The last three months have seen the foreign exchange market faced with unprecedented challenges and while the general sense is that the industry handled matter well, what was it really like in the eye of the storm? Colin Lambert talks to Melanie Cristi, head of FX market structure and client content; Matt Thomas, global head of MSET FX distribution; and Thomas Restout, global head of e-macro trading at Morgan Stanley to get a sense of how clients – and the bank itself – reacted.
Colin Lambert: Can we start by talking about client behaviour around the peak volatility in March, it’s accepted that the market functioned well, but what were the micro behaviours you saw?
Melanie Cristi: As liquidity became sparser, with wider, more volatile spreads and significantly thinner order book depth, we saw our algo volumes more than double. This was in contrast to the prolonged low-vol environment preceding the crisis that saw clients favouring risk transfer. The first clients to rely more heavily on the algos during this period were those who were already familiar with them, as higher spread costs made them a viable alternative to risk transfer. What was most interesting for us was the number of new first-time users leveraging algos to help them navigate these challenging liquidity conditions, rather than exposing themselves to market highs/lows by simply executing via risk transfer.
CL: What type of strategies did clients go for?
MC: We saw higher urgency trading strategies that reduce exposure to risk and volatility in most demand, this was from clients who understood they would have market impact but wanted to get done. For example, our most popular strategy, Seeker, which dynamically adjusts to seek liquidity across varying market conditions, spiked in volume by 180%. There are four urgency settings on seeker – low, medium, high, and ultra-high. Prior to the peak in volatility, High/Ultra high strategies accounted for 33% of total Seeker flow, during peak Covid vol, they accounted for 48%. Seeker does not execute along a set trajectory or end time, and we found that clients were generally happy to leave their orders in the hands of the algo during this time.
In sum, we saw a spike in higher urgency strategies, both from clients who sought to reduce duration risk, but also as a result of a willingness from clients typically executing via risk transfer to increase some duration risk to reduce market impact.
CL: Fusion, your suite of algorithmic strategies, involves elements of bespoke liquidity pools and client-to-client matching?
MC: It does. Fusion Edge is a component of all of our Fusion strategies for spot FX that enables passive client orders to cross with a subset of streaming clients. Fusion Mid is also a component of all of our strategies that seeks opposing interest at mid before crossing the spread. In higher volatility environments, these mechanisms represent significant spread capture, and many of our clients recognised the value our client-to-client matching brought to their execution. In March and April, the total client-to-client matching flow as a proportion of our overall algo business reached nearly 50%.
Thomas Restout: We use similar strategies for our own hedging of our principal business, and the fact that our Fusion client-to-client matching is growing so rapidly speaks to the benefits of this type of internalisation.
CL: So that’s “true” internalisation, I find people can sometimes be loose with their definition?
MC: Internalisation certainly requires a definition around what you include. We are talking about flow that is in addition to what is filled against our principal book, this is flow that is matched against a subset of clients, into liquidity pools that we have curated to help reduce market impact for clients. While clients understood in these conditions there would be market impact, when they accessed these pools via our smart order router (SOR), that impact was reduced beyond what they may have expected based on the liquidity conditions they saw in front of them.
Matt Thomas: The bottom line was that many clients were simply not willing to pay the higher spread for risk transfer unless they really had to. Those clients that had some discretion and were willing to embrace opportunistic strategies found they saw nice improvements in execution outcomes. And to Melanie’s point, clients could easily see the improvement they were receiving, the reduced market impact, and how many client-to-client fills they were getting, in real time on our QSI React tool. This is empowering because it means increased trust in the algos which in turn compounds usage. It really helps when something is that clear cut and is backed up with the empirical analysis we provided.
In addition, as the algo was working, clients could see in real time on React where their child orders were getting filled and whether they were collecting or paying spread. It was interesting to see clients actually start to change the duration of their orders. As they approached the end, they would lengthen duration as they became more comfortable with the results. The thinking was “this is working really well for me so I’ll let it go a little longer and not rush it”. That’s a great sign of trust in the analytics and the algos.
This also provides an opportunity to combine the traders’ intuition and skills with the right analytical tools to make a real difference to the execution quality. Overall, I would say client-to-client matching was a big story of that period of volatility. Thanks to the improved execution outcomes, clients could see the impact it brought and feel confident in the strategy.
CL: You mentioned React there, your analysis tool, can you outline what that delivers?
MT: Sure, React is our pre-trade and real time/in-flight TCA tool. Our desk can work with clients to visualise their fills, passive/aggressive ratios, fills by venue, and various different performance metrics. It also allows clients to review potential strategy changes in real time with live benchmarks.
CL: It has been quite a wait for algos to really step into the limelight as far as FX markets are concerned, and it’s interesting to me that perhaps their value was highlighted in conditions that some thought would act against them.
MC: Often clients simply have a rate in their mind when they want to execute – and the value of an algo is whether or not it beats it. We were often beating that and that, to me, is a catalyst for people who were otherwise sceptical to give algos a chance.
CL: Will it sustain?
MC: After the SNB de-pegging in 2015, we saw a similar change in market structure – uncertainty expressed in the form of thinner liquidity, higher volatility, which I think led to the spike in mainstream usage of algos. When volatility eventually reverted lower, some clients went back to more traditional methods of execution, and thus, some providers stopped investing in their businesses, but we didn’t. We continued to invest, and it paid off in this event. These events are often catalysts for market structure change and I think when we look back at this period, we will see it as one of those catalysts, and that means that yes, interest in algos will be sustained.
