ArticlePDF Available

Reputation and Influence in Charitable Giving: An Experiment


Abstract and Figures

Previous experimental and observational work suggests that people act more generously when they are observed and observe others in social settings. However, the explanation for this is unclear. An individual may want to send a signal of her generosity to improve her own reputation. Alternately (or additionally) she may value the public good or charity itself and, believing that contribution levels are strategic complements, give more to influence others to give more. We perform the first series of laboratory experiments that can separately estimate the impact of these two social effects, and test whether realized influence is consistent with the desire to influence, and whether either of these are consistent with anticipated influence. Our experimental subjects were given the opportunity to contribute from their endowment to Bread for the World, a development NGO. Depending on treatment, “leader” subjects’ donations were reported to other subjects either anonymously or with their identities, and these were reported either before these “follower” subjects made their donation decisions. We find that “leaders” are influential only when their identities are revealed along with their donations, and female leaders are more influential than males. Identified leaders’ predictions suggest that are aware of their influence. They respond to this by giving more than either the control group or the unidentified leaders. We find mixed evidence for “reputation-seeking.” KeywordsCharitable giving–Experimental design–Reputation–Influence–Peer effects–Altruism
Content may be subject to copyright.
ISSN 1755-5361
Discussion Paper Series
Reputation and Influence in Charitable Giving: An
David Reinstein
Gerhard Riener
Note : The Discussion Papers in this series are prepared by members of the Department of
Economics, University of Essex, for private circulation to interested readers. They often
represent preliminary reports on work in progress and should therefore be neither quoted nor
referred to in published work without the written consent of the author.
University of Essex
Department of Economics
No. 688 April 2010
Noname manuscript No.
(will be inserted by the editor)
Reputation and Influence in Charitable Giving: An Experiment
David Reinstein · Gerhard Riener
Abstract Previous experimental and observational work suggests that people act more generously when
they are observed and observe others in social settings. But the explanation for this is unclear. An indi-
vidual may want to send a signal of her generosity in order to improve her own reputation. Alternately
(or additionally) she may value the public good or charity itself and, believing that contribution levels are
strategic complements, give more in order to influence others to give more. We perform the first series of
laboratory experiments that can separately estimate the impact of these two social effects, and test whether
realized influence is consistent with the desire to influence, and whether either of these are consistent with
anticipated influence. We find that “leaders” are influential only when their identities are revealed along
with their donations, and female leaders are more influential then males. Identified leader’s predictions
suggest that are aware of their influence. They respond to this by giving more than either the control group
or the unidentified leaders. We find mixed evidence for “reputation-seeking.
Keywords Charitable giving · experimental design · reputation · influence · peer effects · altruism
1 Introduction
Cooter and Broughman (2005) note that while many organizations publicly acknowledge their donors
(e.g., in bronze plaques and opera programs), there is no systematic public record of an individual’s overall
philanthropy. To remedy this, they suggest that the IRS should start a voluntary “donation registry. Citing
laboratory evidence, they claim that “many Americans would be more generous if other people observed
their level of giving. While no such registry has yet been implemented, a host of nonprofit institutions and
applications on the internet publicize donations and aim to take advantage of social connections to promote
More generally, an Independent Sector report (Kirsch et al, 2000) noted that 77 percent of donors
David Reinstein
University of Essex, Economics, Wivenhoe Park, CO4 3SQ Colchester, United Kingdom
Tel.: +44(0) 1206 87 - 3518
Gerhard Riener
University of Jena GK EIC and MPI of Economics, Carl Zeiss Strasse 3, 07743 Jena, Germany
Tel.: +49 (0) 3641 - 9 43278
E.g., JustGiving (UK-based; Firstgiving in the US), DonorsChoose, Kiva,, Wikipedia, SixDegrees, and Face-
book’s “Causes” employ various versions of this strategy. JustGiving, which reported 10 million users in February 2010,
enables users to start fundraising pages for a particular cause, publicize their own donations, and recruit other donors, whose
donations are sequentially posted on the page with their names or anonymously. A target and a measure of the total amount
raised can also be posted. The US site touts that as of 2009 “supporters have raised over $115 million on Firstgiving for causes
they care about” <>.
contributed as a result of a request from someone they knew well. As Carman (2003) noted, “fundraisers
often design their campaigns to leverage the power of social influences, including via university alumni
networks, “walk-a-thons, and displaying “thermometers that measure dollars raised towards their goal or
the proportion of people who have made contributions. While it is often supposed that making donations
public will lead to an increase in giving, there is still a lack of evidence on whether, how, when, and why
this occurs. Much experimental (lab and field) and observational work has shown that people sometimes
act more generously when they are observed (and observe others) in social settings. For example, Soetevent
(2005) ran a field experiment in churches in the Netherlands, and found that reducing anonymity (moving
from closed bags to open baskets) increases donations for certain church offerings. However, there are
several possible explanations for the increased giving.
An individual may give more in order to send a signal of her generosity (or wealth) and improve
her own reputation (Harbaugh, 1998b; Glazer and Konrad, 1996). We call this the Reputation-Seeking or
Repseek effect. Alternately (or additionally) she may value the charity itself, and, believing that contribu-
tion levels are strategic complements, give more when observed in order to influence others to give more
and and thus magnify her gift; we refer to this as the Influencer effect.
In turn, we call the actual influence
on the follower’s gift F-Influence.
Our experimental design, a modified dictator game in which subjects can donate some, none, or all of
their endowment to a real charity (in the mold of Eckel and Grossman, 1996), is the first that can separately
measure Repseek and Influencer effects. As we observe both leaders
(first-movers) and followers (second-
movers) in a variety of anonymity/reporting conditions, we can measure the actual influence (F-Influence)
of prior observed gifts and examine whether the apparent desire to influence is consistent with the achieved
influence. We differentiate each of these effects by the gender of the leader and the follower, and by
whether the reporting is anonymous or identified. Through eliciting incentivized predictions of what other
subjects will donate, we are also able to impute the leader’s beliefs of their own influence.
Our experiment was run on 170 student subjects at the laboratory at the University of Jena, Germany.
The subjects first participate in a “meet and greet” stage in which they exchange names and their subject of
study with another subject. After this, they read about the charity Brot fur die Welt (in English “Bread for
the World”), an international development organization, and read the rules of the experiment, including a
description and an example of their specific treatment. Next, the subjects, under several treatments, decide
how much of their endowment to give to the charity. Subjects may keep what they do not donate.
The treatments differ in the information transmitted and obtained. Depending on the treatment, the
subject is either told that her donation will not be reported to any subject, told that it will be reported
anonymously to a random subject, or told that it will be reported, along with her identity, to a specific
subject. Similarly, subjects may receive no information about others’ choices, information about some
(anonymous) subject’s gift, or information about an identified subject’s gift.
When the donation is not
going to be reported, there should be neither an Influencer nor a Repseek motivation. When it is reported
anonymously, there may be an Influencer effect (a desire to influence giving either within or after the
We remain open to alternate explanations with a more psychological flavor. People may prefer giving in group settings or
a group setting may cue an altruistic mode of thinking. Furthermore, at least in Soetevent’s church setting, the effect may also
include increased giving by “followers” who have observed previous gifts. If the followers are consistently favorably surprised
by the gifts and gifts are strategic complements this may increase giving in net. However, having consistently incorrect priors
(over others’ gifts) is not an equilibrium phenomenon, and this explanation would not explain the increases in giving by
first-movers when they are observed by others, as in Carman (2003).
The terms “leaders” and “followers” are for the purpose of this discussion only we do not use these terms in the
experiment itself.
As a final dimension of variation this reporting may occur in the middle of the experiment, when it has a chance to
influence another subject’s laboratory giving behavior, or it may occur at the very end of the experiment.
experiment) but no Repseek effect. When it is reported along with the identity, there is the potential for
both an Influencer and a Repseek effect.
In section 2 we review both the theoretical and empirical/experimental literature. We consider the
theoretical case for each of the social effects we are interested in. We survey relevant laboratory and field
experiments in voluntary contribution mechanisms (henceforth VCM) and charitable-giving settings. In
section 3 we describe our experimental design and implementation. Section 4 presents and interprets our
results and tests against alternative explanations. We conclude in section 5, summarizing and interpretating
of our results, relating our findings to previous work, and highlighting the practical implications of our
2 Literature Review and Background
Previous researchers have modeled the Repseek motive, emphasizing that philanthropy can improve the
donor’s reputation as altruistic (Harbaugh, 1998a) or wealthy (Glazer and Konrad, 1996). These authors
model reputation as a direct input into an individual’s utility function. Alternately, reputation may be
cultivated and used for personal gain. For example, in a prisoner’s dilemma in a population with some
conditional cooperators (e.g., reciprocal types as in Sugden, 1984, or agents with fairness norms as in
Rabin, 1993) and asymmetric information over a player’s type, a reputation for altruism may improve a
player’s expected profit.
For the leader to have an Influencer motive she must believe that the magnitude of her (reported) gift
has a positive effect on the amount a follower who observes this gift will donate. There are several possible
explanations for such an effect. First, if there is asymmetric information over the charity’s value, a leader’s
gift may be a signal of the quality of the charity. This is modeled by Vesterlund (2003), and Potters et al
(2005) offer some experimental evidence for this in a standard laboratory VCM setting (rather than with
an actual charity).
