The effect of social norms estimation on non-strategic giving: Discarding the role of numeric
anchoring and extra monetary prospects.
Carlos Maximiliano Senci1, Brenda Elizabeth Ryan1, Natalia Gregorietti1, Esteban Freidin1
1Instituto de Investigaciones Económicas y Sociales del Sur (IIESS), CONICET Bahía Blanca, San
Andrés 800, Bahía Blanca (8000), Buenos Aires, Argentina.
TE +54 291 495138 extension 2715
When participants are incentivized to estimate others´ opinion of the socially appropriate behavior
(i.e., the prescriptive norm) in a Dictator Game (DG), they then make more generous decisions in the
game. Authors have interpreted this result in terms of Norm Focus Theory (Cialdini et al., 1990), that
is, the idea that norms influence behavior when they are attended to. Nonetheless, alternative
explanations have not been discarded yet. First, the incentive present in the estimation exercise
increases earning prospects which might be responsible for the increased generosity. Second, the
estimation exercise makes participants think of a number which might anchor their subsequent
decision closer to the equitable share. The goal of the present study was to replicate the effect of
prescriptive norms on subsequent DG decisions with an Argentine sample, and to clarify the
interpretation of this phenomenon. In experiment 1a, we replicated the original finding with university
students from Argentina. In experiment 1b, we controlled for earning prospects by having a non-
incentivized norm estimation exercise before the DG. We found this treatment to have a similar effect
on subsequent generosity than the condition in which the norm estimation was incentivized (i.e., it
caused greater generosity than the baseline control), suggesting that extra monetary prospects did not
affect DG decisions in this context. Last, in experiment 1c, the estimation exercise involved a DG
with a monetary pie that doubled that of the decision game, implying that the fair share in the
estimation exercise coincided with the most selfish choice in the DG. We found no evidence that the
number associated with the estimated norm anchored DG decisions. We conclude that the Norm
Focus interpretation was supported with added rigor, and we discuss the practical relevance of these
Keywords: prescriptive norm; Dictator Game; generosity; pro-sociality; behavioral insight; nudge
PsycINFO code: 3020
JEL codes: D31; D33; D62; D63; D64; Z130
The socio-psychological literature on social norms makes the distinction between descriptive
and injunctive (or prescriptive) norms (Cialdini & Goldstein, 2004; Aronson, Wilson, & Akert, 2010).
These categories refer to people´s expectations about what most people do and what people believe it
is the socially appropriate thing to do, respectively (Bicchieri, 2006). There is ample evidence that
social norms can influence a myriad of behaviors relevant to vital societal issues, such as, voting
(Gerber & Rogers, 2009), tax collection (Alm, Bloomquist, & McKee, 2016), bribery (Kobis, van
Prooijen, Righetti, & Van Lange, 2015; Abbink, Freidin, Gangadharan, & Moro, 2016), littering
(Cialdini, Reno, & Kallgren, 1990), cheating (Ariely, 2009), and stealing (Cialdini, Demaine, Sagarin,
Barrett, Rhoads, & Winter, 2006), among others. These normative influences, some authors argue, are
particularly effective when people´s attention can be focused towards the norm (Cialdini et al., 1990;
also see Epley & Gilovich, 1999, and Aarts, Dijksterhuis, & Custers, 2003, for subliminal normative
influences on behavior).
More recently, the study of the behavioral effects of directing people´s attention to norms have
permeated into the analysis of decision making in economic experiments (e.g., Fehr, Kirchler,
Weichbold, & Gaechter, 1998; Bicchieri & Xiao, 2009; Krupka & Weber, 2009; Raihani &
McAuliffe, 2014; Banerjee, 2016). There are a myriad of topics involving economic decisions with
social consequences, in which understanding how norms affect behavior could mean a relevant step
towards improving social welfare and adequately orienting policy decisions. Examples can go from
non-strategic distribution of resources (e.g., donations to charities), through cooperation in public
goods, to participation in bribery and corruption. We here focused on the effect of prescriptive norms
on economic decisions in a non-strategic context, namely the Dictator Game (DG).
There are different variants of the DG, but at its core lays a unilateral decision with payoff
consequences for the decision maker and a passive associated participant. In this context, decisions
can be seen as more or less generous depending on the relative payoff each gets (see Engel, 2011 for a
review of DG results). Krupka and Weber (2009) investigated the effect of making participants guess
the prescriptive norm (others´ opinion of the socially appropriate decision) before they played a
variant of the DG. They found that the normative estimation led to a higher proportion of decisions
for the equitable distribution than in a control condition without the former normative exercise.
Similar findings have been obtained by Bicchieri and Xiao (2009), and Raihani and McAuliffe (2014;
we discuss these studies in the next section).
