The rich get richer and the poor get poorer: On risk aversion in behavioral decision-making

Judgment and Decision Making, Vol. 1, No. 2, November 2006, pp. 153–158
The rich get richer and the poor get poorer:
On risk aversion in behavioral decision-making
Ingmar H. A. Franken?, Irina Georgieva, Peter Muris
Institute of Psychology, Erasmus University Rotterdam, The Netherlands
Ap Dijksterhuis
Department of Psychology
University of Amsterdam
Abstract
Some studies have found that choices become more risk averse after gains and more risk seeking after losses, although
other studies have found the opposite. The latter tend to use hypothetical cases that encourage deliberation. In the current
study, we examined the effects of prior gains and losses on a task designed to encourage less re?ective decision making,
the Iowa Gambling Task (IGT). Fifty participants conducted a manipulated decision-making task in which one group
gained money, whereas the other group lost money, followed by the IGT. Participants who experienced a prior monetary
loss displayed more risky choice behavior on the IGT than subjects who experienced a prior gain. These effects were
not mediated by a positive or negative affect, although the sample size may have been too small to detect a small effect.
Keywords: implicit decision-making, reward, punishment, Iowa Gambling Task, monetary choices, risk behavior.
1 Introduction
Some traditional economic studies addressing theories
of decision-making assume that decision-making is based
Kahneman and Tversky (1979) noted that people are of-
on deliberate evaluations of varying option-outcome sce-
ten risk averse for gains and risk seeking for losses.
narios, that is, people weigh the pros and cons of various
Whether people consider a consequence of their choice
choices against each other and base their decision on the
as a loss or as a gain is dependent on their point of ref-
outcome of this comparison. These kinds of choices can
erence. This reference point, which is often equivalent to
be characterized as deliberate, and carefully thought-out.
the current wealth position, plays a key role in the theory
However, some recent psychological studies address-
of choice.
ing decision-making show that decisions can also be
It should be possible to manipulate perceptions of the
driven by less carefully thought-out choices (Dijkster-
domain (gain or loss) with actual prior gains or losses.
huis, Bos, Nordgren, & van Baaren, 2006), are often
People may see their starting point, before the gain or
implicit and automatic (Hastie, 2001), and are based on
loss, as the reference point. If they had lost money, for
“gut-feelings” (Damasio, 1996) or emotions (Loewen-
example, they may see new gambles as in the domain of
stein et al., 2001; Sanfey, Loewenstein, McClure, &
losses, and they therefore might be risk seeking. Earlier
Cohen, 2006). Recently, Sanfey et al. (2006) made a
studies of the effect of gains and losses show con?icting
clear distinction between these two psychological sys-
results. Thaler and Johnson (1990) found the opposite
tems involved in economic decision-making: an emo-
results — which they called a “house money effect” —
tional system, which involves the activation of automatic
although their participants would take risks to gain back
processes and a deliberative system involving controlled
all of their loss. Weber and Zuchel (2005) review this lit-
processes, with each having separate neural substrates.
erature and ?nd some conditions that support the Prospect
In the present contribution, we want to apply this recent
Theory prediction. Aside from their result, however, most
knowledge to risk aversion. Is risk aversion after gains
of the results consistent with Prospect Theory are from
the consequence of people’s deliberate, conscious deci-
studies that use more realistic situations such as invest-
sions to avoid risk? Or is the case that risk aversion
ment, rather than hypothetical tasks.
can largely automatic, whereby people’s current refer-
ence point leads them to pursue less risky options without
?This work was supported by the Netherlands Organization for
deliberately weighting all outcome scenarios?
Scienti?c Research (NWO). Ingmar Franken, Institute of Psychology,
Erasmus University Rotterdam, Woudestein J5-43, P.O. Box 1738, 3000
In the present study we examined the role of reference
DR Rotterdam, The Netherlands, E-mail: franken@fsw.eur.nl
point in a task designed to encourage automatic, emo-
153

