Data and machine learning are transforming pricing models for the future
Pricing investment grade U.S. corporate bonds is a bit like navigating your way to a friend’s summer house for the weekend. They give you directions, and mile markers along the way, but your journey is invariably marked by traffic, potential detours, and other factors holding up your holiday. Fortunately, GPS technology and community driven platforms like Google.
Pricing investment grade U.S. corporate bonds is a bit like navigating your way to a friend’s summer house for the weekend. They give you directions, and mile markers along the way, but your journey is invariably marked by traffic, potential detours, and other factors holding up your holiday. Fortunately, GPS technology and community driven platforms like Google Maps and Waze have transformed how we travel by providing live reference data to find the fastest and easiest route to our destinations.
And where credit traders previously leveraged their relationships to determine pricing and access to liquidity, they’re beginning to benefit from similar technology to navigate markets and achieve best execution without getting caught in gridlock.
Mapping the corporate bond landscape
In many marketplaces, there are enough visible prices to help discern where the market is trading before an investor seeks to process an order or execute. This level of pre-trade transparency is relatively common in treasuries, equities, FX, or futures, but not in more bespoke markets like U.S. corporate bonds.
“It’s partly a precision issue, but I would argue it’s more than that,” says Chris Bruner, Head of U.S. Credit at Tradeweb. “It’s a timing issue, too. For example, it’s not hard to take the public TRACE data and determine what the last trade spread was. But if the last trade spread on a corporate bond was a month ago, what’s the price of that bond now? In the current environment, yields can shift 30 or 40 basis points in a matter of weeks.”
This lack of a generally agreed upon pre-trade benchmark price leaves market participants to proceed in a piecemeal manner, which makes it very difficult to do their job well. “You can’t even do an imperfect transaction cost analysis if you don’t have some sense of what the benchmark price was at the time of execution,” says Bruner.
This is why institutional credit trading is frequently described as part art and part science. But as is the case with GPS and traffic apps, Bruner and his team at Tradeweb are adding a heavier dose of data science to improve the equation for pricing in the U.S. corporate bond market.
A new model not just about more information
Leveraging its expertise in fixed income, Tradeweb has developed a sharper composite pricing model which leverages the relationships between bonds; based on factors such as liquidity, maturity, time since issuance, amongst other things. As a result, data tells you to what degree a bond resembles its “nearest neighbors,” and a clearer picture of the marketplace emerges for traders. The model moves beyond aggregation of more information to a more sophisticated, data-driven approach that empowers traders to be more precise and focused in their understanding of where a bond is trading. It blends the data on an unprecedented scale, supplying market participants with a basis to make decisions in a relative manner and far more quickly than soliciting multiple prices over the phone.
“Typical approaches used in other asset classes, such as bootstrapping curves, do not capture the dynamics of corporate bond price movements. Our data-science model is married with our domain expertise to derive the relationships between bonds and how those relationships evolve through time,” says Bruner. “With data on every corporate bond trade in the world, we take all of that information and the associations between these securities and say, ‘What’s the right price if it were going to trade again right now?’ That’s basically our approach.”
Innovation whose time has come
The components of Tradeweb’s composite model evolved along separate tracks over the past few years; however, unlocking the innovation that brings them all together is a fairly recent development.
“We currently employ TRACE, high quality data from interest rate swaps and treasury prices, alongside pricing analytics and reference data in order to capture the characteristics of all these bonds. Essentially it’s an engine that processes that information in real-time, and delivers market data on over 10,000 bonds. We’re using the sources and bond universe that we know provide value and robust outputs. In the future, additional data sources and bond coverage will be added as appropriate where tangible benefit can be demonstrated,” says Bruner.
As demand for this kind of pricing benchmark grows in the credit market, the potential uses for the model are numerous. It is already making a difference in Tradeweb’s Automated Intelligent Execution (AiEX) functionality, which executes large volumes of trades directly from an OMS using pre-programmed rules.
“Our clients who use AiEX need a price quality check to make sure a price isn’t too far out of bounds, and they’re already using our reference price as that check,” says Bruner. “That is strong confirmation we are headed in the right direction so we will continue to refine. It’s important to note we are not suggesting the model is an executable price. Rather, having a robust reference prices with broad coverage allows new workflows that didn’t exist before. In this case, combining AiEX with the real-time pricing helps fully automate trades within a tolerance, but allows clients to manually inspect those that need closer attention before executing.”
Yet another change the reference price model could bring about is in transaction cost analysis (TCA). In the near future, the model will power transaction cost analytics for Tradeweb credit clients. “We have an opportunity to offer a differentiated solution; the model delivers advanced metrics such as liquidity scores, price confidence values, and flow metrics, along with price estimates. They’ll easily be able to look at whole portfolios of trades and have an efficient route to figure out pricing, and finally be able to move away from the current trade-by-trade process. That’s the value of having a better benchmark.”
As the model continually matures, buy- and sell-side participants alike will reap the benefits. On the buy side, the model will aid in constructing portfolios, understanding the liquidity in portfolios, and making execution more efficient. On the sell side, where appetite abounds to become more effective systematic market makers, a high quality pre-trade credit input source will inevitably be embraced as a trusted, and much needed composite. The impact of this innovation hasn’t been realized in the credit market yet, but with Ai-Price from Tradeweb it will be.