More broadly, the extended version of the model with contrarians fits
with the observation that there are a variety of professional money-
management “styles,” each of which emphasizes a different subset of public
information. Such heterogeneity cannot be understood in the context of the
standard rational model, where there is only one “correct” style, that
which processes allavailable information in an optimal fashion. But it is a
natural feature of our bounded-rationality framework, which allows multi-
ple styles to coexist and earn similar profits.
4 .Comparison to Related WorkAs noted in the introduction, this work shares the same goal as recent work
by BSV (1997) and DHS (1997)—that is, to construct a plausible model that
delivers a unified account of asset-price continuations and reversals. How-
ever, the approach taken here is quite different. Both BSV and DHS use rep-
resentative agent models, while our results are driven by the externalities that
arise when heterogeneous traders interact with one another.^27 Consequently,
many of the auxiliary empirical implications of our model are distinct.
First, it is impossible for a representative agent model to make predictions
linking trading horizons to the temporal pattern of autocorrelations, as we
do in section 3.D. Second, neither the BSV nor the DHS model would seem
to be able to easily generate our prediction that both continuations and rever-
sals are more pronounced in stocks with thinner analyst coverage (sections
3.A and 3.B). A further difference with BSV is that our model allows for a
differential impulse response to public and private shocks (section 3.C),
while theirs only considers public news.
In its focus on the interaction of different types of traders—including
those who behave in a trend-chasing fashion—this work is closer to earlier
models of positive-feedback trading by DeLong et al. (1990) and Cutler,
Poterba, and Summers (1990). However, there are significant differences
with this work as well. For example, in DeLong et al., the positive-feedback
traders are extremely irrational, and get badly exploited by a group of ra-
tional frontrunners.^28 In our model, the momentum traders are very nearly
A UNIFIED THEORY OF UNDERREACTION 529(^27) BSV develop a regime-switching learning model, where investors wind up oscillating be-
tween two states: one where they think that earnings shocks are excessively transitory; and
one where they think that earnings shocks are excessively persistent. DHS emphasize the idea
that investors are likely to be overconfident in the precision of their private information, and
that this overconfidence will vary over time as they learn about the accuracy of their past pre-
dictions.
(^28) In Cutler, Poterba, and Summers (1990), positive-feedback traders canmake money, as
there is background underreaction, like in our model. However, since the feedback behavior is
assumed, rather than derived, their model does not yield many of the predictions discussed in
section 3.