108
S T A R M E R
with a strict inequality for at least one
i
. Monotonicity is the property where by
stochastically dominating prospects are preferred to prospects that they dominate
and it is widely held that any satisfactory theory—descriptive or normative—
should embody monotonicity. I will have more to say about this later.
The shape of the utility function also has a simple behavioral interpretation
whereby concavity (convexity) of
u
(
?
) implies risk averse (prone) behavior; an
agent with a concave utility function will always prefer a certain amount
x
to any
risky prospect with expected value equal to
x
. Modeling risk preferences in this
way does collapse some potentially distinct concepts into a single function: any
attitude to chance (e.g., like or dislike of taking risks) and any attitude toward
consequences (e.g., a diminishing marginal utility of money) must all be captured
by the utility function. That need not imply any weakness of the theory. Indeed it
is precisely the simplicity and economy of EU that has made it such a powerful
and tractable modeling tool. My concern, however, is with the descriptive merits
of the theory and, from this point of view, a crucial question is whether EU pro-
vides a sufficiently accurate representation of actual choice behavior. The evi-
dence from a large number of empirical tests has raised some real doubts on this
score.
3
. Descriptive Limitations of Expected Utility Theory—
The Early Evidence
Empirical studies dating from the early 1950s have revealed a variety of patterns
in choice behavior that appear inconsistent with EU. I shall not attempt a full-
blown review of this evidence.
3
Instead, I discuss one or two examples to illustrate
the general nature of this evidence, and offer a discussion of its role in stimulating
the development of new theories. With hindsight, it seems that violations of EU
fall under two broad headings: those that have possible explanations in terms of
some “conventional” theory of preferences and those that apparently do not. The
former category consists primarily of a series of observed violations of the inde-
pendence axiom of EU; the latter, of evidence that seems to challenge the as-
sumption that choices derive from well-defined preferences. Let us begin with the
former.
There is now a large body of evidence that indicates that actual choice behavior
may
systematically
violate the independence axiom. Two examples of such phe-
nomena, first discovered by Maurice Allais (1953), have played a particularly
important role in stimulating and shaping theoretical developments in non-EU
theory. These are the so-called
common consequence effects
and
common ratio ef-
fects
. The first sighting of such effects came in the form of the following pair of
hypothetical choice problems. In the first you have to imagine choosing between
the two prospects:
s
1
5
($1M,1) or
r
1
5
($5M, 0.1; $1M, 0.89; 0, 0.01). The first
3
Those interested in more thorough reviews are recommended to consult Schoemaker (1982) and,
more recently, Camerer (1995).
Do'stlaringiz bilan baham: |