Professional Documents
Culture Documents
American Economic Association
American Economic Association
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The American Economic
Review.
http://www.jstor.org
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
Risk Aversion and Incentive Effects
Although risk aversion is a fundamental ele- ever, they also obtain dramatically different
ment in standard theories of lottery choice, asset results depending on whether the choice task
valuation, contracts, and insurance (e.g., Daniel involves buying or selling, since subjects tend
Bernoulli, 1738; John W. Pratt, 1964; Kenneth to put a high selling price on something they
J. Arrow, 1965), experimental research has pro- "own" and a lower buying price on something
vided little guidance as to how risk aversion they do not, which implies risk-seekingbehav-
should be modeled. To date, there have been ior in one case and risk aversion in the other.'
several approaches used to assess the impor- Independentof the method used to elicit a mea-
tance and nature of risk aversion. Using lottery- sure of risk aversion, there is widespreadbelief
choice data from a field experiment, Hans P. (with some theoreticalsupportdiscussed below)
Binswanger (1980) concluded that most farmers that the degree of risk aversion needed to ex-
exhibit a significant amount of risk aversion that plain behaviorin low-payoff settings would im-
tends to increase as payoffs are increased. Al- ply absurdlevels of risk aversionin high-payoff
ternatively, risk aversion can be inferred from settings. The upshot of this is that risk-aversion
bidding and pricing tasks. In auctions, overbid- effects are controversialand often ignoredin the
ding relative to Nash predictions has been at- analysisof laboratorydata.This generalapproach
tributed to risk aversion by some and to noisy has not caused much concern because most
decision-making by others, since the payoff theorists are used to bypassing risk-aversion
consequences of such overbidding tend to be issues by assuming that the payoffs for a game
small (Glenn W. Harrison, 1989). Vernon L. are already measuredas utilities.
Smith and James M. Walker (1993) assess the The natureof risk aversion (to what extent it
effects of noise and decision cost by dramati- exists, and how it depends on the size of the
cally scaling up auction payoffs. They find little stake) is ultimately an empirical issue, and ad-
support for the noise hypothesis, reporting that ditional laboratory experiments can produce
there is an insignificant increase in overbidding useful evidence that complements field obser-
in private-value auctions as payoffs are scaled vations by providing careful controls of proba-
up by factors of 5, 10, and 20. Another way to bilities and payoffs. However, even many of
infer risk aversion is to elicit buying and/or those economists who admit that risk aversion
selling prices for simple lotteries. Steven J. may be importanthave asserted that decision
Kachelmeier and Mohamed Shehata (1992) re- makersshould be approximatelyrisk neutralfor
port a significant increase in risk aversion (or, the low-payoff decisions (involving several dol-
more precisely, a decrease in risk-seeking be- lars) that are typically encounteredin the labo-
havior) as the prize value is increased. How- ratory. The implication, that low laboratory
incentives may be somewhat unrealistic and
* Holt: Department of Economics, University of Vir-
therefore not useful in measuring attitudes to-
ginia, Charlottesville, VA 22903; Laury: Department of
Economics, AndrewYoung School of Policy Studies, Geor-
gia State University, Atlanta,GA 30303. We wish to thank This is analogous to the well-known "willingness-to-
Ron Cummingsfor helpful suggestions and for funding the pay/willingness-to-acceptbias." Asking for a high selling
human subjects' payments. In addition, we are especially price implies a preferencefor the risk inherentin the lottery,
grateful to John List and Brett Katzmanfor setting up the and offering a low purchaseprice implies an aversionto the
sessions at the Universities of CentralFlorida and Miami, risk in the lottery. Thus the way that the pricing task is
respectively. This work was also funded in partby the Na- framed can alter the implied risk attitudes in a dramatic
tional Science Foundation (SBR-9753125, SBR-9818683, manner.The issue is whether seemingly inconsistent esti-
and SBR-0094800). We wish to thank Loren Smith for mates are due to a problem with the way risk aversion is
research assistance, and John Kagel, Dan Levin, Andrew conceptualized,or to a behavioralbias that is activated by
Muller, Tom Palfrey, Peter Wakker,seminarparticipantsat the experimentaldesign. We chose to avoid this possible
the Federal Reserve Bank of Atlanta, and an anonymous complication by framing the decisions in terms of choices,
referee for helpful suggestions. not purchasesand sales.