Second, there may be asymmetric information over the distribution of altruism in the population. For
the reasons discussed above, an individual may seek to be known as altruistic relative to her reference
group. Suppose each person knows her own altruism, but does not know the altruism of the others in her
relevant reference group. She only has a consistent Bayesian prior belief of the family of distributions from
which other players’ altruism is drawn. Observing another player’s gift gives her an informative signal of
the altruism of others, and a larger gift will lead her to update her beliefs in a positive direction, and thus
she may give more if she hopes to signal her own relative altruism.
Finally, there are also a host of explanations based on psychological mechanisms that could explain
the (potential) positive influence of a leader’s gift. Conformity may be a primal desire in itself (Bern-
heim, 1994). Alternately, giving “less then the last guy” may lead to embarrassment, another primal. A
reciprocity motivation (Sugden, 1984) can also explain a leader’s influence.
The leader’s belief that she has an influence does not in itself imply that she will contribute more when
she is observed. As a simple example, suppose that the leader gets utility only from private consumption
If my opponent has preferences that imply reciprocity and she believes that I am so altruistic that Cooperate is my
dominant strategy, she will certainly choose to cooperate as well. Whether reputation yields a direct or an instrumental benefit,
as long as giving boosts my reputation I will choose to give more when my contribution is observed than when it is anonymous;
this is the Repseek effect. For example, if two players are playing a game that is a prisoner’s dilemma in its material payoffs,
and one player is altruistic and the other is a conditional cooperator and this is common knowledge, the cooperative outcome
may be the only Nash equilibrium. See Ostrom (2000) for a related discussion of the survival of cooperative types in an
evolutionary framework.
Harbaugh (1998a) makes a similar point, noting that prestige benefits may be “relative to gifts by others”, in particular a
“reference group” of ones peers. He writes: “people should presumably increase their donations after being told that a member
of their group has given a large amount. Romano and Yildirim (2001) offer a related model of “snob appeal.
and from her net impact on the charity’s total receipts. Suppose she also believes that for every dollar
she is observed to give her peers will give an additional dollar. In such a case, making her contribution
observed is the same as offering a 50% price cut or a 100% match rate; these should have the same
effect on her contribution.
In general, if a product’s price decreases then expenditure on this product may
increase or decrease, depending on the price-elasticity. Still, there may be some reason to anticipate a
positive Influencer effect: if this observability is a rare opportunity, she is likely to take advantage of this
temporary price cut by giving more than she otherwise would have (Randolph (1995) makes a similar
argument in the context of tax incentives).
Several lab and field experiments have investigated social influences on cooperative behavior. Charness
and Gneezy (2008) compare giving in dictator games and ultimatum games and find that revealing the
name of the proposer increased the amount transferred in the dictator game, but had no effect in the
ultimatum game, where “strategic considerations crowd out impulses of generosity.
Social effects have been examined in many VCM experiments. Rege and Telle (2004) note that their
subjects take less money out of the common pool when they are observed by other subjects. Levati et al
(2007) and Guth et al (2007) make subjects’ endowments heterogeneous, and have leaders report their gifts
(but not their identities) to followers. They find that leaders’ gifts are larger than followers’ (particularly
in later periods), and having a leader increases the average contribution. Our reanalysis of their data shows
that the marginal influence of the leader on the other group members was close to 0.40 on average, i.e.,
when the leader gave $1 more, the followers gave 40 cents more (details available by request).
Potters et al (2001) consider a VCM where the return to contributions is either complete or asymmetric
information. All of their treatments involve a binary contribution decision, exogenously assigned leaders,
and complete anonymity between subjects. They find that leaders give more and followers respond posi-
tively to these gifts when only the leader is informed, but not when returns are common knowledge. This
is interpreted as evidence that leadership effects come exclusively from informative signaling, rather then
from reciprocity motives.
Andreoni and Petrie (2004) (henceforth A&P) run a series of laboratory experiments, setting up five-
subject groups who play eight rounds of a standard linear public good/VCM game. Subjects are rematched
with new groups five times to play an additional eight rounds each time. The treatments include no-
revelation, revealing identities alone (digital photographs), revealing individual amounts contributed af-
ter each round, and revealing both identity and contribution. They find that revealing contributions with
photographs positively and significantly affects the level of contribution.
There are important differences between A&P’s experiments and our own. A&P have a different focus.
They argue that “identification of other group members may serve to reduce social distance and encourage
some level of social responsibility to the group. They do not consider reputation, and it is difficult to
separately identify a Repseek and Influencer effect from their data. Since their subjects learn their payoffs
after each round, they implicitly learn the group’s net contribution after each period, and thus have a
potential Influencer effect in all of their treatments.
It is unclear when identified subjects are contributing
in order to boost their reputation outside the laboratory (the subjects are economics and business students
Unless she is also altruistic towards the followers (more so than towards the person providing the match), in which case
the match may have more effect on her contribution than the observability.
They also find that the photograph treatment alone reduces the number of “laggards” (those who gave less then five of
20 tokens in the first period of four or more of the five eight-round treatments). This could be interpreted as a Repseek effect;
albeit reputation only in a probabilistic sense (no one knows exactly how much you gave, but they will base their guess on the
average gift).
This is confirmed by their finding that revealing individual contributions without identities has virtually no effect. This
should not be surprising: there is hardly any informational gain from those reports.
at the same university), and when they are doing so in order to convince other people to cooperate with
them in future rounds.
The most important difference between our experiment all of the other laboratory experiments men-
tioned is that our experiment involves giving to an external charity that is widely believed to be socially
valuable, while they use a VCM environment in which all the contributions are returned to participants in
the experiment. Since Eckel and Grossman (1996), economists have differentiated between preferences in
these two environments. For each of the three effects, Repseek, Influencer, and F-Influenced, the funda-
mental issues are different for actual charitable giving.
In VCM settings it is unclear whether people have the “Nash construct” and whether there is common
knowledge of rationality or of selfishness (Sugden, 1983). Even in a one-shot game, some people may
essentially assume (or act as if it is the case) that if they give more that means others are doing so as well
and hence they may simply be acting out of an ultimate desire to maximize their own payoffs.
Desire for reputation in the charity setting is also distinct. People may desire a reputation for being
cooperative in a game or for being generous to someone directly, but they may not care as much (or may
care more) about a reputation for being generous to third party charities. Intuition, as well as most simple
economic models would imply that, in the VCM environment, if subjects thought they had an influence,
they would want to use it to increase their own monetary payoffs (see, e.g., Chaudhuri et al (2006) for
experimental evidence of this). But with charitable giving, this is a distinct and nontrivial question that
depends on preferences. Economic models of giving differ over whether an individual gets utility from
others contributions, whether or not the individual caused others to make these contributions. According
to a strict warm glow model (Andreoni, 1990) others’ donations are inconsequential to an individual’s
utility. On the other hand, the warm glow could be interpreted to depend on an individual’s net impact on
total donations. If charity is a pure public good (“pure altruism”), an individual would gain as much utility
from inducing others to donate as from giving herself. The predictions of a Kantian (mentioned in Sugden,
1983) or reciprocity model (Sugden, 1984) are not obvious.
People may be influenced by previous contributions in the VCM setting but not the charity setting, or
vice-versa. As discussed above, if there is an influence in the charity setting it might be because previous
gifts signal the charity’s value, or signal the average level of altruism. It is important to understand whether
these factors matter for a typical charity rather than for a lab-specific public good.
Several field experiments involving actual charitable giving are relevant (in addition to the aforemen-
tioned Soetevent). At least three such experiments present potential contributors with carefully selected
information on previous contributions and observe how donations respond. Frey and Meier (2004) showed
that students were more likely to contribute (a specified discrete amount) to at least one social fund support-
ing other students when they were told that 64% of their peers donated (in a recent semester) in comparison
to when they were told that 46% of their peers donated (on average over the past ten years). The authors
claim this is evidence of conditional cooperation, as in Fischbacher et al (2001); we interpret it as sugges-
tive of F-Influence.
Shang and Croson (2005) have callers to a radio fund informed that “we had another
member, they contributed $75 [$180 or $300]” and find that the $300 treatment lead to significantly larger
average donations, an apparent F-Influence effect. Finally, Alpizar et al (2008) run a field experiment at a
national park in Costa Rica. They find an F-influence effect using a similar information-selection treatment
as the other two papers. In addition, they also find evidence for a reputation-seeking motive: contributions
made in front of the solicitor are significantly higher than contributions made in private, but donors did not
have a possibility to influence others due to their behavior.
See the discussion of “commitment theories” in Croson (2007).
While conditional cooperativeness could be a force behind F-influence, other models yield this as well.
These information-selection treatments are not explicitly deceptive the mentioned donations had
indeed been made at some point. The also pertain directly to a potential fundraising tool. However, these
findings do not correspond directly to the F-Influence effect we are seeking to identify.
It is unclear what
information a participant should draw from these announcements, as they were deliberately selected from
a large pool of previous donations some participant might have realized this. Furthermore, these papers
can not isolate the influence of an individual’s gift on her peers, and can not differentiate this effect by the
characteristics of the leader and follower.
Carman (2003) examines the workplace giving campaign from a large national bank. She finds some
evidence of a leader effect “captains” and “pace-setters” give more (even after controlling for their
observable characteristics) but she cannot disentangle Repseek and Influencer effects. She also finds ev-
idence for “the existence of social influences, but her estimates need to rely on a plausible but nontrivial
set of assumptions over the error structure and instrumental variable validity but in order to overcome the
usual obstacles to identifying social influences in a natural setting (see, e.g., Manski, 1993). She also notes
that she cannot separate “endogenous” from “contextual” effects, and hence cannot determine whether
contribution behavior responds to the “giving behavior of one’s peers” or to the peers’ inherent charac-
teristics. While our estimates may also be also vulnerable to this problem, we are able to test for at least
some potential contextual effects.
