Hedge funds with artificial intelligence capabilities showed a huge competitive edge over investors that didn’t use AI, new research indicates.
AI-led hedge funds produced cumulative returns of 34 percent in the three years through May, a report Tuesday from consulting and research firm Cerulli shows. That compares with a 12 percent gain for the global hedge fund industry over the same period.
“There has long been suspicion of the ability of AI to react to unexpected events, such as the coronavirus pandemic,” said Justina Deveikyte, associate director of European institutional research at Cerulli, in a statement Tuesday. “But there is now a sense that the technology has advanced to the point where it is better able to adapt to unforeseen scenarios via the ever-growing amount of market data available.”
Machine-learning algorithms in finance tend to recognize patterns in historical data, making it tough to adapt to economic lockdowns caused by a one-in-a-century pandemic, according to the report. Still, factors used in machine learning draw from sources such as newspaper stories to collect information daily, spotting trends for investors amid the Covid-19 turmoil, according to Cerulli.
AI-powered funds in Europe saw “strong growth” in assets from 2016 to 2019. Net new inflows dropped in the first four months of this year as the pandemic shook markets, but the research firm found signs that they fared better than the broader universe of managers. European active equity funds led by AI saw a “less pronounced decline” in market appreciation in March.
Quantitative and discretionary managers may use artificial intelligence to extract information on, for example, retail foot traffic, satellite data, and pandemic spread, according to the report. They may also analyze massive volumes of text through natural language processing to gauge sentiment tied to the economy such as employment and business activity.
“Investors that understand how sentiment steers choice and judgment can better gauge how a specific news item will affect markets in terms of both direction and intensity,” Cerulli said in the report
As part of the research, Cerulli consulted firms including RAM Active Investments, PanAgora Asset Management, and NN Investment Partners. PanAgora’s director of equity investments Mike Chen told Cerulli that constructing machine-learning factors is a “balancing act” as fund managers don’t want them to react to quickly to “noise in the market information” — or so slowly that they miss a trend, according to the report.
NN uses behavioral science, AI, and sentiment data —along with fundamental research — to try and anticipate market movements, Cerulli said. RAM fund manager Emmanuel Hauptmann emphasized that “machine learning still relies on a huge amount of data to be efficiently trained.”
The MarketPsych Indices from Refinitiv is one way that investors extract macroeconomic information from social and news media, according to the report.
“Even though most machine-learning algorithms are open source, it is possible to keep a technological edge thanks to fine-tuning of hyperparameters for specific tasks and agile management of machine-learning pipelines to continuously incorporate technological improvement,” Nicolas Jamet, a senior quantitative analyst at RAM said in the report.