ZurichOpenRepositoryand
Archive
UniversityofZurich
UniversityLibrary
Strickhofstrasse39
CH-8057Zurich
www.zora.uzh.ch
Year:2017
Theethicsofincentivizingtheuninformed:avignettestudy
Ambuehl,Sandro;Ockenfels,Axel
Abstract:Ourrecentworkingpaper(Ambuehl,Ockenfels,andStewart2017)showstheoreticallyand
experimentallythatpeoplewithhighercostsofinformationprocessingrespondmoretoanincreasein
theincentiveforacomplextransaction,anddecidetoparticipatebasedonaworseunderstandingof
itsconsequences. Here,weaddresstheresultingtradeobetweentheprincipleofinformedconsentand
theprincipleoffreecontract.Respondentstoourvignettestudyonoocytedonationoverwhelmingly
favortheformerandsupportpoliciesthatrequiredonorstothoroughlyunderstandthetransaction.This
ndinghelpsdesignmarketsthatarenotonlyecientbutalsoconsideredethical.
DOI:https://doi.org/10.1257/aer.p20171109
PostedattheZurichOpenRepositoryandArchive,UniversityofZurich
ZORAURL:https://doi.org/10.5167/uzh-182660
JournalArticle
PublishedVersion
Originallypublishedat:
Ambuehl,Sandro;Ockenfels,Axel(2017). Theethicsofincentivizingtheuninformed:avignettestudy.
AmericanEconomicReview,107(5):91-95.
DOI:
https://doi.org/10.1257/aer.p20171109
91
American Economic Review: Papers & Proceedings 2017, 107(5): 91–95
https://doi.org/10.1257/aer.p20171109
The Ethics of Incentivizing the Uninformed: A Vignette Study
†
By S A A O*
* Ambuehl: Department of Management, UTSC, and
Rotman School of Management, University of Toronto, 105
St. George Street, Toronto, ON M5S 3E6, Canada (e-mail:
sandro.ambuehl@utoronto.ca); Ockenfels: Department of
Economics, University of Cologne, Universitätsstrasse 22a,
50923 Cologne, Germany (e-mail: ockenfels@uni-koeln.
de). We are grateful to Fulya Ersoy, Annett John, Nico
Lacetera, Matthew Mitchell, Luigi Zingales, as well as par-
ticipants at the University of Toronto’s economic theory
workshop for helpful comments and suggestions. Ockenfels
thanks the DFG research unit “Design & Behavior” (FOR
1371) for nancial support.
†
Go to https://doi.org/10.1257/aer.p20171109 to visit the
article page for additional materials and author disclosure
statement(s).
Economists often espouse incentives, since
they can lead to desirable outcomes simply by
enlarging the set of voluntary choices available.
Becker and Elias (2007), for instance, argue that
allowing incentives for living organ donation
would be a Pareto-improvement.
1
Ethicists, by
contrast, are typically queasy about incentives, in
particular as they apply to transactions like organ
donation, medical trial participation, or surrogate
motherhood. Our aim is to better understand the
empirical nature of the constraints that ethical
concerns place on markets (Roth 2007; Ambuehl,
Niederle, and Roth 2015; Ambuehl 2017).
In a recent working paper (Ambuehl,
Ockenfels, and Stewart 2017), we show both
theoretically and experimentally that when
the acquisition and processing of information
about a transaction is costly, individuals with
higher marginal costs of information often
respond more to a given increase in the incen-
tive. Hence, as incentives rise, people who nd
it more difcult to become well-informed about
the transaction comprise an increasing fraction
of participants. They elect to participate based
on a less complete understanding of the con-
sequences of their choice (see Section IIC for
intuition).
Incentives may thus be at odds with informed
consent. This fundamental principle of bioethics
1
Such incentives are currently prohibited in all countries
except Iran.
maintains that a decision is ethically sound if it
is made not only voluntarily but also, in light of
all relevant information, properly comprehended
(DHEW 1978, The Belmont Report ).
Will people express reservations about
incentives if they lead to the selection effects
we document in our working paper? On the
one hand, such behavior is consistent with
Bayesian rationality. Hence, within a standard
welfare economics framework, it does not give
rise to concern. On the other hand, a mecha-
nism that causes people to participate based on
an inferior understanding of the transaction is
in uncomfortable company with the principle
of informed consent, no matter whether ratio-
nal or not.