1644
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
VOL.92 NO. 5 HOLTAND LAURY:RISKAVERSIONAND INCENTIVEEFFECTS 1645
Expected payoff
Option A Option B difference
1/10 of $2.00, 9/10 of $1.60 1/10 of $3.85, 9/10 of $0.10 $1.17
2/10 of $2.00, 8/10 of $1.60 2/10 of $3.85, 8/10 of $0.10 $0.83
3/10 of $2.00, 7/10 of $1.60 3/10 of $3.85, 7/10 of $0.10 $0.50
4/10 of $2.00, 6/10 of $1.60 4/10 of $3.85, 6/10 of $0.10 $0.16
5/10 of $2.00, 5/10 of $1.60 5/10 of $3.85, 5/10 of $0.10 -$0.18
6/10 of $2.00, 4/10 of $1.60 6/10 of $3.85, 4/10 of $0.10 -$0.51
7/10 of $2.00, 3/10 of $1.60 7/10 of $3.85, 3/10 of $0.10 -$0.85
8/10 of $2.00, 2/10 of $1.60 8/10 of $3.85, 2/10 of $0.10 -$1.18
9/10 of $2.00, 1/10 of $1.60 9/10 of $3.85, 1/10 of $0.10 -$1.52
10/10 of $2.00, 0/10 of $1.60 10/10 of $3.85, 0/10 of $0.10 -$1.85
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
1646 THEAMERICANECONOMICREVIEW DECEMBER2002
to B. Even the most risk-averseperson should selected. The second decision task involved the
switch over by decision 10 in the bottom row, same ten decisions, but with hypotheticalpay-
since Option B yields a sure payoff of $3.85 in offs at 20 times the levels shown in Table 1 ($40
that case. or $32 for Option A, and $77 or $2 for Option
The literature on auctions commonly as- B). The third task was also a high-payoff task,
sumes constant relative risk aversion for its but the payoffs were paid in cash. The final task
computationalconvenience and its implications was a "returnto baseline" treatmentwith the
for bid functionlinearitywith uniformlydistrib- low-money payoffs shown in Table 1. The out-
uted private values. With constant relative risk come of each task was determinedbefore the
aversion for money x, the utility function is next task began. Incentives are likely dilutedby
u(x) = x'-' for x > 0. This specification the random selection of a single decision for
implies risk preferencefor r < 0, risk neutrality each of the treatments,which is one motivation
for r = 0, and risk aversion for r > 0.3 The for runningthe high-payoff condition. Subjects
payoffs for the lotterychoices in the experiment did seem to take the low-payoff condition seri-
were selected so that the crossover point would ously, often beginningwith the easier choices at
provide an interval estimate of a subject's co- the top and bottom of the table, with choices
efficient of relative risk aversion.We chose the neartheir switch point more likely to be crossed
payoff numbers for the lotteries so that the out and changed.
risk-neutral choice pattern (four safe choices To control for wealth effects between the
followed by six risky choices) was optimal for high and low real-payoff treatments, subjects
constant relative risk aversion in the interval were requiredto give up what they had earned
(-0.15, 0.15). The payoff numbers were also in the first low-payoff task in order to partici-
selected to make the choice patternof six safe pate in the high-payoff decision. They were
choices followed by four risky choices optimal asked to initial a statementaccepting this con-
for an interval (0.41, 0.68), which is approxi- dition, with the warning:
mately symmetric around a coefficient of 0.5
(square root utility) that has been reported in Even though the earnings from this next
econometric analysis of auction data cited be- choice may be very large, they may also
low. For our analysis, we do not assume that be small, and differences between people
individuals exhibit constant relative risk aver- may be large, due to choice and chance.