As List (2007) and Bardsley (2008) note, subjects’ behavior in dictator
games is influenced by subjects’ desire to conform to salient social norms. Experiments in this category
(including charitable giving experiments) are thus particularly vulnerable to experimenter-demand effects.
However, we are examining charitable giving in a social context meant to reflect real-world giving sit-
uations. We claim that similar social norms are present and influence subjects’ donation behavior in the
corresponding field settings (e.g., in a church with an open collection-plate) as well. We therefore argue
for the external validity of our results even in the presence of potential demand effects (see Zizzo, 2009,
for a formal discussion of this general argument).
In summary, although several papers identify various social influences on pro-social behavior, none
separately identify Repseek, Influencer, and F-influenced effects in the context of charitable giving, nor
can they link these effects and measure their consistency with each other.
3 Experimental Design
3.1 Implementation
Our experiments were conducted at the University of Jena on 80 male and 112 female subjects. Each
subjects received \euro8, reflecting the local hourly wage level for students, who constituted most of our
subjects. The experiment lasted for about one hour. In total we ran 12 sessions, each session consisting of
16 subjects.
In considering reputation motives, we cannot rule out the possibility that subjects may care about their
reputation vis-a-vis the experimenters, and this may interact with the Repseek effect and Influencer effect
in unknown ways. Thus, we used very strong measures to guarantee subject-experimenter anonymity, and
we made this anonymity common knowledge. The Jena University Experimental Lab was partitioned into
an outer and inner part; it was made clear to all subjects that the inner part could not be seen from the outer
Another distinction between our work and both Frey and Meier (2004)and Shang and Croson (2005) focus on contribu-
tions to smaller-scale causes which offer either concrete benefits to the donor (public radio) or to close peers (fellow students),
whereas our experimental contributions go towards a large-scale humanitarian cause.
We also speculate that contextual effects (in particular, the extent to which an individual’s personality motivates others
to give) are likely to be less important in our experiment than in the context of a giving campaign, as our subjects meet only
briefly and the meeting occurs before they know that the experiment involves charitable giving.
part. The server PC and the experimenter remained in the outer part. Subjects drew a random number in
private (not revealed to the experimenter) to determine their computer terminal. After the door separating
the inner and the outer lab was closed they were instructed to go to the terminals. The session in the inner
part of the lab was administered by a subject volunteer, who otherwise did not take part in the experiment.
This subject was instructed to supervise the meet and greet stage (outlined below), to communicate with the
experimenters in case subjects had questions during the experiment, and to distribute the sealed envelopes
with payments at the end of the experiment.
3.2 Sequence and reporting structure
Subjects were randomly assigned to two groups labeled Phase 1 and Phase 2: Phase 1 subjects made
their donation decisions before the Phase 2 subjects. Phase 1 subjects were given treatments that varied
in the reporting structure (no reporting, anonymous reporting, or identified reporting) and in when this
donation was to be reported (before or after phase 2 choices). The reporting structure is summarized in
figure 1. Phase 2 treatments differed in the information a subject received about another subject’s behavior.
Phase 2 subjects received either anonymous reports about phase 1 subjects’ donations, reports from phase
1 subjects they could identify, or no reports. Each subject participated in either Phase 1 or Phase 2.
3.3 Meet and greet stage
To create some social connection between potential partners and make them identifiable to each other, we
conducted a meet and greet (M&G) stage prior to the treatments. In this stage, subjects meet and exchange
their names and fields of study (or professions). To control for the possibility that this social interaction
has a direct effect on giving behavior (and to be able to test for this), all subjects participated in this stage,
whether or not they were in treatments with identified reporting. Subjects in the anonymous reporting or
no-reporting treatments were told (shortly before making their donation choice) that the person they met
would not see their donation, nor would they see this person’s contribution. Where the subject’s treatment
involved revealing his or her identity and donation to another subject, or receiving such a report, these
subjects were assigned to be “partners” in the meet and greet stage, and were informed of this. Subjects
who received a report at the end of the experiment were not informed that this would occur in advance.
At the point of the initial M&G stage subjects did not know what would follow - the detailed instruc-
tions came subsequently.
We used a subject volunteer both to allow subject-experiment anonymity and to ensure that our promise to donate these
funds to BFDW was credible. The subject volunteer was randomly assigned. As the computer screens were separated by
privacy barriers and the payment envelopes were sealed, the volunteer did not learn individual subjects’ contributions, and this
was made clear to all subjects.
This was implemented as follows: The volunteer monitored the M&G stage and was advised to inform us if something
went wrong. Subjects reported to their desk where they found - on the first page of their instructions - their subject number on
a sticker and an explanation of the meet and greet stage. The computer screen advised them then to stand up again and meet
the assigned partner, whose number was displayed on the computer screen. They then stood up and met the other subject and
exchanged names, ages and fields of study or job. The volunteer then - after five minutes - announced that it was time to return
to the computers and the experiment continued at the terminals. In sessions 10-12, as a robustness check, subjects also met in
a second meet and greet stage that was announced only to subjects whose decision and identity was reported to their partner,
after all decisions had been made. In the analysis we pool these two conditions, as we do not find significant differences in
behavior along our main treatments. (ranksum tests: p-values between 0.17 and 0.97)
Fig. 1 Summary of Experimental Design
3.4 Treatments
In the baseline “No Report” treatment (BL) subjects’ donations remained anonymous - neither the gift nor
the identity were revealed to any other subject within or outside the experiment.
In treatment Reporting Donation before phase 2” (RDb2) the donation but not the identity of the
subject was revealed before phase two to another subject, before this subject made a decision. In treatment
Reporting Donation after phase 2 (RDa2) the donation but not the identity was revealed after phase
two. Treatments RDIb2 and RDIa2 were similar to RDb2 and RDa2, but here both the leader and follower
knew each other’s identities from the M&G stage.
In treatment Received Anonymous Information (RecAI) phase 2 subjects received information on
the donation from a non-identified subject and in treatment “Received Identified Information” (RecIdI)
they received information from an identified donor. In both “received information” treatments a subject’s
second phase donation was not reported to any other subject.
3.5 Eliciting beliefs
After phase 1 subjects made their donation choices we elicited these subject’s to predict the average do-
nation of subjects who had neither received nor given a report. Subjects in a reporting treatment were also
asked to predict the gift their partner (the person to whom their donation was to be reported) made. The
subjects were paid e1 if their first guess was within a range of e1 of the actual donation or their second
guess was within e1 of the average donation.
At the end of the experiment, all payments were distributed
in numbered envelopes by the subject volunteer. Furthermore, subjects were given a report on an A4 page
We are aware that from a theoretical point of view this compensation may not elicit the mean of the distribution of a
subject’s beliefs, but will favor modal predictions. We choose this because it was easier to explain than a more complex rule
such as quadratic scoring, and because a feasible quadratic compensation scheme tends to offer rather flat incentives at points
close to the correct prediction.Sonnemans and Offerman (2001) offer evidence suggesting that even a flat fee leads to good
judgments; hence we do not expect the choice of scoring rule to matter much.
Table 1 Composition of Potential Motives Present in Phase 1 Treatments
Treatment Intrinsic Influence inside Influence Outside Reputation
Baseline X
Report don. (not id) before St. 2 (RDb2) X X X
Report don. (not id) after St. 2 (RDa2) X X
Report don. and id before St. 2 (RDIb2) X X X X
Report don. and id after St. 2 (RDIa2) X X X
on how much they donated (this report did not contain any information that could link the donation to the
subject) which subjects were to check and, if correct, put in a box at the entrance.
This box was then
emptied by the volunteer and the sum of all donations was calculated. The volunteer then observed that
the experimenters transferred the aggregate amount donated to the charity. This procedure was known to
the subjects before they made their donations.
4 Results and interpretation
Table 1 gives an overview of the hypothesized effects of each treatment on the “motives” for giving - i.e.,
the ways in which a subject’s own donation could increase her utility either directly or indirectly. An “X”
is marked in the table where we hypothesize that an effect may be present for some or all subjects. How-
ever, we allow that the presence of each of these motivations may lead to nonlinear effects on donations,
and there may be interactions in utility function itself (e.g., the marginal utility of reputation may decline
as the “impact utility” one gains from influencing others to give more increases). In the baseline treatment
nothing is reported to others; here giving can not be motivated by reputation concerns nor the desire to in-
fluence other subjects. We will label the remaining motivation to make an anonymous donation “intrinsic”;
this could include most of the models presented in recent work, including warm glow (Andreoni, 1990),
the desire for impact (Duncan, 2004), and/or the desire to supplement a public good. In treatments RDb2
and RDIb2, where the donation is “reported before, there also may be a motive to influence the subject
to whom the donation is reported to give more within the lab. When identity is also revealed (RDIb2 and
RDIa2) donating more may also boost a subject’s reputation. While reporting after phase 2 eliminates the
possibility of influencing another subject’s laboratory behavior, these reports (as well as the other reports)
may influence real world behavior after the end of the experiment, and reporters may anticipate this.
4.1 Phase 1 donations: Repseek and Influencer effects
We begin by analyzing donations in the first phase. Table 2 displays average donations by treatment,
broken down by whether and when the donation was reported (before or after phase 2) and according to
whether the identity was revealed when reporting.
A subject donates more on average when her identity and donation will be revealed to another subject,
irrespective of the timing of this reporting. As shown in table 2, this effect is stronger when the the reporting
occurs before the partner has the possibility to donate (see “Compared Treatments” No report - “Report
id. before”) than when the reporting occurs after all decisions have been made (“No report - Report id.
