ArticlePDF Available

Decomposing Desert and Tangibility Effects in a Charitable Giving Experiment


Abstract and Figures

Several papers have documented that when subjects play with standard laboratory “endowments” they make less self-interested choices than when they use money they have either earned through a laboratory task or brought from outside the lab. In the context of a charitable giving experiment we decompose this into two common artifacts of the laboratory: the intangibility of money (or experimental currency units) promised on a computer screen relative to cash in hand, and the distinct treatment of random “windfall” gains relative to earned money. While both effects are found to be significant in non-parametric tests, the former effect, which has been neglected in previous studies, has a stronger impact on total donations, while the latter effect has a greater impact on the probability of donating. These results have clear implications for experimental design, and also suggest that the availability of more abstract payment methods may increase other-regarding behavior in the field.
Content may be subject to copyright.
Decomposing Desert and Tangibility Effects in a Charitable
Giving Experiment
David Reinstein and Gerhard Riener
Several papers have documented that when subjects play with standard laboratory “endow-
ments” they make less self-interested choices than when they use money they have either
earned through a laboratory task or brought from outside the lab. In the context of a chari-
table giving experiment we decompose this into two common artifacts of the laboratory: the
intangibility of money (or experimental currency units) promised on a computer screen rela-
tive to cash in hand, and the distinct treatment of random “windfall” gains relative to earned
money. While both effects are found to be significant in non-parametric tests, the former
effect, which has been neglected in previous studies, has a stronger impact on total dona-
tions, while the latter effect has a greater impact on the probability of donating. These results
have clear implications for experimental design, and also suggest that the availability of more
abstract payment methods may increase other-regarding behavior in the field.
1. Introduction
Several economists have found that when subjects play with standard laboratory “endow-
ments” they make less self-interested choices than when they use money they have either
“earned” through a laboratory task or brought from outside the lab (Cherry et al. 2002; Hoff-
man and Spitzer 1985; Loomes and Burrows 1994). This effect is typically interpreted as a
result of Lockean desert effects (Rutstrom and Williams 2000; see Locke 1988, pp. 287-8.),
fairness concerns (a la Rabin 1993), or a different mental accounting over windfall gains
(Sheffrin and Thaler 1988, Thaler 1985, and Thaler and Johnson 1990). Our experiment does
not differentiate between these models (this is left for future research); we will refer to the
Preprint submitted to Elsevier July 21, 2011
net effect of these as the windfall effect.
There is ample evidence for windfall effects in the context of dictator games. Cherry et al.
(2002) ran a series of dictator game experiments where, in the baseline treatment the dictator’s
endowment was randomly determined, while in their earnings treatment the endowment was
based on performance in a cognitive task (solving GMAT questions), and this was common
knowledge. Their double blind with earnings treatment modified the earnings treatment
to increase subject-experimenter anonymity. Both their earnings and double-blind-earnings
treatments lead to significantly less generous dictator behavior; in the latter treatment the
dictators became almost entirely hardnosed, keeping nearly all of the money.
Oxoby and Spraggon (2008) compare dictator behavior in treatments adapted from Hoff-
man et al. (1996), comparing a case where the funds to be divided are “earned” by the dicta-
tor’s quiz performance to a case where the potential recipient takes the quiz (to determine the
amount available for the dictator to divide). They find that dictators are significantly more
generous in the latter case. This suggests that fairness concerns are important, and that rela-
tive desert may be driving dictator decisions. Ruffle (1998), Mittone and Ploner (2006), and
Cherry and Shogren (2008) find similar results on the importance of the receiver’s effort.1
In contrast to the dictator environment, there is little evidence for windfall effects in
voluntary contribution mechanism (henceforth VCM) experiments. Clark (2002) examines
contribution rates in a VCM game. He finds no significant difference between contributions
in the “own money” treatment, in which subjects are asked to bring $8 from outside the lab
to purchase tokens, and the “house money” treatment, in which subjects are simply given the
tokens.2However, as Clark’s “own money” subjects are also given house money at the end
1In all of the experiments mentioned above the dictator subjects’ “earnings” come from answering GMAT
questions, some of which involve retailers’ and consumers’ decisions, dishonest job applicants, wealth, invest-
ments, money, and marketing. These may be triggering more self-interested behavior through a framing effect
as in Vohs et al. (2006) and Cookson (2000), rather than simply increased legitimacy of the dictators’ own en-
dowments. However, the estimated relationships between the dictators’ gifts and the recipient’s performance,
hence the observations of a fairness (or relative desert) motive is robust to this critique; the variation in the
recipient’s performance does not yield any additional framing effect of this sort.
2Still, Harrison (2007), who reanalyzed Clark’s data to deal with the potentially non-independent error
structure, suggests that a house money effect is present. However, the tangibility and earnings effects are not
of the experiment, they presumably have the same expected “windfall” earnings as the other
subjects. Furthermore, Clark’s subjects use tokens, and the earnings effect may be more
salient when the rewards are more tangible. In related VCM experiments, neither Cherry
et al. (2005) nor Kroll et al. (2007) find that subjects who earned their endowments (through
answering GMAT questions) contribute less than those who did not.
Finally, there is some very recent evidence in the context of charitable giving itself. Carls-
son et al. (2009) find windfall effects in a charitable giving experiment in both a laboratory
and a field setting; subjects in both environments donate less when they have “earned” their
pay by completing a survey.
