Researchers from a Dutch university have developed an algorithm that scans tweets about stocks, successfully predicting price fluctuations.
Ting Li and Jan van Dalen from RSM Rotterdam School of Management, Erasmus University analysed over a million tweets that mentioned stocks listed on the S&P 100 share index.
The researchers developed an algorithm that looked at the sentiment of the tweets and extracted distinct ‘buy’, ‘hold’ and ‘sell’ signals embedded in them - before comparing them to actual price fluctuations on the stocks over the following days.
Stocks that saw bullish tweets such as ‘buy!’ experienced, on average, higher abnormal returns. Meanwhile, the correlation is even stronger for influential Twitter users who are frequently retweeted and often mentioned - while the number of tweets about a particular stock could predict trading volumes, volatility and follow-up return on a stock.
To see if the algorithm could be used as a the basis of a profitable trading strategy, the researchers ran a 21 week simulation using the information from the study and found that, even taking transaction costs into account, the simulated returns beat the market.
Says Li: "This could be used by institutional investors or home-based day traders and proves that twitter isn’t just noise - useful information can be extracted and could help investors make better decisions."
Li and van Dalen are far from the first researchers to link Twitter and stock prices. As far back as 2010, researchers found that analysing the content of daily Twitter feeds using two mood tracking tools could predict with an 87.6% accuracy the daily ups and downs in the closing value of the Dow Jones Industrial Average.
The algorithm used became the basis of a £25 million hedge fund, although the experiment was quickly shuttered despite a steady performance.