Analysing Performance of Company through Annual reports using Text Analytics

Investors and financial analysts mainly rely on Annual Reports to decide on the company's current performance and make calculated plans on their investments. Annual reports provide an overview of firms past performance, how they make use of the external environment for their growth needs, their strategies for growth and their expectations for the future. The financial information presented in the financial statements is a combination of textual and numerical information. The numerical information includes the balance sheet, income statement and cash flow statement, which constitutes about less than 20%. The remaining 80% are textual in nature, which includes footnotes, letters from executive leaderships, strategies, leadership team, shareholder details, and various reports including directors report, sustainability report, corporate governance report. Until recent years, financial analysts use conventional methods like ratio and trend analysis primarily based on numerical information for analyzing the performance of the organization. So far textual data in annual reports were considered as complex and categorized as inaccessible information for the huge number of novice investors. This paper tries to investigate how the textual component of Annual report can be used deduct meaningful information about company's performance by employing text analytics. Text analytics uses statistical pattern learning for understanding the trends and patterns, which in turn provides information of high accuracy. This is attained by working on the unstructured information spread across different section in details, extract meaningful data contained in the text, convert it to numerical, and then use it along with relevant data mining algorithms. The research tries to answer 2 questions; 1. How emotions in the annual reports as a whole determines the current performance of the company, 2. Does these emotions have an impact on the expected returns on the forthcoming year. We conducted a descriptive research, selecting 12 IT Firms functioning in India. The research considered annual reports for a period of 3 years starting from 2015–16 to 2017–18, thereby creating 36 samples. We performed mining on these reports using Bing Lexicon for sentiment analysis as well as NRC lexicon for emotional analysis. The structured data analysis was collected from Bloomberg. Risk and Return analysis were done on the Market performance of the subsequent years. We employed statistical techniques on these data to create multivariate models. The results suggest that companies' current performance plays an important role in the emotions in the annual reports. We also noticed an established relation between the emotions in annual report and future performance of the firm. We concluded that text analytics can be efficiently used on the unstructured data present in annual reports and the results can be effectively used by the stakeholders using these data to make their decisions based on companies' performance.