sion; these calculations will provide the basis Thus we realize that some people may
for a null hypothesis to be tested. In particular, prefer not to participate,and if so, just
indicate this at the top of the sheet... . Let
if all payoffs are scaled up by a constant,k, then me reiterate, even though some of the
this constant factors out of the power function
that has constant relative risk aversion. In this payoffs are quite large, there is no catch
or chance that you will lose any money
case, the numberof safe choices would be un- that you happen to earn in this part. We
affected by changes in payoff scale. A change are preparedto pay you what you earn.
in choice patterns as payoffs are scaled up Are there any questions?
would be inconsistentwith constantrelativerisk
aversion.In this case, we can use the numberof Nobody declined to participate,so there is no
safe choices in each payoff condition to obtain selection bias. For comparability,subjectsin the
risk aversion estimates for other functional high-hypothetical treatment were required to
forms. initial a statementacknowledgingthat earnings
In our initial sessions, subjects began by in- for that decision would not be paid. The hypo-
dicating a preference, Option A or Option B, thetical choice does not alter wealth, but the
for each of the ten pairedlottery choices in Ta- high real payoffs alteredthe wealth positions a
ble 1, with the understandingthat one of these lot for most subjects, so the final low-payoff
choices would be selected at randomex post and task was used to determine whether risk atti-
played to determinethe earningsfor the option tudes are affected by large changes in accumu-
lated earnings. Comparingchoices in the final
low-payoff task with the first may also be used
3 When r = 1, the naturallogarithmis used; division by to assess whetherany behavioralchanges in the
(1 - r) is necessary for increasing utility when r > 1. high-payoff condition were due to changes in
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
VOL.92 NO. 5 HOLTAND LAURY:RISKAVERSIONAND INCENTIVEEFFECTS 1647
TABLE2-SUMMARYOFLOTTERY-CHOICE
TREATMENTS
risk attitudeor from more careful consideration tively. All of these sessions were conducted at
of the choice problem. Georgia State University. The number of sub-
All together,we conductedthe initial sessions jects in these treatmentswas necessarily much
(with low and 20x payoffs) using 175 subjects, smaller due to the large increase in payments
in groups of 9-16 participantsper session, at required to conduct them. All subjects were
three universities (two at Georgia State Univer- presented with both real and hypothetical
sity, four at the University of Miami, and six at choices in these two treatments,allowing for a
the Universityof CentralFlorida).About half of within-subjects comparison. Average earnings
the students were undergraduates,one-third were about $70 in the 20x sessions using real
were MBA students,and 17 percent were busi- payments,$130 in the 50x sessions, and $225 in
ness school faculty. Table 2 presentsa summary the 90x sessions.5 All individual lottery-choice
of our experimental treatments. In these ses- decisions, earnings, and responses to 15 demo-
sions, the low-payoff tasks were always done, graphicquestions (given to subjects at the con-
but the high-payoff condition was for hypothet- clusion of the experiment)can be found on the
ical payoffs in some sessions, for real money in Web at (http://www.gsu.edu/-ecoskl/research.
others, and in about half of the sessions we did htm).
