5.3.3 Strategy and empirical results
As part of constructing a set of industry sentiment indexes, I map each company
belonging to the S&P 500 to their corresponding industry:^5 (1) oil and gas, (2) basic
materials, (3) industrials, (4) consumer goods, (5) health care, (6) consumer services,
(7) telecommunications, (8) utilities, (9) financials, and (10) technology. Each of these
indexes are mapped to an iShares industry exchange traded fund (ETF) for testing
purposes.
The indexes are constructed based on a 90-day trailing window (P¼90), which
ensures similar news flow characteristics in each window. In order to map the calculated
sentiment ratio into a sentiment index value, I consider the empirical distribution over
the period March 2005 through December 2009. Finally, at each month end I decide
which industries to hold long and short in the following 1-month period depending on
anindustry sentiment delta rank(see Definition 5.2 for a definition of sentiment delta).
In Figure 5.3, I show the cumulative log return spread between the top-5 and
bottom-5 industries according to a sentiment ranking. As can be observed, theout-
How news events impact market sentiment 139Figure 5.3.Cumulative industry return spreads covering the out-of-sample period May 2005
through December 2009 for the top-5- and bottom-5-ranked industries. The sentiment indexes
have been constructed based on a 90-day trailing window, and a monthly industry rank was made
based on the monthly index delta.
(^5) Industry mapping has been based on the ICB industry classification.