The present study attempts to build on Krupka and Weber´s (2009) results. We tested the
robustness of their finding with a different variant of the DG and using a sample from a different
country (experiment 1a). In addition, our goal was also to contribute to the interpretation of their
results by incorporating two control conditions. First, we ran a treatment in which the norm estimation
exercise prior to the DG involved no monetary incentive (experiment 1b). The comparison of this
latter condition with the treatment in which the norm estimation was incentivized and with the control
without norm estimation allowed us to dissociate whether participants were more generous because
they focused on the prescriptive norm or because they could earn more money in the session. Second,
the estimation of the prescriptive norm could have been effective, not because it directed attention
towards the norm, but because it led participants to think of a number which anchored their decision
(we further discuss the relevance of anchoring in the coming section). To control for this, we ran a
treatment (experiment 1c) in which participants had to estimate the prescriptive norm in a $100-pie
DG (where the equity norm was the $50-$50 split) before making their decision in a $50-pie DG
(where the equity norm was the $25-$25 split). If a numeric anchoring was responsible for the effect
of the normative estimation, we expected less generosity after the $100-pie estimation exercise than
after the $50-pie estimation exercise.
Present results confirmed the pro-social effect of estimating the prescriptive norm with a
different variant of the DG in an Argentine sample, and allowed us to conclude that this effect is
neither an artifact of extra monetary prospects associated to the estimation exercise nor the
consequence of a numeric anchoring effect.
2. Relevant theoretical and empirical background
Following Cialdini and collaborators´ (1990) Theory of Normative Focus, Krupka and Weber
(2009), Xiao and Bicchieri (2009), and Raihani and McAuliffe (2014) have attempted to test the
relative effects of descriptive and prescriptive norms on non-strategic giving. For present purposes,
we refer here to their methods and results in relation to prescriptive norms and control conditions
Krupka and Weber (2009) did a two-options DG with US university students. In their DG
protocol, participants could choose between only two options: an equitable share ($5 to the sender and
$5 to the receiver) or a selfish socially inefficient option ($7 for the sender and $1 for the receiver).
They found that participants who had to guess previous participants´ opinion about the socially
appropriate option in that game later showed a higher proportion of choices for the fair share than
participants in the control condition without the normative estimation exercise. Note, however, that
participants in the norm condition were incentivized to accurately estimate the prescriptive norm and
thus had higher financial prospects than participants in the control treatment. This feature of the
protocol is problematic to securely infer a normative effect, because participants could have behaved
more generously, not because of thinking about the norm, but because they expected to earn more
money in the session. Indeed, changing the size of the pie in the DG have small or no effect in the
average proportion shared (e.g., Forsythe, Horowitz, Savin, & Sefton, 1994; Carpenter, Verhoogen, &
Burks, 2005; Raihani, Mace, & Lambda, 2013), meaning that absolute amounts shared increase with
In turn, Bicchieri and Xiao (2009), and Raihani and McAuliffe (2014) tested whether
prescriptive information had an effect on DG decisions using US samples as well. In these studies,
DGs were different from the game used by Krupka and Weber (2009) and participants did not have to
guess others´ normative opinions but they were provided with concrete information about others´
prescriptive beliefs (Bicchieri & Xiao, 2009) or with normative suggestions (Raihani & McAuliffe,
In the study by Bicchieri and Xiao (2009), US university students played a DG in which senders
could offer receivers any integer amount from $1 to $9, excluding $7 and $3. They found that
participants made fairer choices in the condition in which the normative message stated that a
majority believed that a fair share should be chosen than in the condition in which the message stated
that own earnings maximization should be pursued. This result goes in the same direction as Krupka
and Weber´s and was the outcome of a similar protocol: Bicchieri and Xiao´s information in the fair
norm condition was similar to participants´ guesses in Krupka and Weber´s experiment. Note,
however, that Bicchieri and Xiao´s result were not compared against a no-message baseline; therefore,
one cannot know whether their normative effect is the consequence of the pro-social, the selfish or
both norms acting together.
Raihani and McAuliffe (2014) did a $1-stake DG with a US sample using Amazon Mechanical
Turk (i.e., an out-of-the-lab online protocol). Relevant for present purposes, they ran two treatments
with messages suggesting a minimum sharing amount of either 20% or 50% of the pie, and two
control conditions in which, before making the DG decision, participants read a message containing
the number either 20 or 50 in a statement unrelated to social norms (i.e., whether the typical age of
MTurk participants was “at least 20” or “less than 50”). Note that the normative messages did not
refer to previous participants´ opinions unlike the protocols described before, but were impersonal
suggestions. Notwithstanding, mean donations were higher in conditions with normative suggestions
than in control treatments, consistent with results from Krupka and Weber (2009) and Bicchieri and
Xiao (2009). However, alike Bicchieri and Xiao´s design, the controls used by Raihani and McAuliffe
(2014) may not be proper baselines against which a normative effect could be safely inferred as we
Raihani and McAuliffe´s intention was to use the age messages to control for a potential
anchoring effect of the numbers 20 and 50. Anchoring occurs when words or numbers from a
previous task bias people’s subsequent judgments or decisions. Kahneman and Tversky’s (1974)
classical example involved asking judges to estimate the number of African countries in the UN.
