Tweets from Donald Trump about “out of control costs” for a new 747 Air Force One being built by Boeing and Lockheed Martin’s F-35 programme sent both stocks rapidly lower in December. The president-elect then started January with a tweeted threat to General Motors about one of its cars: “Make in USA or pay big border tax!”
Executives, protectors of corporate reputation, investors and traders are all having to get used to a politician able to bypass the traditional setpieces and media gatekeepers, to share thoughts on specific stocks, sectors and policy plans.
Indeed, as @realDonaldTrump prepares to take over as US president, his penchant for uncensored tweets has become a hot topic in financial circles.
“Because he uniquely uses this medium to speak directly for the first time to everyone at once and the impact of his voice is so large as the president, it obviously offers unique opportunities,” says Ian Rosen, chief executive of StockTwits, a social media platform to share ideas between investors. “We are getting a tremendous amount of discussion about the effect of this.”
A start-up called Trigger, which sends customised alerts to mobile phones within 30 seconds of Trump tweets already has thousands of users, according to founder Rachel Mayer, an alum of JPMorgan’s electronic trading desk for rates and forex.
Mining social media for market moving sentiment is not new, nor is focusing automated tracking tools on the chatter of famously successful investors, such as George Soros and Carl Icahn.
The idea of using Twitter as a signal for market moves had its coming of age moment in 2013 — in both a good and bad way. Known as the “Twitter crash,” what turned out to be a false tweet from the Associated Press about an explosion at the White House that had injured President Barack Obama sent the Dow Jones Industrial Average down 150 points in two minutes, affirming that Twitter could move markets but also raising questions about the reliability of its signals.
Still, over the years, professional investors, primarily hedge and quant funds, have developed sophisticated algorithms and other procedures to track social media, hoping for the coveted “edge” on the rest of the pack, says Eric Noll, chief executive officer of Convergex.
“It grew out of the early reading of Reuters and Dow Jones tapes and being able to algorithmically and electronically monitor what is going on in the world in order to stay in front of the news,” Mr Noll says.
It is unclear if anyone has come up with automated ways to track and trade Trump tweets specifically — yet. The trading data indicates the response times are slower than the rapid-fire moves of an algo.
Mr Trump’s Lockheed Martin tweet, for example, was issued at 8:26:13 on December 12, according to an FT analysis of the tweet’s metadata. Lockheed Martin shares had been trading at $260.05 before the tweet. The first trade afterwards was 11 seconds after the tweet, for $258.57, according to Bats data compiled for the FT.
But Ian Domowitz, the chief executive of solutions networks, or the analytical arm, at agency broker ITG, argues against reading too much into the reaction time this early on. “If I am setting up an algo, I am not sure I would trust the first tweet,” he says. “Maybe I need some confirmatory information before the trade because there is other input on a continuous basis to that algo.”
On the other hand, perhaps with someone like a US president, the algo would not need anything else.
Trump’s tweets and their effects
1. Dec 6: Trump attacks Boeing costs on Air Force One
2. Dec 12: Says F-35 fighter programme is out of control
3. Dec 22: Attacks Lockheed-Martin again on F-35 costs
4. Jan 3: Threatens border tax if Chevy Cruze is built at Mexico plant
5. Jan 5: Tells Toyota to build plant in US or face taxes
One challenge with developing a Trump-specific algo is that there is little history to go on to draw reliable conclusion about the predictability of the responses, or if x then y. That has some market participants on the sidelines until there is more to go on.
“This new president is generating a good amount of social media data and many companies are looking at ways to make sense of this,” says Sandeep Kumar of Synechron, a technology consultancy for financial services. “The only limitation being that machine learning works best when you have a good amount of data — like data for the last few years. In this case we have only a few weeks and months of history when people are taking these tweets seriously.”
Another obstacle is the issue of fake news. Developers of this type of trading technology must not only find the correlations, but also ways of parsing fact from fiction, an issue that has come to the forefront of late.
James Ross, the head of HedgeChatter, which sells predictive analytics using inputs such as social media, says he has had a request from a hedge fund for tools specifically identifying Mr Trump’s targeting of sectors and companies combined with its other analysis that correlates other comments.
“President-elect Trump will be the first president to utilise Twitter as a financial throne,” he says. “This power will dramatically impact all sectors, companies, and even countries. It will also introduce a new level of volatility for the coming year.”
For Mr Ross and his clients, social media has become the third leg of analysis that once stood primarily on traditional fundamental and technical factors. This has gained steam against a backdrop of lacklustre financial returns of late.
“Look at returns in the hedge fund industry. Everyone’s data are tapped out. We need new data,” he says. “If [Mr Trump] can say something and Boeing’s stock can drop 5 per cent in five minutes, that is pretty powerful. Is that saying something about Trump or Twitter?”