It's Not Just Bluster: Here's A Surprising Way Executives Can Literally Talk Up Their Shares
Posted by Tyler Durden on October 13, 2017 11:20 pm
Tags: Ben Bernanke, Brown University, Business, Dow 30, Dow Jones Industrial Average, Fail, Harvard University, machine-learning algorithm, Market Sentiment, Share price, stock market, US Federal Reserve
Categories: Ben Bernanke Brown University Business Dow 30 Dow Jones Industrial Average Economy Fail Harvard University machine-learning algorithm Market Sentiment Share price stock market US Federal Reserve
As it turns out, it’s not what an executive says during a corporate earnings call that matters, it’s how much time they spend saying it.
Prattle, a market-research company that recently closed a $3 million seed round, claims to have discovered a surprising correlation between the length of time executives spend speaking during a quarterly earnings call, and the performance of a company’s stock over the short- and long term, MarketWatch reports.
The company’s co-founder and CEO Evan Schnidman, a former Brown University economics professor, helped establish the company in 2014 with the goal of parsing market-moving language in the statements and speeches of central bankers to try and categorically determine whether their remarks were hawkish or dovish.
But now Schnidman is expanding into covering quarterly earnings calls, a potentially lucrative line of business – that is, if Schnidman’s data proves to be as valuable as the company says it is.
Prattle’s fundamental assumption is that patterns of speech aren’t random, but rather loaded with subconscious meaning…
The underlying principle of Prattle’s machine-learning algorithm is that “people speak in patterns,” and that those patterns are not random. “Rather, linguistic patterns link specifically to the conscious and subconscious thoughts of the communicator,” Prattle wrote in a white paper. “These patterns manifest in the language of a corporate executive like a poker player’s tell.”
“Traditional sentiment analysis is deeply flawed — it tends to be ‘good buzzword minus bad equals 4,’ ” Schnidman told MarketWatch.
Many systems are unable to remove the bias of the creator and fail to take into account the full flow of sentences and paragraphs and how they are all related, he said.
The Prattle Equities Analytics product aims to produce analysis based on how the market has responded to prior communications.
…And by analyzing these metrics from past earnings calls and linking them to the stock’s performance that day as well as the following weeks. Through this, Prattle claims to have unlocked a valuable insight.
Prattle’s algorithm seeks to link a company executive’s historical speech patterns with the performance of a company’s stock, then it scores the lexicon, including individual words, phrases, sentences and so on, by their past impact on the stock’s price.
“By controlling for common fundamental factors like peer-company performance, Prattle scores represent the price movement that can be directly tied to the sentiment expressed in corporate communications,” according to the white paper.
While Prattles’ analysis has a pretty good track record for predicting short-term moves, it has proven to be much more effective as a leading indicator for a stock’s long-term performance.
A Prattle sentiment score is the expected 10-day cumulative abnormal return (CAR) resulting from the language of a company conference call.
To wit, Prattle’s analysis of Nike’s post-earnings call performance revealed that fluctuations in the stock’s Prattle score portended a drop in the company’s share price.
While the one-off scores can be helpful in predicting short-term stock directions, reading the Prattle scores in a time series can also serve as a leading indicator of company and stock performance.
For example, Nike Inc.’s first-quarter call after the Sept. 26 closing bell was scored 3.07, which is fairly bullish on an absolute basis, and was ranked in the 92nd percentile. The stock ran up as much as 3.2% in after-hours trade, after a big earnings beat and a very slight revenue miss.
In the next session, however, the stock dropped 1.9%, then fell another 2.3% through Tuesday to close at a four-month low. The Dow Jones Industrial Average climbed 2.2% to a record close over the same time span. Although the Prattle score was strongly positive, it was down from a 3.23 score after the fourth-quarter conference call on June 29 and a 3.35 score after the third-quarter call on March 21.
“The Nike signal demonstrated that despite weaker-than-expected revenue, the management team was optimistic about the outlook, likely because [earnings per share exceeded] estimates,” said Schnidman. “Nevertheless, the signal was not as positive as other recent Nike earnings calls, which helps explain why, after an initial pop, [the] stock price has declined a bit since the earnings call.”
Schnidman stumbled on the idea for Prattle when he was a graduate student at Harvard University during the financial crisis. Initially, he was experimenting with using automated analysis to study Fed communications while former Fed Chairman Ben Bernanke was frantically slashing interest rates to zero. After spending several years building the product, Prattle raised a $3.3 million I a seed round led by GCM Grosvenor. New Enterprise Associates, Correlation Ventures, Plug and Play Ventures, Neotribe Ventures, and a group of prominent Silicon Valley and Wall Street angels also signed on as investors, MarketWatch reported.
Prattle launched a beta version of its new Prattle Equities Analytics product in May, which measures call-related metrics including length for 3,000 companies publicly traded in the US.
With third-quarter earnings season having only just started, it may be too soon to expect executives to incorporate Prattle’s findings while strategizing for their earnings calls.
However, since market strategists and investors often have a difficult time differentiating between causation and correlation, will Prattle’s findings inspire desperate executives to ‘prattle’ on? At any rate, investors can now start ignoring metrics like EPS, sales, loss reserves and other trivial minutiae.