Participants’ estimates were lower after responding whether that number was above 10 than after
responding whether that number was above 65. Anchoring is a pervasive cognitive bias (see Furnham
& Boo, 2011 for a review), and has been also found in normatively charged settings such as legal
decision making (Englich, Mussweiler, & Strack, 2006). However, we cannot know whether Raihani
and McAuliffe were successful in controlling for an anchoring effect in their experiment for two
reasons. First, they did not use a baseline control without message against which an anchoring effect
could be safely assessed. Second, results from their anchoring treatments were in the opposite
direction to their number anchoring prediction: senders gave less money to receivers in the treatment
with the “less-than-50” message than in the treatment with the “at-least-20” message. In light of this
unexpected result, the control messages used by these authors may seem particularly troublesome
because of the phrases “less-than…” and “at-least…”. Unfortunately, this wording opens the
possibility of a semantic priming effect (Strack & Mussweiler, 1997) which could have lowered the
shared amounts in control treatments, thus casting doubts on whether there was any normative effect
at all (we further discuss this in experiment 1c).
3. Our study
The main goal of the present study was to build on the work done by the authors mentioned in
the previous section (Krupka & Weber, 2009; Bicchieri & Xiao, 2009; Raihani & McAuliffe, 2014),
testing the robustness of the prescriptive effect on DG decisions using a non-US sample, and
incorporating two controls that allowed discarding alternative (non-normative) explanations.
3.1. Experiment 1a: prescriptive guessing and DG decisions in an Argentine sample
To begin with, the three studies discussed that measured the effect of prescriptive norms on
DG decisions were done with US samples. Trans-cultural economic experiments have shown
extensive cross-population variability (Henrich et al., 2005, 2006; Herrmann et al., 2008), and,
especially relevant for present purposes, such variability in DG offers have been associated with
variation in normative behaviors, such as third-party punishment, and societal norms (Henrich et al.,
2006). Therefore, it is not trivial to assess the extent to which prescriptive norms affect DG decisions
in a non-US sample. We did so using Argentine university students in experiment 1a. To have in
mind, the USA and Argentina contrast in many aspects of their normative culture, as, for example,
indicated in their corresponding Corruption Perception Indexes (whereas the USA ranks 16 out of 167
countries in transparency, Argentina ranks 107; Transparency International, 2015). This contrast adds
an interesting dimension to the present evaluation of the prescriptive norm effect on DG decisions.
Indeed, a recent cross-national economic experiment showed that participants´ counter normative
behavior in a dice game (cheating while reporting the outcome of a roll) was predicted by a proxy
score of the level of corruption in that society (Gächter & Schulz, 2016).
22.214.171.124. Participants and task
Fifty-eight undergraduates participated in experiment 1a (mean age ±1SD = 22.5±3.3; 65%
women) at the Universidad Nacional del Sur, Bahía Blanca, Argentina, in the second semester of
2015 and the first semester of 2016. Participants were recruited via e-mail invitations from a pre-
registered subject pool consisting of undergraduate and graduate students from different educational
institutions in Bahía Blanca.
Participants received AR$25 (approx. u$s 2.5 at the time of the study) as show-up fee
(hereafter, we refer to AR$ simply as $). Experimental sessions were always ran on either Tuesdays,
Wednesdays, or Thursdays at the same time of the day (12:30), using paper and pencil, and lasted for
approximately 45 min.
Upon arrival participants were seated in individual chairs. Sessions began with a brief oral
presentation, including basic details about the DG and how roles were assigned. Participants were told
that they would play with another participant present in the room, but that they would not be able to
identify each other during or after the session. Following the procedure used by Krupka and Weber
(2009), in both conditions, participants were informed that they would make their DG decisions as
senders without knowing their actual role. Participants were informed that they would know the role
randomly assigned to them at the end of the session, when they received their payment. Throughout
the session, however, we emphasized that participants’ decisions would only affect payoffs if they
were later assigned the role A (sender); on the contrary, if they were later assigned the role B
(receiver), decisions made in role A would be inconsequential.
After that, participants received the written instructions. These instructions included every
detail participants needed to know about the game (decision options, payoffs, the consequences of
role assignment, etc.) and how the session would unfold. To ensure understanding of the instructions
we asked participants to answer a series of control questions which were individually checked before
participants could proceed to make their decisions.
We used a DG different than the one used by Krupka and Weber (2009). Instead of a two-
options DG, we implemented a one-shot continuous DG with a $50 pie, which the dictator had to
distribute between him/her-self and a randomly associated participant (he/she could opt for keeping
any integer amount between $0 and $50).
