As news broke this week that TV anchor Bill O’Reilly would be leaving Fox News, investors observed a reaction in the stock market that few could make sense of: The stock of the company that owns O’Reilly Auto Parts suddenly rose, even while the rest of the market was falling.
On Wednesday, as reports of O’Reilly’s imminent departure were cheered by those who pushed for his ouster amid sexual harassment claims against the political commentator, O’Reilly Automotive gained as much as 3%.
O’Reilly Automotive (orly), a Fortune 500 company that sells car accessories and tools, has no relation to the host of Fox’s now-canceled O’Reilly Factor. But without any other news that day to explain the move in the retailer’s stock price, some investors had a different theory: Computerized algorithms that trade stocks based on Twitter and social media alone had picked up on a surge in posts about Bill O’Reilly, and interpreted it as a signal to buy O’Reilly stock.
This effect, of course, would be considered a glitch, illustrating the flaws in the nascent but increasingly popular field of so-called social sentiment-based investing. (For more on that trend, read my Fortune Magazine feature, “Investors Are Using Social Media to Make Money.“)
Quant trader Michael Harris summed up the incident in the headline of a blog post about the incident: “Dumb Social Media Sentiment Algos Target The Stock of O’Reilly Automotive.”
Given the way repeated mentions of “O’Reilly” might make O’Reilly Auto Parts’ familiar jingle stick in your head (“Oh, oh, oh, O’Reillyyy”), it’s not hard to see how computer algorithms might get mixed up. Even Google, which tracks trends in its online searches, suggests “Bill O’Reilly” as the top related query for those searching “O’Reilly Automotive.”
After all, it would hardly be the first time that a Twitter-based trading algorithm got tricked into buying the wrong stock. Quants love to tell the story, now legendary among sentiment traders, of an early experiment in social media trading gone awry: Trawling the Twitterverse, algorithms had picked up on a flurry of tweets about the Devil Wears Prada actress Anne Hathaway. Misinterpreting the chatter, computers using the algorithm bought stock in Berkshire Hathaway (brk-a), Warren Buffett’s bluechip company.
Another time, a hedge fund’s algorithm correctly identified the company, but confused the sentiment, when it bought up shares in Lululemon (lulu) based on tweets about its too-sheer pants. Those pants became a disaster for the company that ended up tanking its stock price, burning the traders who bought shares.
Still, it’s not quite clear if, or to what extent, the Bill O’Reilly tweets triggered the buying of O’Reilly Automotive stock. After all, some social media-based algorithms distinguished between the auto parts company and irrelevant chatter and therefore didn’t fall for the ruse. “It probably did happen to some, but our scores were unaffected,” says Joe Gits, CEO of Social Market Analytics, which measures sentiment on stocks, having found that swings in the positivity or negativity in online postings can predict whether a stock price will rise or fall.
O’Reilly Automotive stock eventually gave back some of its gains for the day, but still finished Wednesday up 1.1% at market close. Meanwhile, shares of 21st Century Fox (foxa), which owns Fox News, were barely impacted, falling 0.9%, more in line with the broader market. O’Reilly was up about 4.5% for the five days through Friday, while the market overall was slightly down.
Not everyone is buying the more mundane version of what happened. “There is a small probability that the two events are unrelated and the stock of O’Reilly Automotive (ORLY) rallied in the past two days because of normal market activity,” Harris wrote in his widely shared blog post, “But this happening while the name O’Reilly was trending in Twitter and social media raises some questions about what is really happening in Wall Street. What are the risks from dumb algos taking over Wall Street?”