Modeling consumption and dividends separately also leads to higher re-
turn volatility in consumption-based models such as that of Campbell and
Cochrane (1999). However, the fact that the equity premium goes up is
unique to our model: as Campbell and Cochrane demonstrate, separating
consumption and dividends has no effect on the equity premium in a
consumption-based model. The reason is that in a consumption-based
model, the canonical measure of a stock’s risk is its covariance with con-
sumption growth. By separating consumption and dividends, the volatility
of stock returns goes up, but the correlation of stock returns with consump-
tion goes down, since dividends are poorly correlated with consumption.
Overall, the covariance with consumption remains unaffected, and the eq-
uity premium remains as hard to explain as before.
The investor in our model cares not only about consumption but also
about fluctuations in the value of his investments. An increase in dividend
volatility makes stocks more volatile, scaring the investor into charging a
higher equity premium. It is true that stocks are now less correlated with
consumption, but this does not matter in our model, since the investor
cares about fluctuations in the stock market per se, not merely about how
those fluctuations covary with consumption growth.
Table 7.4 confirms that stock returns are only weakly correlated with con-
sumption growth in our model. As discussed earlier, this prediction is unique
to our framework: since consumption-based models ascribe a significant
254 BARBERIS, HUANG, SANTOS
Figure 7.5. Price-dividend ratios in Economy II. The price-dividend ratios are plotted
against zt, which measures prior gains and losses: a low ztindicates prior gains. The
parameter b 0 controls how much the investor cares about financial wealth fluctuations.
We fix the parameter kat 3, bringing average loss aversion close to 2.25 in all cases.