The algorithms powering the world's automated investment platforms are going toe-to-toe over trillions of dollars of investment assets versus their human counterparts, who are slower and more prone to mistakes.
But the algorithms, which can speed up financial processes by making millions of calculations faster than any human ever could, are not infallible. The hidden problem of relying on software is that they still have the computational assumptions of their programmers and can be based on outdated and flawed information, according to some critics and computer programmers.
Many of these robo advisors base their underpinnings on modern portfolio theory, which relies on past returns or an analysis of expected returns and risk to determine portfolio allocation. But MPT, as it is known, came under fire for not taking into account black swan events like the global financial crisis of 2008, which took an axe to many investors' supposedly safe and diversified portfolios.
Add in newer and faster trading algorithms and unpredictable and unknown factors, and these robos are navigating an increasingly complex market that they are perhaps not built for, some experts suggest. They have not tested their mettle during an extended bear market or another black swan.
"Robo advisors have been simulated on bear markets in the past, but the challenge is that every market is different. Those market participants are increasingly computerized," says Christopher Thorpe, a computer scientist who has taught computational finance at Harvard University and is co-founder of Blueleaf, an online financial planning software firm.
It would be difficult to simulate future financial markets, Thorpe adds, because "we don't even know who will participate in these financial markets. We don't know what computer programs will be invented and we don't know what crazy things these computer programs will do. If you can't predict a class of things, you can't program a computer to react to it in a rational way. Computer algorithms people invent two years from now will be different from what is used to trade today. SEC rules can be different."
In the event of a black swan, a robo advisor may react faster but the program may make decisions that are surprising, Thorpe says. Humans make mistakes, but if there is a bug in a robo advisor software, there could be a trading mistake that could be harder to discover.
"I know they are definietely not infallible," says James W. Watkins, a Georgia attorney, financial planner, and founder and CEO of InvestSense, an investment education firm. "I think one of the things overlooked is when robo advisors talk about managing someone's assets."
When Watkins consulted with a Charles Schwab representative on its robo advisor, he was displeased to learn they only rebalance the portfolio, he says. They don't provide true asset management especially when bad times happen.
"That creates a problem. The market proves to be cyclical. We don't know when a bear market will happen, but we know it will happen. It's better to reallocate one's portfolio in order to protect assets. And rebalancing doesn't necessarily do that," Watkins says.
(“Schwab Intelligent Portfolios provides asset management that is automated and sophisticated,” says Schwab spokeswoman Alison Wetheim in response to Watkin’s critique. “It uses a simple, straightforward questionnaire to gather insights into an investor’s goals and risk profile. The algorithm creates a portfolio with up to 20 different asset classes, customized to those individual circumstances. Over time the portfolio is re-balanced and tax-loss harvested at the individual account level, providing further customization based on cost basis; the rate, size and timing of additional investments; and the size of the portfolio.”)
Other pitfalls in a fully automated system are service outages that cause delays or an inability to contact the custodian, says Vlad Kobilansky, one of the programmers behind Wealthbot or Webo, an open source robo advisor that caters to human financial advisors. Another issue is that these robo advisors are walled gardens. There is no way to access their programming and determine if they are truly effective.
"We don't know what their rebalancing logic is like. It's a closed system. What if Dow Jones crashed by 50%, the rebalancing may go out of wack," says Kobilansky. "Our rebalancer logic is fully open to the world. Anybody can take a look at the code and scrutinize it."
But some leaders in the industry think robo advisors won't be prone to too much error.
"Am I worried about algorithms making mistakes? Not really. They are mostly static equations," says Joseph Pagano, practice advisor with Cisco's financial services business transformation group. "The algorithm or computer logic makes decisions based on a person's input. As long as there is documentation on what robos are doing. As long as the customers understand what the investment products are ... there is little room for error."
And if the market hits a rough patch, programmers can make adjustments with a few clicks, says JD Singh, senior vice president of business and product development at Vetr. Because of their scale and automation, changes can be made quicker to robo advisors. A human advisor, who would have perhaps around 100 clients, would have to move money around per account.
Singh also points out these algorithms go through rigorous testing. Many companies use live data to test and refine these robots before they go live.
As for any black swans like several years ago, Singh adds: "Everybody was caught with their pants down. Very few advisors accurately predicted the crisis."
Despite misgivings of investment algorithms, they are here to stay and can help retail clients and financial advisors alike, Pagano says. They fill a gap in the market for those starting out in their prime earning years and automate many of the tedious tasks that take up a financial advisor's precious time.
And algorithms have been useful in not only determining what to invest in, but also to make sense of the voluminous amounts of data these programs have generated.
“Unstructured data is hard to make sense of,” Singh says. “They (algorithms) have to sift through so much noise."
The data exhaust coming from these programs will make financial service firms more responsive to clients and advisors, Pagano says. Algorithms can crunch the numbers coming from these programs and suggest areas where adjustments can be made.
“The key to all of this is analytics, the real value of robo,” Pagano says. “It’s the collection and correlation of data to make a service more relevant and to better understand what the client is trying to do.”
“The information about a transaction has more market value,” he adds,” than the transaction fee itself.”
The ideal situation would be the combination of human and robot, many experts say. While algorithms are faster, they don't have the same capacity for creativity or finesse as a financial advisor, says Matt Pistone, Riskalyze chief technology officer.
"There's a general concern when you give a computer a set of instructions that it doesn't capture every decision making factor," Pistone says. "Human beings will make more complex decisions than a robo advisor."
Thorpe echoes that sentiment.
"The advantage of humans is they have the ability to react to new situations they have never seen before and make sense of them. Robo advisors are not trained to be creative," he says. "I think humans will make more sense of a Black Swan than a computer. Make sense of it and react to it."
A possible twist, then, to William Shakespeare: To err is robo.
Register or login for access to this item and much more
All Financial Planning content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access