In experiment 1a, we ran two independent conditions: the baseline control condition (baseline; n=24),
and the social norms condition (SN; n=34). In the baseline condition, participants, first, decided as
senders in $50-pie DG, and, second, responded a post-decision questionnaire. In condition SN,
participants were, first, incentivized with $25 for accurately estimating baseline participants´ modal
opinion of the socially appropriate decision in the $50-pie DG; second, they made their decisions as
senders in the $50-pie DG; and last, they responded the post-decision questionnaire. The initial
estimation exercise in condition SN involved guessing, for each of the ten ranges of distributions
(money the dictator could keep for her/himself in increments of 5: $0-5, $6-10,…, $46-50), the
percentage of previous participants who had chosen it as the most socially appropriate decision range.
If their estimation did not differ in more than three percentage points from the distribution that was
actually chosen as the most socially appropriate by baseline controls, participants won a bonus
payment of $25. Participants were told that the exact answer was written in a sheet which was inside
an envelope that was placed on a desk in the front, and that it was going to be opened and showed to
them at the end of the session.
Last, the post-decision questionnaire asked for the (non-incentivized) estimation of the
descriptive norm in the session (estimation of others´ decisions), each participant´s personal norm
(their opinion about what was the most socially appropriate decision in the DG), and socio-
demographic information (gender, age, discipline of study, among others).
3.1.2. Results and Discussion
As expected, participants in condition SN estimated (mean ±1 sem) that previous participants´
opinion about the most socially appropriate option in the DG would be close to the $25-$25 fair share
(to keep $26.38 ±0.84), which, in fact, was not significantly different from baseline participants´
reported personal norm (to keep $25.29 ±0.5; Mann-Whitney U test, U=378, Z=0.47; P=0.64).
Figure 1 shows the mean amount of money kept by senders and their estimation of other senders´
decisions (descriptive norm) as a function of condition in experiment 1a-c. After the normative
estimation, participants in condition SN decided to keep a smaller amount of money than participants
in the baseline condition (Mann-Whitney U test, U=235, Z=2.79, P<0.01). This finding replicates that
of Krupka and Weber (2009) with an Argentine sample. Therefore, similarly to US university
students, Argentine students increased their generosity in the DG when their attention was focused on
the prescriptive norm before making a decision.
Estimation of others´ decisions followed the same pattern as actual decisions across
conditions (see Figure 1). Participants in condition SN estimated others to keep less money in the DG
than participants in the baseline control (Mann-Whitney U test, U=256, Z=2.25, P<0.025).
In short, we showed that making participants think of the prescriptive norm, not only increased their
generosity, but also increased their estimation of others´ generosity.
Figure 1. Money kept by senders (DECISION) and participants´ estimation of others´ decisions
(ESTIMATION) in a $50-pie Dictator Game. Whereas participants directly made their decision in the
baseline condition, in the other three conditions (SN, SN_WO, and SN_100), before deciding,
participants had to estimate previous participants´ opinion about the most socially appropriate option
in the game.
3.2. Experiment 1b: attention to the prescriptive norm or generosity based on increased
The design used by Krupka and Weber (2009) as well as that used in experiment 1a left open
the possibility that participants became more generous, not because of thinking about the prescriptive
norm, but because the normative exercise involved the possibility of extra earnings (i.e., the
estimation exercise was incentivized). In this sense, we conjectured that the increased generosity in
condition SN could have been the consequence of participants having better financial prospects than
in the baseline condition. This idea would be consistent with results showing higher absolute amounts
shared in DGs with larger pies (Forsythe et al., 1994; Carpenter et al., 2005; Raihani et al., 2013).
To control for the extra incentive present in the SN condition in experiment 1a, we conducted
another social norms treatment (experiment 1b) identical to condition SN but without the monetary
incentive for accurately guessing the prescriptive norm. If thinking about the norm was responsible
for the increased generosity in experiment 1a, then DG decisions in experiment 1b should differ from
the baseline control. Alternatively, if increased prospects after the incentivized estimation exercise
inclined participants to be more generous, then DG decisions in the non-incentivized norm condition
should be less generous than in condition SN.
126.96.36.199. Participants and task
Thirty-six undergraduates participated in sessions of experiment 1b at Universidad Nacional
del Sur (mean age ±1SD: 22.8 ±3.8; 64% women) in the second semester of 2015 and the first
semester of 2016. The recruitment of participants was similar to that described for experiment 1a.
The procedure and experimental task in condition SN_WO (WO refers to “WithOut incentive”) was
identical to condition SN in Experiment 1a, with the exception that participants were not offered any
incentive for accurately guessing others´ opinion about the most socially appropriate behavior in the
3.2.2. Results and Discussion
Participants in condition SN_WO also estimated (mean ±1 sem) that baseline participants´
opinion about the most socially appropriate option in the DG would be close to the $25-$25 fair share
(to keep $28.35 ±0.92). This estimation differed neither from the prescriptive estimation in condition
SN (Mann-Whitney U test, U=1110, Z=-0.55; P=0.58) nor from baseline participants´ reported
personal norm (Mann-Whitney U test, U=739, Z=-0.88; P=0.39).
