‘Did the Model Do Anything Stupid?’

Factor funds are all about following rules, but it’s a myth that humans aren’t involved.

Illustration by II

Illustration by II

Humans don’t always vanish after they program the algorithms for factor-based funds, according to insiders.

At many asset managers, people are still central — for now — to the day-to-day running of funds, which systematically target securities with certain characteristics such as value or low volatility.

Portfolio managers add a big proportion of factor-fund returns at Northern Trust Asset Management, for example. A third of the excess returns of NTAM’s 25-year-old U.S. quality small-cap value strategy comes from the decisions of its portfolio managers, said quantitative chief Michael Hunstad.

Humans can mitigate the risks of quantitative funds, Hunstad told Institutional Investor in an interview. For one, most quant processes are based on historical data. Portfolio managers can respond more quickly to daily events such as the deadly crashes of Boeing’s 737 MAX aircraft and the subsequent grounding of the model globally. “A human can react to certain risk events, well before the event is reflected in historical financial data,” he said.

Likewise, at Robeco, fundamental managers get a final look at all purchases for the quantitative credit funds. The Rotterdam-based asset manager’s global credit team includes both quants and fundamental managers.

Patrick Houweling, a lead Robeco portfolio manager and quant researcher, said the investment process is 95 percent systematic. Human analysts contribute the final 5 percent. “The model is a best-efforts approach,” in Houweling’s view. “It’s right on average, but it’s not good with the exceptions. Humans are better at that.”


“Before we buy a name for a quant portfolio, we ask fundamental analysts to take a last look and ask, ‘Did the model do anything stupid?’ Fraud or lawsuits are more difficult to quantify.” Houweling emphasized, though, that human beings at Robeco only check for risks beyond what the algorithm already measures. There is no double-checking of the algorithm’s work.

[II Deep Dive: The Winners Will Take All in Factor Investing]

Many managers, including Northern Trust, also use humans to help with trade execution.

“Quantitative processes are sometimes known as turnover maximizers in that they tend to take the maximum amount of allowable turnover,” said Hunstad. “Humans can often be a better judge of the trade relative to the cost,” he added.

Jonathan White of Rosenberg Equities — AXA Investment Managers’ quantitative investing arm — concurs. “Some [managers] are slaves to the index, they have an ‘index mentality.’ We’ve gone in the other direction,” said White, Rosenberg’s head of client portfolio management in London.

White said the firm puts people in charge of the big picture.

“We review portfolios to make sure they have the exposure to the factors we are targeting, that the portfolios are behaving in the way we expect in this market environment, and that we don’t see unexpected pockets of risk.”

But White cautioned that there’s a balance to be struck between humans and the quant process. “You need to avoid introducing human biases into these investments. Respect the benefits of a systematic process,” he said.