both in order to obtain a within-subjectscom-
parison. Doing the high-hypothetical choice II. Incentive Effects
task before high real allows us to hold wealth
constant and to evaluate the effect of using real In all of our treatments,the majorityof sub-
incentives. For our purposes, it would not have jects chose the safe option when the probability
made sense to do the high real treatmentfirst, of the higherpayoff was small, and then crossed
since the careful thinking would bias the high- over to Option B without ever going back to
hypothetical decisions. We can compare Option A. In all sessions, only 28 of 212 sub-
choices in the high real-payoff treatmentwith jects ever switched back from B to A in the first
either the first or last low-payoff task to allevi- low-payoff decision, and only 14 switched back
ate concerns that learning occurred as subjects in the final low-payoff choice. Fewer than one-
worked throughthese decisions. fourthof these subjectsswitchedback from B to
In order to explore the effect of even larger A more thanonce. The numberof such switches
increases in payoffs we next ran some very was even lower for the high-payoff choices,
expensive sessions in which the 20x payoffs
were replaced with 50x payoffs and 90x pay-
5 All of the
offs. In the two 50x sessions (19 subjects), the lottery-choice tasks reportedin this paper
were preceded by an unrelatedexperiment.Those sessions
"safe" payoffs were $100 and $80, while the conductedat the Universities of Miami and CentralFlorida
"risky"payoffs were $192.50 and $5. In the 90x followed a repeated individual-decision (tax compliance)
sessions (18 subjects)the safe and risky payoffs task conductedby a colleague, for which earningsaveraged
were ($180, $144) and ($346.50, $9), respec- about$18. The lottery-choicesessions conductedat Georgia
State University followed a different set of (individual-
choice) tasks for which average earnings were somewhat
higher (about $27). We conclude that these differences are
4 Of course, the orderthat we did use could bias the
high probablynot relevant;in the 20x payoff sessions, including
real decision toward what is chosen under hypothetical Georgia State data does not alter the means, medians, or
conditions, but a comparisonwith sessions using one high- modes of the number of safe choices in any of the treat-
payoff treatmentor the other indicates no such bias. ments by more than 0.05.
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
1648 THEAMERICANECONOMICREVIEW DECEMBER2002
?
I1 .----------- lies to the right of the risk-neutralprediction,
0.99
showing a tendency toward risk-averse behav-
0.8 ior among these subjects. The thin lines in the
0.7 i
0.6,
figure show the observedchoice frequenciesfor
the hypothetical(20x, 50x, and 90x) treatments;
0.5
these are quite similar to one another and are
; 0.4
also very close to the line for the low real-
0.3
, 0.2
payoff condition. Actual choice frequenciesfor
the initial (20x payoff) sessions, along with the
0.1!
0---- implied risk-aversionintervals,are shown in the
1 2 3 4 5 6 7 8 9 10 "low real" and "20x hypothetical"columns of
Decision Table 3. Even for low-payoff levels, there is
FIGURE 1. PROPORTION OF SAFE CHOICES IN EACH
considerable risk aversion, with about two-
DECISION: DATA AVERAGES AND PREDICTIONS thirds of subjects choosing more than the four
safe choices that would be predicted by risk
Note: Data averages for low real payoffs [solid line with
dots], 20x, 50x, and 90x hypotheticalpayoffs [thin lines], neutrality.However, there is no significantdif-
and risk-neutralprediction [dashed line]. ference between behavior in the low real- and
high- (20x, 50x, or 90x) hypothetical-payoff
treatments.
althoughthis difference is small (6.6 percentof Figure 2 shows the results of the 20x real-
choices in the last low-payoff task, compared payoff treatments(the solid line with squares).
with about 5.5 percent in the 50x and 90x real- The increase in payoffs by a factor of 20 shifts
payoff treatments). More subjects switched the locus of choice frequencies to the right in
back in the hypotheticaltreatments:between 8 the figure,with more than 80 percentof choices
and 10 percent. in the risk-aversecategory (see Table 3). Of the
Even for those who switched back and forth, 150 subjects who faced the 20x real-payoff
there is typically a clear division point between choice, 84 showed an increase in risk aversion
clusters of A and B choices, with few "errors" over the low-payoff treatment.Only 20 subjects
on each side. Therefore, the total number of showed a decrease (the others showed no
"safe"A choices will be used as an indicatorof change). This difference is significant at any
risk aversion.6Figure 1 displays the proportion standardlevel of confidence using a Wilcoxon
of A choices for each of the ten decisions (as test of the null hypothesis that there is no
listed in Table 1). The horizontal axis is the change.8 The risk-aversioncategories in Table
decision number,and the dashed line shows the 3 were used to design the menu of lottery
predictionsunderan assumptionof risk neutral- choices, but the clear increase in risk aversion
ity, i.e., the probabilitythatthe safe OptionA is as all payoffs are scaled up is inconsistent with
chosen is 1 for the first four decisions, and then constantrelative risk aversion.One notable fea-
this probability drops to 0 for all remaining ture of the frequenciesin Table 3 is that nearly
decisions. The thick line with dots shows the 40 percent of the choice patterns in the 20x
observedfrequencyof OptionA choices in each
of the ten decisions in the low-real-payoff (lx)
treatment.7This series of choice frequencies to the 93 subjects who made choices under real and hypo-
thetical conditions. A Kolmogorov-Smirov test fails to
rejectthe null hypothesisof no differencein the distribution
6 The
analysis reportedin this paperchanges very little if of the numberof safe choices between the full sample and
we instead drop those subjects who switch from B back to the relevant restrictedsample for any of our comparisons.