The market will also support algorithmic providers who invest in the quality of their data. During March and April, clients found they could trust the data we were providing, as the improvement in execution performance supported it. This was a very public stress test and while the concept of “trust” with regards to data in FX is still met with scepticism, often rightly so, our clients have built trust in us based on their experiences.
MT: Events like this certainly help build trust. In a low volatility environment, it is harder to see the quantitative value. At the beginning of the peak stress period traders were able to look at the QSI visualisations and see that they were starting to provide out of range parameters in terms of low liquidity and high cost. We were moving the bounds of previous parameters and this was easy for the clients to see and digest. These were not just small improvements, it was a really interesting story to tell clients and helped build the crucial trust factor in the data and analytics that QSI React provides.
TR: It’s very much about the numbers. The empowering factor for me is how we are able to use these analytics and data to provide expected costs on different amounts. We can show that using Seeker was estimated to be three times cheaper than the risk transfer price on 100 million units in the majors, for example.
Observations of how people use these strategies are also interesting. At first they can see what one standard deviation of risk in a low vol environment would do to the cost of execution and that creates expectations for the clients. They get used to the fact that occasionally it may cost them more to execute but the balance is very much in their favour overall. As they get used to this risk by using the algo and experiencing what it can deliver, then using it becomes part of their nature. There might still be questions raised if the outcome isn’t as good as hoped, but there is less underlying drama in the whole process. That is how you grow and sustain the use of these strategies.
CL: And this, presumably, helps the principal business?
TR: It has made our business better in general. We believe the identity of Morgan Stanley in FX is how we have brought the principal and algo businesses together. We have always seen the value in sharing the same IP and tech stack between the principal and algo businesses. When we on the principal desk think about our algo business, we think of the risk services we can provide to clients. We leverage the same technology in our principal business which has allowed us to step up and provide risk transfer services to those clients that want it.
To us, the most important thing is the amount of liquidity we can source, that we know how each of our sources operate, such as the client-to-client pools. It is important that we thoroughly understand the properties of these pools and how they are operated, so that we can be confident, for example, that there are no externalisers – information leakers – in it.
Another factor for us sharing the IP and stack is that we can add inventory which means tighter pricing and pricing in larger amounts if required – and the risk from this is not just cleared into the market immediately, the algo allows us to tap into a number of unique sources of liquidity. It gives us confidence in holding this larger risk for longer.
MT: That is a real strength of Morgan Stanley, we have a bench of strategists always looking to refine the algos and SORs – we target a lot of resources at this because the results are shared across the whole business.
MC: I also think that innovation is, to a degree, a product of necessity, because we understand we are not the biggest FX shop on the street. We have to be more intelligent in how we create and manage the liquidity we are offering clients. I think we are ahead of our time on this because pre-Covid-19, in a period of sustained low volatility, some clients were less concerned how their liquidity was curated. Even though it was always how we sought to differentiate, it just didn’t matter to them. Now, we increasingly have more clients starting to show interest in the microstructure surrounding our pools, the benefits of client-to-client matching in spot FX and working with us on bespoke ways to interact with it.
CL: A reasonably regular refrain I hear from execution desks involves a reluctance to use algos because they are a threat to their job. Your story seems to refute that, but how involved are the clients in the process?
MT: They have control over the process, it’s as simple as that. We talk to clients about what is happening in the market, we share the QSI visualisations and QSI React data and discuss the type of parameters the clients want to put on their execution. While the order is being executed QSI React delivers real-time analysis so we can talk to them about what is happening to the their order and how the market is absorbing it. As the analytics change so too does the dialogue – that has been a big change over the past year, it’s not just about showing clients the analysis, it’s also about discussing it with them, as it happens.
Clients also want more data in machine readable format, they are plugging in more APIs, and with that comes the need for quality data. We are working on various ways Morgan Stanley can provide normalised access to markets and functions by embedding QSI data and analytics in the clients’ workflow. Data, analytics, and market access is something we have delivered for some time in our equities business, and it means the clients focus on what they are good at and we provide the infrastructure to help them execute more efficiently. In FX now we are getting more requests from clients to expose some of our QSI queries bit by bit, so I see us moving more towards those execution and data service business models.
MC: We try to present data in a matter-of-fact way to help empower clients in their own execution decision making process. For example, we wanted to help clients determine which factors contributed most to algo performance, and we broke it down into the ones they may have some control or discretion over, like duration, size and liquidity sourcing, versus the ones they do not have control over such as volatility and market volume.
CL: I sense there is still a nervousness out there though.
MC: There is, and we see a market repricing as a result; virtually all pairs have reverted back to pre-Covid levels in terms of volumes but the cost of trading is still higher.
TR: Expectations of a return to pre-Covid levels in the short term are unrealistic in my view and Vols are a good 30-40% higher still. I think, though, this is probably the best time for clients to talk about what their strategies are doing and how they are executing in the FX market. Current conditions, with an underlying nervousness meaning trading costs are a little higher, but not seriously high, providing enough savings versus measured risk. This is an ideal time to have a serious conversation about your execution strategies and the liquidity you are accessing.