Active managers under threat from the boom in passive investment funds can be saved from oblivion by using artificial intelligence techniques long used successfully to power everything from Netflix to forecasting hurricanes, argues a new consortium of academics, data scientists, and technology and investment professionals.
With active managers failing to beat their benchmarks after fees for a decade, the group is advocating that the industry use a technology-based approach that leverages the best insights of multiple portfolio managers — humans —as opposed to the single-expert model that has long been used by active managers.
The techniques, called ensemble methods, were originally developed to combat the diminishing returns that people were getting from their predictive algorithms. Ensemble methods combine multiple predictive algorithms into one that is potentially more accurate.
Active managers could use the methods and turn the best ideas of multiple experienced stock pickers into a market-beating strategy, says the group, which calls itself the Ensemble Active Management Consortium. In a white paper to be released later this week, the group examined whether portfolios constructed using the techniques could beat both active and passive funds.
[II Deep Dive: Managers’ New Hustle Is Outsourced R&D]
“This is a technology that has been proven for ten years in other industries,” said Alexey Panchekha, co-founder of technology and software firm Turing Technology Advisors and the consortium’s technology guru.
Panchekha said the methods have not been widely used in asset management. “But the root of the problem they are solving is general and well understood,” he said. “There’s a natural application to the investment industry.”
In the white paper, the consortium applied ensemble methods to high-conviction stock selections of traditional actively managed mutual funds and back-tested 30,000 portfolios over 10 years. The group found that what it calls EAM (ensemble active management) portfolios outperformed the Standard & Poor’s 500 Index 72 percent of rolling one-year periods between July 2007 and December 2017. The average annual excess return was 3.4 percent.
In addition, the portfolios outperformed actively managed mutual funds during 82 percent of the rolling 1-year periods, and 95 percent of the rolling 3-year periods.
“I think of a portfolio manager as a forecasting engine. But a manager can have a confident view on 10 to 20 stocks, not 300,” says Panchekha, who worked at Goldman Sachs and Bloomberg earlier in his career. “They have these convictions for good reason; they’ve built an entire career under a particular approach. Some have a bias toward value, some momentum. These are different views, and are extremely valuable.”
Matthew Bell, president of Bell Family Interests, a private family office management and consulting firm and also a member of the consortium, says, “The root stock of this whole method is the work that all these active managers are doing. They’re distilling down their best picks. Collectively they can be a lot better than separately.”
Bell added, “This brings technology, human intuition and research together in a way that our results show have value. It will be interesting to see how the industry embraces this.”
Some are skeptical about the white paper’s results, however. “The promise of AI is extraordinary in asset management,” says Michael Spellacy, global capital markets practice lead at Accenture. “When I say AI, I mean applied intelligence; there’s nothing artificial about intelligence at all. And it has exponential potential to provide humans with technology and insight advantages.”
But Spellacy stresses that the group is trying to do the impossible. “Alpha can’t be democratized. By its very nature, alpha is a non-democratic return stream.”
Will Kinlaw, head of State Street’s academic affiliate, added, “The ‘wisdom of crowds’ concept has a natural appeal. The challenge in applying it to financial markets is that the crowd, if defined broadly enough, actually is the market. By definition, the entire market cannot be overweight a particular company. So it all comes down to which part of the crowd you want to track and why.”
Added Kinlaw, “The jury is still out on whether AI methods will add value relative to simpler aggregations of best ideas and high conviction positions.”
