Development of disclosure interpretation theory and sentiment analysis techniques for market surveillance via corporate disclosures

Abstract

As a major communication medium between firm and investors, corporate disclosures play an important role in reducing information asymmetry between firms and investors. However, unlike numerical data in financial reports, textual statements in financial reports are difficult to audit and use because (1) their narrative nature hinders the utilization of numerical analysis techniques and tools, and (2) they can be ambiguous and incomplete due to corporate incentives. Recent years have seen major advances made in textual analytics and their widely successful applications in various business disciplines. This work utilizes text analytical and modeling techniques to shed light on the underlying information of narrative corporate disclosures. Specifically, we develop new theory and novel techniques that help aggregate market signals from a large number of financial reports so as to reduce information asymmetry between firms and other stakeholders. • Theoretical advancement. We examine relevant theoretical foundations of corporate reporting by studying several well-known theories and propose a Disclosure Interpretation Theory to guide our technical efforts in business analytics of textual disclosures by analyzing millions of corporate reports. • Technical Innovations. We devise several analytical models to aggregate and analyze the signals from forward-looking disclosures and their applications in market surveillance. • Empirical Validation. We conduct empirical studies on the informativeness of textual corporate disclosures at individual and market levels by examining the statistical and economic significance of the potential informativeness. The contributions of this thesis are threefold. Firstly, we extend the signaling theory from an information aggregation perspective. The textual analytics and sentiment analysis on large-scale corporate reports enable us to investigate the overall effects of forward-looking disclosures. Secondly, we contribute to existing literature by devising new models and algorithms to achieve tone classification and sentiment analysis of corporate reports. Thirdly, we obtain empirical results on informativeness of forward-looking disclosures at different levels, including individual and market level, by examining the statistical and economic significance of the potential informativeness.

Date of Award	2 Oct 2013
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	J Leon ZHAO (Supervisor)