the future dividend stream at a lower rate, giving stock prices an extra jolt
upward. A similar story holds for a negative dividend innovation. It gener-
ates a low stock return, depleting prior gains or increasing prior losses. The
investor is more risk averse than before, and the increase in risk aversion
pushes prices still lower. The effect of all this is to make returns substantially
more volatile than dividend growth.
The same mechanism also produces long horizon predictability. Put simply,
since the investor’s risk aversion varies over time depending on his investment
performance, expected returns on the risky asset also vary. To understand this
in more detail, suppose once again that there is a positive shock to dividends.
This generates a high stock return, which lowers the investor’s risk aversion
and pushes the stock price still higher, leading to a higher price/dividend
ratio. Since the investor is less risk averse, subsequent stock returns will be
lower on average. Price/dividend ratios are therefore inversely related to fu-
ture returns, in exactly the way that has been documented by numerous stud-
ies, including Campbell and Shiller (1988) and Fama and French (1988b).
If changing loss aversion can indeed generate volatile stock prices, then
we may also be able to generate a substantial equity premium. On average,
the investor is loss averse, and fears the frequent drops in the stock market.
He may therefore charge a high premium in return for holding the risky
asset. Earlier research provides hope that this will be the case: Benartzi and
Thaler (1995) analyze one-period portfolio choice for loss-averse investors—
a partial equilibrium analysis where the stock market’s high historical mean
and volatility are exogenous—and find that these investors are unwilling to
invest much of their wealth in stocks, even in the face of the large historical
premium. This suggests that loss aversion may be a useful ingredient for
equilibrium models trying to understand the equity premium.
Finally, our framework also generates stock returns that are only weakly
correlated with consumption, as in the data.^22 To understand this, note that
in our model, stock returns are made up of two components: one due to
news about dividends, and the other to a change in risk aversion caused by
movements in the stock market. Both components are ultimately driven by
shocks to dividends, and so in our model, the correlation between returns
and consumption is very similar to the correlation between dividends and
consumption—a low number. This result distinguishes our approach from
consumption-based habit formation models of the stock market such as
Campbell and Cochrane (1999). In those models, changes in risk aversion
are caused by changes in consumption levels. This makes it inevitable that
returns will be significantly correlated with consumption shocks, in con-
trast to what we find in the data.
PROSPECT THEORY AND ASSET PRICES 245(^22) This is a feature that is unique to Economy II, which allows for a meaningful distinction
between consumption and dividends.