With some large Australian superannuation funds internalising capabilities with the help of systematic approaches, Andrew Chin, AllianceBernstein’s chief data scientist and head of quantitative research, said systematic approaches have the advantage of low barriers to entry as investment teams are able to create strategies from open-source code.
But this is a double-edged sword, Chin said. “You have to make sure that you’re keeping up with everybody else on the other side.”
Speaking at the Investment Magazine Fiduciary Investors Symposium held at Healesville, in a session examining the evolving role of quantitative investing, Chin said challenges such as the cherry-picking of favourable data – known as “p-hacking” and “data mining” – are emerging owing to the signal-to-noise ratio being very low in finance.
Having more data does not necessarily mean there are more insights in that data, Chin said, and teams need to be on their toes to ensure their insights are useful and investable.
He gave the example of “CEO sentiment inflation,” or an upward trend of machine-generated measures of CEO sentiment over time. This is because CEOs have become aware that algorithms are scoring the sentiment of their earnings transcripts and responded by adapting their language.
“So CEOs are thinking: ‘Well, you know what? If I can improve my sentiment score, maybe people will buy my stock, right?” Chin said. “If you’re using sentiment to assess companies, I think analyst sentiment is a much better view because it’s more neutral and it doesn’t have the inflation that you see with CEO sentiment.”
Success in this space depends on a lot of factors, Chin said, but it is critical to have an organisational culture that encourages innovation and creativity, and is open to new techniques for solving investment questions in a fast-evolving space. At the same time, the culture needs to be rigorous to avoid issues such as the above-mentioned p-hacking and data mining.
Bringing in new talent to complement existing experts is critical, he said. “These new techniques and this new data does require new ways of thinking and new techniques to analyse the data,” Chin said.
Organisational infrastructure is also important, with techniques such as graph theory – which measures relationships and linkages – impossible to execute on an Excel spreadsheet.
“To infer relationships and to show how linkages work across all the different entities you need more sophisticated techniques and more sophisticated databases to actually harness that,” Chin said.
Chin said one example of low-hanging fruit for investors in the area of data science was Natural Language Processing or NLP, which can be used to read text and quickly find answers that are buried in long, dense documents.
This can be particularly useful for ESG issues, where an analyst team could, for example, quickly extract the metrics a company is using to describe its ESG goals.
“This speeds up things for you and it makes things much more efficient in terms of how you run your business,” Chin said.