The rise of social media, AI, and NLP technology are helping quant investors turn sentiment into just another data point
According to the conventional wisdom, investors can broadly be divided into two categories: the “smart money,” who make their decisions based purely on rational metrics of risk and returns; and the “dumb money,” whose decisions are easily swayed by their own emotions as well as waves of bullish and bearish market sentiment.
But as they become increasingly sophisticated, the smart money is becoming more comfortable letting their decision-making be influenced by emotion – only it’s not their own.
Grant Wang, who’s senior vice-president and co-CIO of AGFiQ Quantitative Investing as well as head of Research at AGF Investments described how quantitative investors are able to leverage sentiment data thanks to advances in technology and digital communication.
“Perhaps the most obvious example of this evolution is the work being done by asset managers to decipher the emotional core of quarterly earnings calls,” Wang said.
He described how AGFiQ analyses transcripts of earnings calls through a proprietary sentiment-scoring model that uses NLP to aggregate the positive words and negative words used, and see how they correlate with returns. With that model, the firm was able to determine that compared to companies that didn’t mention COVID-19 at all, those that commented on the virus in earnings presentations last summer tended to suffer more severe losses in trading immediately subsequent to the calls.
Beyond that, he said social media platforms are being mined for up-to-the-minute conjecture on topics and themes that have significant influence on investors. In line with this, he said AGFiQ recently began integrating third-party research from Refinitiv and MarketPsych into decision-making for its ESG and infrastructure strategies. The data comes from a tool developed jointly by the two firms, which they say filters millions of articles and social media posts in multiple languages every day.
“The scoring – as described in a recent white paper by Refinitiv – derives from a curated data feed of live and archival content that is pulled from traditional news gatherers like Reuters and credible sources on social media, but shuns what is deemed ‘low-value’ or ‘non-objective’ forms of content such as corporate press releases and promotional article,” Wang said, focusing on the ESG aspect of sentiment analysis. “[S]ome of the sentiment scores resulting from [the analysis] have been strongly correlated to market returns in back-tested studies.”
AGFiQ isn’t alone. A recent article by the Wall Street Journal described how Cindicator Capital, a New York-based trading firm, uses a blend of survey data and artificial intelligence to track sentiment on Reddit and Twitter. Nodari Kolmakhidze, the firm’s chief financial officer, told the news outlet that he now views retail investors as a force that will probably shift markets for the foreseeable future.
“When you see these trades, you are like ‘wow, I could be there,’” Kolmakhidze said.
That’s not to say investors should let the pronouncements of the Reddit army dictate their decision making. As Wang from AGF noted:
“[W]hether it’s this particular sentiment analysis or one of the many others offering their own special take these days, it’s important not rely on them exclusively or, worse, treat them as a panacea that determines buy and sell decisions all on their own. They are not an end in themselves.”