All reports were returned to the box.
Table 2 Summary Statistics: First Stage Contributions by Treatment (in Euros)
No report Report before Report after Total
Identity not reported
Mean 1.69 1.64 2.49 1.95
Std.Dev. 1.92 1.39 1.97 1.8
Median 1.75 1.00 2.00 1.75
Share pos. contr. (%) 60.67 79.17 92.31 82.64
Mean if contr. pos. 2.52 2.08 2.70 2.48
N 24 24 26 74
Identity reported
Mean 2.79 2.14 2.46
Std.Dev. 1.98 1.55 1.79
Median 3 1.5 2
Share pos. contr. (%) 95.83 95.83 95.83
Mean if contr. pos. 2.91 2.23 2.57
N 24 24 48
Mean 1.69 2.22 2.32 2.16
Std.Dev. 1.92 1.79 1.77 1.81
N 24 48 50 122
Test of differences in treatments
Compared Treatments Difference in \euro t-test ranksum test
No report - Report id. before 1.10* 0.056 0.021
No report - Report id. after 0.45 0.376 0.149
Report before: Id. - anon. 1.15** 0.025 0.025
Report after.: Id. - anon. -0.35 0.489 0.568
Pooled: Report before - after- -0.10 0.777 0.773
Pooled: Id. Report - No Id. report -0.38 0.289 0.284
* p<0.10, ** p<0.05, *** p<0.01
When a subject’s donation is reported before phase 2, contributions are significantly higher when the
identity is also reported: the difference is e1.15 (t-test p-value: 0.025, rank-sum p-value: 0.025). However,
there is no significant difference between the identity reporting treatments when the reporting occurs af-
terward (see Report after: Id. - anon”). Pooling across treatments, the difference between Report before
and Report after is small and not significantly different from zero (“Pooled: Report before - after”); nor
is the general difference between Identity reported and Identity not reported (“Pooled: Id. Report - No Id.
report”) significant.
Table 3 estimates Poisson regressions (the Poisson pseudo-maximum likelihood PPML estimator)
and OLS regressions of the first-phase donation decision, estimating a separate effect for each of the
reporting and reporting-timing treatment dummies. In table 3, columns (1) and (2) include the interactions
There is an apparent anomalous result: subjects whose anonymous donation will be reported to another subject at the end
of the experiment donate more on average than those whose anonymous donation will be reported to another subject before
(and hence have a chance to influence) this follower subject’s decision (\euro2.49 versus e1.64). This accords with our result
(Table 4) that anonymous leaders’ gifts have a negative but insignificant relationship to followers’ gifts. On the other hand,
as we show in section 4.3, both anonymous and identified leader subjects implicitly predict a positive relationship between
their donation and the follower’s choice. But such theoretical speculation is probably not necessary: this result (the greater
contribution of RDa2 subject) is only marginally significant in simple tests, and the relevant coefficient is insignificant in
controlled regressions (table 3) and disappears almost entirely when we control for a subject’s prediction for other subject’s
average contribution (table 3, columns 7 and 8). As the latter control may be a proxy for a subject’s own generosity, this
suggests that the apparent anomaly is an artifact of small-sample bias.
of the dummy variables, columns (3) and (4) add controls for gender and gender composition in the meet
and greet stage. In column (5) and (6) we control for whether the subject had already donated to BfdW or
a similar charity (outside the lab) in this year. Finally, columns (7) and (8) include the subjects’ prediction
for the average contribution as a regressor. This term is included to control for what subjects might see
as the appropriate gift and further reduce the potential small-sample bias from random variation in the
subjects assigned to each treatment.
Table 3 Regressions of Phase 1 Donations
Gender ctrls. Other ctrls. Belief ctrls.
(1) (2) (3) (4) (5) (6) (7) (8)
Psn. OLS Psn. OLS Psn. OLS Psn. OLS
Report before phase 2 (d) -0.055 -0.044 0.0037 0.013 0.084 0.11 -0.38 -0.33
(0.597) (0.483) (0.598) (0.499) (0.535) (0.434) (0.491) (0.410)
Report after phase 2 (d) 0.85 0.80 0.74 0.70 0.89 0.85 0.25 0.28
(0.618) (0.549) (0.624) (0.568) (0.594) (0.538) (0.432) (0.402)
Report Id. before 2 (d) 1.33** 1.15** 1.29** 1.13** 0.85 0.79* 1.26* 1.01**
(0.655) (0.494) (0.641) (0.496) (0.539) (0.459) (0.670) (0.507)
Report Id. after 2 (d) -0.31 -0.35 -0.32 -0.37 -0.43 -0.50 0.045 -0.0099
(0.408) (0.499) (0.401) (0.504) (0.403) (0.523) (0.304) (0.332)
Female (d) -0.33 -0.34 0.26 0.29 0.38 0.41
(0.520) (0.562) (0.290) (0.310) (0.257) (0.291)
Partner female (d) -0.60 -0.57
(0.522) (0.515)
Fem. × Par. fem. (d) 1.05 0.97
(0.804) (0.724)
Previously donated (d) 0.95** 0.97**
(0.383) (0.389)
Belief: av. contr. 0.49*** 0.69***
(0.089) (0.154)
Combined Coefficients
Report b4 + Id. Rep. 0.503* 1.104* 0.522** 1.143** 0.405 0.897* 0.344 0.679
( 0.268) ( 0.562) ( 0.267) ( 0.570) ( 0.259) ( 0.536) ( 0.221) ( 0.479)
Report after + Id. Rep. 0.236 0.450 0.183 0.333 0.189 0.352 0.149 0.273
( 0.270) ( 0.503) ( 0.297) ( 0.576) ( 0.265) ( 0.499) ( 0.193) ( 0.319)
Observations 122 122 122 122 122 122 122 122
0.061 0.081 0.126 0.391
Pseudo R
0.024 0.032 0.048 0.135
Regressions are made on contributions of all subjects who donated in the first phase.
Marginal effects; Robust standard errors in parentheses
(d) for discrete change of dummy variable from 0 to 1
* p<0.10, ** p<0.05, *** p<0.01
The additional effect of identification relative to anonymous “Report before phase 2” (the third row of
table 3) is significant and positive in all but one specification; the other marginal effects are not significant.
This is consistent with a greater desire to influence, and/or a greater anticipated influence, when a leader’s
identity is reported.
The coefficient on the binary variable “Previously donated” is positive and significantly different from
zero. This accords with previous evidence (e.g., Landry et al, 2008) that people who have previously
donated to a specific cause donate more when approached again.
While some might argue that this final control could be endogenous, reflecting a subject’s ex-post justification of her own
decision, we note that (i) subjects have a financial incentive to make accurate predictions and (ii) the inclusion of this control
does not substantially affect our key results.
4.2 Phase 2 donations: F-Influenced
We next examine whether subjects actually have an influence on their partner’s contribution (when their
gift is reported). In table 4, columns (3) and (4), we observe a positive and significant relationship between
the level of the reported gift and the donation of the subject to whom this was reported if the reporter’s
identity was observed. Columns (5) and (6) show that this effect is mainly driven by donations of subjects
whose partners are female.
Table 4 Regressions of Phase 2 Donations (subjects who received a report before Phase 2)
Ctrbn. Partner Id. interact Gender ctrls.
(1) (2) (3) (4) (5) (6)
Psn. OLS Psn. OLS Psn. OLS
Info. Id. (d) 0.77 0.76 -0.57 -0.75 -0.58 -0.96
(0.475) (0.488) (0.887) (1.066) (0.769) (0.886)
Partner Contribution 0.13 0.16 -0.48 -0.34 -0.49 -0.39
(0.174) (0.246) (0.398) (0.287) (0.362) (0.278)
Id. inf. × contr. P. 0.73* 0.74* 0.54 0.52
(0.426) (0.431) (0.423) (0.370)
Female (d) -0.40 -0.42
(0.475) (0.547)
Partner female (d) -0.30 -0.52
(0.498) (0.485)
Id. inf. × contr. P. × partner fem. 0.27* 0.55**
(0.149) (0.257)
Combined Coefficients
Contr. partner + id. inf.× contr. partner 0.142* 0.404 0.033 0.129
( 0.085) ( 0.322) ( 0.089) ( 0.227)
Contr. partner + id. inf.× c.p. × fem. 0.192*** 0.676***
( 0.040) ( 0.203)
Observations 48 48 48 48 48 48
0.078 0.170 0.271
Pseudo R
0.039 0.082 0.112
The regressions include all subjects who have received a report of a contribution.
This means that we excluded the 22 No report subjects of Phase 2.
Marginal effects; Robust standard errors in parentheses
(d) for discrete change of dummy variable from 0 to 1
* p<0.10, ** p<0.05, *** p<0.01
When the identities are not observed there appears to be a negative - although not significant - rela-
tionship between the gifts.
