Donald Trump has long had a habit of tweeting about the people and things he doesn't like -- but now that he's president-elect, his missives move markets. Over the last several weeks, his tweets have managed to knock 1% off Boeing's stock price, albeit temporarily, and 2.5% from that of Lockheed Martin.
On Tuesday, it was automakers' turn in the Trump spotlight, as the president-elect suggested G.M. should pay a big tax to build cars in Mexico. At about the same time, rival Ford announced its plans to scrap plans for a Mexican factory and invest the money in Michigan instead. (Ford's shares were up about 3% in afternoon trading.)
It now seems as though an investor could buy and sell stock, and make money, based on what @realDonaldTrump happens to be targeting on any given day. Indeed, that's almost certainly going to happen.
But that enterprising trader is not going to be you.
As it turns out, Wall Street has been watching Twitter, and many other sources of online news and gossip, for quite some time. And unlike casual investors -- who may happen to see the latest tweet, look up the stock, and fumble with their Schwab password before actually making a trade -- professional investors can rely on powerful computers to largely (or sometimes completely) automate the process.
“There are people diligently working to create algorithms for Trump’s tweets, and if he continues to increase the size of the data set then we’ll likely see full automation sooner than later,” Zachary David, a senior analyst at KOR Group, a consulting firm told Politico last month.
Such trading algorithms aren't new, even if the golden-maned subject may be. Wall Street firms have used computers to "scrape" various parts of the Internet for new information for more than decade. Long before Twitter existed, for instance, computer programs were pinging websites of large companies hoping to be the first to detect when a new quarterly earnings release was updated. In the age of social media and high-frequency trading, the scale and scope of those activities has only increased.
How fast are the bots? This Slate article, which quotes, an unnamed options trader, suggests that it takes a matter of moments for the stock market to react to news on Internet. "The speed is unbelievable. They’re buying everything within like 3 seconds of it coming out, which is not possible for a human,” the trader told Slate. Trading robots can react to other information -- like trading orders that appear at stock exchanges -- even faster, within millionths of a second.
Feel like the deck is stacked against the little guy? Well, you can console yourself with this: Trading stocks on the news, much less Twitter, is a dangerous game. And while a lot of money stands to be made, there have been some big disasters too.
One example: In 2013, the Associated Press' Twitter accounted reported that President Barack Obama had been injured amid two explosions at the White House. As it turned out, the report was the result of a hack. It was debunked within minutes by both the White House and the AP itself -- but not before the S&P 500 dropped about 0.9%, erasing roughly $130 billion in stock market wealth.
Here's how TIME summed up that situation at the time, offering a lesson that still holds true.
As for Tuesday’s incident, it’s possible that many firms had the words “White House,” “explosion,” or “Barack Obama” in their databases as key words that could trigger selling given the right circumstances. According to [portfolio manager and author Irene] Aldridge, “If a trusted news source with a lot of followers like the Associated Press sends out those words close together that may have triggered some selling.” Aldridge says the fact that so many people re-tweeted the message -- many of whom were trusted journalists themselves -- would likely make the news appear even more trustworthy to these bots.
In the grand scheme of things, Tuesday’s mini-crash was not a big deal. “No long-term investors lost any money,” says Aldridge. The market recovered almost instantaneously, and an optimist may look at the event as an example of high-frequency algorithms behaving in a more orderly manner than they have in the past.