Keith Black, PhD, CFA, CAIA, FDP, Managing Director of Content Strategy at CAIA Association
Do you want to be in the room where it happens, when “The Smartest Guys in the Room”* are plotting their fraud and trying to prevent the demise of Enron? Look around at how lucky we are to be alive right now, with the use of artificial intelligence, machine learning, and natural language processing. Without hearing a single word that was said, we can investigate the risk of corporate conversations through a look at the characteristics of emails sent within a firm.
CAIA Association and FDP Institute recently had a conversation with Professor Seoyoung Kim from Santa Clara University to discuss her paper with Sanjiv Das and Bhushan Kothari “Zero-Revelation RegTech: Detecting Risk through Corporate Emails.” This is the third in a webinar series where authors of papers that are required readings for the FDP exam explain their research.
Professor Kim warns that employees of US companies should have no expectation of privacy when using their corporate email, as many firms have enacted surveillance programs to remain in compliance with regulations and reduce corporate risk. Of course, no one can or should read everyone’s email, so a technology solution is needed to filter the emails and point out the characteristics of the most risky emails. Natural language processing (NLP) can filter emails to find topics and words that are characteristic of illegal or risky conversations. Some traders may mine alternative data sources and purchased email for signals of corporate risk or to estimate the revenue of ecommerce companies by adding together emailed purchase confirmations.
What patterns of email conversations denote risks to the firm? Professor Kim studied this question through an analysis of emails in the time leading up to the demise of Enron. Her team had access to over 100,000 emails sent by upper management of Enron in the two year period ending December 2001. The emails contained redactions and did not include emails sent by the disgraced CFO Andrew Fastow. The email data is available from Carnegie Mellon University’s Cognitive Agents that Learn and Observe (CALO) project, one of the labs that helped launch the technology behind Apple’s Siri NLP interface.
This paper is designed to investigate the link between signals included in the Enron sent email box and the subsequent return to Enron’s stock. They also compare the signals in the Enron emails to publicly available data such as from newspaper articles. After scrubbing the emails, the first task is to investigate the net sentiment of the emails, counting the portion of positive vs. negative words when matched to a sentiment dictionary. The net sentiment of emails was a significant predictor of Enron’s stock price. Predictably, as sentiment became more negative, the stock price declined. Not surprisingly, the change in the net sentiment of the internal emails was more valuable for predicting future stock prices than the publicly available news items.
An even stronger signal than news or sentiment came from the length of the emails, as emails contained fewer and fewer words as the scandal deepened. In fact the signal from tracking the number of characters in emails was so statistically significant that the previously significant news and email sentiment indicators became insignificant when included in regressions with email length. This is an important conclusion, as the corpus or text of emails can remain private while a RegTech surveillance system can focus on an easily calculated and non-invasive count of characters in the body of each email. The decline in email length leads stock price declines, so knowing that emails were getting shorter was a strong signal that shorting the stock would be profitable in the coming weeks.
What does talk less, smile more mean? In the context of regulatory technology, those who are talking less may be sporting a devious smile, trying to not put their risky or fraudulent plans into writing. When the emails start saying “see me” or “let’s talk,” regulators and compliance officers need to be worried about what is going on during these off the record conversations.
* “The Smartest Guys in the Room: The Amazing Rise and Scandalous Fall of Enron,” Bethany McLean and Peter Elkind, Penguin, 2004.