20, 21
Carman (2003) notes the possibility of a “contextual” influence from the “characteristics of one’s
peers” rather than through their giving behavior. As all of our subjects take part in the Meet and Greet, we
can test (in Table 5) for such “contextual” effects of M&G partners. Such effects may occur via character-
istics that we the researchers do not observe. For example, having an M&G partner with a sympathetic face
might cue a higher level of contribution. If this (partner’s) characteristic is also correlated to the partner’s
Subjects in the anonymous reporting condition may see their own gift as a substitute for the partner’s gift and therefore
donate less if their partner donated more; a “crowding-out” effect. The scale of this charity makes it doubtful seem that the
partner’s gift will significantly affect the marginal value of a contribution (as a public good). Still, subjects may gain warm
glow from their partner’s contribution, and this may have rapidly diminishing returns. Similarly, the subject may set a giving
target for the whole experimental group and the reported gift may lead them to increase their expectations over others’ total
contributions and hence lower their own contribution. In the case where the partner’s identity has been revealed, this crowding-
out effect might be replaced by a feeling of reciprocity and/or solidarity towards the other subject and therefore the gifts may
be complements
Note that the Phase 1 subject never observes the Phase 2 subject’s choice; there is no “back-reporting. Subjects in the
second phase did not know that there would be a second meet and greet stage (in sessions 10-12) after all subjects made their
donations, so they were not likely to fear that they might be asked by their partner what they donated (although we can not
rule out this expectation).
contribution, our estimate of the influence of a reported gift may be entangled with this contextual effect
and hence biased upwards. Columns 1 and 2 of table 5 measure contextual effects of M&G partners who
do not send or receive a donation report to one another.
We find no evidence of a “spurious” contextual
effect correlated to the M&G partner’s contribution the estimated parameters on “ContrMaGPartner”
are close to zero with small estimated standard errors.
We find some evidence of a contextual effect
through gender having a same-sex M&G partner seems to increase donations, although this effect is
only significant at the 10% level.)
Table 5 Contextual Effects
(1) (2) (3) (4)
M&G Contact (Poisson) M&G Contact (OLS) Id’d. report (Psn.) Id’d. report (OLS)
Contr. MaG Partner 0.014 0.029 0.11 0.33
(0.036) (0.077) (0.070) (0.302)
Female -0.13 -0.25 0.11 0.090
(0.228) (0.436) (0.653) (1.550)
Partner female -0.26 -0.47 0.47 1.13
(0.261) (0.457) (0.411) (1.082)
Fem. × Par. fem. 0.56* 1.08* -1.15 -2.02
(0.327) (0.604) (0.754) (1.717)
Constant 0.68*** 1.97*** 0.34 1.15
(0.171) (0.352) (0.471) (1.083)
Observations 144 144 24 24
0.035 0.242
Pseudo R
0.016 0.111
Robust standard errors in parentheses
* p<0.10, ** p<0.05, *** p<0.01
4.3 Predictions: perceived influence
If the increased giving by leaders is consciously driven by the desire to influence followers, this implies
that the leaders believe they have an influence. For leaders who “report before, we would like to know
what the leader believes is the follower’s response function the expected contribution as a function of the
amount reported to the follower. Suppose that the leaders’ gifts were determined exogenously, or that the
factors that determine how much a leader gives are uncorrelated to her beliefs about the follower’s response
function. In such a case the leader’s average expected response function could be consistently estimated as
the empirical slope of the “report before” leaders’ predictions in these leaders’ own contributions. Given
this exogeneity, for leaders who report at the end, the equivalent empirical slope should be zero, as their
ex-post reports can have no influence on prior behavior.
In the RDb2 and RDIb2 treatments, a leader’s predictions for his partner’s gift indeed does increase
in the leader’s own donation, as seen in the lower right and left scatterplots of figure 2.
This slope
is positive and significant in regression analysis across a variety of specifications (see appendix C.1).
That is, at least not until after the follower subject makes her own choice; we remove subjects in RDIb2 and RecIdI
treatments only.
Although we can not separate this contextual effect in the case where a leader’s identified gift is reported, these results
suggest that this should not be a concern.
In columns 3 and 4 we examine the 24 subjects who received an identified report from their M&G partner here we find
no significant contextual effects (of gender), but the small subsample size and consequent large standard errors implies that
this is not a very powerful test.
The scatterplot points are given a small random “jitter”. These are fit with lowess smoothed lines.
However, the slope is also moderately positive in the upper left and right plots, for the RDa2 and RDIa2
treatments, where no influence is possible. This suggests a potential latent variable bias: the expected
response functions are not orthogonal to own contributions. Those who contribute more may have a more
rose-colored view of the world, trusting charitable organizations and believing others to be generous
(and with a reciprocity motive belief that other subjects will donate may itself increase donations).
However, there is still an important difference between the Report before and Report after plots. While
22 of 50 subjects in the Report after treatments predict their partner will contribute zero, only 1 of 48
Report before subjects makes this prediction. This difference is strongly significant across a range of
parametric and nonparametric tests (available by request). To the extent that leaders predict an influence,
it seems to be at the extensive margin: some leaders believe that revealing their own contribution lead their
partner to give a positive amount rather than give nothing.
These results can be made robust to certain framing effects. It is conceivable that the Report be f ore
treatments have an independent effect on a subject’s beliefs over the contributions of others (but-for any
influence of reported donations), an effect that is distinct from that of the Report a f ter treatments, and an
effect that may vary according to a subject’s own generosity, potentially affecting the relative slopes in the
above figure. To the extent that this framing effect is the same for beliefs over “average” donations, we can
use the subject’s predictions of this as a control.
Figure 3 is similar to figure 2, but here the variables on both axes are expressed relative to (i.e.,
differenced from) the leader’s prediction of the average contribution. Here the slope is less pronounced;
naturally, predictions for partners are strongly positively related to predictions for the average subject (who
gets no report), and both are positively related to own contributions. Still, the slope appears somewhat
positive in the “report before” treatments and close to flat in the “report after” treatments. The significance
of these results are confirmed by regression analyses (in appendix C.1) that control for the “predicted
average” variable. Again, the results are driven by the predicted zero contributions in the “report before”
We do not claim to have eliminated all possible sources of bias. If there is indeed a Repseek and
an Influencer effect then leaders’ contributions themselves are likely to be higher (for a given level of
underlying preferences and beliefs) under these reporting treatments, and hence a particular size gift will
tend to come from a less generous person under these reporting treatments. To the extent that those who
donate more tend to predict that their partner would give more (but for the influence of their own reporting)
relative the mean, this would lead towards a bias against finding a predicted influence.
We see no evidence that predicted influence differs substantially between identified and unidentified
reporting. Although, as demonstrated in section 4.2, the influence appears stronger when identities are
reported, the leaders predictions do not reflect this.
This effect is known in the psychological literature as the “false consensus effect”. Subjects use their own type (in this case
preference for charity) to make inferences over the generosity of the population. Insofar as the charitable giving interaction
resembles a VCM game, this would also reflect the common observation that “those who believe others will cooperate in
social dilemmas are more likely to cooperate themselves” (Ostrom, 2000). The same patterns are consistent with the idea that
(in spite of the monetary incentive to make a correct prediction) elicited beliefs are in part a “justification” of a players own
This interpretation is further supported by the significant positive slope of predicted average contributions in own contri-
butions (details by request).
As shown in appendix C.1, this “zero-driven” difference in relative average slope leads to a significant interaction coef-
ficient in “one-stage” regression specifications but not in the conditional-on-positive part of two-part specifications such as
5 Conclusion
This experiment is the first (to the best of our knowledge) to disentangle three components of the mo-
tivation for charitable giving in social settings. We find that when a leader’s donation and identity are
revealed to a follower (before phase 2) the latter’s donation increases in the former’s. When the donation
is reported without the leader’s identity no such “influence” is observed. The “identified leader” subjects,
the only subjects that have a significant influence, give significantly more than the subjects in the no re-
porting (baseline) treatment, and also significantly more than the subjects whose donation and identity is
be reported after the follower subject’s choice. These findings provide some evidence that individuals’
utility comes not only from their own contributions, but also from their impact on the total amount con-
tributed. The “identified leaders” correctly perceive that their contribution will have a positive influence
on the subject it is reported to, and thus donate more to take advantage of what is, in effect, a temporary
price reduction.
The evidence for a reputation-seeking effect is mixed. While subjects whose donations and identities
are reported to another subject only at the end of the experiment do donate significantly more than baseline
subjects, there is no significant difference from subjects who report only their donations at the end of the
experiment. Further, we do not find a significant effect for pairs with any particular gender composition.
However, reputation seeking is likely to depend on the nature and closeness of individual relationships;
hence, we cannot rule out the possibility that reputation-seeking drives charitable giving in certain real-
world environments.
A follower’s gift tends to tends to increase in a leader’s gift only when the leader is identified. This
suggests that influence is a complex psychological phenomenon, not the product of a consistent Bayesian
updating (of beliefs over the average contribution nor over the worth of the charity); on average, such
updating should be the same for identified and unidentified reports, as both are random draws from the
population of subjects.The influence is found particularly when the leader is female.
This relates to
earlier findings that female solicitors are more successful then males in some contexts (Lindskold et al,
1977, Landry et al, 2006), and contrasts with evidence that men tend to be more influential in general
social settings (Carli, 2001).
Leader subjects’ predictions suggest that they believe they have an influence on their partner’s dona-
tion; in particular, they seem to predict that they will have an influence on their partner’s extensive margin
decision (to donate versus not donate).
Our paper makes an important contribution to the literature on the motivation for altruistic decision-
making in a social context, as well as to the evidence on gender roles in the arena of other-regarding
behavior. Our findings are also relevant to the study of group dynamics, leadership, and influence (see,
e.g., Hoyt et al, 2006). Finally, our results have direct implications for practitioners. Fundraising cam-
paigns that employ peer-driven strategies such as “leadership giving”
and “conspicuous compassion”
(Bekkers and Wiepking, 2008) must do so carefully: in our experiment receiving a report of a “leader’s”
decision does not increase followers’ contributions on average; only reports of larger gifts have a posi-
tive effect. Furthermore, “disembodied” anonymous reports may not be influential: follower’s gifts only
Furthermore, given subjects may choose to discuss their choices outside the lab (and in the end meet and greet stage),
reputation-seeking may be present in all treatments; but if so, this relies on the subjects’ reluctance to lie (see for example
Vanberg, 2008); the enforced reporting does not significant contribute to this.