A. The average numberof safe choices increases slightly in Moreover,the actualdifferencein distributionsis very small
some treatmentswhen we restrictour attentionto those who in all cases.
never switch back, but typically by less than 0.2 choices. 8
Following Sydney Siegel (1956), observations with
7 For this no change were not used. In addition, a one-tailed Kol-
figure, and other frequencies reportedbelow,
the full sample of available observations was used. For mogorov-Smirnovtest applied to the aggregate cumulative
example, in Figure 1, the choices of all 212 subjects are frequencies, based on all observations, allows rejection of
reportedin the low-payoff series. This includes those in the the null hypothesisthatthe choice distributionsare the same
20x, 50x, and 90x sessions. Similarly,when choices involv- between the low (either first or last) and 20x real-payoff
ing 20x payoffs are reported,we do not limit our attention treatments(p < 0.01).
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
VOL.92 NO. 5 HOLTAND LAURY:RISKAVERSIONAND INCENTIVEEFFECTS 1649
BASED ON LOTTERYCHOICES
TABLE3-RISK-AVERSION CLASSIFICATIONS
relativerisk
of relative Proportionof choices
Number ~Range
Number
Range
of safe aversion for Risk preference Low 20x 20x
choices U(x) = xl-r/(1 - r) classification reala hypothetical real
0-1 r < -0.95 highly risk loving 0.01 0.03 0.01
2 -0.95 < r < -0.49 very risk loving 0.01 0.04 0.01
3 -0.49 < r < -0.15 risk loving 0.06 0.08 0.04
4 -0.15 < r < 0.15 risk neutral 0.26 0.29 0.13
5 0.15 < r < 0.41 slightly risk averse 0.26 0.16 0.19
6 0.41 < r < 0.68 risk averse 0.23 0.25 0.23
7 0.68 < r < 0.97 very risk averse 0.13 0.09 0.22
8 0.97 < r < 1.37 highly risk averse 0.03 0.03 0.11
9-10 1.37 < r stay in bed 0.01 0.03 0.06
a
Average over first and second decisions.
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
1650 THEAMERICANECONOMICREVIEW DECEMBER2002
aN = 118.
bN = 150.
For payoff scales of 20x, 50x, and 90x the percent) in the 20x hypothetical-payoffcondi-
mediansare, respectively, (6.0, 7.0, 7.5) and the tion thanis the case in the othertreatments(6-8
modes are (6.0, 7.0, and 9.0). This increased percent). A Kolmogorov-Smirnovtest on the
tendency to choose the safe option when pay- change in hypothetical distributionsshows no
offs are scaled up is inconsistentwith the notion change as payoffs are scaled up from 20x to 50x
of constantrelativerisk aversion(when utility is to 90x. Behavior is a little more erratic with
written as a function of income, not wealth). hypothetical payoffs; for example, one person
This increase in risk aversion is qualitatively chose Option A in all ten decisions, including
similar to Smith and Walker's (1993) results. the sure hypothetical$40 over the hypothetical
However, unlike the subjects in their auction $77 in decision 10. The only other case of
experiments, our subjects exhibit much larger Option A being selected in decision 10 also
(and significant)changes in behavioras payoffs occurredin the 20x hypotheticaltreatment.
