6.5 REFINING FILTERS USING INTERACTIVE
EXPLORATORY DATA ANALYSIS AND VISUALIZATIONThere are many ways to slice and dice event studies, and for advanced content-based
filters the ability to drill down to individual news events is desirable. We developed an
interactive exploratory data analysis and drill-down system, the Event Study Explorer,
using the TIBCO Spotfire visualization tool. The sliders and selection boxes seen in
Figure 6.12 illustrate this capability.
This approach to exploratory data analysis (EDA) was first suggested by John Tukey
(Tukey, 1977), and refined by Tufte inVisual Display of Quantitative Information(Tufte,
2001). These ideas were greatly advanced as computational tools by Ben Schneiderman’s
Human Computer Interface Lab (Schneiderman and Plaisant, 2009).
The Event Study Explorer allows great flexibility in filter selection parameters, study
period, sector, capitalization, and pre-event return. It provides the ability to drill down
to news content as the basis for further natural language processing (NLP) or machine
learning (ML) filtering.
The Event Study Explorer allows the researcher to consider the subsequent cumula-
tive return for specific subsets of events. Events are keyed by date and RIC (security
158 News and abnormal returns
(a)(b)Figure 6.8.Simple unscreened news event studies, showing pre-event returns as well in (a), post-
event returns in (b) (note vertical scale change). The ‘‘forecast’’ subsequent excess return is the
return gained after a specific time period chosen prior to initiating the study, either 1 week or
1 month. This is similar to Tetlock, Saar-Tsechansky, and Macskassy’s (2008, fig. 1) result for
1984–2004, with larger pre-event returns than what one might expect for 2003–2008, which is fully
in the web era of faster moving information. It also suggests the target for advanced news signal
methods using analytics based on novelty and news metadata.