This suggests that charities may prefer to focus on using females as leaders and pacesetters.
Andreoni (2006) notes “The importance of the leadership phase of fund-raising is emphasized in almost all handbooks for
have a positive relationship to reports from identified leaders. However, reports need not come from close
connections to be influential even the brief acquaintance our subjects have with their “meet and greet”
partner appears to matter. Finally, our results suggest that charities may prefer to focus on using females
as leaders and pacesetters.
Acknowledgements We would like to thank the British Academy and the Deutsche Forschungsgemeinschaft under the grant
RTG 1411 for providing funds for our research and experiments. We are also grateful for comments and advice from Emmanuel
Saez, Joao Santos-Silva, Christian Ghiglino, James Andreoni, Adriaan Soetevent, Daniel Zizzo, and participants in seminars
at the University of Essex, the University of Jena and the Max Planck Institute of Economics Jena, the Economic Science
Association conference, Erasmus University Rotterdam, the University of Bristol, and the IAREP conference in Canterbury.
Alpizar F, Carlsson F, Johansson-Stenman O (2008) Anonymity, reciprocity, and conformity: Evidence
from voluntary contributions to a national park in Costa Rica. Journal of Public Economics 92(5-
Andreoni J (1990) Impure altruism and donations to public goods: A theory of warm-glow giving. The
Economic Journal 100:464–477
Andreoni J (2006) Leadership Giving in Charitable Fund-Raising. Journal of Public Economic Theory
Andreoni J, Petrie R (2004) Public goods experiments without confidentiality: a glimpse into fund-raising.
Journal of Public Economics 88:1605–1623
Bardsley N (2008) Dictator game giving: altruism or artefact? Experimental Economics 11:122–133
Bekkers R, Wiepking P (2008) Generosity and Philanthropy: A Literature Review. Working paper
Bernheim B (1994) A Theory of Conformity. Journal of Political Economy 102(5):841
Carli L (2001) Gender and social influence. Journal of Social Issues: Gender, Hierarchy, and Leadership
Carman K (2003) Social influences and the private provision of public goods: Evidence from charitable
contributions in the workplace. Manuscript, Stanford University
Charness G, Gneezy U (2008) What’s in a name? anonymity and social distance in dictator and ultimatum
games. Journal of Economic Behavior and Organization
Chaudhuri A, Graziano S, Maitra P (2006) Social learning and norms in a public goods experiment with
inter-generational advice. Review of Economic Studies 73(2):357
Cooter R, Broughman B (2005) Charity, Publicity, and the Donation Registry. The Economists’ Voice
Croson R (2007) Theories of commitment, altruism and reciprocity: Evidence from linear public goods
games. Economic Inquiry 45(2):199–216
Duncan B (2004) A theory of impact philanthropy. Journal of Public Economics 88(9-10):2159–2180
Eckel C, Grossman P (1996) Altruism in anonymous dictator games. Games and Economic Behavior
Fischbacher U, Gächter S, Fehr E (2001) Are people conditionally cooperative? Evidence from a public
goods experiment. Economics Letters 71(3):397–404
Frey B, Meier S (2004) Pro-social behavior in a natural setting. Journal of Economic Behavior and Orga-
nization 54(1):65–88
Frey and Meier (2004) Social comparisons and pro-social behavior: Testing "conditional cooperation" in
a field experiment. AER
Glazer A, Konrad K (1996) A signaling explanation for charity. American Economic Review 86(4):1019–
Guth W, Levati MV, Sutter M, van der Heijden E (2007) Leading by example with and without exclusion
power in voluntary contribution experiments. Journal of Public Economics 91(5-6):1023–1042
Harbaugh WT (1998a) The prestige motive for making charitable transfers. The American Economic
Review 88(2):277–282
Harbaugh WT (1998b) What do donations buy? a model of philanthropy based on prestige and warm glow.
Journal of Public Economics 67(2)(2):269–84
Hoyt C, Goethals G, Riggio R (2006) 5. Leader-follower relations: group dynamics and the role of leader-
ship. The quest for a general theory of leadership p 96
Kirsch A, Hume K, Jalandoni N (2000) Giving and Volunteering in the United States: Findings from a
National Survey, 1999 Edition
Landry C, Lange A, List J, Price M, Rupp N (2006) Toward an Understanding of the Economics of Charity:
Evidence from a Field Experiment*. Quarterly Journal of Economics 121(2):747–782
Landry CE, Lange A, List JA, Price MK, Rupp NG (2008) Is a donor in hand better than two in the bush?
evidence from a natural field experiment. NBER Working Papers 14319, National Bureau of Economic
Research, Inc
Levati MV, Sutter M, van der Heijden E (2007) Leading by example in a public goods experiment with
heterogeneity and incomplete information. Journal of Conflict Resolution 51:793–818
Lindskold S, Forte R, Haake C, Schmidt E (1977) The effects of directness of face-to-face requests and
sex of solicitor on streetcorner donations. Journal of Social Psychology 101(1):45–51
List J (2007) On the Interpretation of Giving in Dictator Games. Journal of Political Economy 115(3):482–
Manski CF (1993) Identification of endogenous social effects: The reflection problem. The Review of
Economic Studies Vol. 60, No. 3.
Ostrom E (2000) Collective action and the evolution of social norms. The Journal of Economic Perspec-
tives pp 137–158
Potters J, Sefton M, Vesterlund L (2001) Why Announce Leadership Contributions? An Experimental
Study of the Signaling and Reciprocity Hypotheses. University of Pittsburgh mimeo
Potters J, Sefton M, Vesterlund L (2005) After youâendogenous sequencing in voluntary contribution
games. Journal of Public Economics 89(8):1399–1419
Rabin M (1993) Incorporating Fairness into Game Theory and Economics. The American Economic Re-
view 83(5):1281–1302
Randolph WC (1995) Dynamic income, progressive taxes, and the timing of charitable contributions. The
Journal of Political Economy 103(4):709–738
Rege M, Telle K (2004) The impact of social approval and framing on cooperation in public good situa-
tions. Journal of Public Economics 88(7-8):1625–1644
Romano R, Yildirim H (2001) Why charities announce donations: a positive perspective. Journal of Public
Economics 81(3):423–447
Shang J, Croson R (2005) Field Experiments in Charitable Contribution: The Impact of Social Influence
on the Voluntary Provision of Public Goods. Working paper
Soetevent A (2005) Anonymity in Giving in a Natural Context: An Economic Field Experiment in Thirty
Churches. Journal of Public Economics 89(11-12):2301–2323
Sonnemans J, Offerman T (2001) Is the quadratic scoring rule really incentive compatible? University of
Amsterdam manuscript
Sugden R (1983) On the economics of philanthropy: Reply. The Economic Journal 93(371):639
Sugden R (1984) Reciprocity: The supply of public goods through voluntary contributions. The Economic
Journal 94(376):772–787
Vanberg C (2008) Why Do People Keep Their Promises? An Experimental Test of Two Explanations.
Econometrica 76(6):1467–1480
Vesterlund L (2003) The informational value of sequential fund-raising,. Journal of Public Economics
Zizzo DJ (2009) Experimenter demand effects in economic experiments. Experimental Economics
A Screenshots
Fig. 4 No report
Donation Phase
Your endowment: e8.00
Your decision will remain completely anonymous.
Please enter the amount, that you would like to donate to Brot fï¿œr die Welt:
Note to editor: For space considerations we recommend that this entire appendix (except perhaps part D) be put as an
online supplement.
Fig. 9 Treatments Phase 2
Donation Phase
Your endowment: e8.00
Another participant in this experiment has donated to Brot fï¿œr die Welt: \euro 0.00
Your decision will remain completely anonymous.
Please enter the amount, that you would like to donate to Brot fï¿œr die Welt:
Fig. 10 Treatments Phase 2
Donation Phase
Your endowment: e8.00
Participant 8 your partner in the meet and greet stage has donated to Brot fï¿œr die Welt: e0.00
Your decision will remain completely anonymous.
Please enter the amount, that you would like to donate to Brot fï¿œr die Welt:
B Protocols
B.1 Experimenter oral presentation: Wording and protocol
“Thank you for showing up for this experiment. We first will determine a volunteer who will help us in administrating the
[Determination of Volunteer]
“The laboratory is - as you can see - divided into two parts. The inner part [SHOW THE INNER PART, WHERE SUB-
JECTS ARE SEATED] can not be observed from outside. This assures the anonymity of your decisions. The experiment starts
with a Meet and Greet Stage, in which you will introduce yourself to another participant. This event will take place in the
inner part of the lab and will take 5 minutes. To find your assigned partner easily you will get a sticker with a number on it.
Please place this sticker visibly on your cloths. Further instructions will follow. Please do not forget to take off your sticker
before leaving the lab. Now please draw a sticker and go to the place indicated by the sticker number in the inner part of the
If there a no further questions, please go into the inner part of the lab the last person please closes the door [EXPERI-
MENTERS ARE IN THE OUTER PART] Your payment will be distributed in envelopes by the volunteer at the end of the
B.2 Payment Procedure
After all decisions have been made a file with subject numbers and payoffs is generated. The experimenters put the amount
in a labeled envelope and close it. The envelopes have all roughly the same weight. After that, the volunteer is called in to
distribute the envelopes to the subjects. Subjects are then asked to sign that they have received the money and put the receipt in
a closed box before leaving the lab. This box was then emptied by a research assistant who was not familiar with this particular
experiment to do the accounting.
