Skip to Content, Navigation, or Footer.
Support the IDS in College Media Madness! Donate here March 24 - April 8.
Friday, March 29
The Indiana Daily Student

British hedge fund invests 25 million pounds for IU professor’s Twitter research

Illustration

Can Twitter predict the stock market? An IU research paper said yes, and a hedge fund is staking $40 million on the answer.

Having made national headlines in October after publishing his paper, “Twitter mood predicts the stock market,” School of Informatics and Computing professor Johan Bollen could experience a wealthy future as a result of his findings.

A London-based hedge fund, Derwent Capital Markets will be placing 25 million pounds — more than $40 million — on the line in hopes that Bollen’s research can accurately predict the stock market in a real-world setting.

Bollen found that when looking at mood data extracted from millions of tweets, he could accurately predict the changes in the Dow Jones Industrial Average three to four days later with an astonishing 87.6 percent accuracy.

After this insight, Bollen put his new research online and there was an almost instant response.

“That same weekend my phone was ringing off the hook,” Bollen said. “Colleagues, academics, investors, friends.”

Bollen and Huina Mao, co-researcher and post-doctoral student in the School of Informatics, will be looking at 35 percent of the income received from the University’s licensing contract with Derwent Capital Markets, which is pending negotiation.

Derwent Capital Markets’ founding director Paul Hawtin brought on Bollen and Mao to be private consultants.

“It is very much a partnership,” Hawtin said. “We’ll equally share in the success of the business.”

He said there is potential for Bollen and Mao to become very wealthy.

Hawtin stressed confidence when he expressed his calculations that his hedge fund will be successful.

“I’m talking a consistent 15 to 20 percent absolute returns,” Hawtin said. “If the markets are down a whole year, we’ll still be up 15 to 20 percent.”

If they have a successful first month of trading, which begins in May, it could lure in millions more from investors.

“I expect by the summer months to be at 50 million pounds,” Hawtin said.

HOW IT BEGAN

In 2009, Bollen and Mao created a mood mechanism, based on an algorithm which extracts and places Twitter users’ tweets into six mood categories: tension, depression, anger, vigor, fatigue and confusion.

Bollen put his research paper online after it was rejected from several publishers.

“People have been using blog analysis and news analysis, so people have been using this,” Bollen said. “It could also be that we were the only fools that published their results.”

Despite the rejections, the value of his work was proved through citations from eight academic papers who all found interest in his research, although the work remained unpublished.

It’s been dubbed “The Twitter Predictor” by CNBC.
 
“That’s what they call it,” Bollen said. “In the paper, at least, it’s not intended to predict anything. The only thing we said was this contains information that seems to have predictive power towards the stock market.”

Bollen, who has a Ph.D. in experimental psychology, is working in a field of research labeled sentiment analysis. He said it’s a new science aimed at developing mechanisms that can extract sentiment or mood from text. 

“I chose psychology because I was interested in mechanics of behavior and thought,” Bollen said.

After his 2009 paper on Twitter mood analysis was rejected, he said he saw the stock market as a possible way to validate his mood extraction mechanism research, not use it as a predictive tool.

“If you have a new measurement tool, you prove that the measurements that tool produces are valid by comparing it to other existing measurements you know to be valid,” Bollen said. “If they match, then you know the new thing is probably also valid. One of the best correlations we found was actually with the stock market.”

The sentiment data from Twitter was a precursor, not an expression, of stock market performance.

“It wasn’t a day-to-day correlation; the correlation was only significant when we shifted the mood graph forward by three to four days,” Bollen said. “Which means we were correlating the mood of three or four days ago with today’s stock market.”

MAKING IT BIG
Hawtin had been working at Derwent Capital Markets for more than two years, amassing relationships with wealthy investors across Europe. One day in October, he was sitting in his London office gazing into his trading screen and saw a headline describing Bollen’s findings.

“Don’t get me wrong, there are a lot of things that come around like this all the time,” Hawtin said.

But Hawtin said he felt little skepticism after studying Bollen’s research. In December, only two months later, he found himself meeting with Bollen to discuss his research.

“When I read the paper and the depth they had gone through, I knew there was some credibility to this,” Hawtin said. “I really believed the underlying theory as well.”

Several emails later, Derwent Capital Markets contacted IU to begin negotiations to exclusively license the intellectual property that Bollen developed.

Despite the prediction power of his Twitter sentiment extraction tool that excited everyone else, Bollen said he wasn’t expecting such a gamut of
coverage.  
“I never anticipated this much attention,” Bollen said. “I thought it was a cool finding but did not expect everyone to go berserk over this.”

The hedge fund, which Hawtin said will become open for investment April 1, already has 25 million pounds pledged from a network of investors.

“I’m getting inquiries every day from interested parties,” Hawtin said.
Mao said she believes ongoing work by other academics further validates their research.

“Especially in the past year there is more and more research trying to prove the predictive power of social media,” Mao said. “So it seems there’s more and more evidence to prove the idea that we already have.”

Get stories like this in your inbox
Subscribe