As reported by Wall Street Journal’s Michelle Price:
Some firms are experimenting with messages on Twitter. The scale and velocity of data produced by Twitter, with its 175 million users, has struck some traders and academics as a potential goldmine of information.
Although tweets are an unstructured cacophony of thoughts, some studies have claimed that the data can be sifted for trading signals. A ‘buy’ trading signal might include positive chatter about a company, or broadly positive sentiment towards a sector.
Studies by Mines ParisTech and the Technische Universität München found a correlation between Twitter traffic and trading patterns. Using methods derived from computational linguistics, the TUM researchers found a correlation between the content of 250,000 stock-related Tweets and movements in the market. The study showed the level of bullishness detected in the sampled stock-related Tweets was associated with “abnormal” returns in some stocks, and that changes in message volume could predict next-day trading volume.
Another study, published by the University of Manchester and Indiana University last October, focused on the relationship between broad Twitter sentiment and market movements. The researchers argued that by reading and categorizing millions of Tweets on how individuals were feeling into one of six mood states—including calm, alert, sure, vital, kind and happy—it is possible to predict with 88% accuracy how liquid stocks will behave.