in unilateral interactions; an
assumption consistent with the current findings. A parallel class of models has
sought to explain prosocial decision-making using an intention-based framework,
according to which people are sensitive to others’ intentions (Falk et al., 2008;
McCabe et al., 2003; Rabin, 1993). These models have been useful in explaining
prosociality observed in interactions with more than one active player. However,
they are of limited use in the case of unilateral interactions – which are the focus of
the current study – where beliefs about the intentions of others do not apply.
Reputation-based models (e.g., Heck & Krueger, 2017; Jordan et al., 2016; Nowak
& Sigmund, 2005) also appear of limited use for our results because choices were
anonymous and one-shot.
Our findings are consistent with work in social psychology that suggests
individuals are motivated to maintain a positive moral self-image (Dunning, 2007).
Indeed, an open question at the intersection of this research is whether individuals
who assign greater value to a moral self-image (Aquino & Reed, 2002) show
stronger choice framing effects in the TOG, and/or a stronger association between
TOG choice and prosociality in the DG/PD. That said, we note that it is unlikely
individuals are solely motivated by what they perceive to be the right thing,
especially across different choice contexts. In line with Capraro and Rand (2018),
we do expect that these results extend to the PD. There is some recent evidence,

12
however, that behavior in other economic games is driven by outcome-based
preferences, and not by a general morality preference (Capraro, 2018). An
interesting avenue for future work is thus to further explore the boundary conditions
of the morality preference account. One avenue might be situations in which people
have to trade-off conflicting moral principles, like in the case of “altruistic” lying; that
is, lying to benefit others (Biziou-van-Pol et al., 2015; Erat & Gneezy, 2012). An open
empirical question is whether and how preferences for morality – as revealed by
choice in the TOG – predict moral trade-offs of this kind.
We found that a moral frame emphasizing “good” affected TOG choice as
strongly as a moral frame emphasizing “bad”. For example, the proportion of
individuals switching from the efficient choice to the equitable choice was
essentially identical (approx. 30%) whether the latter choice was labelled “fair” or
the former choice “unfair”. We thus found little evidence that “bad” (framing) was
stronger than “good” (framing) in prosocial choice (cf. Baumeister et al., 2001).
An explanation for this discrepancy is found in recent work that has criticized
the dominance of “bad” over “good” (termed the negativity bias) on theoretical and
empirical grounds (Corns, 2018). A particular criticism concerns the credible
alternative explanations for much of the evidence base. For example, asymmetries
in perception that are taken to support the stronger effect of negatively-valenced
stimuli may instead be explained by differences in the informativeness of negative
vs. positive stimuli (Corns, 2018). In research on impression formation, as a case in
point, the greater weight assigned to immoral (vs. moral) traits may be explained by
the fact that immoral traits tend to be more informative of others’ character
(Kellermann, 1984). In the case of self-perception, in contrast, it is likely this
informativeness bias does not hold; for example, because individuals can introspect
on their full “moral history”, unlike when they are judging other people. Absent a
difference in informativeness, immoral and moral traits may yield similar impact on
judgment and behavior.
This provides a plausible explanation for our results, and is consistent with a
recent critique of the negativity bias hypothesis (Corns, 2018). This explanation also
clarifies why the observed association between choosing the morally-framed option
in the TOG, and prosociality in the DG, was similar under the “good” vs. “bad”
moral frame (i.e., rejection of H4). Specifically, because the frames were equally
motivating, individuals choosing the morally-framed option in either frame were
equally liable to donate more in the DG. It is important to highlight that our “bad”
frame label was symmetric vis-à-vis our “good” frame label (e.g., “unfair” vs. “fair”,
respectively). Had we used frames with stronger, but asymmetric labels – such as
“steal” in place of “unfair” – we may have elicited an asymmetric moral frame effect
in line with the negativity bias hypothesis. As noted above, however, in that case it
may have been unclear whether the moral frame asymmetry was due to the valence
of the frames per se (i.e., bad vs. good), or other differences such as asymmetric
strength or salience of the labels (cf. Corns, 2018).
Finally, our results add to a growing body of research on framing effects in
prosocial choice. Prior work finds that situational labels sometimes impact prosocial
decision-making (e.g., Capraro & Vanzo, 2018; Kay & Ross, 2003; Krupka & Weber,
2013; Larrick & Blount, 1997; Liberman et al., 2004), but not always (Dreber et al.,
2013; Goerg et al., 2017). This suggests the effect of labels may depend on the
specific interaction or game type. Our findings replicate recent work showing that