This literature has ignored a second component of the bias that may limit the external
validity of many laboratory results: people may treat money they are promised (or are given
in the form of tokens) differently then cash they physically hold – we call this the tangibility
effect We hypothesize three potential reasons why this may occur. First, psychology exper-
iments demonstrate that subjects given “reminders of money” are both less helpful and less
likely to ask for help in a variety of non-remunerated tasks (Vohs et al., 2006).In addition,
Cookson (2000) finds that subjects subtly motivated with an “I” frame contribute less in a
VCM setting than those motivated with a “we” frame, and Oberholzer-Gee and Eichenberger
(2004) find that offering an unattractive lottery option to the choice set leads dictators to
give less to other subjects.Second, using cash may cause subjects to more carefully consider
the consumption they are sacrificing. Along similar lines, Oberholzer-Gee and Eichenberger
(1999) argue that subjects do not fully consider the opportunity costs of the funds they give
away in experiments, and Mazar et al. (2008) find that people cheat more when using ex-
changeable tokens then when they use cash. Finally, parting with cash may itself bring some
disutility, perhaps through an attachment to this money similar to the “endowment effect” of
Kahneman et al. (1991). For all of these reasons, we might therefore expect that subjects
holding cash will be less likely to contribute this to a public good or a charitable cause.
separable in this context.
To the best of our knowledge, there is no economic evidence on the effect of the medium
of exchange on generosity. However, at least one prominent experiment varies this in con-
junction with other variations in the treatment, leading to a potential confound.3In the present
paper we implement a real charitable giving experiment to provide the first salient economic
evidence that the tangibility of the choice medium affects other-regarding decisions.
2. Experimental design
We use a charitable giving experiment with a 22 design to differentiate two distinct arti-
facts of laboratory endowments. Firstly, the treatments vary according to the extent to which
subjects should see the money as earned; we compare giving behavior after compensation
based on performance on a five minute task to behavior with a randomly assigned payment.
The second dimension of variation involves the tangibility of the payment: we either give
cash to the subjects before they decide how much to donate (and they physically place any
donations they make into envelopes) or they allocate their donation from an endowment on
the computer screen and they are paid cash at the end of the experiment. Thus, we separately
test whether earning the money and having cash in hand affect giving behavior in the lab.4
Unlike many of the experiments previously mentioned, our subjects make decisions over
donations to charitable foundations – institutions outside the laboratory. In line with Eckel
and Grossman (1996), we see this as a more obvious and typical expression of other-regarding
behavior than donations to a laboratory public good or towards another laboratory subject.
Our environment also provides a more demanding test for tangibility and windfall effects. In
the real world it is rare to be asked for a gift from a random non-needy stranger (or to re-
ceive such a gift); hence, it is not surprising that standard dictator games should be sensitive
to framing effects. On the other hand charitable appeals and charitable giving are regular-
3E.g., in Hoffman et al. (1996) the "single blind 2" treatment combines both a decreased social distance from
the experimenter (relative to "single blind 1") and "a decision form for making the decision, instead of money".
4In addition to the treatments mentioned above, we also vary the choice set. As a robustness check, we offer
three charities instead of two in the expanded choice set treatment. This allows us to demonstrate that our results
are not sensitive to a variation in the choice set.
ities, so subjects will have more experience with such decisions and their decisions should
be less easily perturbed. While dictator giving to other subjects is highly sensitive to the
level of social isolation, falling to very low levels in “double-blind” environments Hoffman
et al. (1996)), charitable giving persists at significant levels even under highly anonymous
conditions (Eckel and Grossman, 1996). Our setting may also better isolate the effect of asset
legitimacy: intuition suggests that in the charitable giving context, subjects will focus less
on their desert relative to the recipient(s) then they would in deciding how much to give to a
fellow subject.
All treatments are assigned orthogonally; we have a (nearly) fully balanced design.5By
construction, the distribution of initial endowments is the same for each treatment. Finally,
all of our treatments involve the same strong level of anonymity.6The time spent in each
treatment of our experiment is approximately the same, so subjects in each of our treatments
should have the same earnings expectations.
The sessions were run at the University of Jena Experimental Economics lab using the
standard subject pool. In total 190 subjects participated in the experiments of which 54.2%
were female.7The sessions were conducted in October 2008 (39 subjects), February 2009
(79 subjects), and September 2009 (72 subjects). While we ran each of the four payment
regime treatments in a separate session, the participants were from the same subject pool and
the times and dates of the experiment were stratified by treatment.8To avoid mixing payment
types, we did not give subjects any pre-experiment “show-up fee.”
5Because the treatments were run in separate sessions and their were some no-shows, the actual observations
are very slightly off-balance, and the endowments are not precisely identically distributed by treatment, nor is
the “choice set” treatment. However, these slight differences are controlled for in our multivariate regressions
and in our balanced bootstrapped rank-sum tests. The lack of balance does not measurably affect any of our
results. Our treatments are also not perfectly balanced over time. To test for session-specific effects, in online
Appendix 5 Table 5 we also report regressions with standard errors clustered by session, and controls for time-
of-day and date of session effects; our results are robust to all of these, and none of these are significant.
6See the protocol in online Appendix 1 for a full description of our careful procedure to insure subject-
experimenter and subject-subject anonymity.
7We did not collect extensive demographics on our subjects in order to preserve subject-experimenter
8Appendix 5, Table 4 illustrates this balance, and shows our results are insensitive to the time and date of
the session.