---

Judgment and Decision Making, Vol. 1, No. 2, November 2006
Risk aversion in the IGT
154
tional driven decision-making, the Iowa Gambling Task
had no in?uence on these gains or losses). Note that this
(IGT). During the IGT participants have to select cards
experimental set-up comes close to real-life situations in
from four decks that range in probability and magnitude
which a person’s reference point (real or perceived) is of-
of rewards and punishments (Bechara, Damasio, Dama-
ten the result of their prior choices.
sio, & Anderson, 1994; Bechara, Damasio, & Damasio,
It is known that individual differences can in?uence
2000). To translate our hypothesis pertaining to risk aver-
behavioral decision-making. These individual difference
sion to the IGT, it is necessary to explain the IGT in some
variables include reward sensitivity (Franken & Muris,
detail. In the IGT, participants can repeatedly choose
2005), gender (Overman, 2004), and age (Wood, Cox,
(usually up to 100 times) between four decks of cards.
Davis, Busemeyer, & Koling, 2005). In line with previ-
Two of the decks (e.g., A and B) are disadvantageous.
ous research (Peters & Slovic, 2000), we expected that
They produce large immediate gains, but these gains are
our experimental manipulation would have an effect on
followed by large losses, leading to an overall loss in the
participants’ affect. More precisely, a prior gain would
long run. The other remaining decks (e.g., C and D) are
yield an increase of positive affect, whereas an earlier loss
advantageous. The gains are modest but consistent and
would result in an increase of negative affect. It has been
the losses are small. Consistently choosing these decks
suggested that affect might in?uence decision-making
leads to gains in the long run. This means that people who
(Ashby, Isen, & Turken, 1999; Loewenstein et al., 2001).
are risk seeking would be predominantly choose decks
Positive affect can promote increased sensitivity to losses
A and B, leading to losses in the long run. Conversely,
(Isen, Nygren, & Ashby, 1988). In the present study, we
people who are risk averse will predominantly choose
investigated whether the above-mentioned individual dif-
decks C and D, leading to overall gain. This means that
ferences and affect may have an additional effect on the
risk aversion translates into better performance (overall
participants’ decision-making.
gains) on the IGT, whereas risk seeking would translate
The main hypothesis was that people who experienced
into poor performance (overall losses).
a prior gain on a gambling task performed better (i.e.,
The psychological process that determines people’s be-
made more advantageous choices as a consequence of
havior in the IGT is crucial to our hypothesis that risk
risk aversion) on the IGT as compared to persons who
aversion is not only based on deliberately weighting all
experienced a prior loss. Furthermore, we asked whether
outcome scenarios. A general consensus is that people
this effect was in?uenced by subjective affect, and vari-
performing the IGT at some point steer towards certain
ous other individual differences.
(pro?table) decks, in rather automatic way. Whether this
automatic behavior is entirely unconscious is still subject
to debate; see Maia and McClelland (2004), and Dunn
2 Method
et al. (2006). Behavior on the IGT can be seen as a
form of implicit learning (Reber, 1993), whereby behav-
2.1 Participants
ior changes before people can verbalize why they do what
Fifty undergraduate psychology students (11 males) were
they are doing. Therefore, the IGT can be regarded as
recruited to participate in the present study. Their mean
an instrument capable of assessing intuitive and emotion-
age was 20.6 years (SD = 3.2). All participants received
based decision-making processes.
course credits for participating and could gain additional
In addition to our central aim – to test the relative au-
money depending on their performance on the IGT, rang-
tomaticity of risk aversion – we have another goal. Eco-
ing between 1 and 6 e. Participants were randomized into
nomic studies addressing theories of decision-making of-
two groups: a Prior Loss (PL) group (n = 25; 5 males) or
ten rely on hypothetical situations and choices in which
a Prior Gain (PG) group (n = 25; 6 males). All subjects
participants are confronted with monetary gambles with-
signed informed consent prior to the beginning of the ex-
out any real consequences. Although the use of real in-
periment.
centives is often not crucial for the outcome of experi-
ments, using real incentives has an important role to play
2.2 Instruments
in establishing the quality, credibility, and generalizabil-
ity of experimental data (Beattie and Loomes, 1997).
For the present study we used the computerized version
In the present study, we addressed this point by using
of the IGT to measure decision-making (Bechara, Tranel,
real monetary remunerations in order to mimic real-life
& Damasio, 2000; we used the same monetary outcomes
decision-making more closely. For the purpose of the
but substituted Euros for dollars). This task consists of
present study, we experimentally manipulated the refer-
100 successive trials, which were split into ?ve 20-trial
ence point. That is, participants ?rst performed a ma-
blocks for analysis, in which subjects are instructed to
nipulated gamble-task in which they either gained or lost
try to gain as much money as possible by drawing cards
money as a result of their performance (in actuality, they
from one of four decks. The decisions to choose from the

