light of this concern, we use as a second—and hopefully purer—proxy for in-
formation flow a stock’s residual analyst coverage, after controlling for size.^20
The basic findings from Hong, Lim, and Stein are reproduced here as
tables 14.1–14.3, and figure 14.5. These findings can be briefly summarized
as follows. With respect to size, we find that once one moves past the very
smallest-capitalization stocks (where price discreteness and/or very thin
market-making capacity are issues) the profitability of Jegadeesh-Titman-
style six-month momentum strategies declines sharply with market cap.
With respect to residual analyst coverage, not only are momentum strate-
gies substantially more profitable at a horizon of six months in low-analyst-
coverage stocks, they are also profitable for longer—there is pronounced
positive correlation of returns for up to about two years in these stocks, as
opposed to less than one year in high-coverage stocks. Size and residual
coverage also interact in an interesting and economically plausible fashion:
the marginal impact of analyst coverage is most pronounced in smaller
stocks, which have fewer analysts to begin with. While it may be possible
to come up with alternative interpretations, all these pieces of evidence
would seem to be strongly consistent with our emphasis in this work on
gradual information flow as the root cause of underreaction.
B. Linking Momentum to Overreaction in the Cross SectionThere is also a second, more subtle cross-sectional implication of our model
pertaining to the rate of information flow. In figure 14.3 we saw that not only
does slower information diffusion lead to higher short-run return correla-
tions, but by making stocks more attractive to momentum traders, it also (for
a wide range of parameter values) leads to more pronounced overshooting
and stronger reversals in the longer run. In other words, the same stocks that
we find in Hong, Lim, and Stein to be most “momentum-prone”—small
stocks with relatively few analysts—should also be most “reversal-prone.”
Although this prediction has not to our knowledge been subjected to de-
tailed investigation, it is broadly consistent with recent work which finds
that much of the long-horizon predictability that has been documented in
the stock market is attributable to smaller-cap companies.^21 As noted
above, there is the caveat that size may be proxying for a number of other
factors so, as in Hong, Lim, and Stein, it would be desirable to create a
sharper test, perhaps using analyst coverage or some other nonsize measure
of momentum-proneness.
522 HONG AND STEIN
(^20) Of course, analyst coverage is not an ideal proxy either, as it may be endogenously re-
lated to a number of other stock-specific factors besides size. So in various sensitivity tests, we
also control for the correlation between analyst coverage and share turnover, industry factors,
beta, and market-to-book.
(^21) Fama (1998) argues that this evidence is problematic for existing behavioral models, as
they do not clearly predict that overreaction should be concentrated in smaller stocks.