How Hedge Funds Use Twitter to Gain an Edge in Trading
Investors and stock traders are paying close attention to President Donald Trump’s Twitter account for good reason: His tweets can move markets. They can even use computer algorithms to capture these tweets and make more money.
Defense contractor Lockheed Martin’s stock took a hit in December last year, losing nearly $4 billion in market value, shortly after Trump criticized the company’s F-35 contract on Twitter. Companies including Boeing, General Motors, and Ford were also recent victims of Trump’s tweets.
Trump prefers Twitter to other media, as he gets his messages out more efficiently and honestly, he says.
Fund managers have been using Twitter data for years to make investment decisions. According to a recent study published by NPR for example, 25 percent of institutional investors said they use social media for research.
Fortunately, investors themselves do not have to monitor 500 million tweets posted daily. They can rely on advanced machine learning and artificial intelligence techniques to analyze millions of messages.
Among social media platforms, Twitter is by far the largest source of data, according to Valerie Bogard, an equities research analyst at Tabb Group, a capital markets consulting firm.
With growing demand, the number of companies that mine and analyze social media data, like Dataminr and TickerTags, has exploded in recent years.
“Traders have been using Twitter since its launch [in 2006],” but gained more traction in 2013, said Bogard.
In that year, the Securities and Exchange Commission (SEC) approved a rule allowing companies to use social media sites like Twitter and Facebook to disclose corporate news.
Then, the Associated Press (AP) Twitter account was hacked, resulting in a false tweet that said two explosions in the White House had injured President Barack Obama. The event, known as “hash crash,” sparked a selloff in U.S. stocks and showed how fast a tweet could move the markets.
Such events changed the way investors viewed social data, said Bogard.
Alternative data can be used in sentiment analysis. It can identify the market’s opinion on a particular product, a stock, or the mood of traders. It can also be used to detect events like breaking news that move the markets as well as long-term trends, ideas, or cultural movements important for certain stocks.
“Having alternative data does not guarantee that you are going to make money. But not having alternative data almost guarantees that you are going to lose money, because investing is highly competitive. It is like an arms race,” said Gene Ekster, who advises hedge funds on alternative data strategies.
Twitter is only one component of so-called alternative data or non-traditional data. Hedge funds also use other alternative data like satellite imagery and public website data to make investment decisions.
People think only quantitative funds use alternative data, but this is not true, according to Ekster.
Fundamental long-short investors above a billion dollars in assets under management are using the non-traditional data sets to make their investment decisions, he said.
Depending on the dataset, funds can pay anywhere between $50,000 and $3 million per year, Ekster said.
Shortcomings of Twitter Data
For traders, having an information edge is critical. However, there are some challenges in using Twitter as a data source.
The White House “hash crash” incident, for example, was a costly lesson for some investors who traded on that fake news. The S&P 500 index briefly lost $136.5 billion in value before it bounced back, according to Reuters.
Another example was the Lululemon Athletica case, said Ekster.
Shares of Lululemon Athletica Inc. (ticker symbol: LULU) fell sharply in March 2013, after the athletic apparel retailer said it would recall pants that were too transparent. Following the news, many people on Twitter made jokes about the see-through pants.
“They were very sarcastic when talking about Lululemon. They were using smiley faces and words that made tweets seem like positive news,” Ekster said.
The automated learning tools interpreted this as positive news and created a strong buy signal for the company’s stock. And some investors lost money because of that.
“The problem is that a computer is not good at dealing with sarcasm. … Their algorithms were not picking up on the sarcastic remarks,” Ekster said.
Such incidents are less frequent but they do happen. Hedge funds are constantly improving their data filters to avoid trading on fake news alarms and to gain an edge over their competition.