This content is from: Portfolio
The ‘Alternatives Turing Test’: How BlackRock Is Using AI to Crack Opaque Private Markets
Investors have relied on intuition and loose allocation rules to build private asset portfolios — but that’s about to change, according to BlackRock.
Using artificial intelligence and data science, investors will be able to bring the portfolio construction and risk modeling techniques that they have long used for public securities to alternatives like private equity and venture capital, according to research published by BlackRock.
Private market investors have faced barriers to improving their portfolio construction, including illiquidity, opaque valuations, and a lack of public information. But the BlackRock paper, expected to be published Friday, offers a portfolio framework investors can use for alternatives. By using AI and data science techniques, investors can forecast the performance of potential investment opportunities in different asset classes over time and optimize various combinations of funds according to their risk appetite, BlackRock said.
Pam Chan, chief investment officer and global head of BlackRock’s Alternative Solutions Group, which has $8 billion in client assets, said it didn’t matter quite as much in the past if investors used rules of thumb or other approaches to build alternatives portfolios because they were ultimately a small part of the whole. But now investors are increasing their allocations to alternatives, as well as putting money into new varieties of funds, including music royalties, aviation finance, and agriculture.
“The appetite to increase the amount of alternatives is very high today — and that’s an understatement,” said Chan, in an interview. “Alternatives have become much more versatile. That has increased the importance of thinking through one’s alternatives portfolio as portion of the whole and not as something off to the side.”
Chan stressed that alternatives are often meant to be high returning and are a big part of the risk exposure for most investors. As such, it’s important for investors to understand how alternatives specifically contribute to risk, as well as the impact of illiquidity.
“As allocations to private markets have increased, many investors have struggled to develop portfolio construction methodologies that account for the unique aspects of private market investing,” wrote the authors of the paper, which includes professors from Stanford University, who advise BlackRock AI Labs. “Some have ignored quantitative portfolio construction altogether, opting for a more qualitative approach, while others have attempted to use standard quantitative methods developed for liquid investments.”
The authors pointed out that some of the standard quant methods use metrics like internal rates of return, which don’t fully represent the behavior of private assets.
[II Deep Dive: Artificial Intelligence Isn’t Just for Quants Anymore]
But there are huge benefits to a quant approach, including investors being able to tactically allocate to undervalued or more attractive opportunities at certain times in the economic cycle. Just like publicly traded stocks and bonds, private equity, private credit, infrastructure, and other alternatives have delivered returns that often vary significantly from year to year.
BlackRock’s paper addresses limitations including illiquidity, which prevents investors from easily rebalancing their portfolios. Because of that, BlackRock argued that it’s even more important to bring a new approach to these portfolios.
But BlackRock is careful to acknowledge the current limits, even with AI.
“Unlike traditional marketable securities, private asset classes typically require specialized modeling techniques and carefully curated data,” the report stated. “Importantly, any quantitative approach to modeling should be tempered by the judgment of an experienced private market investor.”
Chan described BlackRock’s approach as an “Alternatives Turing Test,” referencing the test proposed by mathematician and computer scientist Alan Turing to judge the results of a computer by determining whether it’s indistinguishable from that of a human being.
“We work with people who have been investing in these asset classes for a long time,” she said. “If they say, ‘I wouldn’t recognize that,’ then we go back to the drawing board and make sure the model doesn’t get carried away with itself.”