13
labels are highly effective in the Trade-Off Game (Capraro, 2018; Capraro & Rand,
2018).
In summary, recent experimental work advanced the hypothesis that
prosociality in anonymous, one-shot interactions is not driven by outcome-based
social preferences for equity or efficiency per se, but by a generalized morality
preference for “doing the right thing” (Capraro & Rand, 2018). We identified
prominent methodological confounds and open theoretical questions in this work,
and, consequently, conducted a refined and extended test of the morality
preference hypothesis. Consistent with this hypothesis, our findings indicate that
prosociality in the anonymous, one-shot Dictator Game is driven by preferences for
doing the morally right thing. Furthermore, consistent with a recent critique of the
negativity bias hypothesis, our results suggest the preference to do “good” was as
potent as the preference to avoid doing “bad” in this case.

References
Alger, I., & Weibull, J. W. (2013). Homo moralis—preference evolution under
incomplete information and assortative matching. Econometrica, 81(6), 2269-
2302.
Aquino, K., & Reed II, A. (2002). The self-importance of moral identity. Journal of
Personality and Social Psychology, 83, 1423-1440.
Arechar, A. A., Gächter, S., & Molleman, L. (2018). Conducting interactive
experiments online. Experimental Economics, 21, 99-131.
Auguie, B, & Antonov, A. (2017). Package ‘gridExtra’. Available at https://cran.r-
project.org/web/packages/gridExtra/gridExtra.pdf
Baron, J., & Ritov, I. (2004). Omission bias, individual differences, and normality.
Organizational Behavior and Human Decision Processes, 94, 74-85.
Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is
stronger than good. Review of General Psychology, 5, 323-370.
Bicchieri, C. (2005). The grammar of society: The nature and dynamics of social
norms. Cambridge University Press.
Biziou-van-Pol, L., Haenen, J., Novaro, A., Occhipinti-Liberman, A., & Capraro, V.
(2015). Does telling white lies signal pro-social preferences? Judgment and
Decision Making, 10, 538-548.
Bolton, G. E., & Ockenfels, A. (2000). ERC: A theory of equity, reciprocity, and
competition. The American Economic Review, 90, 166-193.
Brekke, K. A., Kverndokk, S., & Nyborg, K. (2003). An Economic Model of Moral
Motivation. Journal of Public Economics, 87, 1967-1983.
Camerer, C. F. Behavioral game theory: Experiments in strategic interaction
Princeton, NJ: Princeton University Press (2003).
Capraro, V. (2013). A model of human cooperation in social dilemmas. PLoS ONE,
8, e72427.
Capraro, V. (2018). Social versus Moral preferences in the Ultimatum game: A
theoretical model and an experiment. Available at SSRN.

