Managing ‘data risk’ for bond traders

Published on 27 January 2020

If data is mismanaged on the trading desk, it can trigger individual trading errors or even flash crashes. Jon Williams, head of fixed income at Refinitiv, identifies key stages of risk management for data including getting the right amount of information; assessing whether it will be consumed by human or machine, and how it is cross-referenced against the benchmark for the investment idea.

Dan Barnes Welcome to Trader TV – your insight into the trading climate for professional investors. I’m Dan Barnes. Bond traders must have a keen understanding of policy, range and reliability of data when using it to support price and liquidity formation, or they may introduce risk into portfolios. Jon Williams, head of fixed income at Refinitiv, and he’s here to help us navigate a fixed income data universe. Jon, welcome to Trader TV.

Jon Williams Thank you very much for having me. I’m very happy to be here.

Dan Barnes What are the different levels of data that bond traders typically need?

Jon Williams Yeah, that’s actually a really good place to start. Fixed income traders are usually fairly expansive users of data. And I think the first question that you need to think about when you’re talking about data for a fixed income trading desk, whether buy-side or sell-side, is, how is that data going to be consumed in the immediate? Meaning, is it going to be consumed by a human being or is it going to be consumed by a machine? There are different types of data, different types of pricing data that affirm both buy-side and sell-side will take in. Are you looking at trade generated data? Are you looking at indicative pricing data? Are we focusing on evaluated data for instruments that are truly illiquid, where basically we’re using engines and we’re using models to basically create prices that represent our expectations based on where similar instruments are trading in the marketplace?

Dan Barnes If the data is consumed by a human trader rather than by a machine, what bearing does that have on the data itself?

Jon Williams So, when we start talking about how data is consumed on the trading desk, it’s certainly consumed by the physical trader, but then it’s also consumed by a broad swathe of systems and applications that the trader uses, to execute his or her business every day. So, the most obvious ones would be things like pricing engines. So, I’m taking data in. I’m using that as fuel to basically generate the next series, the next iteration of prices upon which I’m going to execute. But then traders also have, again, a wide array of applications and systems that they use to manage their business day in and day out.

Dan Barnes How do you see this consumption feeding into the wider investment process?

Jon Williams Each and every one of these trades has to occur within the context of the overall portfolio that they’re managing. So, the systems that they’re utilizing, from an order management perspective, carry out a number of different functions, before they ever go through and generate the actual order. So there needs to be an ability to understand the current context, the current overall position of the portfolio. Virtually every fixed income portfolio manager measures his or her performance relative to a benchmark or a bogey, some index. And so, you need to understand, ‘how is my portfolio positioned vis a vis my benchmark?’ ‘Am I happy with that relative positioning on my long duration or on my short duration or am I overly exposed in a particular industry?’ And that is kind of the first step in this overall trade idea generation process. These order management systems then will apply a series of models to whatever the potential trade idea is, to make it understand; ‘if I trade this particular bond by swapping one for the other, or if I just add a particular security to the portfolio, how does it then change the makeup of the portfolio?’ ‘How am I now positioned vis a vis my benchmark?’ And just as importantly, ‘how is my portfolio positioned relative to what our own internal models are expecting for future interest rate performance?’ Once that’s carried out, the final check is the compliance check. So, every one of these order management systems employs a fairly robust capability to ensure that the portfolio manager doesn’t inadvertently violate any compliance issues. Some of these compliance issues are fairly rigid and they may well be prescriptive from a regulatory perspective. A US portfolio manager certainly has to think about things like 40 Act issues.

Dan Barnes We can see how data feeds into the process and there is risk within the process that potentially something may go wrong. Is there such a thing as data risk?

Jon Williams We’ve seen actually fairly meaningful events in markets that were ultimately driven by bad data. “The 2010 Flash Crash”, everyone remembers the “US Flash Crash” that basically came about because of a series of misreported executions that triggered a number of automated processes and then led to a fairly significant sell-off in a meaningful, potentially meaningfully catastrophic, kind of loss of value across a number of different portfolios.  We talk about risks to data, but there’s always risk to data. I think that’s the reason, getting back to some of our earlier questions, that’s why the ability to access such a broad swathe of data is so critical to portfolio managers. To understand the sources of the data, to understand the types of prices that they’re taking in, to be able to gauge the relative reliability, accuracy, and the relative utility of an individual series of prices helps them as much as possible to mitigate those risks.

Dan Barnes John, thank you very much.

Jon Williams No, thank you for having me. It was great to be here.

Dan Barnes I’d like to thank Jon Williams and of course you for watching. To catch our monthly reports on other markets, or to subscribe to our newsletter, go to TraderTV.NET or ETFTV.NET.