123
N O N E X P E C T E D - U T I L I T Y T H E O R Y
“the most natural and useful modification of the classical expected utility formula”
and, as testament to this, it has certainly proved to be one of the most popular
among economists. In this type of model the weight attached to any consequence
of a prospect depends not only on the true probability of that consequence but
also on its ranking relative to the other outcomes of the prospect. With conse-
quences indexed as before such that
x
1
is worst and
x
n
best, we can state rank-
dependent expected-utility theory as the hypothesis that agents maximize the de-
cision-weighted form with weights given by
p
i
5
• (
p
i
1 ? ? ? 1
p
n
)
2
• (
p
i
1
1
1 ? ? ? 1
p
n
) for
i
5
1, . . . ,
n
2
1,
p
i
5
• (
p
i
)
for
i
5
n
.
In this model there is a meaningful distinction between decision weights (
w
) and
probability weights (•). Richard Gonzalez and George Wu (1999, p. 135) suggest
an interpretation of the probability-weighting function as reflecting the underly-
ing “psychophysics of risk,” that is, the way in which individuals subjectively
“distort” objective probabilities; the decision weight then determines how the
probability weights enter the value function
V
(
?
). Notice that • (
p
i
1 ? ? ? 1
p
n
) is
a subjective weight attached to the probability of getting a consequence of
x
i
or
better, and • (
p
i
1
1
1 ? ? ? 1
p
n
) is a weight attached to the probability of getting a
consequence better than
x
i
, hence in this theory • (
?
) is a transformation on cumu-
lative probabilities. This procedure for assigning weights ensures that
V
(
?
) is
monotonic. It also has the appealing property that, in contrast to the simple deci-
sion-weighting models that assign the same decision weight to any consequence
with probability
p
, the weight attached to a consequence may vary according to
how “good” or “bad” it is. So in principle this would allow for, say, extreme out-
comes to receive particularly high (or low) weights. A less appealing feature of
the model is that a small change in the value of some outcome of a prospect can
have a dramatic effect on its decision weight if the change affects the rank order
of the consequence; but a change in the value of an outcome, no matter how large
the change, can have no effect on the decision weight if it does not alter its rank.
The predictions of the rank-dependent model rely crucially on the form of • (
?
).
If • (
?
) is convex, this generates a set of concave indifference curves (implying
aversion to randomization) that are parallel at the hypotenuse but fan out as we
move left to right across the triangle and fan in (i.e., become less steep) as we
move vertically upwards. Aside from the hypotenuse parallelism that holds for
any • (
?
) (see Camerer 1989), the reverse pattern of indifference curves (i.e., con-
vex curves, horizontal fanning in, and vertical fanning out) is generated with a
concave • (
?
).
Curvature of • (
?
) in the rank-dependent model has been interpreted as
reflecting “optimism” and/or “pessimism” with respect to probabilities (see
Quiggin 1982; Yaari 1987; Diecidue and Wakker 1999). Consider, for example,
the prospect
q
5
(
x
1
, 0.5;
x
2
, 0.5). Assigning weights to the consequences of
q
according to the rank-dependent method above gives
p
1
5
1
2
• (0.5) and
p
2
5
• (0.5). With • (
?
) convex, • (0.5)
,
0.5, hence the weight attached to the
Do'stlaringiz bilan baham: |