Systematic tools gain favor in fixed income

In 2012, fixed-income trading venue Tradeweb began developing its use-case for automated trading. In partnership with a small group of clients, the team developed a guiding principle they called the 80/20 rule, which states that 80% of actions—determining who is placed in competition, receiving quotes, checking prices against references—should take up 20% of a user’s time.

This rule underpinned the first iteration of AiEX, Tradeweb’s automated execution tool that has since been integrated into 26 of Tradeweb’s products. The question, says Charlie Campbell-Johnston, managing director, AiEX and workflow solutions, was what consistent actions were users performing that could be codified to affect the same behaviors without the need for clicking?

But 12 years ago, users met Tradeweb, its question, and its answer with resistance.

“When we started the automation initiative, there was a certain amount of reluctance,” Campbell-Johnston says. “Our client base is pretty unforgiving of things that don’t generate results in a short period of time.”

But technology would come for the change-averse fixed-income traders anyway, as electronification remains a pertinent trend in the space. And while Tradeweb has strictly focused on execution automation within AiEX thus far, it’s enabled the rise of more advanced technology uses for fixed income, such as algo trading and quantitative, or systematic, strategies.

The rise of systematic trading

Four years ago, London-based investment bank Jefferies began rebuilding Jefferies Fixed Income in a quantitative context. Much of the franchise is focused on its own credit trading, but its quantitative approach extends to Jefferies’s position as a market-maker within various securities, including fixed-income ETFs, corporate bond trading, and portfolio trading.

The business is staffed by a blend of technologists, quants, and traders, who have built an infrastructure that couldn’t have existed 10 years ago, says Alexis Besse, head of international fixed-income quantitative strategies at Jefferies. But with the influx of so much data, the choice to become more systematic was not really a choice but a requirement for survival.

Automated trading actually creates trading activity. Because I can automate and trade in this certain way, it creates the opportunity to access markets more efficiently, more systematically—opportunities that didn’t exist before

Charlie Campbell-Johnston, Tradeweb

The age of big data has yielded many things, and in finance, it has led to automation and increased trading activity. Higher activity, in turn, begets more data. The cycle has yet to break. Tradeweb’s Campbell-Johnston agrees.

“Automated trading actually creates trading activity. Because I can automate and trade in this certain way, it creates the opportunity to access markets more efficiently, more systematically—opportunities that didn’t exist before,” he says.

Those changes led clients to ask Jefferies for faster pricing, electronic connectivity, and systematic trading strategies—and the bank decided to create a resilient platform that could meet those demands from scratch.

“Given the number of securities we trade across bonds and ETFs, there are hundreds of millions of data points that we need to consume every day. A lot of it might just be noise where it’s hard to extract valuable signals,” Besse says. “So, we invested a lot here at Jefferies to build this large infrastructure to be able to generate meaningful insights.”

Such meaningful insights include assessing where the market is going, which corners of the market people care about—and predicting what they’ll care about tomorrow. To do so, extensive machine learning to deal with non-linearity was required.

“The thing about automated trading, in general, is that the system cannot fail. And it needs to alert you when something goes wrong,” Besse says.

As he explains, in fixed income, there are many possible scenarios at any given moment, but they are not always in play. For example, a particular sentiment on a security is changing, and the bank wants to capture exactly the nature of the shift. Factors that seem to be influencing the change may not be particularly relevant; the bank needs to prove that.

That meant Jefferies needed an efficient way to test its hypotheses; the firm also decided that sample back-testing, Monte Carlos, and linear regressions wouldn’t cut it.

“At its core, what we’re trying to build are predictive models. And we test the quality of the predictions on things that are not being seen by the model. We do this day in, day out,” Besse says. “This layer of complexity reflects the world of fixed income. I would love to use simple techniques if I could, but it just doesn’t work.”

While the thinking is novel for this particular market, Jefferies isn’t alone in its pursuit of quantitative applications for it. In September, asset manager AllianceBernstein called systematic fixed income a “breakthrough in bond investing” in a blog post that followed its initial whitepaper from March 2023. Like Jefferies, it has hinged its perspective on discovering factors that have predictive power to repeatably find securities with the best risk-adjusted return potential.

A look at liquidity  

AllianceBernstein’s whitepaper identified liquidity as the number one issue affecting fixed-income managers’ ability to outperform or find alpha. While neither entity has asserted that quant trading is the only direction of travel for this market, both the asset manager and Besse contend that bringing more of these systematic approaches to fixed income offers one solution to this issue.

Fixed-income participants want increased liquidity and greater market transparency—but not in every instance.

“In our market, we want more transparency, we want more churn—more trading activity—but there are still cases when you’re working on something more sensitive or something story-driven,” Besse says. “Then, you probably want to work with trusted dealers and not have this information available to the rest of the market straight away. So not even the subject of transparency is completely clear-cut.”

Tradeweb’s Campbell-Johnston says that just having the ability to trade more systematically in fixed income creates a valuable new skill set, another tool in the belt that alters a person’s behavior. And those behavioral changes, he says, create fractions of basis points that add up over time.

“Ironically, where there is more meat on the bone, where the spreads are wider and liquidity is less, the accepted wisdom would be that the benefit of more automation is on the more liquid end of the spectrum,” he says. “But actually, if you can prove that you can reduce your transaction cost when the spreads are wider, the cost-benefit is significantly higher.”