UK asset manager: AI in macro trading ‘very overblown’; useful for nowcasting

Whether for quantitative investment strategies, corporate FX hedging, data extraction, or portfolio management research, there are a plethora of examples of how banks and asset managers are exploring AI to improve investment processes. Still, not everyone is convinced that the coming AI revolution is already here in the capital markets—at least for certain aspects of trading.

Joseph Davidson is the managing partner of Fulcrum Asset Management, a London-based firm that manages £6.4 billion (~$8bn) of assets. While speaking at the Future Industry Association’s International Derivatives Expo on June 19, he said that all the talk around AI in trading is “probably very overblown in terms of the actual impact on the ground as we see it” in the realm of macro investing.

“We have a fiduciary duty to clients, and we therefore cannot experiment with this stuff. … I think that’s a good thing, but it’s also quite a retarding force on the actual market,” Davidson said.

I don’t know what other people here think, but certainly speaking to my trading team, they’ve seen some fits and starts in terms of AI [in algorithmic trading] but nothing has really resonated

Joseph Davidson, Fulcrum Asset Management

In research, however, Davidson said AI is providing new information and has refined long-held investment beliefs, such as the optimal balance between equity and fixed income in a portfolio or the duration of certain investment strategies.

In a number of examples, Davidson detailed a developing idea around nowcasting, a portmanteau of “now” and “forecasting” that uses unstructured and structured data to predict possible scenarios happening at present or in the very near future. He notes a lot of economic indicators are backward-looking. Fulcrum, he said, has built out a large nowcasting research effort, made up of a team of Ph.D. econometricians. Using neural networks, the emerging answer to the question about equity and fixed income is that “it’s dynamic through time. A lot of the modeling assumes fixed perspectives in terms of risk.”

Davidson said the team has consulted with several central banks, which are now using the methodology after taking inspiration from the firm. He adds that neural networks are “a fascinating advance in terms of the way things are looked at.”

“I remember when I was doing the CFA (chartered financial analyst), I remember in the final year, we did quite a lot of work around what investment advisers in the US do: So, Jane Doe works at X. She’s got three kids. She’s going to retire in 20 years. She’s got this asset in her house. This is her savings. What should she do?”

He continued: “The answer typically was, two months, [do] the same thing. … But what the reality is, is that that should be changing all the time based on new facts coming in. … It’s kind of a Bayesian approach to how you work out what’s happening now—how you use data that is coming through to change the actual model.”

The other area of research showing results is in natural language processing, where Davidson jibed that the firm’s head of research, Juan Antolin-Diaz, is “very, very optimistic … [that] he may be able to bring about the death knell for fed watchers” by using NLP to analyze words. “Obviously, our discretionary portfolio managers don’t quite see it that way.”

Finally, Fulcrum is using AI to write “easy-ish bits” of code. “We would never let AI go loose in terms of the actual programming,” though, noting the firm doesn’t use a black box. That, he said, is an appropriate segue to the challenge of AI in the market.

“I don’t know what other people here think, but certainly speaking to my trading team, they’ve seen some fits and starts in terms of AI [in algorithmic trading] but nothing has really resonated, and that seems to be in the sort of equity space.” He notes a similar pattern around settlement and blockchain, which people have talked about “forever … but has it actually happened? Not as far as I’m aware.”

To his earlier point about Fulcrum having a fiduciary responsibility and thus “cannot experiment with this stuff”, he continued: 

“People don’t have the interest, I think, to be trailblazers in this regard. … If you create the black box in the market infrastructure, the regulator is going to be extremely nervous. And there is also this fiduciary duty. I’m sure everyone here who works in financial markets, despite what the media say about us, actually we wear it quite heavily on our shoulders.”