means that you make the decision about whether to take the gamble by
comparing the average of w(−1200) and w(−800) with w(−1000), where w
is defined in equation (12). Of course, under this assumption, convexity in
the region of losses leads to risk-seeking after a loss.
The idea that people integrate the outcomes of sequential gambles, while
appealingly simple, is only a hypothesis. Tversky and Kahneman (1981)
themselves note that prospect theory was originally developed only for ele-
mentary, one-shot gambles and that any application to a dynamic context
must await further evidence on how people think about sequences of gains
and losses. A number of papers, including Thaler and Johnson (1990), have
taken up this challenge, conducting experiments on whether people inte-
grate sequential outcomes, segregate them, or do something else again.
That these experiments have uncovered increasedrisk aversion after prior
losses does not contradict prospect theory; it simply rejects the hypothesis
that people integrate sequential gambles.^16
Thaler and Johnson (1990) do find some situations where prior losses
lead to risk-seeking behavior. These are situations where after the prior
loss, the subject can take a gamble that offers a good chance of breaking
even and only limited downside. In conjunction with the other evidence,
this suggests that while losses after prior losses are very painful, gains that
enable people with prior losses to break even are especially sweet. We have
not been able to introduce this break-even effect into our framework in a
tractable way. It is worth noting, though, that outside of the special situa-
tions uncovered by Thaler and Johnson, it is increased risk aversion that
appears to be the norm after prior losses.
4 .Equilibrium PricesWe now derive equilibrium asset prices in an economy populated by in-
vestors with preferences of the type described in section 2. It may be helpful
to summarize those preferences here. Each investor chooses consumption
Ctand an allocation to the risky asset Stto maximize
(13)subject to the standard budget constraint, where
Xt+ 1 =StRt+ 1 −StRf,t, (14)E(,,),ρ
γργ
t t
tt
ttt
tC
bvXSz1
01
1
0 1−
− +
+
=∞−+
∑ C
γPROSPECT THEORY AND ASSET PRICES 239(^16) There is another sense in which integration of sequential outcomes is an implausible way
to implement prospect theory in a multiperiod context. If investors did integrate many years of
stock market gains and losses, they would essentially be valuing absolute levels of wealth, and
not the changesin wealth that are so important to prospect theory. We thank J. B. Heaton for
this observation.