Morningstar embraces artificial intelligence for fund ratings

Morningstar Inc., in an effort to increase the number of funds and ETFs it rates, rolled out new quantitative ratings Monday that will cover more than 10,000 funds and ETFs.

The new quantitative ratings use machine learning and artificial intelligence to learn from the ratings produced by Morningstar’s analysts. Essentially, the program will adjust its ratings to learn from mistakes — and incorporate improvements — made by its human counterparts.

The move comes, in part, because Morningstar’s analysts are swamped by the sheer number of funds and exchange-traded funds. The population now stands at 7,950 funds and 1,853 ETFs, according to the Investment Company Institute. Morningstar currently covers 1,800 funds and ETFs; with the addition of the quantitative ratings, it will cover more than 10,000 funds.

“They get grief because they are only covering funds that don’t need to be covered,” said David Snowball, publisher of The Mutual Fund Observer. “They tend to cover large old funds whose records are reasonably well-known without close analysis.”

Morningstar has been working to increase its coverage, but acknowledges that its staff has been swamped.

“We’ve been expanding coverage of funds through our analyst team for the past two years, and because of increasing demand for more coverage in the U.S., we realized that at some point we can’t hire 1,000 more analysts,” said Timothy Strauts, director of quantitative research at Morningstar. “The model has been in testing the past four years, and we’re comfortable launching it to the world.”

Unlike human-based fund analytics, the quantitative analysis won’t involve text. Nor, for obvious reasons, will it include manager interviews as part of its ratings. Investors will see a rating of gold, silver or bronze with a superscript “q” — for quantitative — for ratings produced by machine. “It does a very good job of replicating the analyst ratings,” Mr. Strauts said. The new ratings will look back five years on a rolling basis.

Unlike other fund ratings, such as CFRA, Morningstar’s ratings don’t include an analysis of the fund’s underlying holdings.

“Our equity research has a view of the world as to what’s a good stock and a bad stock,” Mr. Strauts said. “But there are lots of other views that are totally valid. In the fund world, managers might do things differently, and we don’t want to penalize them because their view of stock world is different from ours.”

The new analytics won’t ease controversies over whether Morningstar’s ratings have predictive value. A recent review of the ratings’ success found that they were a good starting point for investors, in that they pointed them toward lower-cost funds that are likelier to outperform in the future, albeit by often slender margins for stock funds.

“It’s a bit better than nothing, and every bit is worth celebrating,” Mr. Snowball said. “From our perspective, it tends to miss the point that people often misuse funds, because most people overestimate their tolerance for risk.” Some indication of the fact that people could find certain funds extremely hard to hold in a downturn might help, he said.

Related Topics: Morningstar Inc.