Environmental, social, and governance (ESG) investing is one of the most exciting emerging interfaces between hedge funds and machine learning, according to the latest The Cerulli Edge-Global Edition.
Hedge funds are finding new ways to use machine learning and ultimately artificial intelligence (AI) in the investment process to further capitalise on the ever-growing quantities of data.
“One area of potential growth for hedge funds is applying quant techniques to ESG integration. Traditionally, hedge funds have focused on generating alpha and providing decorrelated returns, but our recent survey showed that 46% of investors believe integrating responsible investments into hedge funds will be ‘very important’ in two years’ time,” says Justina Deveikyte, associate director, European institutional research at Cerulli Associates, a global research and consulting firm.
For example, long/short ESG funds can now allow investors to profit from companies going in the wrong direction on climate change or governance, increasing the cost of capital for polluters or companies with, say, excessive executive renumeration.
Deveikyte says that, despite the vast capabilities of machine learning, quant hedge fund managers have yet to determine exactly how to integrate ESG factors into their investment processes and algorithms. “AI is transforming data gathering and fund managers can now access vast amounts of information from objective sources.
However, it takes considerable effort to identify material ESG signals and shift the investment process in order to accommodate ESG integration across a range of hedge funds strategies. Quantifiable ESG metrics are what matter. Nonetheless, hedge funds are increasingly working to develop repeatable processes that can accommodate custom ESG requirements.”
Cerulli believes AI will be especially useful in short-term, high-frequency trading, but notes that the complexity of financial markets means that AI will not inform long-term financial predictions just yet. Long-term financial data is relatively scarce—GDP figures, for example, are released only once a year. In addition, financial data tends to have large amounts of irrelevant data for every piece of useful data, which can make finding meaningful patterns challenging.