---

Judgment and Decision Making, Vol. 1, No. 2, November 2006
Risk aversion in the IGT
155
decks are motivated by reward and punishment schedules
ipants made. This manipulated IGT was programmed to
inherent in the task. Two of the decks (i.e., A and B)
yield a gain of four e in the PG group and a loss of 10
are disadvantageous, producing immediate gains (large
e in the PL group. Irrespective of the card choice, there
rewards) but these are accompanied by larger losses in
was always a pre-determined pattern of gains/losses. The
the long run (larger punishments). The C and D decks
proportion of cards with losses were in all tasks and all
are advantageous: gains are modest but more consistent
decks 50%. There were no differences among the A, B,
and losses are smaller. See Bechara, Tranel, & Damasio,
C, and D decks, they were all equal. The difference be-
2000, for the payoff and probability scheme of the IGT.
tween the PL and PG condition was were the amount of
The net-score (the number of advantageous decks choices
losses, which were of course larger in the PL condition.
minus the number of disadvantageous decks choices) was
In order to make the reference point (i.e., gain or loss)
used as dependent variable. A higher score indicates that
more salient (Heath, Larrick, & Wu, 1999), participants
a subject is more often choosing advantageous decks.
in the PG group were told that they gained money above
There is general consensus that the “IGT has proved to
average on this task, whereas participants in the PL group
be a sensitive, ecologically valid measure of decision-
were told that they lost more than average on this task. In
making” (Dunn et al., 2006).
addition, participants were instructed that complete new
The BIS/BAS Scales (Carver & White, 1994) were
rules applied to the second game, that they needed to em-
presented as a self-report questionnaire that has been
ploy other decision-strategies in order to gain money, and
constructed to assess individual differences in personal-
that other decks would be advantage and disadvantage.
ity dimensions that re?ect the sensitivity of two motiva-
Again, they were told that some decks would be more
tional systems, the aversive and appetitive system (BIS
advantageous than others. Furthermore, all participants
and BAS; Gray, 1987). The BIS/BAS Scales consist of
were told that their prior loss or gain would be the start-
20 items that can be allocated to two primary scales: the
ing point for the second task. In other words, the PG
Behavioral Inhibition System scale (BIS; 7 items) and
group started with an initial credit of four e, and the PL
the Behavioral Approach System scale (BAS; 13 items).
group started with an initial debt of 10 e. After the ma-
The BAS can be divided into 3 subscales: Fun Seeking
nipulated IGT, subjects completed the PANAS for a sec-
(4 items), Reward Responsiveness (5 items), and Drive
ond time in order to measure whether the experimental
(4 items). The Dutch version of the BIS/BAS Scales
manipulation resulted in a change of affect. Finally, par-
has been described in previous studies (Franken, 2002;
ticipants carried out the “real” IGT, which measured their
Franken, Muris, & Rassin, 2005). Cronbach’s alphas for
actual behavioral decision-making.
various scales were found to range from .61 to .79.
The Positive and Negative Affect Scales (PANAS;
2.4 Analysis
Watson, Clark, & Tellegen, 1988) were administered as
a measure of positive and negative affect. The PANAS
In order to test the main hypothesis, an hierarchical re-
is a 20-item bidimensional mood inventory with a 5-
gression analysis was carried out with the IGT net-score
point Likert-scale response format. Positive affect re-
as dependent variable and age, gender, group, affect (pre
?ects the extent to which a person feels enthusiastic, ac-
minus post affect scores1), and BIS, and BAS as covari-
tive, and alert, whereas negative affect is a general di-
ates. We entered gender and age in the ?rst block of the
mension of subjective distress and unpleasurable engage-
regression, group in the second, positive and negative af-
ment that subsumes a variety of aversive mood states, in-
fect in the third, and BIS and BAS in the fourth block.
cluding anger, contempt, disgust, guilt, fear, and nervous-
Additionally, differences on affect (pre versus post) were
ness (Watson et al., 1988). Psychometric properties of the
tested using a 2 (time) x 2 (group) ANOVA. Further, in or-
PANAS scales are good (Boon & Peeters, 1999; Watson
der to investigate the performance of the two groups per
et al., 1988).
block (i.e., 20 cards), a multivariate ANOVA (MANOVA)
was performed with the scores on the ?ve subsequent
blocks as dependent variables.
2.3 Procedure and manipulation
Participants were told that they participated in a gam-
2.5 Results
bling study and that we aimed to investigate decision-
making qualities. First, participants completed all ques-
Figure 1 displays the mean IGT net scores of both groups
tionnaires. Subsequently, half of the subjects carried out
over the ?ve blocks. As can be seen in Table 1, the
the “loss” version of the manipulated IGT, while the other
group variable made a unique and signi?cant contribu-
half conducted the “gain” version of the manipulated
tion to IGT-scores. Age, gender, affect, BIS, and BAS
IGT. For both groups, we used a ?xed, pseudo-random,
1Using pre-manipulated IGT and post-manipulated IGT affect
gain/loss schedule irrespective of the choices that partic-
scores in the regression model yielded similar results.