We examine this question using a vignette
study. Respondents judge the ethics of incen-
tives for human egg donation when potential
participants differ in cognitive ability, and of
various policies to increase the supply of egg
donors. We design the survey with two goals in
mind. First, we separate concerns about incen-
tivizing people who differ in cognitive ability
from concerns with incentivizing the poor. This
distinction is relevant for a policymaker aiming
for political feasibility. If, empirically, ethical
concerns center around a lack of comprehension
about the transaction, then the moral acceptance
of an incentive system can be improved by inter-
ventions such as stringent informed consent
requirements. By contrast, informational inter-
ventions will not ease concerns that primarily
relate to economic inequality. Second, we aim
to determine whether concerns about incen-
tives for people with heterogeneous ability are
related to the mechanisms we document in our
working paper. We test a necessary condition:
How do respondents think incentives affect the
selection of participants, and how do they think
information acquisition responds depending on
the ability of the incentivized? The theoretically
predicted mechanisms are a plausible reason for
concerns with incentives only if respondents
anticipate them.
MAY 201792
AEA PAPERS AND PROCEEDINGS
I. Vignette
Our vignette briey describes the transaction
of paid egg donation and highlights the need
for information acquisition that arises from its
complexity. We introduce two women thinking
about donating eggs in exchange for $8,000, and
explicitly describe their cognitive ability and
level of education, as well as their nancial situ-
ation. For each respondent, the two women vary
along one of the dimensions and are equal along
the other, yielding the four treatments (ability
varies, rich), (ability varies, poor), (high ability,
nances vary), and (low ability, nances vary).
2
To mute the association between ability and
income, respondents consider nancial resources
that vary due to an inheritance of $500,000,
rather than due to differential wage rates.
We conducted the survey in Fall 2016 on
Amazon Mechanical Turk with a total of 502
US-based respondents. We paid $5 for participa-
tion, and attrition was independent of treatment
(see the online Appendix for details and survey).
II. Results
A. Ethical Judgments
How do respondents ethically judge a rise
in the incentive if it leads to a disproportion-
ate increase in the participation of low-ability
women? We elicited respondents’ judgment of
an increase in the incentive by $4,000, and asked
them to assume that the additional donors drawn
by the higher incentive are mostly low-ability
women. A striking 59.2 percent of our respon-
dents think that the clinic should not raise the
incentive, compared to 10.8 percent who think
the opposite, and 30.0 percent who are indiffer-
ent, as panel A of Table 1 shows. In stark con-
trast, many fewer respondents disapprove of
raising the incentive when asked to assume that
it leads to negative selection regarding nancial
means. In this case, only 32.1 percent think the
clinic should not raise the incentive, whereas
21.0 percent think it should, and 46.8 percent are
indifferent. Hence, respondents are concerned
about incentivizing those who might not easily
understand what they sign up for, and this is not
2
We randomized the order of presentation and the assign-
ment of names to women on the individual level.
merely an implication of concerns with incentiv-
izing the poor.
Both of these results are reected in the
answers to two additional questions. First, we
separately elicited respondents’ judgments
about who should be incentivized if a single
additional donor is needed. On the one hand,
28.4 percent think it is more ethically sound
to incentivize the high-ability woman, 3.2 per-
cent would incentivize the low-ability woman,
and the remaining 68.4 percent are indifferent,
as panel B shows. On the other hand, respon-
dents feel less strongly, and less unanimously,
about targeting women depending on nancial
resources. 86.1 percent are indifferent, and those
who are not fall about equally on either side.
Second, we asked about the extent to which
respondents consider an $8,000 incentive for egg
donation ethical. As panel C shows, 15.6 per-
cent think that incentivizing low-ability women
is unethical, but only 4.8 percent think this way
about incentivizing high-ability women, a dif-
ference of 10.8 percentage points ( p = 0.01 ,
averaged over nancial resources). Varying
nancial resources by half a million dollars, by
contrast, changes the fraction of respondents
who consider the incentive unethical only by a
statistically insignicant 2.4 percentage points
(averaged over ability).