14
Capraro, V., & Rand, D. G. (2018). Do the right thing: Experimental evidence that
preferences for moral behavior, rather than equity or efficiency per se, drive
human prosociality. Judgment and Decision Making, 13, 99-111.
Capraro, V., & Vanzo, A. (2018). Understanding moral preferences using sentiment
analysis. Available at SSRN.
Charness, G., & Rabin, M. (2002). Understanding social preferences with simple
tests. The Quarterly Journal of Economics, 117, 817-869.
Corns, J. (2018). Rethinking the Negativity Bias. Review of Philosophy and
Psychology, 1-19.
Cushman, F., Young, L., & Hauser, M. (2006). The role of conscious reasoning and
intuition in moral judgment: Testing three principles of harm. Psychological
Science, 17, 1082-1089.
DellaVigna, S., List, J. A., & Malmendier, U. (2012). Testing for altruism and social
pressure in charitable giving. The Quarterly Journal of Economics, 127, 1-56.
Dowle, M., et al. (2017). Package ‘data.table’. Available at https://cran.r-
project.org/web/packages/data.table/data.table.pdf
Dreber, A., Ellingsen, T., Johannesson, M., & Rand, D. G. (2013). Do people care
about social contexts? Framing effects in dictator games. Experimental
Economics, 16, 349-371.
Dunning, D. (2007). Self‐image motives and consumer behavior: how sacrosanct
self‐beliefs sway preferences in the marketplace. Journal of Consumer
Psychology, 17, 237-249.
Engel, C. (2011). Dictator games: A meta study. Experimental Economics, 14, 583-
610.
Engelmann, D., & Strobel, M. (2004). Inequality aversion, efficiency, and maximin
preferences in simple distribution experiments. The American Economic
Review, 94, 857-869.
Erat, S., & Gneezy, U. (2012). White lies. Management Science, 58, 723-733.
Eriksson, K., Strimling, P., Anderson, P. A., & Lindholm, T. (2017). Costly
punishment in the ultimatum game evokes moral concern, in particular when
framed as payoff reduction. Journal of Experimental Social Psychology, 69,
59-64.
Falk, A., Fehr, E., & Fischbacher, U. (2008). Testing theories of fairness – Intentions
matter. Games and Economic Behavior, 62, 287-303.
Feldman, G., & Albarracín, D. (2017). Norm theory and the action-effect: The role of
social norms in regret following action and inaction. Journal of Experimental
Social Psychology, 69, 111-120.
Fehr, E., & Gächter, S. (2002). Altruistic punishment in humans. Nature, 415, 137-
140.
Fehr, E., & Schmidt, K. M. (1999). A theory of fairness, competition, and
cooperation. Quarterly Journal of Economics, 114, 817-868.
Gintis, H., Bowles, S., Boyd, R., & Fehr, E. (2003). Explaining altruistic behavior in
humans. Evolution and Human Behavior, 24, 153–172.
Goerg, S. J., Rand, D. G., & Walkowitz, G. (2017). Framing effects in the Prisoner’s
dilemma but not in the dictator game. Available at SSRN:
https://ssrn.com/abstract=2912982
Heck, P. R., & Krueger, J. I. (2017). Social Perception in the Volunteer's Dilemma:
Role of Choice, Outcome, and Expectation. Social Cognition, 35, 497-519.

15
Horton, J. J., Rand, D. G., Zeckhauser, R. J. (2011). The online laboratory:
Conducting experiments in a real labor market. Experimental Economics, 14,
399-425.
Huck, S., Kübler, D., & Weibull, J. (2012). Social norms and economic incentives in
firms. Journal of Economic Behavior and Organization, 83, 173-185.
Janoff-Bulman, R., Sheikh, S., & Hepp, S. (2009). Proscriptive versus prescriptive
morality: Two faces of moral regulation. Journal of Personality and Social
Psychology, 96, 521-537.
Jordan, J. J., Hoffman, M., Nowak, M. A., & Rand, D. G. (2016). Uncalculating
cooperation is used to signal trustworthiness. Proceedings of the National
Academy of Sciences, 113, 8658-8663.
Karlan, D., & List, J. A. (2007). Does price matter in charitable giving? Evidence from
a large-scale natural field experiment. American Economic Review, 97, 1774-
1793.
Kay, A. C., & Ross, L. (2003). The perceptual push: The interplay of implicit cues
and explicit situational construals on behavioral intentions in the Prisoner’s
Dilemma. Journal of the Experimental Social Psychology, 39, 634-643.
Kimbrough, E. O., & Vostroknutov, A. (2016). Norms make preferences social.
Journal of the European Economic Association, 608-638.
Klein, N., & Epley, N. (2016). Maybe holier, but definitely less evil, than you:
Bounded self-righteousness in social judgment. Journal of Personality and
Social Psychology, 110, 660-674.
Klein, N., & Epley, N. (in press). Less evil than you: Bounded self-righteousness in
character inferences, emotonial reactions, and behavioral extremes.
Personality and Social Psychology Bulletin.
Krueger, J. I., & Acevedo, M. (2007). Perceptions of self and other in the prisoner’s
dilemma: Outcome bias and evidential reasoning. The American Journal of
Psychology, 120, 593-618.
Krueger, J. I., & DiDonato, T. E. (2010). Perceptions of morality and competence in
(non)-interdependent games. Acta Psychologica, 134, 85-93.
Krupka, E. L., & Weber, R. A. (2013). Identifying social norms using coordination
games: Why does dictator game sharing vary? Journal of the European
Economic Association, 11, 495-524.
Larrick, R. P., & Blount, S. (1997). The claiming effect: Why players are more
generous in social dilemmas than in ultimatum games. Journal of Personality
and Social Psychology, 72, 810-825.
Lazear, E. P., & Malmendier, U., & Weber, R. A. (2012). Sorting in experiments with
application to social preferences. American Economic Journal: Applied
Economics, 4, 136-163.
Levitt, S. D., & List, J. A. (2007). What do laboratory experiments measuring social
preferences reveal about the real world? Journal of Economic Perspectives,
21, 153-174.
Liberman, V., Samuels, S. M., & Ross, L. (2004). The name of the game: Predictive
power of reputations versus situational labels in determining prisoner’s
dilemma game moves. Personality and Social Psychology Bulletin, 30, 1175-
1185.
Liebrand, W. B. G., Jansen, R. W. T. L., Rijken, V. M., & Suhre, C. J. M. (1986).
Might over morality: Social values and the perception of other players in

