Feature Stories

10.07.2017

FinTech Wave Revolutionizes Financial World

FinTech Wave Revolutionizes Financial World

You have heard the word “FinTech” on various occasions: TV, social media, or even friends’ parties. Is it just a buzzword floating around nowadays? The answer is likely no. More and more people are starting to believe that FinTech is a new wave of digital revolution that could fundamentally change the financial industry and, therefore, our lives in many aspects.

Already, peer-to-peer (P2P) lending websites provide a platform that efficiently matches borrowers with investors from anywhere in the world, a new financing model that has shortened the loan approval process to minutes compared to weeks or months at traditional banks. It also opens up the opportunity for anyone wanting to become an entrepreneur, by lowering financing barriers. According to Morgan Stanley, online loan volume in the US market is expected to reach US$120 billion in 2020, up from US$20 billion in 2015. In investment management, many big companies such as BlackRock Inc. and Vanguard Group Inc. are using computer algorithms called “robo-advisers” to automatically adjust portfolios, according to a customer’s risk preferences. Some investment institutions have used artificial intelligence methods to automate trading decisions, or have even started to make algorithms self-learning. In capital markets, blockchain, the freely available database that underpins the digital currency bitcoin, has prompted much debate as to whether it will replace existing methods of transmitting assets and currencies. Blockchain also has the potential to simplify the way securities are traded, settled and recorded. All these phenomena fall under the umbrella of what we called “FinTech”.

Among others, one important promise of FinTech is that there will be greater reliance on algorithmically-determined financial decisions in areas such as loan, insurance and stock picking. The advancement of artificial intelligence methods has been the propeller facilitating the transition in such a direction. Although humans have the advantage in intuitive and creative thinking, machines are better at weaving through the data and finding hidden connections among different variables. Given the exponential growth in the size of data, the computational advantage of machines begins to outweigh their weakness in sense-making, enabling them to play more roles in business decision-making. Below, two examples are presented to show the power of computational methods in extracting new information from traditional financial documents.

Uncover new info from financial documents

In a collaborative research project by Allen Huang and Amy Zang from the Department of Accounting at HKUST, we use the naïve Bayes machine learning approach to address the challenge of extracting information from a large volume of textual data in a financial analyst report. A typical analyst report contains both quantitative summaries including stock recommendation, earnings forecasts, and target price, and a detailed, mostly textual analysis of the company. Our research found that the textual discussions in analyst reports provide information to investors beyond that contained in those quantitative summaries. We achieved the information extraction by using the naïve Bayes algorithm that can quantify analysts’ sentiment about the covering firms from their written text.

The overall implication here is that a machine can replace a human in processing large amounts of text in a much more efficient way. This information extraction procedure also helps us understand more about the interplay between investors and various types of information. Interestingly, we find that investors react more strongly to negative than to positive text, and that analyst report text is more useful when it places more emphasis on non-financial topics, is written more assertively and concisely, and when the perceived validity of other information signals in the same report is low. Finally, analyst report text is shown to have predictive value for future earnings growth in the subsequent five years. As the first large-sample evidence on the information content of analyst report text, this research study shows good potential in applying computational methods in the domain of financial documents. Thanks to the help of computer-based information processing techniques, the information in the text is no longer hidden or hard to extract.

As an information intermediary, financial analysts interpret information that has been already in the public domain, or discover new information that is not readily available. In a separate study, we try to quantify analysts’ two information roles by comparing and contrasting the thematic content of analyst reports to that of corporate disclosures. To extract the “topics” of the two types of financial documents, we use an advanced topic-modeling methodology from machine learning and natural language processing research called Latent Dirichlet Allocation (LDA).

This methodology allows us to construct novel measures of the thematic content extracted from analyst reports and conference calls, and to identify analysts’ information discovery and interpretation roles. We show that analysts do play the two roles by discussing exclusive topics beyond those from conference and interpreting topics from conference calls. In addition, we find that investors place a greater value on the analyst information discovery role when managers face greater incentives to withhold value-relevant information. Analyst interpretation is particularly valuable when the processing costs of conference call information increase.

Data beyond human’s ability to process

One common feature of the above two research studies is that computer algorithms are used to extract and quantify some otherwise fuzzy concepts: analyst sentiment in the first study, and analyst information discovery and interpretation effort in the second one. The computer achieves it by aggregating a huge amount of data which is surely beyond any human’s ability to process. Even though humans can understand intuition through very limited observations, it is hard for them to transfer the intuition or knowledge to other people. The computational limitation and the qualitative nature of the human knowledge are the underlying reasons why computers will eventually outperform humans in more and more settings.

FinTech does not come as a free lunch, however. Algorithm-based decisions are not immune to anomalies and manipulations. On 6 May, 2010, the Dow Jones Industrial Average dropped 998.5 points (about 9%), mostly within minutes. This sudden market crash was later attributed to the algorithm trading systems being manipulated by a trader. Technology evolution in the financial industry is particularly challenging because it might affect the financial system’s safety and soundness as a whole. Many regulation issues arise with the new financial systems such as P2P lending and bitcoin. We are certainly in the early stages of learning how to protect consumers and the financial system without stifling innovation.

Assistant Professor Zheng Rong, Department of Information Systems, Business Statistics & Operations Management, HKUST Business School

References

Huang, A.H., Zang, A.Y., and Zheng, R., "Evidence on the Information Content of Text in Analyst Reports," The Accounting Review, 89, 6, 2014, 2151-2180

Huang, A.H., Lehavy, R., Zang, A.Y., and Zheng, R., "Analyst Information Discovery and Information Interpretation Roles: A Topic Modeling Approach", Management Science, forthcoming