To guarantee anonymity, the lab was divided into an outer partition - which serves as a
meeting room before the experiment and as a room for the administrators during the experi-
ment - and an inner partition with computer terminals on which the subjects make decisions
and answer questions. These were separated so that it was impossible to see the inner parti-
tion from the the outer partition and vice-versa. For administrative purposes, a volunteer from
the participants helped with the procedures whenever communication between the inner and
the outer part of the lab was necessary. Furthermore, to ensure our credibility, this volunteer
supervised the online donations made by the experimenters after the other participants had
been dismissed.
At the beginning of the experiment all subjects were assured that we would not be able to
connect their name to the decisions they made. Next we asked for a volunteer to help us with
administrative issues, mainly allocating the sealed envelopes with payouts at the end of the
The task. Subjects in the performance treatments (PA and PC, described below) were told
that their endowment would depend on their performance on a simple task. They were asked
to add up five two-digit numbers9using only scratch paper and a pencil. The numbers were
randomly drawn and presented to the subjects as in the example below:
12 77 34 55 62 __
The participants were given five minutes to solve as many tasks as possible. We argue that
this task was sufficiently tedious to make subjects feel that they earned the money recieved.
This task, although numerical, is less likely to cue self-interested “economic” thinking than
the GMAT questions used in many previous studies.
The charitable giving stage (donation decision). The subjects were not given any indication
that this experiment would involve an opportunity for charitable giving until they reached the
9This task has been used in various occasions for testing competitiveness (e.g., Niederle and Vesterlund,
“charitable giving stage”. This stage was a one-shot dictator game in which subjects could
donate none, some, or all of their endowment to any combination of the available charities
in units of 50 Euro cents. All subjects were presented with Brot für die Welt (BfdW) –
“Bread for the World”, a German development aid agency and the World Wild Life Fund for
Nature (WWF), a nature conservancy charity. For the expanded choice set treatment we also
included Deutsches Rotes Kreuz (DRK) - the German Red Cross - which operates in similar
areas as BfdW. Subjects were given information about each of the charities on the computer
screen and next had to decide how much (if anything) to donate to each available charity and
enter this into the computer.10 By using multiple charities we reduced the noise surrounding
heterogeneous tastes for charities, and gained more useful data on a wider range of subjects.
Treatment 1: Performance / (on-screen) Account (PA). Subjects in the performance treat-
ments were told that the probability of higher earnings increased in the number of tasks cor-
rectly completed, but we did not specify exactly how performance translated into payoffs.11
After completing the task they were told how much this earned them. In account treatments
they were endowed C5, C7.50 or C10 (shown on their computer screen) but were not yet
given cash.They next made their donation decisions. At the end of the experiment they were
(anonymously) given envelopes containing their earnings minus their total donations.
Treatment 2: Performance / Cash (PC). As in PA, subjects first completed the task and
learned how much they earned. However, unlike in the account treatments, subjects in cash
treatments were paid in cash before they made giving decisions. After the task stage, the
volunteer was prompted to come outside and bring the numbered envelopes containing the
10The order of the presentation of the charities, both on the description screens and on the actual donation
screen are stratified over subjects, in order to balance any potential order effects.
11We did not tell them that their pay was based on relative performance. They were instead told that “the
more sums you solve, the more likely it is that you will get a higher payment” (in German: “Je mehr Aufgaben
Sie lösen desto wahrscheinlicher ist es, dass sie mehr verdienen.”), because we did not want them to compare
themselves to other subjects in making their charitable contributions. Such a comparison might have lead them
to believe that that subjects who earned more had a greater obligation to donate. In the treatments of October
2008, the subject who solved the most sums received C10, the second C7.50 and the rest of the subjects in the
same session got C5. In the sessions conducted in February, March, and September 2009, the participants who
were in the upper tercile of solved tasks received C10, in the middle tercile C7.50 and in the lower tercile C5.
cash earnings into the inner part , where they hand each subject the envelope with his or her
subject number on it. The payment envelopes were carefully assembled to look identical and
have similar weights.12 Subjects were instructed to inspect and count the money in private at
their computer desks. Next, they made their donation decision(s) by entering these choices
on the computer screen. Finally, subjects were asked to put the chosen contributions (in cash)
into the donation envelope and seal it.
Treatment 3: Random / on screen Account (RA). In this treatment, subjects were endowed
C5, C7.50 or C10 randomly on their computer screen. The donation stage followed, and
payments were distributed as in PA.
Treatment 4: Random / Cash (RC). In the RC treatment the endowments were randomly
determined (as in RA), and given to the subjects in identical envelopes as in PC. The donation
and payments procedure also followed PC.
3. Results
3.1. Summary Statistics
Table 1 compares the proportion of the endowments donated to any of the two (or three)
charities, pooling across choice set treatments.13 Subjects donated significantly less14 when
they were paid in cash than when their endowment was only shown on the computer screen
(13% versus 23% of the total funds, pooling across all other treatments).15
Figure 1 shows the cumulative distribution of the share of earnings donated over the earn-
ings and payment treatments. The distribution of contributions under on-screen entitlements
12We did this by using coins of different increments. To the extent that small coins are less desirable then
bills this would lead to a bias against our finding of a tangibility effect. Since payments in performance and
random treatments had the same distribution, this should not impact our “earnings effect” findings.
13We performed robustness checks and found no significant differences in contribution behavior between the
two and three charity choice set treatments. Details available by request.
14These differences are significant in Wilcoxon rank-sum tests, as well as in familiar parametric tests (avail-
able by request). Because of the aforementioned lack of balance (stemming from no-shows), we also report
bootstrapped rank-sum tests in brackets, with each of the 1000 random draws (with replacement) exactly bal-
anced by payment treatment, earnings treatment, choice set treatment, and stake size.