As shown in Figure 1, DG decisions were more generous in condition SN_WO than in the
baseline control (U=286.5, Z=2.228, P=0.02). In turn, the comparison of the amount of money kept
between conditions SN_WO and SN showed no significant differences (Mann-Whitney U test, U=
594.5, Z=-0.90, P=0.36). Alike results from experiment 1a, decisions and estimation of others´
decisions followed a similar pattern across conditions: participants´ estimation of the descriptive norm
did not differ between conditions SN_WO and SN (Mann-Whitney U test, U= 1100, Z=0.62, P=0.54),
and did differ between conditions SN_WO and the baseline control (Mann-Whitney U test, U=477.5,
Z=-2.92, P<0.005) (see Figure 1).
In conclusion, results from experiment 1a-b imply that the normative estimation exercise
increased generosity regardless of whether the estimation was incentivized.
3.3. Experiment 1c: normative effect or number anchoring?
Given that thinking of irrelevant numbers have been shown to exert an influence upon judgment
and decisions in many diverse situations, including normatively charged contexts (Englich et. al.,
2006), it seems reasonable to suspect of a number anchoring effect being responsible for results from
experiment 1a-b. More specifically, participants that estimated that the socially appropriate decision
in the $50-pie DG was a $25-$25 split or close to that, could have gotten their decision anchored by
their numeric estimate, thus choosing to keep an amount closer to the fair share than controls. In this
sense, previous results do not allow us to firmly conclude that thinking about the prescriptive norm
affects generosity through a normative influence.
In experiment 1c, we ran a condition in which we directly pitted the normative effect and the
number anchoring predictions against each other. In condition SN_100, prior to making their
decisions in a $50-pie DG, participants were incentivized to estimate the prescriptive norm of a $100-
pie DG. Because we expected the estimated norm to be close to a $50-$50 split in the $100-pie DG, if
the estimated number anchored the subsequent DG decision, we would see senders keeping more of
the $50 after the $100-pie-DG normative estimation (condition SN_100) than after $50-pie-DG
normative estimation (condition SN). In contrast, if thinking about others’ opinion of the most
socially appropriate behavior played its role upon decisions by focusing participants´ attention on the
fair share norm, we predicted decision in condition SN_100 to be more generous than in the baseline
control and similarly so than in condition SN.
As mentioned in the introduction, Raihani and McAuliffe (2014) did a test of the anchoring
hypothesis in this context, and did not find evidence to support it. Nonetheless, the present test
improves that of Raihani and McAuliffe (2014) in many respects. First, these authors changed the
wording between normative (“20 or more” and “50 or more”) and anchoring (“at least 20” and “less
than 50”) conditions, thus making them less comparable. Not only that, but they might have
unintentionally induced a semantic priming effect by the wording used. This would explain why
participants were more selfish in the “less than 50” condition than in the “at least 20” condition,
contrary to their numeric anchoring prediction (Raihani & McAuliffe, 2014). In contrast, the present
anchoring test relies on treatments which only differ on the pie size of the estimation exercise prior to
the DG decision. This, jointly with counting on a proper baseline control (i.e., a treatment without
estimation or message before the DG decision), allowed us to pit the normative and the anchoring
predictions directly against each other. Last but not least, we tested a stronger version of the
anchoring hypothesis than that tested by Raihani and McAuliffe (2014). According to the selective-
accessibility model (Strack & Mussveiler, 1997), anchoring works by making anchor-related
information more accessible in memory, and thus more readily available to other cognitive operations.
In consistency with this model, it has been shown that conditions that lead to greater thought about the
anchor produce greater bias (Mussweiler & Strack, 1999; Bodenhausen, Gabriel, & Lineberger,
2000). Whereas the anchor was exogenously and effortlessly provided in the study by Raihani and
McAuliffe (2014), the anchoring information was self-generated through an incentivized thought-
intensive estimation exercise in the present treatment instead.
188.8.131.52. Participants and task
Thirty-six undergraduates participated in sessions of experiment 1c at Universidad Nacional
del Sur (mean age ±1SD: 21.5 ±4.3; 53% women) in the second semester of 2015. The recruitment of
participants was similar to that described before.
The procedure and experimental task were similar to those in condition SN from experiment 1a, with
the exception that the estimation exercise before the DG decision involved a DG with a $100 pie.
More specifically, participants from a previous group had given their opinion about which was the
most socially appropriate option from a list of ten possible ranges of distributions (money the sender
could keep for her/himself in increments of 10: $0-10, $11-20,…, $91-100). Participants in
experiment 1c (condition SN_100) were asked to estimate, for each of the ten ranges of distributions,
the percentage of those previous participants who had chosen it as including the most socially
appropriate option. If their estimation did not differ in more than three percentage points from the
distribution that was actually chosen as the most socially appropriate, participants won a bonus
payment of $25 (as was done in condition SN). Payment and disclosure of the right answer followed
similar procedures to those described before.
3.3.2. Results and Discussion
To make the estimation of the prescriptive norm in condition SN_100 comparable to that of
conditions SN and SN_WO, participants´ scores were divided by two. Participants in condition
SN_100 estimated that previous participants´ opinion about the most socially appropriate option in the
DG would be close to the fair share as well. Indeed, their estimation did not significantly differ from
such estimations in conditions SN and SN_WO (Kruskal Wallis ANOVA by Ranks, χ2=0.55, df=2,
In terms of actual decisions, participants showed greater generosity in condition SN_100 than
in the baseline condition (Kruskal Wallis ANOVA by Ranks, χ2=13.39, df=3, P<0.005; SN_100 vs.
baseline: Mann-Whitney U test, U=233.5, Z=2.99, P=0.002), whereas the amount of money kept in
condition SN_100 did not significantly differ from decisions in conditions SN and SN_WO (Mann-
Whitney U test, U=594.5, Z=0.21, P=0.84; Mann-Whitney U test, U=548.5, Z=1.12, P=0.26;
respectively; see Figure 1).
Estimations of others´ decisions again followed a similar pattern across conditions to that of
actual decisions (see Figure 1). The estimation of the descriptive norm in condition SN_100
significantly differed from that of the baseline condition (Kruskal Wallis ANOVA by Ranks,
χ2=15.62, df=3, P<0.005; SN_100 vs. baseline: Mann-Whitney U test, U=820, Z=3.38, P<0.001),
whereas it did not differ from the other conditions that presented a normative estimation exercise
before the decision (SN_100 vs. SN: Mann-Whitney U test, U=592, Z=-0.04, P=0.98; SN_100 vs.
SN_WO: Mann-Whitney U test, U=586.5, Z=0.50, P=0.62).
To check for the robustness of present results, we did a multiple regression with the decision
(money kept) in the DG as dependent variable and several factors as predictors (see Table 1). We
confirmed the generosity effect of thinking about the prescriptive norm, and we found it to be robust
to the addition of several controls such as increased monetary prospects, gender, age, discipline, the
number of known participants in the session, and previous participation in economic experiments. In
addition, the regression showed that previous participation in economic experiments was associated
with more selfish decisions (see Table 1 for regression coefficients).
Table 1. Multiple regression of decisions in the DG. Prescriptive_Estimation and Monetary_Prospects
are dummies that test for the effects of having the estimation exercise before the DG decision (1=yes;
0=no) and an increased monetary prospect (1=yes; 0=no) due to the incentivized estimation exercise.
Gender: 1=female; Discipline: 1=economics, business administration and accountancy;
Known_participants: number of known participants in the session; Previous_participation: number of
previous participation in economic experiments. * P<0.05
Predictors Coefficients (Std. Errors)
Perspective_estimation -0.20 (0.10)*
Monetary_prospects -0.08 (-0.10)
Gender -0.10 (0-09)
Age 0.01 (0.09)
Discipline 0.09 (0.09)
Known_participants 0.01 (0.09)
Previous_participation 0.18 (0.09)*
4. General discussion
In this study, we replicated the finding that focusing participants´ attention on prescriptive
norms increased generosity in the DG (Krupka & Weber, 2009; Bicchieri & Xiao, 2009; Raihani &
McAuliffe, 2014). Previous reports of this effect were, however, inconclusive because of their lack of
controls. Authors had claimed a normative effect while their results could be explained in non-
normative terms. More specifically, focusing participants on norms could have increased their
generosity because of the increased monetary prospects associated to the incentivized normative
estimation exercise. Moreover, a non-exclusive possibility implied that the effect of prescriptive norm
information on non-strategic giving could be the consequence of a numeric anchoring effect (see the
introduction for references).
Here we implemented a condition in which the normative estimation exercise was not
incentivized, and compared DG decisions in this condition against a condition in which correct
normative guesses were monetarily rewarded and a baseline control without any prior estimation
exercise. We found that generosity in both conditions with normative exercises prior to the DG
decision was higher than in the baseline, while they did not differ between each other. This result
allowed us to discard a possible effect of increased monetary prospects on generosity in this context.
In turn, we ran a condition in which the normative exercise prior to the decision involved guessing
others´ opinion about the appropriate behavior in a DG with a pie ($100) that doubled the size of the
pie to be shared in the decision DG ($50). The comparison of this condition against the condition in
which the guessing exercise was done with the $50-pie DG allowed us to test whether the number of
the estimated norm anchored the subsequent decision. In fact, it did not. Again, both normative
guessing conditions presented similar levels of generosity, both higher than the baseline. This result
suggests that thinking about the fair share norm underlied participants´ increased generosity.