3
B. Policy Judgments
How can one increase the number of partici-
pants in a transaction like egg donation in a way
that respondents will view as ethically sound?
Grant (2006, p. 33) suggests that persuasion “on
the basis of reason alone might be considered
the morally exemplary form of power.”
To test this intuition, we asked respondents
to explicitly compare two policies, assuming
that they both raise the expected number of par-
ticipants by the same amount, and generate an
additional $4,000 in expenses per donor. The
rst policy simply increases the incentive pay-
ment. The second policy leaves the incentive
3
There is an interesting discrepancy between the large
fraction of respondents who disapprove of a rise in the
incentive, and the small fraction who consider the original
incentive unethical. Many respondents seem to subscribe
to the view that once somebody has declined an offer, one
should not attempt to “bribe” them into changing their mind
(Grant 2006).
VOL. 107 NO. 5
93
THE ETHICS OF INCENTIVIZING THE UNINFORMED
payment unchanged, but uses the funds to pro-
vide information such as meetings with previous
donors, and psychological counseling. Panel
A of Table 2 shows that 67.7 percent of the
respondents feel that the information policy is
more ethical than the higher incentive. In stark
contrast, only 10.4 percent predict that poten-
tial donors would prefer the information policy;
78.9 percent predict donors would prefer the
higher incentive.
These results suggest a demand for policies
to ensure that participants in a transaction like
egg donation are sufciently well-informed.
Hence, we asked respondents to assess two
such interventions. The rst requires potential
egg donors to attend mandatory information
sessions and to interview ve previous donors.
As panel B shows, 62.4 percent of respondents
support this policy; only 24.3 percent oppose it.
Support recedes only mildly (to 56.4 percent)
for a more heavy-handed intervention that addi-
tionally requires potential donors to pass a thor-
ough exam about the possible consequences of
egg donation. (Opposition rises to 28.3 percent.)
Respondents’ attitudes are consistent, on
the individual level, with their predictions of
behavior. We elicited, for each woman and
for each participation decision she could have
made, how likely the respondent thought that her
decision was in her own best interest, given the
information she had acquired. As the rst col-
umn of panel C shows, the less likely a respon-
dent thinks a woman’s decision to participate
is in her own best interest, the more likely they
support mandatory information sessions, and the
more likely they consider increasing participa-
tion through informational interventions ethi-
cally superior to higher incentives.
4
The second column shows that whether a
woman’s decision to abstain is deemed in her
best interest has no predictive power; only
judgments about the decision to participate
do. Respondents’ policy judgments depend on
4
This result refutes the alternative hypothesis that
respondents’ reservations about incentivizing low-ability
egg donors primarily concern the potential offspring.
T 1—E J
Do not Do Indifferent
Panel A. Raise incentive?
Selection: ability 59.2 10.8 30.0
(3.1) (2.0) (2.9)
Selection: income 32.1 21.0 46.8
(2.9) (2.6) (3.1)
Low High Indifferent
Panel B. Target whom?
Ability 3.2 28.4 68.4
(1.6) (1.6) (2.2)
Income 6.7 7.1 86.1
Low High Difference
Panel C. Incentivizing unethical?
Ability 15.6 4.8
−10.8
(1.9) (1.9) (2.7)
Income 10.7 8.3
−2.4
(1.8) (1.8) (2.6)
Notes: Panels A and B show the percentage of respondents
selecting into each column. Panel C shows the percentage of
respondents considering incentivizing the respective woman
unethical. All panels only use the respondents for whom the
respective attribute was varied and average over the other
attribute. Standard errors in parentheses.
T 2—P J
Information Pay Neither
Panel A. Which policy?
More ethical 67.7 16.7 15.5
(2.1) (1.7) (1.6)
Donors prefer 10.4 78.9 10.8
(1.4) (1.8) (1.4)
Oppose Support Neither
Panel B. Law
Mandatory information 24.3 62.4 13.3
(1.9) (2.2) (1.5)
Exam 28.3 56.4 15.3
(2.0) (2.2) (1.6)
Choice in best interest
Participate Abstain
Panel C. Individual consistency
Support for
Mandatory information
−0.095
0.020
(0.049) (0.052)
Mandatory information
−0.081
0.040
and exam (0.053) (0.056)
Information more ethical
−0.100 −0.022
than incentive
(0.048) (0.050)
Notes: Panels A and B show the percentage of respondents
selecting into each column, using those for whom the respec-
tive attribute was varied and averaging over the other attri-
bute. Panel C shows how participants’ responses depend on
their beliefs about women’s choices. Jointly estimated using
seemingly unrelated regression on all respondents. Standard
errors in parentheses.
