good reasons to hypothesize that investors can take a longer time to process new
information that has extreme sentiment.
Behavioral response to information is what is happening here. Issues of herding,
cross-validation, overreaction and underreaction, cognitive dissonance, and attention
all come into play.
It is instructive to attempt realistic portfolio construction using only RNSE event data
for extreme sentiment days. Of course, a serious portfolio manager would incorporate
RNSE event data into her arsenal of other investment ideas and information sources,
but it is instructive to see if even simple portfolio construction techniques using only
news data can perform well.
6.7.2 Portfolio construction
We chose to consider an extreme sentiment day for a specific security in the S&P 1500 to
be one on which there were at least four novel news items on the RNSE feed prior to
3 :30 pmnyctime, which were (on average) extremely positive or negative (top 5% of
average daily positivity or negativity). We made distributional observations between
2003 and 2005 regarding which sectors had the most predictable response to extreme
RNSE events and thus restricted our investable universe to securities in the technology,
industrials, health care, financials, or basic materials sectors. We assumed daily portfolio
re-balancing, entering, and exiting positions at market close.
Positions were held for 20 days, subject to a stop loss rule set at 5% and a profit take
rule set at 20%, and constrained the portfolio to be very loosely dollar- and beta-neutral,
while allowing it to opportunistically take advantage of particularly positive or negative
news regimes.
164 News and abnormal returns
Figure 6.14.RNSE Extreme Sentiment Day portfolio cumulative return from January 2006
through November 2009, compared with the S&P 500.