are scaled up. Kachelmeierand Shehata(1992) This result raises questions about the validity
also observed a significant change in behavior of Kahneman and Tversky's suggested tech-
when the payoff scale was increased, although nique of using hypothetical questionnaires to
their subjects (who demandeda relatively high addressissues that involve very high stakes. In
price in orderto sell the lottery) appearedto be particular,it casts doubt on their assumption
risk preferringin their baseline treatment.As that"peopleoften know how they would behave
noted earlier, our design avoids any potential in actual situations of choice" (Kahnemanand
willingness-to-acceptbias by framingthe ques- Tversky, 1979, p. 265).
tion in a neutralchoice setting. To summarize: We can also addresswhetherfacing the high-
increases in all prize amountsby factors of 20, payoff treatment affected subsequent choices
50, and 90 cause sharpincreasesin the frequen- under low payoffs. Looking at Table 4, the
cies of safe choices, and hence, in the implied roughly comparablechoice frequencies for the
levels of risk aversion. "before"and "after"low-payoff conditions (an
In contrast,successive increases in the stakes average of 5.2 versus 5.3 safe choices for 20x
do not alter behavior very much in the hypo- payoffs, and 5.3 versus 5.5 for the 50x and 90x
thetical payoff treatments. Subjects are much treatments)suggests that the level of risk aver-
more risk averse with high real-payoff levels sion is not affected by high earnings in the
(20x, 50x, and 90x) than with comparablehy- intermediate high-payoff condition that most
pothetical payoffs. The clear treatmenteffect subjects experienced. This invariance is sup-
suggestedby Figure2 is supportedby the within- ported by a simple regression in which the
subjects analysis. Of the 93 people who made change in the number of safe choices between
both real and hypotheticaldecisions at the 20x the first and last low-payoff decisions is re-
level, 44 showed more risk aversionin the real- gressed on earningsin the high real-payoffcon-
payoff condition, 42 showed no change, and 7 dition that were obtained in between. The
showed less risk aversion.The positive effect of coefficient on earnings is near zero and insig-
real payoffs on the number of safe choices is nificant.If we only considerthe subset who won
significant using either a Wilcoxon test or a the $77 prize, 21 people did not change their
Kolmogorov-Smirov test (p < 0.01). How- number of safe choices, 11 increased, and 14
ever, there is more risk-seeking behavior (15 decreased. We observe similar patterns in the
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
VOL.92 NO. 5 HOLTAND LAURY:RISKAVERSIONAND INCENTIVEEFFECTS 1651
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
1652 THEAMERICANECONOMICREVIEW DECEMBER
2002
L.
(9
q
far more than the treatmentaverage of 6.0 safe
&l
choices. The intuitionfor this "absurd"amount
of predictedrisk aversioncan be seen by recon-
sidering the utility when payoffs, x, are scaled -- -- r
up by 20 underconstantabsolute risk aversion: 2 3 4 5 6 7 8 9 10
u(x) = -exp(-a20x). Since the baseline Decision
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
VOL.92 NO. 5 HOLTAND LAURY:RISKAVERSIONAND INCENTIVEEFFECTS 1653
aversion model. This can be done with a hybrid This paper presents the results of a simple
"power-expo" function (Atanu Saha, 1993) lottery-choiceexperimentthatallows us to mea-
that includes constant relative risk aversion sure the degree of risk aversion over a wide
and constant absolute risk aversion as special rangeof payoffs, rangingfrom severaldollarsto
cases: several hundreddollars. In addition, we com-
pare behavior under hypothetical and real
1 - exp(- ax -r)
incentives.
(2) U(x) = Although behavior is slightly more erratic
under the high-hypotheticaltreatments,the pri-
mary incentive effect is in levels (measuredas
which has been normalizedto ensurethatutility the numberof safe lottery choices in each treat-
becomes linear in x in the limit as a goes to 0. ment). Even at the low-payoff level, when all
It is straightforwardto show that the Arrow- prizes are below $4.00, about two-thirdsof the
Prattindex of relative risk aversion is: subjects exhibit risk aversion. With real pay-
offs, risk aversion increases sharplywhen pay-
offs are scaled up by factors of 20, 50, and 90.