15This rate of giving is fairly consistent with results of previous experiments. E.g., in Eckel and Grossman
(1996) subjects give 30% of their $10 cash endowment (they were also given a $5 show-up fee).
Table 1: Average proportion contributed by payment regimes
Endowment Random Performance Total N
Account 0.27 0.18 0.23 99
Cash 0.14 0.12 0.13 91
Total 0.21 0.15 0.18
N102 88 190
Wilcoxon rank sum tests
P(Account> Cash) 0.57+ (0.06); [0.05]
P(Random> Performance) 0.58* (0.05); [0.03]
P(Account/Random > Cash/Performance) 0.64** (0.01); [0.00]
P(Account/Random > Cash/Random) 0.61+ (0.06); [0.05]
P(Account/Performance > Cash/Performance) 0.53 (0.59); [0.57]
P(Account/Performance > Cash/Random) 0.49 (0.86); [0.85]
p-values for simple rank sum tests in parentheses, +: p< 0.10, *: p<0.05, **: p<0.01
In square brackets: p-values for bootstrapped rank sum tests, 1000 draws, balanced by all treatments and stake sizes.
(RA and PA) stochastically dominates the distribution under cash payments (RC and PC).
Similarly, the distribution under random payments (RA and RC) stochastically dominates the
distribution under performance-based earnings (PA and PC).
Figure 1: Cumulative distribution functions of share of earnings donated
On the other hand, as Table 2 demonstrates, the performance treatment has a stronger
extensive margin effect; subjects are significantly less likely to donate at all if they have
earned their endowment through their performance. To give an intuitive spin, some people
may feel more comfortable keeping all of their money if they think they have earned it and
thus deserve it, while those who do feel compelled to donate find that giving away cash feels
Table 2: Number of subjects who donated by treatment
Payment Endowment
Random Performance Account Cash
Donated N(column %) N(col. %) N(col. %) N(col. %)
No 37 (36%) 44 (50%) 42 (42%) 39 (43%)
Yes 65 (64%) 44 (50%) 57 (58%) 52 (57%)
Total number 102 88 99 91
p-values of tests
Pearson c20.06 0.95
Fisher’s exact 0.07 1.00
more “costly” in terms of sacrificed consumption than giving away money on a computer
screen. This result is confirmed by our Probit regressions with controls for endowment size,
gender, and the larger choice set (see Appendix 4).
3.2. Multivariate Analysis
To control for observable (random) differences in treatment assignment we regress total
donations on controls for observable heterogeneity and treatment interactions.16 These re-
gressions (Table 3) again suggest that cash treatments reduced generosity. The effect of cash
is negative and significant in the Account treatments, and negative but insignificant in Per-
formance treatments (summed coefficient: Cash+cash perform).17 Donations were also
lower when subjects were paid according to their performance, but this effect was not statisti-
cally significant here. The coefficients on higher stake sizes ( C7.5 or C10) are small and not
significant: subjects who earn more do not tend to donate more. In line with some previous
work, (e.g., Eckel and Grossman, 1998, List, 2004) women donated more than men. Total
donations were not significantly different when a third charity was included. The interaction
16We use a standard OLS specification for familiarity and comparability reasons. We also use a Poisson spec-
ification, both because our data resembles count data (in increments of 50 cents) and because this specification
deals with corner-solution (non-negative) data without being as sensitive to non-normality and heteroskedastic-
ity as a standard Tobit regression Gourieroux et al. (1984); Arabmazar and Schmidt (1981). In online Appendix
5 (table 5) we find similar results using a fractional regression specification. The cash and performance results
are similar in zero-inflated Poisson regressions (available by request).
17The “Cash” dummy is also strongly significant in a univariate linear (-0.66; p-value: <0.01) or Poisson
regression (-0.66, marginal effect; p-value: <0.01) (details by request).
effects are not significant, although their positive sign and magnitude suggest that the treat-
ments have a sub-additive effect – the summed coefficient (Cash+perform+cash perform)
representing the effect of cash and performance combined is very close to the coefficient on
cash alone.
Table 3: Poisson and OLS regression of total donations
Add. contr. Gender contr.
(1) (2) (3) (4) (5) (6)
Psn. OLS Psn. OLS Psn. OLS
Pay cash -0.84* -0.84* -0.68* -0.84* -0.88* -0.89**
(0.33) (0.33) (0.31) (0.34) (0.38) (0.34)
Pay by performance -0.54 -0.54 -0.44 -0.54 -0.56 -0.58
(0.39) (0.39) (0.33) (0.40) (0.42) (0.41)
Cash performance 0.51 0.44 0.41 0.44 0.69 0.56
(1.02) (0.49) (0.82) (0.50) (1.14) (0.52)
Third charity 0.34 0.26 0.42 0.26
(0.33) (0.25) (0.40) (0.24)
Stake: 7.5 -0.18 -0.13 -0.14 -0.08
(0.36) (0.29) (0.44) (0.28)
Stake: 10 0.10 0.093 0.10 0.10
(0.37) (0.32) (0.44) (0.31)
Female 0.63* 0.53*
(0.32) (0.25)
Combined coefficients (sums raw coefficients, not marginal effects)
Cash+perform+cash perform -0.76* -0.94** -0.76* -0.94** -0.73* -0.92**
( 0.32) ( 0.37) ( 0.32) ( 0.37) ( 0.31) ( 0.36)
Cash+cash perform -0.39 -0.40 -0.39 -0.40 -0.35 -0.34
(0.36) (0.36) (0.36) (0.36) (0.36) (0.38)
Observations 190 190 190 190 190 190
R20.048 0.056 0.079
Pseudo R20.033 0.039 0.055
Heteroskedasticity-robust standard errors in parentheses.
