Using the skew symmetry of • (
?
,
?
), the term in square brackets is equal to [•
(
x
,
y
)
1
• (
y
,
z
)
2
• (
x
,
z
)]. Assuming regret-aversion, this will be negative, hence
regret theory implies a strict preference
A
2
s
A
1
. It is easy to see that the same
reasoning applied to the other two possible pairwise comparisons implies
A
3
s
A
2
and
A
1
s
A
3
. Hence, regret theory also implies a
cycle of preference
of the form:
A
2
s
A
1
,
A
3
s
A
2
,
A
1
s
x A
3
. Now consider adding some small positive amount •
to one consequence of action
A
1
. The resulting action, call it
A
1
*, stochastically
dominates each of the original actions. But since regret theory implies
A
2
s
A
1
we should expect
A
2
s
A
1
* for at least some •
.
0. Hence regret theory also im-
plies violations of monotonicity.
Relative to the conventional approach then, preferences in regret theory are not
at all well behaved: they satisfy neither monotonicity nor transitivity, and the the-
ory allows strict preferences between stochastically equivalent acts. While such
properties may seem peculiar to the eye of the conventional economist, from the
descriptive angle, the crucial question is whether such implications of the theory
are borne out by actual behavior. Shortly after proposing regret theory, Loomes
and Sugden (1983) argued that at least one might be. Consider the following three
acts labeled $, P, and M with monetary consequences
x
.
y
.
m
.
0 defined (for
the sake of simplicity) over three equiprobable states:
s
1
s
2
s
3
$
x
0
0
P
y
y
0
M
m
m
m
The actions labeled $ and P have the structure of typical $- and P-bets: they are
binary gambles where $
has the higher prize, and P the higher probability of
“winning”; the third act gives payoff
m
for sure. Loomes and Sugden show that,
given regret-aversion, pairwise choices over acts with this structure may be cycli-
cal, and if a cycle occurs, it will be in a specific direction with P
s
$, M
s
P and
$
s
M. Now recall
that in a standard experiment, subjects reveal P
s
$ in a
straight choice between options but place a higher value on $
relative to P in
separate valuation tasks. If we interpret choices from {$, M} and {P, M} as ana-
logues of valuation tasks asking “is $ (or P) worth more or less than
m
,” then
the cycle predicted by regret theory can be interpreted as a form of preference
reversal.
So, regret theory offers the tantalizing opportunity of explaining violations of
independence and preference reversal within a theory of preference maximiza-
tion. Of course, since observation of preference reversal predates the development
of regret theory, that phenomena offers only weak support for the unconventional
predictive content of regret theory. More recent research has aimed at testing
some novel predictions of regret theory and some of the results from this line of
research are discussed in Starmer (2000).
Do'stlaringiz bilan baham: