The Wiley Finance Series : Handbook of News Analytics in Finance

we introduce the notion of anentry intervalas a certain time period after the event date
such that, using any trading day within the interval as an entry point, there are still good
forward excess returns. For example, if we were willing to settle with forward 1-year
excess returns approximately within the range 14%–16%, sentiment reversals could be
bought up to four months post event date (see Figure 9.4).
It remains to be seen whether such a concept even improves portfolio returns. In any
case, from a pragmatic point of view, we should be familiar with the results of running
portfolios using the smallest possible amount of cash. The notion of an entry interval
likewise addresses practical problems such as trading friction. If a largish position size in
a name has to be taken, multiple days might be required to implement the trade order,
and an entry interval could help estimate the effect on performance. The same logic
applies to exit points.
Given the event distribution of sentiment reversals inU 17 and the desire to capture as
much portfolio performance as possible, 10-stock portfolios are simulated and annual-
ized returns are computed over the timeframe 2000–2008.^4 This, of course, would be a
bad time to begin and end a portfolio (the S&P 500 had an annualized return of
5.65%).

Remark 4.When returns are ‘‘annualized’’ we simply take a geometric average. For
example, ifr 1 ;:::;rnare different calendar year returns for a portfolio, the annualized
return is the quantity

Yn

i¼ 1

ðriþ 1 Þ

!

1 =n

 1 :

Sentiment reversals as buy signals 241

Figure 9.4.Forward 1-year winsorized excess S&P 500 returns for sentiment reversals as a
function of trading days post event date.

(^4) There are 67 starting points pre 2000 that yield different, fully invested 10-stock portfolio constructions over the timeframe
2000–2008.