onexcess S&P 500returns for individual stocks.^3 That is, if a sentiment reversal with
event datedhas returnrover the holding periodP, it has excess S&P 500 returnrr^0
wherer^0 is the return of the S&P 500 index fromduntil the expiration ofP.
The forward 1-year to 3-year holding period winsorized excess S&P 500 returns for
each of the 24 universes are recorded in Figure 9.2. Focusing on the forward 1-year
return curve, there is clearly a piecewise-monotonic relationship between the extension
of the trailing net sentiment measurement periods and forward 1-year performance. In
particular, the curve peaks at a measurement period of 17 months. On the other hand,
the 2-year and 3-year holds are not particularly interesting. Not only do the returns
taper off significantly (think in terms of degrading annualized returns), but there is no
attractive pattern such as the glaring monotonicity in the 1-year hold case. Our attention
shall thus be directed towards the forward 1-year returns from the universeU 17.
When implementing a strategy based on an event study, it is important to know the
frequency with which the events occur. Perhaps, for example, all of the opportunities
occur at market bottoms, and we would be left with nothing to buy in more bullish
scenarios. The annual event distribution of sentiment reversals fromU 17 is pictured in
Figure 9.3.
It appears there are more sentiment reversals in bull markets, as can be seen in the
sharp spike at 2003. Of course, this is to be expected. When investor sentiment is turning
positive relative to the market as a whole, one would anticipate the net news sentiment
around individual companies to gradually move in the same direction. However, there
238 News and abnormal returns
Figure 9.2.Forward 1-year to 3-year holding period winsorized excess S&P 500 returns for
sentiment reversals as a function of trailing net sentiment measurement periods.
(^3) It might be objected that the S&P 500 is not an accurate proxy of market performance. The equally-weighted or value-
weighted CRSP market indices are often used instead, having the advantage of being non-managed. On the other hand, the
S&P 500 is probably the most commonly used benchmark in the money management industry. Therefore, using such an index
has the virtue of simulating the real-world experience of a money manager, which is our primary interest here.