Sweet-Talking CEOs Are Starting to Outsmart the Robot Analysts
(Bloomberg) -- A new era of robots reading financial statements and machines monitoring earnings calls means analysts no longer need to sweat a company’s every last word. For corporate leaders, it’s quite the opposite.
Company executives have started to adapt their statements and even their delivery to cater to the algorithms parsing text and speech for trading signals, researchers at Georgia State University and Columbia University have found.
Managers are emphasizing positivity and avoiding words or phrases known to be perceived by machines as negative. So it’s out with things like “claimants” and “cease” and in with the likes of “innovator” and “improving.”
“Increasing AI readership motivates firms to prepare filings that are more friendly to machine parsing and processing,” a team of academics argues in a new paper. “Technological progress and the sheer volume of disclosure make the trend inevitable.”
Tailoring corporate communications for the investing audience is as old as Wall Street, of course, and the practice of doing it for a human audience is well documented. But the research claims to be the first to tackle the “feedback effect” of companies adjusting how they talk for listening machines.
Researchers Sean Cao, Baozhong Yang, and Alan L. Zhang at Georgia State and Wei Jiang at Columbia believe the trend started in 2011, when groundbreaking work for measuring sentiment in financial contexts was published. Companies -- in particular those whose filings see a high number of machine downloads -- began reducing the use of certain words at that point.
Natural language processing consists of converting text or speech into numbers so it can be fed into a computer program which calculates statistics that aren’t obvious at first glance. It may detect correlations between word choices and a company’s profitability, for example.
For the alert CEO, however, the predictability of the programs offers a potential way to ensure company statements receive a warm reception.
“Because such rules are transparent, observable, or reverse-engineerable to at least some degree, agents who are impacted by the decisions have the incentive to manipulate the inputs to machine learning in order to game at a more desirable outcome,” the paper says.
©2020 Bloomberg L.P.