In sum, we are able to conclude, with added rigor, that focusing participants on the
prescriptive norm increases generosity through a normative effect. Therefore, the Norm Focus
interpretation of the effect of prescriptive norms on non-strategic giving was supported after
controlling for non-normative explanations. In addition, whereas previous studies had shown this
effect on US samples (Krupka & Weber, 2009; Bicchieri & Xiao, 2009; Raihani & McAuliffe, 2014),
we did so with university students from Argentina. Despite differences between US and Argentine
social norms (e.g., see Transparency International, 2015 for differences in the levels of perceived
corruption), we found a consistent normative effect in these two societies. This result might be seen as
somehow unexpected given trans-cultural studies showing that DG decisions in a society were
positively associated with normative behaviors in the lab as well as with societal norms (Henrich et
al., 2006). Nonetheless, the similarity between cultures detected here is only qualitative. Both in the
US and in Argentina, reminders of the prescriptive norm increased generosity; however, we cannot
easily compare the degree to which generosity was increased in different studies because of several
methodological differences among them (Krupka & Weber, 2009; Bicchieri & Xiao, 2009; Raihani &
McAuliffe, 2014; see the introduction for details).
In any case, the robustness of the effect of prescriptive norms on economic behavior might
serve as a behavioral insight to address other issues of public interest (Sousa Lourenço, Ciriolo,
Rafael Almeida, & Troussard, 2016). D´Adda, Drouvelis, and Nosenzo (2016) have recently tested
the effect of prescriptive norm estimation in a collusive Bribery Game (BG) done at the University of
Birmingham, UK. In their study, bribery was however not affected by whether norm estimation
occurred before or after bribery decisions (D´Adda et al., 2016). In our own laboratory in Bahía
Blanca, Argentina, we also did a simple one-shot collusive BG and, alike D´Adda et al. (2016), we
did not find evidence of any effect of prescriptive norms on bribery (Abbink et al., 2016).
Why would the effect of norm estimation be robust in variations of the DG, whereas have no
effect in BGs? We do not have a definitive answer to this question, but we discuss several
possibilities. First, it could be the case that the effect of prescriptive norms has a more prominent
effect in non-strategic than in strategic contexts. This could be the case if strategic considerations
overshadow the effect of norms. However, as present results showed, focusing attention on norms not
only changed participants´ behavior but also affected participants´ beliefs about others´ behavior.
Based on this finding, we should expect an effect of prescriptive norms on strategic considerations as
well. Second, there could be a difference in the level of consensus between DG norms and norms in
the BG (Rauhut & Winter, 2010). Maybe, the simplicity of the DG leads to greater agreement in terms
of what represents a socially appropriate behavior in that context (i.e., the fair share norm). If that
were the case, we could expect a stronger effect of norms in the DG than in the BG simply because of
a more homogeneous aggregated outcome in the former than in the latter game. However, the
evidence does not support this view either. According to results from the study by D´Adda and
collaborators (2016), more than 90% of participants in their BG evaluated bribery (offer or
acceptance) as socially inappropriate, thus suggesting a high level of normative consensus. Last, we
consider the possibility that the norm may be more salient in the BG than in the DG. This could be the
case because the BGs used by D´Adda et al. (2016) and by Abbink et al. (2016) both relied on a
loaded frame with words that directly or indirectly referred to bribery (e.g., “bribe” in D´Adda et al.,
2016; “private payment” in Abbink et al., 2016). In this sense, even when participants were not asked
to think about the prescriptive norm in these BGs, the mere frame of the games could have focused
participants´ attention on normative considerations. This could explain the absence of difference in
bribery decisions between conditions with and without a normative estimation exercise (D´Adda et
al., 2016; Abbink et al., 2016). To what extent this conditions the applicability of Norm Focus Theory
to deter bribery remains an open issue.
To finish, we would like to highlight the potential relevance of present and related findings as
behavioral insights that may contribute to the design of public policies (see Sousa et al., 2016). This
requires further research, both in the lab and in the field, to acknowledge the conditions under which
the effect of focusing people´s attention to relevant norms actually affects their behavior in a pro-
We would like to thank Fabricio Carballo for his help in collecting data in some of the
sessions. This work was supported by The Argentine Council of Science and Technology (Spanish
acronym: CONICET; grant number PIP 2014-2016 N° 112-201301-00600-CO) and The Argentine
Agency of Science and Technology Promotion (Grant code: PICT 2013-1582).
Aarts, H., Dijksterhuis, A., & Custers, R. (2003). Automatic normative behavior in environments: The
moderating role of conformity in activating situational norms. Social Cognition, 21, 6, 447-464.
Abbink, K., Freidin, E., Gangadharan, L., & Moro, R. (2016). The effect of social norms on bribe
offers. IIESS CONICET Bahía Blanca Working Paper nº 2. http://www.iiess-
All, J., Bloomquist, K. M., & McKee, M. (2016). When you know your neighbour pays taxes:
Information, peer effects, and tax compliance. The Journal of Applied Public Economics, in press.
Aronson, E., Wilson, T. D., & Akert, R. M. (2010). Social psychology. Upper Saddle River, NJ:
Banerjee, R. (2016). On the interpretation of bribery in a laboratory corruption game: moral frames
and social norms. Experimental Economics, 19, 1, 240-267. doi:10.1007/s10683-015-9436-1
Bicchieri, C. & Xiao, E. (2009). Do the right thing: But only if others do so. Journal of Behavioral
Decision Making, 22, 2, 191-208.
