Skip to main content

Horizons ETFs Management Canada Inc. is recalibrating its artificial intelligence (AI) exchange-traded fund (ETF) after a year of underperformance that was a “complete disappointment,” according to the provider’s chief executive officer.

The Horizons Active A.I. Global Equity ETF (MIND), which uses a proprietary AI-directed selection process to invest in major global equity indexes through a basket of ETFs, has trailed the market over the past year, down 1.04 per cent on a return basis, versus a positive total return, in Canadian dollar terms, of 5.76 per cent for the MSCI World Index.

“We have been looking at MSCI World, and we’re significantly underperforming on a year-to-date basis,” Horizons Canada CEO Steve Hawkins said.

“When we first launched this product, we had very high hopes. People would ask how well this could outperform the market. All the back-testing we did provided us with a lot of comfort that an unbiased AI system making investment management decisions would outperform the market when you’re dealing with a lot of investor bias.”

With a management fee of 0.55 per cent, the AI ETF monitors dozens of metrics, including the news, money flows, moving averages and volatility, to assess index exposure and rebalances on a monthly basis to revise its asset mix.

But earlier this year, the ETF’s AI faltered when it expected the December stock sell-off to continue into January, according to Mr. Hawkins. Instead, the market swung sharply in the opposite direction as investors snapped up shares at a bargain.

Since then, Horizons and Korea-based Qraft Technologies Inc., which developed the AI in the ETF, have been re-evaluating the way that the AI makes decisions, considering incorporating new factors to improve how the machine interprets news and hedges for currency fluctuations. With the changes set to launch at the end of October, Mr. Hawkins said the technology needs an opportunity to learn from its mistakes.

“The AI systems will continue to adapt and to the extent that they may have underperformed today, there is a strong possibility that they can outperform in the future,” Mr. Hawkins said.

AI-powered ETFs that launched in recent years to much fanfare are showing cracks in their performance and, with the funds consistently trailing their benchmarks, investors are growing skeptical about whether they should trust the mechanics under the hood to make investment decisions.

Some question AI’s ability to predict how broad issues will affect markets. Anything from a controversial statement from a world leader to a transport truck crash could cause a swing in the global economy on any given day, and the AI-powered ETFs have yet to prove they’re capable of predicting how markets will react, said Jason Pereira, partner and senior financial consultant at Toronto-based Woodgate Financial Inc.

“The Horizons ETF had its ups and downs and dips, but essentially you’re not much better off than when you started,” Mr. Pereira said. “It’s an interesting experiment. I’m not holding out a lot of hope that in the early stages it will be sufficient enough to outperform a benchmark. As an academic experiment, it’s great. But getting people to invest their money in it, it’s bad.”

San Francisco-based firm Equbot LLC’s AI Powered Equity ETF (AIEQ) has followed a similar pattern. While the fund is outperforming the S&P 500 year to date, it’s trading below its value when it opened in 2017 and has lost 0.7 per cent over the past year on a total return basis, versus the benchmark’s 4.4-per-cent gain.

The IBM Watson-powered ETF, which charges a management fee of 0.77 per cent, analyzes data on thousands of U.S. stocks, including regulatory filings, social-media posts and quarterly results. It hasn’t gained much investor attention because of its inconsistent performance, said Todd Rosenbluth, senior director of ETF and mutual-fund research at New York-based investment research group CFRA.

“Investors will give these strategies a look because of the appeal of using a computer-powered approach, but anything that is offered at a premium price, you need to deliver strong performance for investors,” Mr. Rosenbluth said. “So until they show consistent outperformance that’s in-line or better than the broader market, they’ll be below the radar for many investors.”

Andreas Park, director of research at the University of Toronto’s Rotman Financial Innovation Hub in Advanced Analytics, suggests the use of AI in investing products could be more of a marketing tactic than technology that can outperform humans in predicting markets.

“Everyone wants to say they use AI,” Mr. Park said. “It sounds cutting edge, but it’s data analytics applied to a problem to make a prediction of what to do next, and that’s what we’ve been doing for quite a while. This isn’t transformational or different from what we’ve been doing for decades in investment.”