The next round of corporate results will soon be with us highlighting yet more progress by many, affecting stock market prices. However, not all investors have the time, the inclination or the expertise to study bulky corporate reports, leaving this to so-called experts to deduce how a company is performing.
This is short-sighted, as the lifeblood of financial markets is relevant information published in annual reports, press releases, web pages, analyst’s research and financial media. The volume of data can range from sparse to vast, and over the past few years, this reporting has expanded rapidly in response to greater business complexity, environmental, social, and governance compliance, evolving national and international regulations and social media.
Translating qualitative information contained in financial reporting into quantitative outputs such as investment recommendations and earnings forecasts pose a challenge to those focused on hard numbers. However, detecting potential fraud from narrative information, rather than pure data analysis can now be undertaken by the use of techniques from computer science and linguistics.
The basic approach involves developing computer algorithms that harvest texts from various sources, which can be used in applications to try to predict future performance or highlight red flags regarding a firm’s current activities.
The key is that financial narratives often contain material information above that contained in current stock prices and hard accounting data, especially if news stories highlight negative information about a company’s stated earnings, above analysts’ forecasts and historical accounting data. Sometimes commentaries on specialist media platforms like Seeking Alpha are also useful for predicting future stock returns and earnings surprises.
Key research has focused on whether studying financial narratives help identify deceptive or fraudulent behavior by company management, especially when discussing future results.
Sometimes it is not the written reports but the choice of words used during investor conference calls that can help to identify violations of accounting rules and other unacceptable reporting practices.
Such examples of deceptive management practices might be the more frequent use of references to general knowledge, as opposed to specific company information, like “investors well know, others know well, you know,” as well as extremely positive language such as “fantastic, great, definitely” when describing forecasts.
Some might argue that the type of language used is often a reflection of the cultural setting of the organization and the use of such words should not be taken as potential fraud, and that a more scientific approach is required.
In the US, the Securities and Exchange Commission’s Electronic Data Gathering, Analysis and Retrieval is one such tool. It is designed to automatically download and analyze large company financial data.
Obtaining such information in a timely and standardized format was an important objective of the SEC, which now uses linguistic tools to screen financial statements to help identify potential fraud cases, especially from qualitative disclosures. Other major regulators still have some way to go to follow the SEC, especially if submitted reports are published as PDF documents, making automated access to text problematic.
This has not deterred researchers from advancing solutions through web-based software tools to break down PDF reports and highlight key linguistic features. It takes time and effort, and sometimes the bigger and better known the corporate brand name, the less effort is made to carry out such linguistic detections, with investors hoping for the best and leaving such forensic auditing to specialists.
In the meantime, the game continues to be splitting the difference between what is presented to the public by companies and what is left unsaid.
• Dr. Mohamed Ramady is a former senior banker and Professor of Finance and Economics, King Fahd University of Petroleum and Minerals, Dhahran.
