Artificial intelligence is this year’s hot investing theme, and technology and semiconductor funds are up sharply.
What isn’t hot? The few ETFs that use AI for portfolio construction and have three- or five-year track records. Even newly launched or repurposed ETFs using AI to pick stocks are lagging behind broader markets.
The oldest is the $103.6 million actively managed
AI Powered Equity
ETF (AIEQ), which is up 6.5% year to date and has an annualized 4% five-year return. Its annual expense ratio is 0.75%. The largest by assets under management is the $1.6 billion index-based
SDPR S&P Kensho New Economies Composite
ETF (KOMP), which will turn five later this year. It’s up 0.08% year to date and down 4.3% on a three-year annualized return basis. It costs 0.20% annually. Meanwhile, the $16 million passively managed
Merlyn.AI Bull-Rider Bear-Fighter
ETF (WIZ) is down 1.1% for the year and down 3% on a three-year average. It has a 1.2% annual fee.
The funds also are lagging behind the S&P 500 and Nasdaq Composite, and, according to Morningstar, their respective peers.
Chida Khatua, CEO and co-founder of EquBot, a partner in the AI Powered ETF, says the fund uses IBM Watson’s language-processing ability to scour millions of financial and nonfinancial data points for over 6,000 U.S. companies, targeting a risk-adjusted return versus the broader U.S. equity market.
The team trained its AI on historical market performance to make probabilistic forecasts of how different securities will perform, using much bigger data sets than ever possible before. Currently the fund is overweight financial services and industrials, while significantly underweight technology versus its Morningstar peer group.
Morningstar considers AI Powered to be large-cap, but Khatua says it also contains small- and mid-cap equities, and that a closer benchmark is the small-blend
iShares Russell 2000
ETF (IWM), which is up 3.1% on a five-year average, slightly under the AI Powered ETF’s return. The AI ETF’s underperformance versus the S&P stems from the latter being a large-cap index, which has outperformed smaller-cap indexes, he says. “AIEQ was meant to be a complement to the S&P 500, not a replacement,” he says.
The AI Powered ETF takes a page from actively managed methods used by quantitative or systematic investors, says Komson Silapachai, a partner at Sage Advisory. The fund’s struggles may be due to active management underperforming indexes in general, not the fund’s methodology or process.
“It’s still pretty early for an AI-focused actively managed fund. But I commend them for doing it,” he says.
Matt Bartolini, head of SPDR Americas Research, says the New Economies ETF uses natural language processing to find innovative companies, categorizes them into themes, and uses a modified equal-weighting to balance them. It doesn’t fit into specific style, market cap, or sector categorization. Compared with its Morningstar mid-cap peers, it’s overweight technology and communication services, and underweight consumer cyclicals.
With a more evenly distributed market-cap profile, “naturally we’re going to underperform the broader market,” Bartolini says.
The biggest misconception he sees with retail investors’ interest in AI is that they want a pure-play fund that only invests in companies exclusively creating the technology. “The reason it doesn’t exist is because there’s a lack of consensus, and AI is very widespread within different economic use cases,” he says.
As interest rates have increased, speculative, innovation-focused ETFs have suffered, Silapachai says. But outperforming broader markets isn’t the point of the fund. “The AI use is not necessarily to beat the market, but trying to define the market,” he says.
Chris Berkel, investment advisor and founder of AXIS Financial, has experimented with AI to create portfolios. He says investors tend to think AI means that the computer is thinking for itself, rather than quickly gathering data and following human-created guidelines.
“The way we think of AI right now is like, there’s a tree in my yard and I want to turn that into a desk,” he says. “It doesn’t do that.”
Khatua concurs that investors need to remember the purpose of using AI, which is to see if it can improve over time against a preset objective, not to tell the future.
“AI is not a crystal ball,” he says.
As with any fund, investors need to look beyond the name to see what it holds and if it fulfills their objective, Berkel says. But he has other concerns about AI funds: “They actually scare me a little bit. And the reason they scare me is because they don’t share what they’re looking at.”
Sometimes AI’s use in portfolio methodology isn’t what contributes to returns, like the Merlyn.AI fund. It’s an adaptive momentum strategy that uses genetic algorithms to cut hindsight selection bias that can occur with momentum strategies. The AI is more “window dressing,” says Scott Juds, chairman & CEO of Merlyn.AI.
“It’s not intended to contribute to performance,” he says, attributing the weak returns to the momentum factor underperforming.
AI stock-picking may improve over time, Silapachai says. However, Brett Manning, senior market analyst at Briefing.com, says the current use of AI for portfolio construction proves the efficient market hypothesis, which suggests that current stock prices reflect all publicly available information.
“I don’t think it will ever beat the market,” he says. “There’s no ability to move outside of that model and see things differently from the rest of the market.”
Write to Debbie Carlson at [email protected]
Read the full article here