“I would still prefer a hybrid model where you have a quantitative approach combined with a human in charge,” said Matthias Weber, partner at ifund Services, a provider of fund due diligence.
Ifund screens around 2,000 active and 1,000 passive funds using an algorithm to come up with only five “most promising candidates” for its clients. However, analysts still have a role in the whole process, scrutinising the results at different stages, according to Weber.
In addition, part of that algorithm replicates the analysts’ behaviour. “[Our analysts] do not work so much on selecting funds, [but] also on improving the algorithm.”
Echoing Weber, Alex Ypsilanti, CEO and co-founder of Quantifeed, which provides digital solutions to wealth management firms in Asia, said that robotising fund selection is fairly nascent.
“[At] many of the institutions we work with, 90% of fund selection is done manually,” he said during the panel discussion.
Steven Seow, principal and head of wealth management for Asia investments at Mercer, also stressed the importance of human judgment.
Besides quantitative measures, Mercer has a qualitative review of funds, he said. “Part of that process is looking into the [eyes of the fund manager] and questioning them in their investment strategies and how they work together as a team.”
Mercer’s analysts ask how fund managers generate their ideas, construct their portfolios, implement their strategies and how the whole firm is managed, he said.
Risks in robotising fund selection
There is a danger that fund selectors will be buying the same funds if fund selection process is fully robotised, according to JB Beckett, UK lead of the Association of Professional Fund Investors (APFI), who also spoke during the discussion.
Ifund’s Weber acknowledged that risk. “If you are selecting Russian markets you always have the same top 10 [funds],” as long as the metrics and managers don’t change, he said.
Individual clients, however, can affect the outcome of the process.
“As soon as the clients tell us to tweak the selection, it can change the algorithm,” Weber said, adding that the algorithm could assign different weights to different metrics, such as investment processes, asset classes, manager gender and others.
Another potential risk is that if fund managers were to learn more about the algorithm, they could manipulate the metrics of their funds to get into the fund selectors’ system, APFI’s Beckett said.
An example of such manipulation is “window dressing”, whereby a fund manager changes the asset mix of their portfolio during the last quarter of the year to change the fund’s performance, according to Beckett.
“There is potential for fund managers to exploit the fact that many fund buyers today actually buy on standard time periods, such as the one-three-five years. These are statistically very weak reasons to buy funds,” he said.
The solution for fund selectors is due diligence and understanding what happens underneath the fund, he said. In particular, they need to understand the drivers of returns – whether they are factor-based or economic-based – and to get away from standard statistics such as outperformance and volatility, he added.
