What can 48 million stock market tweets teach you?

Researchers from the University of Edinburgh and Cardiff University analyzed seven years of posts from StockTwits and found two big conclusions:

–Social media posts predict market performance
–Specialists do a better job than generalists

For the first conclusion, the researchers used machine learning and natural language processing to detect positive or negative sentiment. They then calculated alpha using using a basic long/short strategy.

Their second insight was to analyze the performance of people who post about a wide variety of companies versus traders who focus on a small group. The results showed the “specialists” did much better.

What struck me about the work by researchers, Gabriel Wong, Woon Sau Leung, Woon Wong, was how social posts could be used to detect and quantify engagement.

An investor may have a large portfolio, but posting about specific securities would be a sign that they were more focused on those.

Its also a good reminder of the insights that alternative data sets like social media can offer.