---

Judgment and Decision Making, Vol. 1, No. 2, November 2006
Risk aversion in the IGT
156
8
The MANOVA showed a signi?cant multivariate ef-
fect, Wilks’ lambda = .75, F(5,44) = 2.89, p = .024.
6
Follow-up MANOVAs performed on the participants per-
formance on the separate blocks revealed a signi?cant
4
difference in the IGT scores for block 2, F(1,42) = 8.41,
p = .006, and block 3, F(1,42) = 11.05, p = .002. The fact
2
that only in blocks 2 and 3 participants in the PG group
0
made more advantageous choices than participants in the
Net IGT score
PL group is consistent with our theorizing. In block 1,
?2
people are generally oblivious to the nature of the decks,
leading to rather random choice behavior. In blocks 2 and
?4
3, people are developing preferences for certain decks,
Prior gain
Prior loss
leading to more systematic choices. Later during the task
?6
(blocks 4 and up), more and more people start to under-
1
2
3
4
5
stand the nature of the decks, leading to consistent fa-
vorable (risk averse) choices irrespective of experimental
Block
condition.
Figure 1: IGT score over the ?ve blocks per group (with
standard errors).
3 Discussion
did not predict IGT-scores, indicating that these variables,
Our results show that a reference point manipulation us-
including affect, had no in?uence on decision making.
ing prior gains or losses affected decisions with mone-
tary consequences. The study adds further experimental
evidence that people who “have” make more risk-averse
Table 1: Results of hierarchical regression analyses pre-
decisions, while the “have-nots” make more risk-seeking
dicting performance on the net score of the Iowa Gam-
decisions. This phenomenon has frequently been ob-
bling Task.
served from studies using hypothetical decision-making
situations (Thaler & Johnson, 1990) and agrees with the
B
S.E. B
?
?2 R
increased risk aversion principle of Prospect Theory. This
Step 1
.01
theory predicts exactly what we found, that is, prior losses
Gender
0.49
1.27
?0.06
put the subject in the domain of losses and prior gains
Age
?2.94
9.54
?0.05
have the opposite effect.
Step 2
.15*
Insofar as the IGT is, as hypothesized, sensitive to non-
Group
?21.32
7.35
?0.40*
deliberative mechanisms of decision making, our results
Step 3
.00
show that risk seeking and risk aversion as a function of
Positive Affect
?0.17
0.62
?0.04
prior gains and losses does not need to be the result of a
Negative Affect
0.24
0.70
0.06
deliberate, well-considered choice strategy: risk seeking
and risk aversion can be automatic and non-deliberately,
Step 4
.02
it can be seen as a spontaneous process, steering people
BIS
?1.19
1.25
?0.15
towards or away from risk.
BAS
?0.18
0.85
?0.03
A secondary goal was to investigate the role of emo-
BIS = Behavioral Inhibition System. BAS = Behavioral
tions (affect). We successfully induced positive affect
Approach System.
in the PG group and negative affect in the PL group.
* p < .001.
However, affect variables did not in?uence the relation
between prior loss/gain and decision-making.
Addi-
There was a signi?cant group x time effect for posi-
tional correlation analysis between positive/negative af-
tive affect, F(1,48) = 10.32, p = .002, and negative affect
fect scores (i.e., pre, post, and pre-post difference scores)
F(1,48) = 31.54, p = .000001. More speci?cally, the ma-
and IGT score showed that there were no signi?cant links
nipulated IGT resulted in an increase in positive affect in
between affect and decision-making (all p’s > .05). Ac-
the PG group and an increase in negative affect in the PL
cordingly, from the present ?ndings, it can be concluded
group. However, in the regression analysis, this change
that the effect of a reference point on behavioral decision
in affect did not in?uence the relation between the prior
was not mediated by positive or negative affect. This is in
loss or gain and “real” IGT performance.
contrast with earlier ?ndings of Peters and Slovic (2000),

