What's in a tweet? The power of Twitter to predict bond market performance

An award-winning research team expands their analysis of Twitter from stock returns to the corporate bond market.

Social media was once, not too long ago, a source of great technological optimism. Their brands are shown as beacons of a brighter, more connected tomorrow in which we transcend global and cultural differences to join in our shared humanity. Now social media is roundly blamed for most, if not all, of what ails us as a society. Information, misinformation, and outright disinformation that battle for our ever-shrinking attention span seemingly only add to our ignorance rather than relieving it.

But new research by Associate Professor of Accountancy Lucile Faurel and her colleagues, Partha Mohanram of the Rotman School of Management at the University of Toronto and Eli Bartov of the Leonard N. Stern School of Business at New York University, points to real value in perhaps the most reviled social media platforms, Twitter, at least for those busy investing in the corporate bond market.

"We usually think of social media as being full of misinformation, but in the case of financial markets, we find that overall, the information is relevant and useful," says Faurel.

Specifically, Faurel and her team show associations between the aggregate Twitter opinion and future changes in bond returns, CDS spreads, bond yields, credit ratings, and default risk. "This doesn't imply causation," she says. "It means that future changes are associated with what is communicated on Twitter." So, in this case, Twitter knows what it is talking about.

Social media isn't always stupid

To prove this, Faurel and her co-authors gathered approximately 11 million tweets from Twitter. To be considered relevant, each tweet had to feature a dollar sign and a specific stock symbol from one of the top 3,000 firms on the market.

The tweets were then filtered further to those carrying just one stock symbol tweeted just before quarterly earnings announcements for the company they mentioned. "We melted it down to a very targeted sample," Faurel says.

This reduced the sample size by about 90% to 1 million tweets, which were then exposed to textual analysis to capture the sentiment of each message. By aggregating these according to the company they referenced and the reported earnings quarter, Faurel and her team could assess how the tweets stacked up to the reality in the markets.

The analysis of these tweets showed that when taken at an aggregate level, Twitter's opinion could help predict business performance that would impact bond markets. Specifically, this study demonstrated that aggregate Twitter opinion (OPI) does expect upcoming announcements of bond returns. And that OPI is associated with future changes in bond yield spreads and credit ratings, meaning Twitter is providing economically important information to the bond market about such things as changes in credit risk and the likelihood of credit being downgraded.

This is surprising, given that the corporate bond market is supposed to attract more sophisticated investors than the equity market. "You might think they would already know all the needed information to price these instruments correctly," says Faurel, "and yet we find this new source that provides relevant information that they haven't fully priced into the market."

Wisdom of crowds revisited

Faurel sees this as another example of the wisdom of crowds, which is the idea that the collective opinion of a diverse group of individuals is more reliable and accurate than that of a single expert or a small group of experts. It's a democratic notion that was perhaps more popular before the current era of QAnon and other conspiracy theories that seem to testify more to the ignorance of the crowd, or at least certain crowds.

Fortunately, Faurel and her team have the data to back them up. A robust sample is a million tweets sent by people who are not necessarily professionals or employees of companies or institutions. "We didn't know the users," Faurel says, "There was no selection for expertise, only timing and subject matter."

In their analysis, the team even went so far as to differentiate between tweets providing original information or those re-tweeting current information. They studied them separately to see if the results would hold for both, and they did.

"This is cool," says Faurel, "it shows the role of Twitter in both sending new information and disseminating existing information."

Along the way, Faurel read many Tweets and ended up "laughing in front of my computer more than once. People say crazy stuff, including using the f-word — no filters exist. But it's the only way to get a real feel for the data you're dealing with. Who is tweeting? How many followers do they have? You can only tell by looking at the tweets."

Twitter as a source for stock performance

If it sounds like Faurel, Mohanram, and Bartov know what they're doing, it's because they've tested Twitter opinion before. Before this research, they set out to discover whether information shared on Twitter is relevant in the equity market.

They started by analyzing StockTwits, a social media platform designed for sharing between investors, traders, and entrepreneurs, because it focused more on their topic. But with only 6 million users total, compared to over 350 million Twitter users, the sample size couldn't measure. "We always want the largest sample possible to make our findings credible," says Faurel.

Even though Twitter isn't dedicated to financial markets, it found its users still collectively knew what they were talking about. Their research provided evidence that the aggregate opinion from individual Twitter messages successfully predicted a firm's upcoming quarterly earnings and announced stock returns.

That work, one of the first to focus on Twitter, was very well received and won a research award. So the team sought to expand their work into the corporate bond market. "We spent so much time on our first project, we were hoping for similar results with bonds," says Faurel, "but it wasn't apparent because the audiences are so different."

In contrast to the stock market, where retail stock investors actively trade, corporate bond investors are almost exclusively institutions, such as broker-dealers and large financial entities. They look for very different kinds of information. Stock investors are primarily searching for earnings and growth, but bond investors are interested in credit and default risks. Nevertheless, the results were similar, showing that Twitter's opinion was also relevant to bond trading.

What can you do with this kind of information? You could incorporate it into your trading strategy, according to Faurel. "If you had a way to aggregate and analyze such information efficiently enough, which is not too hard to do," according to Faurel, you could use it to trade more effectively. She says hedge funds are looking at this type of analysis as a primary trading strategy.

And even if you never use it to help you invest, you might regain some modicum of respect for Twitter and the social media ecosystem it represents because of this study.