C Supplementary Statistics and Data
C.1 Perceived influence: regressions
We can implicitly measure the leaders’ beliefs about their influence using their elicited beliefs about others’ contributions.
Bl f pn, a subject’s reported belief of the contribution of her partner,
Ctbn, a subject’s own contribution,
Bl f av a subject’s reported belief of the average contribution of a “subject who does not get a report”,
Rpb2, a dummy indicating whether the subject’s contribution was reported to another subject before that subject’s contri-
bution decision,
Idd, a dummy indicating whether the subject’s identity was reported with this contribution.
Table 6 Perceived influence
(1) (2) (3) (4) (5) (6) (7) (8)
Poisson OLS Psn OLS Psn, donors Psn Probit Probit
Dependent var.: BLFPN BPGEN
BLFAV 0.15** 0.33* 0.15** 0.34* 0.14** 0.19** -0.094 -0.097
(0.054) (0.14) (0.049) (0.14) (0.048) (0.050) (0.089) (0.090)
RPT B4 0.20 -0.043 -0.071 -0.28 -0.14 0.088 -0.058 -0.049
(0.25) (0.39) (0.29) (0.44) (0.32) (0.18) (0.36) (0.41)
CT BN 0.041 0.021 0.038 0.020 0.021
(0.093) (0.19) (0.091) (0.19) (0.093)
RPT B4 × CT BN 0.097 0.40* 0.18+ 0.44* 0.19+
(0.091) (0.19) (0.10) (0.21) (0.11)
IDD × RPB4 0.48* 0.60 0.53* -0.019
(0.23) (0.43) (0.25) (0.49)
IDD × RPB4 × CT BN -0.12+ -0.11 -0.13+
(0.066) (0.16) (0.069)
BGEN -0.35 -0.085 -0.085
(0.30) (0.36) (0.36)
RPT B4 × BGEN 0.81* 1.25* 0.99
(0.32) (0.55) (0.70)
IDD × RPB4 × BGEN 0.46
Summed (net) Coefficients
IDD × RPB4 .41 .33 .38 -.07
(.26) (.46) (.27) (.46)
IDD × RPB4 × CT BN .06 .33+ .06
(.09) (.20) (.09)
IDD × RPB4 × BGEN 1.44*
Observations 98 98 98 98 89 98 98 98
(Pseudo) R
0.105 0.363 0.111 0.371 0.095 0.102 0.099 0.102
Robust standard errors in parentheses
+ p<0.10, * p<0.05, ** p<0.01
Subset: First-stage subjects who report their donations at some point.
In columns 1 and 2 we estimate versions of the following equations with both Poisson and OLS regressions:
Bl f pn = f (
RPB4 +
Bl f av +
Ctbn +
Rptb2 × Ctbn) +
Equation 1 models a homogeneous predicted influence, pooling identified and unidentified reports. In columns 3 and 4
we allow an interaction with the identified treatment, estimating:
BLFPN = f (
Rptb2 +
× Idd × Rptb2 +
Bl f av
Ctbn +
Rpb2 × Ctbn +
Idd × Rpb4 × Ctbn) +
coefficients are intended to capture effects other than the leader’s believed influence of her donation on the
follower. The Rpb2,Idd, dummy variables and the interactions of these serve as controls for any independent framing effect
of these treatments on beliefs, and also differences away what the leader perceives to be the secular effect of receiving an
(identified) report on the follower’s donation, irrespective of the amount reported. Bl f av acts as a control for the leader’s
overall belief over what followers will donate; inclusion of this control serves to reduce the possibility of small-sample bias
(as latent beliefs may not be evenly spread across the randomized treatments), and also controls for any potential effect of
the leader’s treatment on her own overall belief of others’ contributions. As mentioned in the main text, Ctrbn removes the
potential bias from a correlation between (latent determinants of) one’s own generosity, and one’s beliefs over one’s partner.
This variable also captures a possible direct effect a leader’s contribution may have on her predictions for her partner (relative
to the prediction for the average gift), e.g., if the prediction is adjusted as an ex-post justification of her own choice.
Column 5 runs the same specification as column 3, but limits the estimator to those leaders who made a nonzero donation.
The predicted influence may be nonlinear; leaders may believe that reporting a gift that is above average (or higher than
the gift the donation follower otherwise would have made) will have a strong positive effect on the follower’s gift. To detect
this, we define the dummy variables BGEN = 1(CT BN > BLFAV ) and BPGEN = 1(BLFPN > BLFAV ) and estimate the
specifications below (leaving out the continuous “own contribution” variables to reduce multicollinearity). In column 5 we
BLFPN = f (
RPB4 +
IDD × RPB4 +
IDD × RPB4 × BGEN) +
We also measure the predicted extensive margin influence in columns 7 and 8 using the specifications:
BPGEN = f (
RPB4 +
RPB4 × BGEN) +
BPGEN = f (
RPB4 +
IDD × RPB4 +
IDD × RPB4 × BGEN) +
C.2 Data from the Questionnaire
Additionally to the experimental data, we have information from a short survey at the end of the experiment, where we asked
about subjects motivation to donate. The options we presented to them were in line with the questions we want to answer
with the experiment: Desire to influence, warm glow, being influenced and reputation. Contrary to our experimental results
only 5% of the subjects said that reputation did play a role in their giving behavior. They also did not acknowledge that
they were influenced by their partners donation. This speaks for the necessity of complementing survey data with carefully
collected experimental data, as policy relevant behavior that is either influenced by subconscious processes or is socially not
fully acceptable is hardly revealed by surveys only.
Fig. 2 Predictions by Own Contribution
Fig. 3 Predictions by own contribution; relative to predicted average
Fig. 5 Report donation only before phase 2
Donation Phase
Your endowment: e8.00
Your decision, but not your identity, will be shown to another participant for information purposes, be for
he or she makes a donation.
This participant will see your donation again, after all decisions in this experiment have been made.
Please enter the amount, that you would like to donate to Brot fï¿œr die Welt:
Fig. 6 Report Donation Only After Phase 2
Donation Phase
Your endowment: e8.00
Your decision, but not your identity, will be shown to another participant for information purposes, be
for he or she makes a donation. The other person will see your donation again, after all decisions in this
experiment have been made.
Please enter the amount, that you would like to donate to Brot fï¿œr die Welt:
Fig. 7 Report Donation and Identity Before Phase 2
Donation Phase
Your endowment: e8.00
Your decision and your identity will be shown to participant 16 for information purposes, before he or she
makes a decision.
Participant 16 was your partner in the “Meet and Greet” stage. You will meet this participant again at the
end of the experiment.
This participant will see your donation again, after all decisions in this experiment have been made.
Please enter the amount, that you would like to donate to Brot fï¿œr die Welt:
Fig. 8 Report Donation and Identity After Phase 2
Donation Phase
Your endowment: e8.00
Your decision and your identity will be shown to participant 12 for information purposes after all decisions
have been made.
Participant 12 was your partner in the “Meet and Greet” stage. You will meet this participant again at the
end of the experiment.
Please enter the amount, that you would like to donate to Brot fï¿œr die Welt:
... Past studies have found charitable behavior to be influenced by perceptions of what others are doing or giving, and that such conditional cooperation is able to increase both contribution rates and donation levels (Wiepking & Heijnen, 2011). This is because charitable actions can accrue benefits to social reputation, while uncharitable actions can incur costs to social reputation (Milinski et al., 2006;Reinstein & Riener, 2012). For instance, a natural field experiment found that manipulating the social information available to potential donors by altering what was visible in a donation box had a significant impact on donation composition (coins or bills), frequency, and value (Martin & Randal, 2008). ...
... Croson et al. (2009) also found that donation behavior is influenced by norm perceptions, with donors making higher contributions if they are under the impression that others are also making large donations. Increasing the visibility of acknowledgements of monetary contributions to conservation efforts by hard-cooperators may also play positively into the psychology of charitable giving and ultimately generate more funds for conservation (Reinstein & Riener, 2012). ...
Full-text available
Conservation needs adequate support and funding to address our ecological crises. People support conservation in different ways, from social media engagement to donating money. Various factors influence how people choose to support conservation, including social norms and ecological status. The rise of social media has provided people with an easy and low‐cost way to support conservation: sharing information online. How valuable is social media engagement and activism for conservation funding and outcomes? Here, we develop an evolutionary game‐theoretic framework to understand the complex interactions between individuals in the context of social media information sharing, conservation philanthropy, and how these interactions ultimately impact ecological outcomes. From a game theory perspective, we can consider donors to be hard‐cooperators, sharers of information on social media to be soft‐cooperators, and those who do nothing to be non‐cooperators. Our model shows that soft‐cooperators can help stabilize conservation funding flows and develop social norms. Supporting conservation through social media sharing can ultimately contribute to conservation success. Our study conceptualizes the complex decision‐making processes of conservation funding and affirms the importance and value of mobilizing all types of supporters in conservation. In this manuscript, we develop a novel game‐theoretic framework to understand the complex interactions between individuals in the context of information sharing on social media, conservation philanthropy, and how these interactions ultimately impact ecological outcomes. Our results demonstrate the importance of mobilizing people who are weakly engaged in conservation action to stabilize funding flow and ecological status.
... There are several experimental [40][41][42] and theoretical [43][44][45] studies on charitable activities. For instance, there are studies about the social network supporting charity in New York City [46], about the optimal location of charity boxes for maximizing public donations [47], and about gamification of donation-based crowdfunding campaigns to improve the engagement of users [48]. ...
