Three major trends have converged to drive growing appeal in new alternative data classes of quantitative investing, according to a leading quant researcher.
“Quants like us who were in the right place at the right time in history can take advantage of the confluence of these three major secular trends,” said Mike Chen, head of alternative alpha research at Robeco in the United States.
Speaking about finding alternative alpha at Conexus Financial’s Fiduciary Investors Symposium in Singapore, Chen said the amount of data in the world is growing exponentially. Algorithms have become very powerful, with some highly sophisticated algorithms free for consumers to use, such as artificial intelligence chatbot ChatGPT. And the computing power required to run these algorithms has arrived and is getting faster.
Chen gave examples of some of the developments in the market as quant investors seek to stay ahead of the game.
Company executives, aware their conference calls with investors are being fed into algorithms, have long been coached to use positive and bullish key words to trick the quants.
But vocal chords, made up of 47 separate physiological mechanisms, are much harder to train, Chen said. Some algorithms are now converting audio recordings into spectrograms and using this to detect a person’s underlying emotional state.
“You can compare that against the words that they say,” Chen said. “Are they in agreement or are they not? Their intonation, pitches, volume, pauses, all that information can be analysed.”
Patterns of interaction
Machine learning is also detecting patterns of interaction between market participants and stock prices, such as decoding the mysterious ‘reversal effect’ where stocks rebound or ‘bounce’ somewhat after sharp inclines or declines. The fact that they do not do this on some rare occasions was long thought to be a “statistical fluke,” Chen said, but it is actually related to the news volume surrounding the event that caused the rise or fall.
“When there’s a huge amount of news that’s happening related to given company, when that company’s price is going up or down, the price does not reverse,” Chen said.“What this means is that the price movements in those situations where there is a very high or abnormal news volume are actually being supported by factual information, not just being pushed in a vacuum by speculators or FOMO people.”
Language processing can also not only check whether company executives are using bullish language, but also whether they are answering analyst questions directly or evasively, he said.
Also on the panel was Charles Wu, chief investment officer at State Super in Australia. Around seven years ago, Wu began looking at machine learning to complement State Super’s investment process by providing more information to back up investment decisions.
Insights from data can help investment professionals challenge the judgements they make based on the limited experience of their careers when long-term paradigm shifts take place in markets, Wu said.
“It tells us things such as that interest rate differentials may not be your best determinant for a currency movement,” Wu said. “That’s something that we learned during this machine learning process, and that in itself gets us to more useful questions.”
For investors who want to add elements of quant to their investment process, it is important to start small, with clear and well-defined goals, he said. An advisory board of experts from academia can help bridge the communication gap between the board and internal stakeholders who are skeptical of quant, he said.