MAY 201794
AEA PAPERS AND PROCEEDINGS
beliefs about decisions that may cause a woman
to be ex post worse off than before the transac-
tion, but not about those that may cause her to
forgo a potential benet.
C. Predictions of Behavior
To test whether moral concerns are plausi-
bly related to the selection effects described
in Ambuehl, Ockenfels, and Stewart (2017),
respondents predicted the behavioral effects of
incentives. They did so before we elicited ethical
judgments.
We presented the following scenario. A
woman interested in donating eggs in exchange
for $8,000 has informed herself by talking to
a previous donor. That donor has encouraged
participation. The woman considers searching
for one more donor to interview, but she is not
quite sure whether it is worth the effort. Now, she
learns that the compensation for egg donors has
increased to $12,000. How will this change the
likelihood that she contacts an additional donor?
And how will this effect differ if the previous
donor had instead discouraged participation? For
each of the women, each respondent saw both
these questions.
Panel A of Table 3 shows the results.
Conrming the theoretical and experimental
result in Ambuehl (2017), respondents predict
that women who have talked to a discouraging
donor will become more likely to contact another
donor as the incentive rises. Indeed, if the oppor-
tunity cost of nonparticipation rises by $4,000, a
Bayesian should exert more effort to ensure that
the decision to abstain is not a mistake. At the
same time, respondents predict women who have
met an encouraging donor will become less eager
to obtain a second opinion as the incentive rises.
This is also consistent with rationality. The addi-
tional incentive provides partial insurance against
ex post undesirable outcomes, causing a Bayesian
to reduce the acquisition of costly information
that may prevent ex post-mistaken participation.
Most crucially, respondents believe that
incentives affect the information acquisition of
low-ability women more strongly. Specically,
they predict that after receiving encouraging
information, the higher incentive signicantly
deates the propensity of low-ability women
to contact another donor, but barely affects
high-ability women. The effect of ability is
signicantly smaller if the initial information
was discouraging.
5
Such information acquisi-
tion behavior implies that low-ability women
will respond more strongly to an increase in
the incentive, and will therefore be selected
disproportionately.
Indeed, respondents anticipate these selection
effects. We asked whether women who would
participate for $12,000 but not for $8,000 are
more frequently high- or low-ability women.
Panel B shows that a 64 percent-majority pre-
dict that the marginal participant would more
frequently be a low-ability woman, whereas
18.4 percent predict the opposite, and the
remainder predict no selection effects.
5
Similarly, respondents predict that the information
acquisition of women with lower nancial resources reacts
more strongly to an increase in the incentive.
T 3—P B
Low High
ability ability Difference
Panel A. ∆P (ask another donor)
First donor
encouraging
−0.55 −0.09
0.46
(0.05) (0.05) (0.07)
discouraging 0.37 0.50 0.13
(0.05) (0.05) (0.07)
Low High None
Panel B. Predicted selection
Ability 64.0 18.4 17.6
(3.0) (2.5) (2.4)
Income 75.8 19.0 5.2
(2.7) (2.5) (1.4)
∂
ability
∂
incentive
P(nd second)
after rst donor was
encouraging discouraging
Panel C. Individual consistency
Selection effect of
−0.129
0.048
incentive on ability
(0.053) (0.047)
Notes: Panel A jointly estimated using seemingly unrelated
regression. Dependent variable coded as 1 = more likely,
0 = just as likely, −1 = less likely. The difference-in-differ-
ences of 0.33 is statistically signicant ( p < 0.01). Panel B
shows the fraction of respondents selecting into each col-
umn. Panel C shows the individual-level relation between
predicted effects on information acquisition and predicted
selection effects. Panels B and C only use respondents for
whom the respective attribute was varied and average over
the other attribute. Standard errors in parentheses.