+ p<0.10, * p<0.05, ** p<0.01 for tests using heteroskedasticity-robust standard errors (for all columns)
All regressors are dichotomous (0,1) variables, dy/dx for discrete change of dummy variable reported.
Marginal effects evaluated at Account/Random, Female, Stake = 7.5, two charity choice set.
As we show in online Appendix 5, table 3, the tangibility and windfall effects on dona-
tions are similar across charities, and our results also hold for a fractional response (Papke
and Wooldridge, 1996) regression of “share donated”.
4. Conclusion
Our experiment is the first to document the tangibility effect; its magnitude appears at
least as strong as the windfall effect, although the latter has a stronger effect at the extensive
margin. Furthermore, by using a charitable giving context and a relatively neutral real-effort
task, we add to the evidence that the legitimacy (absolute desert) of experimental subjects’
own assets affects their other-regarding behavior.
Our findings do not imply that experimenters should always use “tangible” cash. In the
context of our experiment, we cannot say which contribution level is more externally valid.
Whether the differences are because seeing money cues self-interest, because cash causes a
more careful consideration of trade-offs, or because parting with cash is more painful, either
frame (cash or endowment) may have external validity.18 In the field many decisions are
made without physical cash, as credit cards and electronic payments have become dominant
in many markets. However, researchers must be aware of this framing effect and take it
into account. This distinction is important: economic experiments vary greatly along both
dimensions, often simultaneously. As noted in Hoffman et al. (1994), comparisons between
experimental results must take into account differences in the decision medium.19 To the
extent that future income is less tangible than present income, our results agree with Breman
Breman (2006), who offers field experimental evidence that people are more generous with
the former than with the latter. Similarly, our result that the payment instrument matters
may also be generalizable to real-world decision making, particularly over intangible “warm-
glow” goods such as charitable donations.
18On the one hand, cash is obviously better if it leads to greater experimenter credibility. However, this is
unlikely to have been a driver of our results, as all of our subjects had previously participated in economic
experiments. On the other hand, cash may lead to other extraneous effects; e.g., it may cue subjects to consider
the nuisance carrying around earnings that include “small change”.
19In comparing their dictator results to those of Forsythe et al. (1994), Hoffman et al note that "other aspects
of the double blind procedures require experimental examination to identify what is driving the outcome; an
envelope containing the cash might be an important factor." Our analysis confirms this speculation.
Arabmazar, A. and P. Schmidt (1981). Further evidence on the robustness of the Tobit esti-
mator to heteroskedasticity. Journal of Econometrics 17(2), 253–258.
Breman, A. (2006). Give more tomorrow: Evidence from a randomized field experiment.
Unpublished manuscript.
Carlsson, F., H. He, and P. Martinsson (2009). Easy come, easy go-The role of windfall
money in lab and field experiments. rapport nr.: Working Papers in Economics 374.
Cherry, T., P. Frykblom, and J. Shogren (2002). Hardnose the Dictator. American Economic
Review 92(4), 1218–1221.
Cherry, T., S. Kroll, and J. Shogren (2005). The impact of endowment heterogeneity and
origin on public good contributions: Evidence from the lab. Journal of Economic Behavior
and Organization 57(3), 357–365.
Cherry, T. and J. Shogren (2008). Self-interest, sympathy and the origin of endowments.
Economics Letters 101(1), 69–72.
Clark, J. (2002). House Money Effects in Public Good Experiments. Experimental Eco-
nomics 5(3), 223–231.
Cookson, R. (2000). Framing effects in public goods experiments. Experimental Eco-
nomics 3(1), 55–79.
Eckel, C. and P. Grossman (1996). Altruism in anonymous dictator games. Games and
Economic Behavior 16(2), 181–191.
Eckel, C. and P. Grossman (1998). Are women less selfish than men?: Evidence from dictator
experiments. The Economic Journal 108(448), 726–735.
Forsythe, R., J. Horowitz, N. Savin, and M. Sefton (1994). Fairness in simple bargaining
games. Games and Economic Behavior 6(3), 347–69.
Gourieroux, C., A. Monfort, and A. Trognon (1984). Pseudo maximum likelihood methods:
applications to Poisson models. Econometrica 52(3), 701–720.
Harrison, G. (2007, December). House money effects in public good experiments: Comment.
Experimental Economics 10(4), 429–437.
Hoffman, E., K. McCabe, K. Shachat, and V. Smith (1994). Preferences, property rights, and
anonymity in bargaining games. Games and Economic Behavior 7(3), 346–380.
Hoffman, E., K. McCabe, and V. Smith (1996). Social Distance and Other-Regarding Behav-
ior in Dictator Games. American Economic Review 86, 653–660.
Hoffman, E. and M. Spitzer (1985). Entitlements, rights, and fairness. Journal of Legal
Studies 14(2), 259–297.
Kahneman, D., J. Knetsch, and R. Thaler (1991). Anomalies: The endowment effect, loss
aversion, and status quo bias. The Journal of Economic Perspectives 5(1), 193–206.
Kroll, S., T. Cherry, and J. Shogren (2007). The impact of endowment heterogeneity and
origin on contributions in best-shot public good games. Experimental Economics 10(4),
List, J. (2004). Young, Selfish and Male: Field evidence of social preferences*. The Eco-
nomic Journal 114(492), 121–149.
