Can Textual Analysis of Corporate Filings Predict Business Related Risks?

Abstract

This article presents an overview of the textual analysis of companies’ regulatory filings. More specifically, it provides a discussion on readability indices, and new developments in this area of research. Next, the article discusses six different risk sentiment metrics, one by Feng Li (2006) that is based on the six risk-related words, and the other five risk sentiment metrics, financial risk, litigation risk, tax risk, idiosyncratic risk, and systematics risk, based on the risk factors disclosed in Item 1A of 10K (annual report filed with the US Security Exchange Commission). Using several examples, the article contrasts these risk sentiments for companies that have been involved in financial fraud and the companies that have not been involved in financial fraud. Finally, the article introduces cosine similarity metric that determines how different two documents are by considering the two documents as multi-dimensional vectors where each word represents a dimension. Several examples are used to contrast the pattern of changes over time in cosine similarity metric for companies involved in fraudulent financial reporting and not involved in such activities.

Keywords: Textual Analysis, Business Risk, Risk Sentiments, Risk Assessment, Cosiine Similarity

JEL Classification: G, M