Textual analysis of corporate disclosures: prediction model for stock price movement

Abstract

Accounting is fundamentally the measurement, processing, and communication of financial and non-financial information about different economic entities. An organization uses both quantitative and qualitative information to communicate with its capital market stakeholders. Accountants track and aggregate various economic transactions into financial statements: Income Statement, Balance Sheet, and Cash-Flow Statement, which can be analyzed to evaluate the firm's performance. Apart from the numerical information on these financial statements, the corporate disclosures comprise a large amount of unstructured textual information about the firm's performance and prospects. Thus, quantitative financial information in isolation provides an incomplete notion about the firm's economic performance. In contrast, qualitative financial disclosures such as annual and quarterly reports, press releases, Management Discussion, and Analysis play a critical role in helping stakeholders to process the quantitative information given in financial reporting. Due to this complementary nature, many studies have shown that the textual tone of qualitative corporate disclosure is informative to stakeholders and significantly influences returns in the capital market. This research work aims to understand the applications, determinants, and measurements of the textual tone of corporate disclosures. The study helps analyze the qualitative unstructured data and examines whether any correlation exists between stock price movement and sentiment. Further, different machine learning algorithms are used to develop models to predict the stock price movement based on the sentimental analysis of annual reports of Nifty 50 companies.