Locke, J. (1988). Two Treatises of Government. Cambridge: Cambridge University Press.
Loomes, G. and P. Burrows (1994). The impact of fairness on bargaining behaviour. Empir-
ical Economics 19, 201–221.
Mazar, N., O. Amir, and D. Ariely (2008). The Dishonesty of Honest People: A Theory of
Self-Concept Maintenance. Journal of Marketing Research 45(6), 633 – 644.
Mittone, L. and M. Ploner (2006). Is it just legitimacy of endowments? an experimental
analysis of unilateral giving. CEEL Working Paper N. 02/2006.
Niederle, M. and L. Vesterlund (2007). Do women shy away from competition? do men
compete too much? Quarterly Journal of Economics 122(3), 1067–1101.
Oberholzer-Gee, F. and R. Eichenberger (1999, May). Focus effects in dictator game experi-
Oberholzer-Gee, F. and R. Eichenberger (2004). Fairness in extended dictator game experi-
ments. Manuscript.
Oxoby, R. and J. Spraggon (2008). Mine and yours: Property rights in dictator games. Jour-
nal of Economic Behavior and Organisation 65, 703–713.
Papke, L. and J. Wooldridge (1996). Econometric methods for fractional response variables
with an application to 401 (k) plan participation rates. Journal of Applied Economet-
rics 11(6), 619–632.
Rabin, M. (1993). Incorporating Fairness into Game Theory and Economics. American
Economic Review 83(5), 1281–1302.
Ruffle, B. (1998). More is better, but fair is fair: Tipping in dictator and ultimatum games.
Games and Economic Behavior 23(2), 247–265.
Rutstrom, E. and M. Williams (2000). Entitlements and fairness: an experimental study of
distributive preferences. Journal of Economic Behavior and Organization 43(1), 75–89.
Sheffrin, H. and R. Thaler (1988). The behavioral life-cycle hypothesis. Economic In-
quiry 26, 609–643.
Thaler, R. (1985). Mental accounting and consumer choice. Marketing Science 4, 199–214.
Thaler, R. and E. Johnson (1990). Gambling with the house money and trying to break even:
The effects of prior outcomes on risky choice. Management Science 36, 643–660.
Vohs, K., N. Mead, and M. Goode (2006). The psychological consequences of money. Sci-
ence 314, 1154–1156.

Supplementary resource (1)

... Substantial gender inequities in various economic outcomes persist in our societies and receive much scholarly attention, such as the gender wage gap or the glass ceiling phenomenon (Bertrand 2011;Blau and Kahn 2017). However, an area in which we know less about potential gender differentials relates to two-way interactions between entrepreneurs and potential investors (Garrett 2020;Recalde and Vesterlund 2020). The corresponding industry is so male-dominated that 90% of US-based venture capital firms do not feature any women on their investment team (Brush et al. 2018). ...
... One avenue of research focuses on the entrepreneurs' side-a narrative sometimes labeled as 'fix-the-women' (Recalde and Vesterlund 2020). For instance, Roper and Scott (2009) show female entrepreneurs perceive greater barriers to accessing finance, which may discourage them from seeking funding. ...
... Other lines of research consider the investor (or employer) side, sometimes labeled as a 'fix-the-institution' narrative (see Recalde and Vesterlund 2020), to seek explanations for the stark gender gap in funding. For example, Kanze et al. (2018) explore interactions between investors and entrepreneurs at pitch competitions to find female entrepreneurs are often asked prevention-focused questions (concerned with the return of capital), while male entrepreneurs receive promotion-focused questions (concerned with the growth of the venture; also see Becker-Blease and Sohl 2007, andTurban 2012). ...
Full-text available
Using comprehensive data of 4893 interactions from the popular television show Shark Tank, we test whether gender match with entrepreneurs correlates with investors’ likelihood to extend funding offers. We find female investors are 35% more likely to engage with female (rather than male) entrepreneurs, while less systematic gender preferences emerge for male investors. Heterogeneity analyses suggest this result remains exclusive to non-male-dominated product categories, lending support to the industry representation hypothesis. We also find it is exclusive to ventures with lower asking valuations. Estimates are robust to the inclusion of a comprehensive set of control variables (such as asking valuation, investor-, and season-fixed effects) and a range of alternative specifications.
... Cherry, Kroll and Shogren (2005) find that public good contribution levels decline when subjects are given heterogeneous endowments and observe how their endowment compares to that of others in their collective group. Reinstein and Riener (2012) test for the effect of different earning conditions (i.e., accumulating experimental currency units on a computer versus cash in hand) on charitable giving. Tamai (2018) uses a theoretical model to show how economic uncertainty can reduce donations toward a public good. ...
... The second main objective of this paper is to test for, and generalize, the relationship between charitable giving in the lab and the effort required to obtain one's income. Several previous studies have explored that relationship by comparing giving under the two extremes of windfall gains and earned income (Carlsson, He & Martinsson, 2013;Li, Liang, Xu & Liu, 2019;and Reinstein & Riener, 2012). The experiment described in this paper similarly includes a windfall income treatment and one in which subjects earn income by performing a real-effort task (i.e., solving math problems), but it also adds a third income source: the return from a stochastic investment. ...
... Explanations for this result generally relate to the idea that real effort tasks legitimize earnings (Cherry, 2001;Cherry et al., 2002;Oxoby & Spraggon, 2008) and elicit a stronger sense of ownership or property rights (Dasgupta & Mani, 2015;Hoffman et al., 1995;Korenok et al., 2017), causing individuals to exhibit more self-regarding behavior. 4 In the context of charitable giving, a few previous studies have similarly shown that individuals are indeed more generous with windfall gains than income that was earned performing some menial task (Carlsson et al., 2013;Li et al., 2019;Reinstein & Riener, 2012). ...
