On using text analytics for event studies

Event studies seek convincing evidence connecting behavior with a known or conjectured event. Originating in finance, event studies have spread to and become established in many other fields, including law. Multiple regression modeling, the standard methodology for doing event studies, may not be ideally suited for event studies based on data derived from bodies of text, since the required distributional assumptions may be problematic. This paper reports on an exploratory study that uses text analytic methods to discern events. The study draws upon and extends a previously published study of the speeches made by the CEO of a tobacco company in the years surrounding the tobacco settlement in 1998. Examining SEC filings from three major tobacco companies, the study reported in this paper finds cause for optimism in the use of text analytics to discern and investigate events of interest in law.