106
S T A R M E R
first statement of EU—solved the St. Petersburg puzzle, it did not find much favor
with modern economists until the 1950s. This is partly explained by the fact that,
in the form presented by Bernoulli, the theory presupposes the existence of a
cardinal utility scale; an assumption that did not sit well with the drive toward
ordinalization during the first half of the twentieth century.
Interest in the theory was revived when John von Neuman and Oskar Morgenstern
(1947) showed that the expected utility hypothesis could be derived from a set of
apparently appealing axioms on preference. Since then, numerous alternative
axiomatizations have been developed, some of which seem highly appealing,
some might even say compelling, from a normative point of view (see for exam-
ple Peter Hammond 1988).
2
To the extent that its axioms can be justified as sound
principles of rational choice to which any reasonable person would subscribe,
they provide grounds for interpreting EU
normatively
(as a model of how people
ought to choose) and
prescriptively
(as a practical aid to choice). My concern,
however, is with how people actually choose, whether or not such choices con-
form with a priori notions of rationality. Consequently, I will not be delayed by
questions about whether particular axioms can or cannot be defended as sound
principles of rational choice, and I will start from the presumption that evidence
relating to actual behavior should not be discounted purely on the basis that it
falls foul of conventional axioms of choice.
For the purpose of understanding alternative models of choice, it will be useful
to present one set of axioms from which EU can be derived. In the approach that
I adopt, at least to begin with, preferences are defined over
prospects
, where a
prospect is to be understood as a list of consequences with associated probabili-
ties. I will assume throughout that all consequences and probabilities are known
to the agent, and hence, in choosing among prospects, the agent can be said to
confront a situation of
risk
(in contrast to situations of
uncertainty
in which at
least some of the outcomes or probabilities are unknown). I will use lowercase
letters in bold (e.g.,
q
,
r
,
s
) to represent prospects, and the letter
p
to represent
probabilities (take it that
p
always lies in the interval [0,1]). A given prospect may
contain other prospects as consequences, but assuming that such compound
prospects can be reduced to simple prospects following the conventional rules
of probability, we can represent any prospect
q
by a probability distribution
q
5
(
p
1
, . . . ,
p
n
) over a fixed set of pure consequences
X
5
(
x
1
, . . . ,
x
n
) where
p
i
is the probability of
x
i
,
p
i
$
0 for all
i
, and •
p
i
5
1. Hence, the elements of
X
are
to be understood as an exhaustive and mutually exclusive list of possible conse-
quences which may follow from a particular course of action. While this notation
allows a prospect to be written simply as vector of probabilities (as
q
above) it
will sometimes be useful to be explicit about the consequences too—e.g., by writ-
ing
q
5
(
x
1
,
p
1
, . . . ,
x
n
,
p
n
).
Given these preliminaries, the expected utility hypothesis can be derived from
three axioms: ordering, continuity, and independence. The ordering axiom requires
2
Such arguments, while widely accepted, are nevertheless controversial. See, for example, Anand
(1993) and Sugden (1991).
Do'stlaringiz bilan baham: |