Can Twitter predict the stock market? So one PhD student believes.
The BBC reports that Timm Sprenger of the Technical University of Munich found that buying and selling in response to an analysis of 250,000 tweets over six months would have produced a 15% return.
Sprenger looked at comments that had been marked with the hashtag relating to a company’s stock market symbol. He also paid attention to which comments were widely retweeted.
The logic seems to be along similar lines to the “wisdom of the crowd” theory, best known from a book by James Surowiecki. The idea is that a single Twitter user’s comment, hunches or “insider info” about a stock may be unreliable, but taken in aggregate Twitter users correctly indicate the mood of the market, which is effective because stock prices ultimately come down to nothing but the balance between demand and supply.
A similar study that looks at the market as a whole (and uses the wonderful phrase “Self-Organizing Fuzzy Neural Network”) claimed that looking through Twitter posts and measuring six elements of positivity or negativity (Alert, Calm, Happy, Kind, Sure and Vital) could predict whether the Dow Jones Industrial Average would move up or down on a particular day — based on the mood of the previous three days — with 87.6% accuracy.
Of course, to take advantage of that “prediction” you’d have to trade in market tracker funds rather than individual stocks, which may limit your ability to cash in on a day-to-day basis.
There’s even a website, TweetTrader (pictured), dedicated to tracking tweets about markets, industry sectors and specific stocks. It’s tough to say it’s effective though: of the top six stocks that were moving up in “sentiment” and the top six that were moving down, barely half had price patterns that matched this sentiment. It’s also hard to say in this specific set-up how much of the sentiment was genuinely a predictor, and how much was a self-fulfilling prophecy, with the sentiment a reaction to existing movements.