-u"(x)x This result is qualitatively similar to that re-
(3) =r+ a( - r)xr,
u'(x) portedby Kachelmeierand Shehata(1992) and
Smith and Walker (1993) in different choice
which reduces to constantrelative risk aversion environments. In contrast, behavior is largely
of r when a = 0, and to constant absolute risk unaffectedwhen hypotheticalpayoffs are scaled
aversion of a when r = 0. For intermediate up. This paper presents estimates of a hybrid
cases (both parameters positive), the utility "power-expo"utility function that exhibits: (1)
function exhibits increasing relative risk aver- increasingrelativerisk aversion,which captures
sion and decreasingabsoluterisk aversion (Mo- the effects of payoff scale on the frequency of
hammed Abdellaoui et al., 2000).
Using the proportionof safe choices in each 12
of the ten decisions in the four real-payofftreat- If we restrictour attentionto those subjectswho never
ments, we obtained maximum-likelihood pa- switch back to Option A after choosing Option B, the noise
rameterestimates for this "power-expo"utility parameteris smaller, and both risk-aversionparametersare
larger.The estimates (and standarderrors)from this sample
function: /u = 0.134 (0.0046), r = 0.269 are I. = 0.110 (0.0041), r = 0.293 (0.017), and a = 0.032
(0.017), and a = 0.029 (0.0025), with a log- (0.003), with a log-likelihood of -247.8.
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
)
1 i--t--- xv.
0.9 - 0 .9 -
0.8 -Low 0 .8 - 20x Real Payoffs
Payoffs
0.7 - 0.7 - ',
', \\
o 0.6 - S 0..6-
? 0.5 - .5 - ',
= 0.4- '
0.4-
0.3 - . 0 .3-
v 0.2-, o .2 -
.9 6-
0.9 0.9 - I
,I\
' \ 50x Real Payoffs 11 Ox Real Payoffs
0.8 0.8 - I
II
0.7 0.7 - II
' 0.6
II .<
i \\ io 0.6 - II
. 0.5
I'
II ~~ ~ II
^ :_
0.5 - II
II
D 0.4 0.4 - II
II~ ~ \ _, II
I
i 0.3 , I \ 0.3- I
v 0.2 ,II,
9L( 0.2 - 1
"A
II
0.1 0.1- II
I
0 0v
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Decision Decision
FIGURE 6. PROPORTION OF SAFE CHOICES IN EACH DECISION: DATA AVERAGES AND PREDICTIONS
Note: Data [thick lines], risk neutrality [thin dashed lines], and predictions [thick dashed lines] with noise, for the hybrid
"power-expo"utility function with r = 0.269, a = 0.029, and noise = 0.134).
safe choices, and (2) decreasing absolute risk tive Utility with Revealed Preference: Ex-
aversion, which avoids absurdamounts of risk perimental Arguments Based on Prospect
aversion for high-stakes gambles. Behavior Theory." Working paper, CREED, Univer-
across all treatmentsconforms closely to the sity of Amsterdam, 2000.
predictionsof this model. Arrow, Kenneth J. Aspects of the theory of risk
One implicationof these results is that, con- bearing. Helsinki: Academic Bookstores,
trary to Kahnemanand Tversky's supposition, 1965.
subjects facing hypothetical choices cannot Bernoulli, Daniel. "Specimen Theoriae Novae de
imagine how they would actually behave under Mensura Sortis." Comentarii Academiae Sci-
high-incentive conditions. Moreover, these dif- entiarum Imperialis Petropolitanae, 1738, 5,
ferences are not symmetric: subjects typically pp. 175-92 [translated by L. Sommer in
underestimate the extent to which they will Econometrica, January 1954, 22(1), pp. 23-
avoid risk. Second, the clear evidence for risk 36].
aversion, even with low stakes, suggests the Binswanger,HansP. "AttitudeTowardRisk: Ex-
potential danger of analyzing behavior under perimental Measurement in Rural India."
the simplifying assumptionof risk neutrality. American Journal of Agricultural Econom-
ics, August 1980, 62, pp. 395-407.
REFERENCES Bosch-Domenech, Antoni and Silvestre, Joaquim.
