By the SMU Social Media Team
Computer-driven, ‘quantitative’ investment strategies have long been deployed by hedge funds and institutional investors, but as artificial intelligence (AI) becomes increasingly widespread in the financial markets, they are becoming available to ordinary, retail investors.
Singapore has been at the forefront of the development of so-called ‘robo-advisory’ platforms in Asia. These platforms use artificial intelligence to provide financial advice and to manage investors’ portfolios with minimal human intervention. Because they are automated, and require fewer expensive humans to run, they can offer the same financial returns as traditional platforms, but without the high fees that financial advisors charge.
The Monetary Authority of Singapore (MAS) has been trying to encourage the development of these platforms, which in theory could allow retail investors access to more sophisticated investment strategies. In June 2017, MAS began a consultation on whether regulations should be relaxed to allow digital advisory platforms to operate without all of the compliance requirements of their traditional counterparts.
Associate Professor Paul Robert Griffin at the SMU School of Information Systems, says that these products are unlikely to be “a panacea for financial health”, and that investors should see them as one of many tools available to them as they try to achieve their own, individual investment objectives.
“I don’t think anyone should look for a robo-advisor product per se, but should look for the best performing financial product with an open mind for robo-advisors. Investors should be clear about their financial needs and risk tolerance and then look for the best financial products that suit their situation. When choosing any financial product the considerations are always to compare cost, performance and flexibility,” Assoc Prof Griffin says.
He further explains that robo-advisors should offer lower costs alongside flexibility, but their relative novelty means that they may lack a track record on which investors can properly assess their performance. Providers often ‘back-test’ their algorithms on historical data, but that is not necessarily a good measure of how well the product would have performed.
“Historical performance [information] will be limited,” he says, “And any back-testing which may be offered should be taken with caution.”
Investors should also be aware of their own limitations where there is no human advisor to help them to make decisions about how to react to market events.
“‘Garbage in, garbage out’ could be expensive if the investor is not really sure of what they are doing,” Assoc Prof Griffin says, referring to the old adage that no matter how sophisticated a process may be, it is only as good as the information that goes into it.
Computer-driven investment strategies have, in the past, been blindsided by market events that were not foreseen by the people who wrote the algorithms that drive them. In the early stages of the global financial crisis, several high profile “quant” funds suffered huge losses, dumping stocks despite there being no clear reason for them to do so.
Robo-advisors are a slightly different prospect, but the calculations that they make to build and balance a portfolio could also be vulnerable to “black swans”— events that are not just unlikely, but have never happened before and so are not included in the model.
“There has been significant ‘testing’ of computer-driven strategies in unusual or volatile market conditions over the past years with many lessons learnt from automated trading in investment companies,” Assoc Prof Griffin says.
“Of course, ‘unusual’ means that future conditions won’t be the same as in the past. Risks should be considered in the products but not all risks can be covered and there should be fail-safe procedures in place.”
As AI becomes increasingly advanced, more and more of the investment process can be automated. Advances in machine learning mean that automated systems should be able to become more intuitive and more customisable to individual investors’ needs. Currently retail investors really only have access to ‘off-the-shelf’ robo-advisor products, unlike the more tailored offerings that are reserved for wealthy clients or institutions.
“Whilst the management process is likely to be able to be more fully automated, the main gap in AI is on understanding the nuances of people’s emotions. Assuming that an investor can provide accurate and complete answers to investment questions then the robo-advisor can do a good job to provide relevant and good advice and products,” Assoc Prof Griffin says.
“However, if the investor is not sure of something, say their risk tolerance, but gives an answer that the robo-advisor assumes is accurate and complete, then the wrong advice and product may be given. A human may likely pick up that there is uncertainty in the voice, or maybe conflicting answers with the person’s demeanour, and can then ask more questions to ensure the correct advice.”
Griffin thinks that robo-advisors are likely to complement, rather than completely replace, human financial advisors.
“As the process of investing becomes more automated, more sophistication becomes available to more people. This means greater opportunity for those that understand the markets but more risk to those who don’t,” he says. If customers do not fully understand financial markets, or cannot easily articulate their risk tolerance then they may end up buying the wrong products through an AI-based system.
“From the perspective of human advisors, they will need to show how they compete with robo-advisors,” Assoc Prof Griffin says. “But I believe that AI is best augmenting humans rather than replacing them.”