Full-text available
This paper evaluates the effect of income volatility and income source on charitable giving, in a laboratory setting. Using three income sources (windfall, investment-based, and earned), we find that the proportion of income subjects donate decreases monotonically with the level of effort required to obtain the income. We also find that income volatility causes an increase in giving, both in terms of donation amounts and the proportion of income donated. These effects are driven primarily by subjects who received their income through windfall gains. Relative to the baseline, subjects are more generous when they face highly-volatile income, even when they experience a negative income shock, although they are more sensitive to positive shocks; we estimate income elasticities of charitable giving of 0.8 for positive income shocks and 0.4 for negative shocks.
... In a second manipulation, participants earn their endowment for the gambling task by completing questionnaires, while being fully informed that they would be compensated for their effort with a fixed payment. Although the effect of earned vs. windfall money has been studied in various domains, including charitable giving (Carlsson et al., 2013;Reinstein & Riener, 2012), dictator games (Cherry & Shogren, 2008;Cherry et al., 2002), public goods games (Cherry et al., 2005;Kroll et al., 2007) and experimental asset markets (Corgnet et al., 2015), 3 evidence on how it influences individual decisions under risk is surprisingly scarce. Zeelenberg and van Dijk (1997) contribute to this question by presenting results from hypothetical choices that "behavioral sunk costs", i.e., effort that has been exerted to earn money, can decrease subsequent risk proneness. ...
... Davis et al. (2010) report that paying a show-up fee after the experiment rather than before decreases risky behavior, measured as refraining from information purchases in their experiment. Reinstein and Riener (2012) show that donations in a dictator game decrease substantially if the endowment to be allocated is handed to recipients in cash instead of being shown on a computer screen. Imas (2016) observes individuals to take on greater risk after a paper loss while taking on less risk after a realized loss. ...
Full-text available
In a seminal contribution, Thaler and Johnson (1990) detected the existence of a house money effect which is defined as an increase in risk tolerance after previous gains resulting from a risky activity. Subsequent studies used the term house money effect also in case of windfall gains, i.e., easily acquired money like show-up fees or initial endowments in experiments which does not result from a risky investment. The present study is to the best of our knowledge the first that disentangles the house money effect and windfall gains. We find a clear and systematic pattern that windfall gains increase risk tolerance. In contrast, the house money effect is far less ubiquitous and seems to require skewed lotteries and/or a large number of rounds played. We, therefore, conclude that a careful distinction between windfall gains and the house money effect is warranted in future research.
... We would expect that endowment effects are larger when participants can physically hold the money. We believe our results are relevant to most experimental settings since participants make decisions in anticipation of payment at the end of the session. 2 The literature that examines whether form of payment affects willingness to spend or save money finds mixed evidence (Bushong, King, Camerer, and Rangel, 2010;Reinstein and Riener, 2012;Runnemark, Hedman, and Xiao, 2015;Spantig, 2021;Luccasen III and Grossman, 2018;Mishra, Mishra, and Nayakankuppam, 2006;Svirsky, 2014). There is also some evidence that there are differences in behavior with payment in cash versus virtual experimental currency in laboratory experiments (Reinstein and Riener, 2012;Shen and Takahashi, 2017;Wadhwa and Zhang, 2015). ...
... We believe our results are relevant to most experimental settings since participants make decisions in anticipation of payment at the end of the session. 2 The literature that examines whether form of payment affects willingness to spend or save money finds mixed evidence (Bushong, King, Camerer, and Rangel, 2010;Reinstein and Riener, 2012;Runnemark, Hedman, and Xiao, 2015;Spantig, 2021;Luccasen III and Grossman, 2018;Mishra, Mishra, and Nayakankuppam, 2006;Svirsky, 2014). There is also some evidence that there are differences in behavior with payment in cash versus virtual experimental currency in laboratory experiments (Reinstein and Riener, 2012;Shen and Takahashi, 2017;Wadhwa and Zhang, 2015). Our paper differs from this literature in that our participants do not physically handle the cash, thus weakening any endowment effects that may occur. ...
We use an experiment to examine whether form of payment (cash or mobile money) affects estimates of intertemporal choice and risk taking. We find that form of payment does not affect temporal discounting and risk taking. Given that participants prefer payment via mobile money, the results suggest that there are minimal concerns with using mobile money to pay participants in experimental studies.
... A commonly accepted mechanism driving the effect of the payment medium on spending is the "pain of paying", which is defined as the psychological cost of parting with money (Raghubir and Srivastava, 2008;Soman, 2001Soman, , 2003Thomas et al., 2011). Other studies, such as Reinstein and Riener (2012) and Shen and Takahashi (2013), investigate the effect of the tangibility of money in giving behavior. ...
Full-text available
Employers reward workers and researchers pay human subjects using different payment mediums. These payment mediums can have various psychological effects on workers/research participants, and thereby influence their effort provision. In this paper, we investigate the effect of payment medium on participants' effort in a lab experiment. We find evidence that payment mediums affect participants' productivity when a fixed payment scheme is used, but not under a performance‐based payment scheme.