---

Judgment and Decision Making, Vol. 1, No. 2, November 2006
Risk aversion in the IGT
157
who found that high negative affect was associated with
not task complexity. Child Neuropsychology, 11, 245-
more avoidance of high-loss options and high positive af-
263.
fect was associated with more choices from high-gain op-
Damasio, A. R. (1996). The somatic marker hypothe-
tions. An explanation for these different results might be
sis and the possible functions of the prefrontal cortex.
that Peters and Slovic used a different version of the Iowa
Philosophical Transactions Of The Royal Society Of
gambling task. Whereas we used the original task, Pe-
London. Series: B-Biological Sciences, 351, 1413-
ters and Slovic used a gambling task that was on several
1420.
points different from the original task. In addition, the
Dijksterhuis, A., Bos, M. W., Nordgren, L. F., & van
present sample size may have insuf?cient power to detect
Baaren, R. B. (2006). On making the right choice:
a signi?cant result concerning the in?uence of affect.
the deliberation-without-attention effect. Science, 311,
Although it is conceivable that, by the ?fth block, PL
1005-1007.
participants might have thought that the risky decks could
Dunn, B. D., Dalgleish, T., & Lawrence, A. D. (2006).
undo their prior loss, this could not occur in the second
The somatic marker hypothesis: A critical evaluation.
block, and the difference between PL and PG conditions
Neuroscience and Biobehavioral Reviews, 30, 239-
was already present. Thus, we conclude that the PL does
271.
increase risk seeking in the IGT, as predicted by Prospect
Franken, I. H. A. (2002). Behavioral approach system
Theory.
(BAS) sensitivity predicts alcohol craving. Personality
and Individual Differences, 32, 349-355.
Franken, I. H. A., & Muris, P. (2005). Individual differ-
References
ences in decision-making. Personality and Individual
Differences, 39, 991-998.
Ashby, F. G., Isen, A. M., & Turken, U. (1999). A neu-
Franken, I. H. A., Muris, P., & Rassin, E. (2005). Psycho-
ropsychological theory of positive affect and its in?u-
metric properties of the Dutch BIS/BAS Scales. Jour-
ence on cognition. Psychological Review, 106, 529-
nal of Psychopathology and Behavioral Assessment,
550.
27, 25-30.
Beattie, J., & Loomes, G. (1997). The impact of incen-
Gray, J. A. (1987). The psychology of fear and stress.
tives upon risky choice experiments. Journal of Risk
Cambridge: Cambridge University Press.
and Uncertainty, 14(2), 155-168.
Hastie, R. (2001). Problems for judgment and decision
Bechara, A., Damasio, A. R., Damasio, H., & Anderson,
making. Annual Review of Psychology, 52, 653-683.
S. W. (1994). Insensitivity to future consequences fol-
Heath, C., Larrick, R. P., & Wu, G. (1999). Goals as
lowing damage to human prefrontal cortex. Cognition,
reference points. Cognitive Psychology, 38, 79-109.
50, 7-15.
Hertwig, R., & Ortmann, A. (2001). Experimental prac-
Bechara, A., Damasio, H., & Damasio, A. R. (2000).
tices in economics: a methodological challenge for
Emotion, decision making and the orbitofrontal cortex.
psychologists? Behavioral and Brain Sciences, 24,
Cerebral Cortex, 10, 295-307.
383-403.
Bechara, A., Damasio, H., Tranel, D., & Damasio, A. R.
Huber, O., Wider, R., & Huber, O. W. (1997). Active
(1997). Deciding advantageously before knowing the
information search and complete information presen-
advantageous strategy. Science, 275, 1293-1295.
tation in naturalistic risky decision tasks. Acta Psycho-
Bechara, A., Tranel, D., & Damasio, H. (2000). Char-
logica, 95, 15-29.
acterization of the decision-making de?cit of patients
Isen, A. M., Nygren, T. E., & Ashby, F. G. (1988). In?u-
with ventromedial prefrontal cortex lesions. Brain,
ence of positive affect on the subjective utility of gains
123, 2189-2202.
and losses: it is just not worth the risk. Journal of Per-
Boon, M. T., & Peeters, F. P. (1999). Affectieve dimen-
sonality and Social Psychology, 55, 710-717.
sies bij depressie en angst (Dimensions of affectivity in
Kahneman, D., & Tversky, A. (1979). Prospect Theory:
depression and anxiety). Tijdschrift voor Psychiatrie,
An analysis of decision under risk. Econometrica, 47,
41, 109-113.
263-292.
Carver, C. S., & White, T. L. (1994). Behavioral inhi-
Loewenstein, G. F., Weber, E. U., Hsee, C. K., & Welch,
bition, behavioral activation, and affective responses
N. (2001). Risk as feelings. Psychological Bulletin,
to impending reward and punishment: The BIS/BAS
127, 267-286.
scales. Journal of Personality and Social Psychology,
Maia, T. V., & McClelland, J. L. (2004). A reexamina-
67, 319-333.
tion of the evidence for the somatic marker hypothesis:
Crone, E. A., Bunge, S. A., Latenstein, H., & Van Der
What participants really know in the Iowa gambling
Molen, M. W. (2005). Characterization of children’s
task. Proceedings of the National Academy of Sciences
decision making: Sensitivity to punishment frequency,
of the United States of America, 101, 16075-16080.

