Our specification of this additional source of utility captures two ideas
we think are important for understanding investor behavior. First, our in-
vestor is much more sensitive to reductions in his financial wealth than to
increases, a feature sometimes known as loss aversion. Second, how loss
averse the investor is, depends on his prior investment performance. After
prior gains, he becomes less loss averse: the prior gains will cushion any
subsequent loss, making it more bearable. Conversely, after a prior loss, he
becomes more loss averse: after being burned by the initial loss, he is more
sensitive to additional setbacks.
By extending the traditional asset pricing framework in this way, we find
that we are able to understand many of the hitherto perplexing features of ag-
gregate data. In particular, starting from an underlying consumption growth
process with lowvariance, our model generates stock returns with a high
mean, high volatility, significant predictability, and low correlation with con-
sumption growth, while maintaining a low and stable riskless interest rate.
In essence, our story is one of changing risk aversion. After a run-up in
stock prices, our agent is less risk averse because those gains will cushion
any subsequent loss. After a fall in stock prices, he becomes more wary of
further losses and hence more risk averse. This variation in risk aversion al-
lows returns in our model to be much more volatile than the underlying
dividends: an unusually good dividend raises prices, but this price increase
also makes the investor less risk averse, driving prices still higher. We also
generate predictability in returns much like that observed in the data: fol-
lowing a significant rise in prices, the investor is less risk averse, and subse-
quent returns are therefore on average lower.
The model also produces a substantial equity premium: the high volatility
of returns means that stocks often perform poorly, causing our loss-averse
investor considerable discomfort. As a result, a large premium is required to
convince him to hold stocks.
Our framework offers a distinct alternative to consumption-based mod-
els that attempt to understand the high mean, high volatility, and significant
predictability of equity returns. Campbell and Cochrane (1999) explain
these empirical features using an external habit level for consumption that
generates time-varying risk aversion as current consumption moves closer
to or farther from habit. Although our model is also based on changing risk
aversion, we generate it by introducing loss aversion over financial wealth
fluctuations and allowing the degree of loss aversion to be affected by prior
investment performance.
The differences between our framework and a consumption-based ap-
proach like Campbell and Cochrane (1999) are highlighted by the distinct
predictions of each. In the consumption-based model, a large component of
stock return volatility comes from changes in risk aversion that are ulti-
mately driven by consumption. It is therefore inevitable that stock returns
and consumption are significantly correlated, although this is not the case
PROSPECT THEORY AND ASSET PRICES 225