16
 experimental games. Journal of Experimental Social Psychology, 22, 203-
215.
McCabe, K. A., Rigdon, M. L., & Smith, V. L. (2004). Positive reciprocity and
intentions in trust games. Journal of Economic Behavior and Organization,
52, 267-275.
Nowak, M. A. (2006). Five rules for the evolution of cooperation. Science, 314,
1560-1563.
Nowak, M. A., & Sigmund, K. (2005). Evolution of indirect reciprocity. Nature, 437,
1291-1298.
Ogilvie, D. M. (1987). The undesired self: A neglected variable in personality
research. Journal of Personality and Social Psychology, 52, 379-385.
Paolacci, G., & Chandler, J. (2014). Inside the Turk: Understanding Mechanical Turk
as a participant pool. Current Directions in Psychological Science, 23, 184-
188.
Rabin, M. (1993). Incorporating fairness into game theory and economics. The
American Economic Review, 83, 1281-1302.
Rand, D. G., Brescoll, V. L., Everett, J. A. C., Capraro, V., & Barcelo, H. (2016).
Social heuristics and social roles: Intuition favors altruism for women but not
for men. Journal of Experimental Psychology: General, 145, 389-396.
Rand, D. G., & Nowak, M. A. (2013). Human cooperation. Trends in Cognitive
Sciences, 17, 413-425.
R Core Team (2017). R: A Language and Environment for Statistical Computing.
https://www.R-project.org/
Rozin, P., & Royzman, E. B. (2001). Negativity bias, negativity dominance, and
contagion. Personality and Social Psychology Review, 5, 296-320.
RStudio Team (2016). RStudio: Integrated Development for R. Boston, MA: RStudio,
Inc.
Samuelson, W., Zeckhauser, R. (1988). Status quo bias in decision making. Journal
of Risk and Uncertainty, 1, 7-59.
Spranca, M., Minsk, E., & Baron, J. (1991). Omission and commission in judgment
and choice. Journal of Experimental Social Psychology, 27, 76-105.
Tappin, B. M., & McKay, R. T. (2017). The illusion of moral superiority. Social
Psychology and Personality Science, 8, 623-631.
Teper, R., & Inzlich, M. (2011). Active transgressions and moral elusions: Action
framing influences moral behavior. Social Psychological and Personality
Science, 2, 284-288.
Thomas, K. A., & Clifford, S. (2017). Validity and Mechanical Turk: An assessment of
exclusion methods and interactive experiments. Computers in Human
Behavior, 77, 184-197.
Tomasello, M. (2014). The ultra‐social animal. European Journal of Social
Psychology, 44, 187-194.
Torchiano, M. (2017). Package ‘effsize’. Available at https://cran.r-
project.org/web/packages/effsize/effsize.pdf
Trivers, R. L. (1971). The evolution of reciprocal altruism. The Quarterly Review of
Biology, 46, 35-57.
Vaish, A., Grossmann, T., & Woodward, A. (2008). Not all emotions are created
equal: The negativity bias in social-emotional development. Psychological
Bulletin, 134, 383-403.

17
Weber, J. M., Kopelman, S., & Messick, D. M. (2004). A conceptual review of
decision making in social dilemmas: Applying a logic of
appropriateness. Personality and Social Psychology Review, 8, 281-307.
Wickham, H. (2007). Reshaping data with the reshape package. Journal of
Statistical Software, 21, 1-20.
Wickham, H. (2009). ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag:
New York.
Wickham, H. (2011). The Split-Apply-Combine Strategy for Data Analysis. Journal of
Statistical Software, 40, 1-29.
Wickham, H., et al. (2017). Package ‘dplyr’. Available at https://cran.r-
project.org/web/packages/dplyr/dplyr.pdf
Zeelenberg, M., van den Bos, K., van Dijk, E., & Pieters, R. (2002). The inaction
effect in the psychology of regret. Journal of Personality and Social
Psychology, 82, 314-327.

18