Bodenhausen, G. V., Gabriel, S., & Lineberger, M. (2000). Sadness and susceptibility to judgmental
bias: The case of anchoring. Psychological Science, 11, 4, 320-323.
Carpenter, J., Verhoogen, E., & Burks, S. (2005). The effect of stakes in distribution experiments.
Economic Letters, 86, 393-398.
Cialdini, R. B. & Goldstein, N. J. (2004). Social influence: Compliance and conformity. Annual
Review of Psychology, 55, 591-621.
Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative conduct:
Recycling the concept of norms to reduce littering in public places. Journal of Personality and Social
Psychology, 58, 1015–1026. doi:10.1037/0022-5184.108.40.2065.
Cialdini, R. B., Demaine, L. J., Sagarin, B. J., Barrett, D., W., Rhoads, K., & Winter, P. L. (2006).
Managing social norms for persuasive impact. Social Influence, 1, 1, 3-15.
D´Adda, G., Drouvelis, M., & Nosenzo, D. (2016). Norm elicitation in within-subject designs:
Testing for order effects. Journal of Behavioral and Experimental Economics, 62, 1-7.
Engel, C. (2011). Dictator games: a meta-study. Experimental Economics, 14, 583–610.
Englich, B., Mussweiler, T. & Strack, F. (2006). Playing dice with criminal sentences: The influence
of irrelevant anchors on experts’ judicial decision making. Personality and Social Psychology
Bulletin, 32, 2, 188-200.
Epley, N. & Gilovich, T. (1999). Just going along: Nonconscious priming and conformity to social
pressure. Journal of Experimental Social Psychology, 35, 578–589.
Fehr, E., Kirchler, E., Weichbold, A., & Gächter, S. (1998). When social norms overpower
competition: Gift exchange in experimental labour markets. Joumal of Labor Economics, 16, 2, 324-
Forsythe, R., Horowitz, J. L., Savin, N. E., & Sefton, M. (1994). Fairness in simple bargaining
experiments. Games and Economic Behavior, 6, 347-369.
Furnham, A. & Boo, H. C. (2011). A literature review of the anchoring effect. The Journal of Socio-
Economics, 40, 35–42.
Gächter, S. & Schulz, J. F. (2016). Intrinsic honesty and the prevalence of rule violations across
societies. Nature, 531, 496–499. doi:10.1038/nature17160
Gerber, A. S. & Rogers, T. (2009). Descriptive social norms and motivation to vote: Everybody’s
voting and so should you. The Journal of Politics, 71, 1, 178-191.
Henrich, J., Boyd, R., Bowles, S., Camerer, C., Fehr, E., et al. (2005). “Economic man” in cross-
cultural perspective: Behavioral experiments in 15 small-scale societies. Behavioral and Brain
Sciences, 28, 795-855.
Henrich, J., McElreath, R., Barr, A., Ensminger, L., Barrett, C., et al., (2006). Costly punishment
across human societies. Science, 312, 1767-1770.
Herrmann, B., Thöni, C., & Gächter, S. (2008). Antisocial punishment across societies. Science, 319,
Kahneman, D. & Tversky, A. (1974). Judgment under uncertainty: Heuristics and biases. Science,
185, 4157, 1124-1131.
Kobis, N. C., van Prooijen, J. W., Righetti, F., & Van Lange, P. A. M. (2015). “Who doesn´t?” – The
impact of descriptive norms on corruption. PLoS ONE, 10, 6: e0131830. doi:
Krupka, E. & Weber, R. A. (2009). The focusing and informational effects of norms on pro-social
behavior. Journal of Economic Psychology, 30, 307-320.
Mussweiler, T. & Strack, F. (1999). Hypothesis-consistent testing and semantic priming in the
anchoring paradigm: A selective accessibility model. Journal of Experimental Social Psychology, 35,
Raihani, N. J. & McAuliffe, K. (2014). Dictator Game giving: The importance of descriptive versus
injunctive norms. PLoS ONE, 9, 12: e113826. doi:10.1371/journal.pone.0113826
Raihani, N. J., Mace, R., Lamba, S. (2013). The effect of $1, $5 and $10 stakes in an online Dictator
Game. PLoS ONE, 8, 8: e73131. doi:10.1371/journal.pone.0073131
Rauhut, H. & Winter, F. (2010). A sociological perspective on measuring social norms by means of
strategy method experiments. Social Science Research, 39, 1181-1194.
Sousa Lourenço, J., Ciriolo, E., Rafael Almeida, S., & Troussard, X. (2016). Behavioural insights
applied to policy: European Report 2016.
Strack, F. & Mussweiler, T. (1997). Explaining the enigmatic anchoring effect: Mechanisms of
selective accessibility. Journal of Personality and Social Psychology, 1997, 73, 3, 437-446.
Transparency International (2015). Corruption Perception Index 2015.