4 Comments

Summary:

Twitter may not yet be in the frame for its own stock market debut, but some investors think they can make millions by using Twitter to pred…

Crystal Ball
photo: Corbis

Twitter may not yet be in the frame for its own stock market debut, but some investors think they can make millions by using Twitter to predict market trends.

Last October, Indiana University researcher Johan Bollen, along with Xiao-Jun Zeng from University of Manchester, published a paper claiming sentiment analysis run on tweets can predict swings in the Dow Jones (NSDQ: NWS) Industrial Average with up to 87 percent accuracy.

Now Paul Hawtin, a London city trader at Derwent Capital Markets is forming a £25 million hedge fund that will use Bollen’s thesis as the basis for investment. Bollen and his co-researchers will get a 35 percent of the fees for which Hawtin has licensed the work from Indiana.

{tweet_id=”48335442906595328″}

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

So how does it work? And does it really work? The method revolves around machine-identified mood analysis. Bollen’s paper puts stock-related tweets in to one of six emotional buckets: tension, depression, anger, vigor, fatigue and confusion. Analysts then project that mood curve forward a few days to make a prediction.

The concept is not unlike existing comptuer-led investment strategies and, for that reason, we might expect growing interest from major financial media and information services.

And this is not the only instance we’ve seen lately. Pace University of New York doctoral student Arthur O’Connor has crunched Famecount social media data comprising Facebook Likes Twitter followers and YouTube (NSDQ: GOOG) views for the most popular brands on said destinations – Starbucks (NSDQ: SBUX), Coca Cola and Nike (release).

After running ANOVA and linear regression analysis, O’Connor claims to have found a statistically significant correlation between social media popularity and the companies’ stock prices, during the study period from April 2010 to February 2011.

We were able to reliably predict their respective daily stock prices over a 10-month period,” O’Connor says, “– during which, the stocks of the companies experienced radically different returns, with Starbucks climbing 29 percent, Nike appreciating by 14 percent, and yet Coke declining by nearly six percent – even when the social media data was lagged by as much as 30 days.”

If this holds true, it makes it doubly as incumbent on listed companies to combat negative consumer sentiment in social media as soon as possible.

O’Connor advises you watch Viacom (NYSE: VIA), Walmart and Sony (NYSE: SNE) stock – which have all seen recent spikes in social media popularity – for upcoming movement.

  1. Folks who are interested in the sentiment angle should consider attending a conference I’ve organized, the Sentiment Analysis Symposium, April 12 in New York: http://sentimentsymposium.com . We have a couple of speakers on capital-markets applications.

  2. AbsolutWealth Thursday, March 24, 2011

    Doesn’t seem like Derwent Capital Markets the only one with the crystal ball. I came across another one at http://nnatch.com. When everyone starts using this type of sentiment analysis, the edge will be gone sooner or later.

  3. I doubt how reliable this social network sentiment detection/prediction is. Isn’t that many banks/hedge funds block out social network access, i.e. FB, twitter at work? And how big is the sample these study are looking at? Are they representative to the financial professionals population?

    But technically, it is doable. And the technologies are already there – many companies in online advertising industries have offer similar services/products to help marketers monitoring brand reputation. I myself have run similar analysis for clients.

  4. Funny that an article about the use of Twitter cannot be shared using Twitter!

Comments have been disabled for this post