Many people worldwide depend on charitable donations to survive. Consider a social network engaged in charitable activities. Let this network be composed of individuals with heterogeneous donation behaviors: there are individuals that are influenced by the neighborhood activity and individuals that act by themselves. In this article, a model formulated in terms of a spatial iterated game is proposed to investigate the time evolution of the number of donors in this social network. In the proposed model, the population spatial structure is represented by a three-dimensional torus and the social interactions are represented by a payoff matrix. Here, the payoffs are a measure of the feeling of personal satisfaction for making (or not) a charitable contribution. Numerical simulations are performed in order to find out how the population size, the payoff for simultaneous donations, the probability of spontaneous donation, and the proportions of the distinct charitable behaviors affect the number of donors. The simulations reveal that an increase in the proportion of impure altruistic behavior improves the engagement in charitable projects if the donations motivated by pure altruism occur with low probability. The practical implications of this study can be especially relevant in times of economic crisis, like the one we are experiencing.
... This naming service costs $8.80 and is a way for donors to reveal their identity. Past literature indicates that identity plays an important role in charitable giving as it is associated with additional extrinsic motivations such as reputation [Reinstein and Riener, 2012]. The availability of ENS in Ethereum and its unavailability in Bitcoin could potentially make these two blockchains incomparable. ...
Full-text available
Extrinsic incentives such as a conditional thank-you gift have shown both positive and negative impacts on charitable fundraising. Leveraging the crypto donations to a Ukrainian fundraising plea that accepts Ether (i.e., the currency of the Ethereum blockchain) and Bitcoin (i.e., the currency of the Bitcoin blockchain) over a seven-day period, we analyze the impact of crypto rewards that lasted for more than 24 hours. Crypto rewards are newly minted tokens that are usually valueless initially and grow in value if the corresponding cause is well received. Separately, we find that crypto rewards have a positive impact on the donation count but a negative impact on the average donation size for donations from both blockchains. Comparatively, we further find that the crypto rewards lead to an 812.48% stronger donation count increase for Ethereum than Bitcoin, given that the crypto rewards are more likely to be issued on the Ethereum blockchain, which has higher programmability to support smart contracts. We also find a 30.1% stronger decrease in average donation amount from Ethereum for small donations (<=$250); the rewards pose similar impacts on the average donation size for the two blockchains for large donations (>$250). Our study is the first work to look into crypto rewards as incentives for fundraising. Our findings indicate that the positive effect of crypto rewards is more likely to manifest in donation count, and the negative effect of crypto rewards is more likely to manifest in donation size.
We investigate whether a desire to influence others’ choices affects our own behavior. To separate such influence motives from social pressure, we study participants’ willingness to register for a COVID-19 vaccination in a field experiment in Germany. We vary whether participants’ registration decisions are shared with a peer, to activate social pressure, and whether peers are informed before their own decision, to isolate influence motives. We find that influence motives double participants’ registration likelihood, an effect driven by individuals with ex ante trust in the vaccine. Despite anticipating to influence their peers, participants cannot alter peer behavior in the experiment. (JEL C93, D12, D82, D91, I12, I18)
Corporate charitable giving (CCG) is increasingly pervasive in the global hospitality industry. Among various subtypes of CCGs, many hospitality companies request donations from customers during the service delivery or payment process to encourage customer engagement and co-created social responsibility. Karma—the implicit belief that the universe rewards good deeds and punishes wrongdoings—resonates with consumers’ prosocial behaviors, yet little is known about how karma beliefs, along with various donation elements, affect donation likelihood. Study 1 demonstrates that karma believers are more likely to donate gifts-in-kind than money, as it signals altruism that engenders karmic rewards. Study 2 shows that consumers with weak (vs. strong) karma beliefs are more likely to donate in a public (vs. private) setting due to greater reputational benefits. This research enriches the literature on hospitality CCG marketing and provides insights to managers regarding how to effectively execute donation programs.
Donation has an important role in bridging the rich and poor gap and maintaining social stability. With the development of Fintech, selfless support in crowdfunding has become a new channel of donation. This paper uses Python to capture the data from and studies the donation behavior in crowdfunding. The results of our study show that the higher financing performance of a crowdfunding project is the result of the higher number of people who support the crowdfunding project selflessly and anonymously. The donation behavior of crowdfunding investors stems from social needs and thus shows the characteristics of “pro-sociality”. The anonymous donation behavior matches the specific target audience through signal burying. Further research shows that the impact of donation behavior on financing performance is significantly heterogeneous. In agricultural and public welfare projects, the effect of selfless donation and anonymous donation on financing performance is more significant. The donation behavior prompts the crowdfunding project to complete the financing target more quickly and get more time for over-funding.
Charitable fundraisers frequently announce giving by others, and research shows that this can increase donations. However, this mechanism may not put information about peers to the most efficient use if it is costly to inform individuals who are indifferent to peer actions or causes some individuals to give less. We investigate whether a simple mechanism without incentives can predict heterogeneity in charitable responses to peer decisions. We elicit beliefs about donations in a baseline solicitation, and in subsequent solicitations we randomly assign information about others’ donations. We find that elicited beliefs are often logically inconsistent and that many subjects fail to update beliefs when treated. Elicited beliefs did not predict heterogeneous treatment effects on beliefs or donations, making the strategy unlikely to succeed unless individuals are engaged and the information is salient.
Full-text available
A persistent puzzle in the social and biological sciences is the existence of prosocial behav- ior, actions that benefit others, often at a cost to oneself. Recent theoretical models and empirical studies of indirect reciprocity show that actors behave prosocially in order to develop an altruistic reputation and receive future benefits from third parties. Accordingly, individuals should stop investing in reputations via prosocial behavior when a future bene- fit (via indirect reciprocity) is unlikely. The conclusion that the absence of reputational incentives necessarily leads to egoistic behavior contrasts sharply with models of heteroge- neous social preferences. Such models demonstrate the theoretical plausibility of popula- tions composed of egoists and altruists. Results of Study One show that actors classified a priori as egoists respond strategically to reputational incentives, whereas those classified a priori as altruists are less affected by these incentives. Egoists act prosocially when reputa- tional incentives are at stake but not when opportunities for indirect reciprocity are absent, while altruists tend to act prosocially regardless of whether reputational incentives are pre- sent. These results suggest that altruistic behavior can result from non-strategic altruism or reputation-building egoism. Study Two replicates these results and explores indirect recip- rocation of others' prosocial acts. We found that altruists indirectly reciprocate at higher levels than egoists, and individuals tend to discount others' prosocial behaviors when they occur in the presence of reputational incentives. As a result, public prosocial behaviors are indirectly reciprocated less than private prosocial behaviors. In line with our argument that altruists pay less attention to reputational incentives, egoists showed a greater tendency than altruists to discount others' public prosocial behaviors. The results support the grow- ing focus on heterogeneity of individuals' social preferences in models of altruism and indi- rect reciprocity.
Full-text available
In this paper we study the effect of social influence in the voluntary provision of public goods in two field experiments. In the first field experiment we demonstrate the existence of a social influence effect on individual contributions. We explore the effectiveness of different levels of social information, and find the most influential to be information drawn from the 90th to 95th percentile of previous contributions. In our experiment, social influence increases contributions on average 12% ($13) for all donors in the most effective condition. Further, these increased contributions do not crowd out future contributions. In our second field experiment we demonstrate the boundary conditions of the effect. The results highlight the social cause of our results rather than an alternative cognitive cause (anchoring-and-adjustment or reference points).
Experimenter demand effects refer to changes in behavior by experimental subjects due to cues about what constitutes appropriate behavior. We argue that they can either be social or purely cognitive, and that, when they may exist, it crucially matters how they relate to the true experimental objectives. They are usually a potential problem only when they are positively correlated with the true experimental objectives’ predictions, and we identify techniques such as non-deceptive obfuscation to minimize this correlation. We discuss the persuasiveness or otherwise of defenses that can be used against demand effects criticisms when such correlation remains an issue.
Charities publicize the donations they receive, generally according to dollar categories rather than the exact amount. Donors in turn tend to give the minimum amount necessary to get into a category. These facts suggest that donors have a taste for having their donations made public. This paper models the effects of such a taste for "prestige" on the behavior of donors and charities. I show how a taste for prestige means that charities can increase donations by using categories. The paper also discusses the effect of a taste for prestige on competition between charities. (C) 1998 Elsevier Science S.A.
We compare a partners condition, where the same small group of subjects plays a repeated public good game, to a strangers condition, where subjects play this game in changing group formations. From the first period onward, subjects in the partners condition contribute significantly more to the public good than subjects in the strangers condition. Strangers' contributions show continual decay, while partners' contributions fluctuate on a high level prior to decreasing in the final periods. We interpret subjects' behaviour in terms of conditional cooperation which is characterized by both future-oriented and reactive behaviour.
Approximately 3000 male and female pedestrians (mostly college students) were solicited for a donation to a charitable organization. Three male and three female undergraduates served as solicitors and made either a direct, face-to-face appeal or a less direct, impersonal appeal. As hypothesized, the direct appeal was more successful than the impersonal appeal. Also, donors receiving the direct appeal gave larger amounts when the traffic was comparatively light, but the proportion giving was unaffected by traffic density. The sex of the solicitor made a difference only with the impersonal appeal. Females may be viewed as more trustworthy in situations in which suspicion of the genuineness of the request is most relevant; face-to-face requests of a relatively nonsuspicious kind may appear equally trustworthy whether made by male or female solicitors.