In a policy paper, the ECB attempted to determine the predictive powers of new technologies on gauging the short-term trading direction of the stock markets. One of the main things the paper looked at was Twitter, the popular social network founded in 2007 that currently boasts more than 300 million active members.
The researchers' methodology was simple — they looked at 310,000 tweets between 2010 and 2012 that contained one of two words: "bullish" and "bearish." There were 1,081 days in that timeframe, which works out to an average of 280 such tweets every day.
But activity tends to spike on days of extreme volatility.
Focusing on two specific phrases differentiates the study from previous ones because unlike others where the use of a word could be positive or negative (or not related to the stock market at all) the researchers made the assumption that any use of those two words would be related to stock markets.
The research also included tweets of those terms in other languages including the Mandarin ideograms 牛市 (bull market) and 熊市 (bear market) and while the scope of tweeters was global, the researchers only included stock market returns from four countries: the United States, the United Kingdom, Canada and China.
For every one per cent increase in what they call their Daily Sentiment Index — a relative change in the number of "bullish" references to "bearish" ones — corresponded with a 2.26 per cent increase in the performance of the Dow Jones that day. Even stranger still, that was followed by a 12.56 per cent increase on the second day.
But for whatever reason, the correlation has petered out within five days, when there appears to be no relationship between Twitter bullishness and the stock market.
"Twitter bullishness has a statistically and economically significant predictive value in respect of share prices," the ECB paper said. "This impact is statistically significant at the 99 per cent confidence level."
That's a much bigger impact than some of the other indicators the researchers looked at. They took a similar view to Google searches for two terms: "bull market" and "bear market" but found that was a much less useful indicator. They attribute part of that to the fact that Google's data on popular search terms is available in weekly increments, which makes it more difficult to gauge it for individual trading days.
"Both the level and change of Google bullishness are not predictive of the weekly returns of the Dow Jones," the paper said.
The paper noted, however, that while the researchers found a clear correlation between bullishness on Twitter and stock performance, that doesn't necessarily imply any causation is at play.
"Causal inference is important for result interpretation, robust prediction and policy-making," the authors said, recommending that more research on the topic is needed.