Full-text available
Economic preferences are often taken as given, yet evidence shows that preferences respond to life events and change over time. We examine the evolution of other-regarding preferences for a cohort of university students over 5 years, starting before they matriculate and extending one year beyond graduation. Using survey and incentivized measures of preferences, we show that altruism declines over the university years. This decline is reflected in changes in charitable giving over three donation opportunities. We rule out several alternative explanations for the observed change, including cohort differences, perceptions of the charities, and experience with experiments. We show evidence of a ‘giving type’ in charitable giving, with consistency in behavior across giving opportunities. Methodologically, we also show that the incentivized and survey measures are similar at predicting giving types. We conclude that preferences reflect common tendencies over time, while simultaneously showing an overall decline in generosity during the university years.
Full-text available
The introduction of new payment methods has resulted in one of the most significant changes in the way we consume goods and services. In this paper, I present results of a field and a laboratory experiment designed to determine the effect of payment method (cash vs. mobile payment) on spending, and a meta-analysis of previous literature about payment method effect. In the field experiment, I collected cashier receipts from Chinese supermarkets. Compared to cash payment, mobile payments significantly increased the amount purchased and the average amount spent on each item. This effect was found to be particularly large for high price elasticity goods. In the laboratory experiment, participants were randomly assigned to one of four groups that varied with respect to the kind of payment and the kind of incentives, eliminating the potential endogeneity problem from the field experiment. I found that compared to cash, mobile payments lead to a significantly higher willingness to pay (WTP) for consumption. In contrast to others, I found that pain of paying does not moderate the payment method effect; however, other psychological factors were found to work as potential mechanisms for affecting WTP.
In the United States, most charitable donations go to religiously affiliated organizations, yet the impact of a charity’s affiliation on donor behavior is currently unclear. To better understand this impact, this article uses a laboratory experiment to explore how a charity’s religious affiliation drives donor behavior. In the experiment, participants select one charity from a list of eight, with each charity varying in religious affiliation. Masked and unmasked sessions differ in the inclusion of religious affiliation from half the charities, with masked sessions omitting religious affiliation of the charities. This article finds that adding religious language decreases donation frequency and average donation amounts for Christian charities competing against other religious charities. This drop is primarily driven by participants that are politically liberal. Participants prefer charity religious affiliation to match their own religious identity; however, participant strength of religiosity is more predictive in charity choice than religious affiliation.
The dictator game has become a celebrated workhorse of experimental economics and social psychology. In the standard version of the game an individual is given a sum of money and must choose how to split this money between themselves and some other individual. In a variant of the game the individual must split the money between themselves and a charitable cause. This charity version of the dictator game has now been used in well over fifty studies and has provided critical insight on the motives behind giving. It also provides a simple tool that policy makers and practitioners can use to test the effect of interventions. In this paper we explain the different ways in which charity dictator games can and have been used. We also look at the external validity of charity dictator games and discuss the research questions that can be appropriately studied using them.
Meta-analysis techniques are used to analyse behaviour in a set of 136 experimental dictator game conditions. The aim is to find the motivating factors of dictators’ generosity under diverse experimental treatments. For that purpose, a specific meta-analysis is performed on gender differences in giving decisions. It is found that gender differences exist, being women on average significantly more generous than men, even if controlling for several moderator variables that are included in the study. The gender effect may change over several experimental conditions and locations. On the one hand, women are more generous than men for moderate and large social distance, while they are less generous than men when playing with close friends or family members. On the other hand, women give more than men in South America, North America and Oceania, while they give less than men in South Africa. The location seems to play an important role in the overall dictators' behaviour. A general result is that, compared to the baseline, women change their behaviour significantly more than men when they are exposed to any experimental condition.
Full-text available
I. INTRODUCTION T. HE research reported in this paper arose from a previous experimental study conducted by the authors.1 That study involved bargains struck between two subjects who had opposing payoff functions and full infor- mation of one another's payoffs. By a flip of a coin one participant could choose a noncooperative outcome, unilaterally: the winner of the coin toss could simply choose an outcome which gave him $12 and left the other subject nothing, whether or not the loser agreed. However, if the two subjects cooperated, they could obtain from the experimenter $14, which could be split between the subjects in any mutually agreed-on manner. Cooperative game theory predicts that the subjects will cooper- ate and divide the rewards $13 to $1 (the Nash bargaining solution: an even division of the $2 gain from trade). Under no circumstances should the winner of the coin flip settle for less than $12, according to game
We develop attractive functional forms and simple quasi-likelihood estimation methods for regression models with a fractional dependent variable. Compared with log-odds type procedures, there is no difficulty in recovering the regression function for the fractional variable, and there is no need to use ad hoc transformations to handle data at the extreme values of zero and one. We also offer some new, robust specification tests by nesting the logit or probit function in a more general functional form. We apply these methods to a data set of employee participation rates in 401 (k) pension plans.
When the error terms in a Tobit model are heteroskedastic, the MLE which assumes homoskedasticity is inconsistent. For the special case of a constant-term-only model, we investigate the size of the inconsistency. The inconsistency is greater the greater the heteroskedasticity and the greater the degree of censoring (i.e., the greater the number of limit observations). However, the inconsistency is much smaller than in the corresponding truncated-normal model considered by Hurd.
Under three different rules for allocation of initial income we elicit experimental subjects’ preferences for income redistribution using an incentive compatible elicitation mechanism. The three income allocation rules are designed to capture preferences for distributive justice among subjects. The concern is motivated by claims in some of the experimental economics literature that non-self-interested motives often underlie individual behavior. We cannot reject self-interest in favor of any redistribution motives based on our observations. Almost all individuals chose the income distribution which maximized their own income — high income individuals chose no redistribution and low income individuals chose perfect equality in income distribution.