---

Judgment and Decision Making, Vol. 1, No. 2, November 2006
Risk aversion in the IGT
158
Overman, W. H. (2004). Sex differences in early child-
of prior outcomes on risky choice. Management Sci-
hood, adolescence, and adulthood on cognitive tasks
ence, 36(6), 643-660.
that rely on orbital prefrontal cortex. Brain and Cogni-
Watson, D., Clark, L. A., & Tellegen, A. (1988). De-
tion, 55, 134-147.
velopment and validation of brief measures of positive
Peters, E., & Slovic, P. (2000).
The springs of ac-
and negative affect: The PANAS Scales. Journal of
tion: Affective and analytical information processing
Personality and Social Psychology, 54, 1063-1070.
in choice. Personality and Social Psychology Bulletin,
Weber, M. & Zuchel, H. (2005). How do prior outcomes
26, 1465-1475.
affect risk attitude? Comparing escalation of commit-
Reber, A. S. (1993). Implicit learning and tacit knowl-
ment and the house-money effect. Decision Analysis,
edge: An essay on the cognitive unconscious. New
2, 30–43.
York: Oxford University Press.
Wood, S., Cox, C. R., Davis, H., Busemeyer, J., & Kol-
Sanfey, A. G., Loewenstein, G., McClure, S. M., & Co-
ing, A. (2005). Older adults as adaptive decision mak-
hen, J. D. (2006). Neuroeconomics: cross-currents in
ers: Evidence from the Iowa Gambling Task. Psychol-
research on decision-making. Trends in Cognitive Sci-
ogy and Aging, 20, 220-225.
ence, 10, 108-116.
Thaler, R. H., & Johnson, E. J. (1990). Gambling with
the house money and trying to break even: The effects

---