"Does Risk Aversion or Attraction Depend
Abdellaoui, Mohammed; Barrios, Carolina and on Income?" Economics Letters, December
Wakker, Peter P. "Reconciling Introspec- 1999, 65(3), pp. 265-73.
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions
VOL.92 NO. 5 HOLTAND LAURY:RISKAVERSIONAND INCENTIVEEFFECTS 1655
Campo, Sandra; Perrigne, Isabelle and Vuong, Isaac, R. Mark and James, Duncan. "Just Who
Quang."Semi-ParametricEstimationof First- Are You Calling Risk Averse?" Journal of
Price Auctions with Risk Aversion." Work- Risk and Uncertainty,March2000, 20(2), pp.
ing paper,University of SouthernCalifornia, 177-87.
2000. Kachelmeier, Steven J. and Shehata, Mohamed.
Chen, Kay-Yut and Plott, Charles R. "Nonlinear "Examining Risk Preferences Under High
Behaviorin Sealed Bid First-PriceAuctions." MonetaryIncentives:ExperimentalEvidence
Games and Economic Behavior, October from the People's Republic of China."Amer-
1998, 25(1), pp. 34-78. ican Economic Review, December 1992,
Cox, James C. and Oaxaca, Ronald L. "Is Bid- 82(5), pp. 1120-41.
ding Behavior Consistent with Bidding The- Kahneman, Daniel and Tversky, Amos. "Prospect
ory for Private Value Auctions?" in R. M. Theory:An Analysis of Choice Under Risk."
Isaac, ed., Research in experimentaleconom- Econometrica, March 1979, 47(2), pp. 263-
ics, Vol. 6. Greenwich,CT: JAI Press, 1996, 91.
pp. 131-48. Luce, Duncan.Individualchoice behavior. New
Cox, James C. and Sadiraj, Vjollca. "Risk Aver- York: John Wiley & Sons, 1959.
sion and Expected-UtilityTheory:Coherence Pratt,John W. "Risk Aversion in the Small and
for Small- and Large-Stakes Gambles." in the Large." Econometrica, January-April
Workingpaper,University of Arizona, 2001. 1964, 32(1-2), pp. 122-36.
Eckel, Catherine;Grossman,Philip;Lutz, Nancy Rabin, Matthew."Risk Aversion and Expected
and Padmanabhan, V. "Playing it Safe: Gen- Utility Theory: A Calibration Theorem."
der Differences in Risk Aversion." Working Econometrica, January 2000, 68(5), pp.
paper, Virginia Tech, 1998. 1281-92.
Goeree,JacobK. and Holt, CharlesA. "A Model Saha, Atanu. "Expo-PowerUtility: A Flexible
of Noisy Introspection."Workingpaper,Uni- Form for Absolute and Relative Risk Aver-
versity of Virginia, 2000. sion."AmericanJournal of AgriculturalEco-
Goeree,Jacob K.; Holt, CharlesA. and Palfrey, nomics, November 1993, 75(4), pp. 905-13.
Thomas."QuantalResponse Equilibriumand Shupp, Robert S. and Williams, Arlington W.
Overbidding in Private-Value Auctions." "Risk Preference Differentials of Small
Working paper, CaliforniaInstituteof Tech- Groupsand Individuals."Workingpaper,In-
nology, 1999. diana University, 2001.
. "Risk Aversion in Games with Mixed Siegel, Sydney. Nonparametric statistics. New
Strategies." Working paper, University of York: McGraw-HillBook Company, 1956.
Virginia, 2000. Smith, Vernon L. and Walker, James M. "Re-
Harrison,GlennW. "Theoryand Misbehaviorin wards, Experience, and Decision Costs in
First-Price Auctions." American Economic First Price Auctions." Economic Inquiry,
Review, September1989, 79(4), pp. 749-62. April 1993, 31(2), pp. 237-44.
This content downloaded from 132.210.236.20 on Wed, 28 Oct 2015 16:00:48 UTC
All use subject to JSTOR Terms and Conditions