How does automation in fixed income differ from other asset classes in practice?
The accelerating growth of automation in fixed income markets presents unique challenges and opportunities compared to other asset classes. Equities markets – which have long benefited from electronic trading and automation – are highly transparent, largely trading on order books and on an agency basis. Fixed income, by comparison, has historically been far more dependent on risk warehousing, is more opaque, and uses request-for-quote (RFQ) as the dominant execution protocol.
These factors, paired with voice trading remaining prevalent, have resulted in slower adoption of automation. However, in recent years, the rise of passive, low-cost/low-margin investment products, and shrinking bid/ask spreads make the demand for efficiency in fixed income as significant as other markets. This has led to considerable innovation in fixed income automation and rapid growth in adoption, particularly in areas such as price making/liquidity provision and order execution.
A key catalyst of this progress has been the rise of predictive pricing models – such as MarketAxess’ AI powered pricing tool, CP+. This solution delivers the same functionality as a ‘top-of-book’ price feed in equities, providing a reliable benchmark of the best available price that enables market participants to automate their decision-making processes.
How has buy-side sentiment shifted towards automated solutions recently?
There is a growing sense of confidence and optimism towards automated solutions on the buy-side with demand increasing year-on-year. Large asset managers and private banks have been the quickest adopters, with the majority of client orders now executed using automation. Until recently, however, their focus has been liquidity taking via RFQ workflows using our rules-based automation solution, Auto-X RFQ.
We are now seeing increasing appetite for more sophisticated automation workflows that include smart order routing, liquidity provision and aggregation.
How far could trading automation realistically go? Is there a point at which it becomes unfeasible or is the sky the limit?
I don’t think it’s really a question of feasibility anymore – it’s a question of how fast the transition will happen.
At MarketAxess, we believe technology can have a positive impact on the entire trade lifecycle by allowing traders to focus on spotting opportunity, solving problems and managing risk. In that sense, I do feel ‘the sky is the limit’ when it comes to trading automation, but I don’t think that is necessarily a single protocol or workflow.
We often talk about the ‘last 50%’ to refer to voice trading still happening in US and EU IG credit. I think it is inevitable that electronification will shrink that number over the next few years, but I also think workflows are likely to be more nuanced and may require different approaches. For example, in some cases, this might involve using automation to break up blocks of liquid bonds and trade throughout the day, taking advantage of the increased turnover generated by the ETF ecosystem. In other instances, it may be about smartly sweeping up natural liquidity in the form of dealer axes or buy-side matching.
Markets outside of IG are also experiencing a shift. For example, in 2024, we saw an 8% increase in automated emerging market trading volumes compared to 2023, showing increased adoption. This momentum is set to continue, particularly as investors look beyond traditional bond markets for new growth opportunities.
What should be front of mind for the industry when adopting increasingly automated solutions?
Automation creates incredible potential for operational scale but that scale carries increased risks that need to be carefully managed. Helping clients do this is front and centre in our product design approach. We focus on providing automated and manual risk controls, observability, transparency and process auditability. This ensures the user is always in control, even during periods of high volatility.
Quality of data is also paramount: automation outcomes are only as good as the quality and reliability of the data you use for decision making. Machine learning can play an important role here in making sense of the vast amount of fragmented data in fixed income markets, but this needs to be underpinned by a large, well-cleansed data set and robust model validation. These models take time and experience to get right.
I think the potential for increased efficiency in fixed income is immense, but it’s all about choosing the right technology and data partners to help you on the journey.