Financial Analytics -- Quantifying Risk Factors in Financial Annual Reports
Dr Ke-Wei Huang is an Assistant Professor with the Department of Information Systems, at the School of Computing, National University of Singapore. Prior to joining NUS, he got his BS in electrical engineering and MBA (ranked 1st in class) in finance from National Taiwan University, a MS degree (with honors) in Information Systems from Stern School of Business NYU, and a Ph.D. degree in Information Systems from Stern School of Business NYU.
His specialization of research includes the economics of pricing information goods, e-commerce personalization strategies, analytical modeling in IS, and text mining applications in accounting and finance. His works have been published or accepted for publication in Information Systems Research, Quantitative Marketing and Economics, Journal of Economics & Management Strategy, ACM Transactions on MIS, International Conference on Information Systems (ICIS), Workshop on Information Systems and Economics (WISE), Conference on Information Systems and Technology (CIST), INFORMS, and Marketing Science.
This project applies text classification algorithms and econometrics models to investigate the correlation between risk factors reported in corporate annual reports and firm performance of publicly listed USA companies. Corporate annual reports are regarded as one of the most important sources of information about a company. After 2006, publicly listed companies in USA are required to report risk factors in their annual reports (SEC Form 10K) in a separate section (Item 1A). This regulation change allows us to employ text mining techniques to quantify risk factors. Specifically, text classification is used to identify different types of risk factors reported by each company in each year. In other words, the output of text classification will tell us the Xth risk factor in each annual report belongs to type Y.
For example, the first three risk factors in the 2010 annual report of Oracle are:
1. "Economic, political and market conditions, including the recent recession and global economic crisis, can adversely affect our business, results of operations and financial condition, including our revenue growth and profitability, which in turn could adversely affect our stock price."
2. "We may fail to achieve our financial forecasts due to inaccurate sales forecasts or other factors."
3. "We may not achieve our financial forecasts with respect to our acquisition of Sun or our entrance into a new hardware systems business, or the achievement of such forecasts may take longer than expected. Our profitability could decline if we do not manage the risks associated with our acquisition and integration of Sun."
In order to classify this kind of textual information, we have developed a multi-label text classification algorithm tailored for this task. Our algorithm is shown to outperform existing multi-label algorithms (Huang and Li 2011).