It is easy to embellish our model so that it also generates short-run un-
derreaction to public news. For example, one might argue that although the
news announcement itself (e.g., “earnings are up by 10 percent”) is public,
it requires some other, private, information (e.g., knowledge of the stochas-
tic process governing earnings) to convert this news into a judgement about
value. If this is true, the market’s response to public news involves the ag-
gregation of private signals, and our previous underreaction results con-
tinue to apply.
On the one hand, this sort of patch adds an element of descriptive real-
ism, given the large body of empirical evidence on postevent drift. But the
more interesting and subtle question is this: If we augment the model so as
to deliver short-run underreaction to public news, what does it have to say
about whether there is overreactionin the longer run to this same news? Is
the impulse response function hump-shaped as before, or do prices drift
gradually to the correct level without going too far?
Unlike with private news, the answer is now less clear. This is because the
inference problem for momentum traders is simplified. Recall from above
that with private news, a momentum trader never knows whether he is buy-
ing early or late in the cycle—that is, he cannot tell if a price increase is the re-
sult of recent news or of past rounds of momentum trade. But if momentum
526 HONG AND STEIN
Figure 14.5. Cumulative beta-adjusted returns in event time: Momentum profits for
low and high coverage stocks. We assign stocks to performance categories based on
six months’ prior beta-adjusted returns, and do an independent sort based on the
analyst coverage residuals. We then track the cumulative beta-adjusted momentum-
portfolio returns (P3-P1) on a month-by-month basis, out to thirty-six months for
low coverage (SUB1) and high coverage (SUB3) stocks.