Volatility surged higher after touching an historic low on January 17. According to the JP Morgan Global FX Volatility Index, the low point at the start of the year was 5.18 and remained below 8.00 until March 6 when it started moving higher, eventually hitting 15.04 on March 19.
Liquid currency pairs such as EUR/USD and USD/JPY went hit one-year highs and lows within days of each other. For example, EUR/USD went from 1.0785 on February 20 up to 1.1450 on March 9, before plummeting back to 1.0688 11 days later. USD/JPY was trading at 112.10 on February 20 before crashing to 102.36 on March 9. It then climbed back to 111.23 on March 23.
Interventionist actions from major central banks including the Bank of England, the U.S. Federal Reserve and the European Central Bank have helped calm markets since then, but volatility still remains elevated versus where it was pre-Covid-19 – range bound between 8.50 and 10.00 on the JP Morgan index. The last time volatility hit such highs was in the aftermath of the financial crisis of 2008, as the European sovereign debt crisis was raging. And since then, how asset managers trade FX has changed quite dramatically.
As a result of regulation shaping how many markets trade, shrinking margins at asset managers, and a need to prove best execution, electronic trading has become much more commonplace. And algorithmic trading in particular has been a fixture of the markets for many participants. According to a Greenwich Associates report published in October, 2019, adoption of FX algos was up 25% year-on-year, with about one fifth of participants in FX markets using an algo to trade across Europe and North America. The report notes that usage is also increasing in Asia.
T.Rowe Price
Toby Baker and Brendan McMurtray
One such firm that is a user of FX algos is Baltimore-based asset manager, T.Rowe Price. With $1.12 trillion in assets under management, the firm has 136 mutual funds across its portfolio, including stock, bond, target-date, asset allocation and money-market funds.
Toby Baker, a Senior FX Trader at T.Rowe Price in London, has been trading FX for the firm for 19 years, and has been in the market for 25 years.
Brendan McMurtray has been at the asset manager for nearly 5 years working as an FX Electronic Trading Analyst.
Given the extremely volatile conditions, this was in many ways one of the first major tests for FX algos considering the very range-bound nature of trading in prior years. The Covid-19 pandemic has helped educate both Baker and McMurtray about what is working and what may need improving with FX algos to help combat these bouts of volatility in the future. We asked them a few questions about this.
In what ways have FX execution algos (and associated TCA toolsets) been influenced by the current market turbulence and increased volatility?
BM: The most significant change that we’ve seen as a result of the current market turbulence is the fact that strategies that primarily favor passive execution (i.e., market tracking or pegging strategies) have been underperforming. This is because we found that they had an asymmetric payoff in times of volatility. If these trades are going against you, they will fill too slowly, as no one on the other side is crossing the spread, and if they go in your favor, they will fill too quickly (i.e., you will get adversely selected all the way up or down) and not benefit from market movement.
On the TCA side, our approach has certainly been influenced by the market conditions. One of the main objectives we have right now is to further incorporate market-condition-related metrics, such as volatility and spread, within our framework to determine when it is the most optimal to use an algo and if so, what strategy to select. This was a long-term goal of ours that has been accelerated by the market turbulence.
T. Rowe Price has always been headquartered in the same few blocks of downtown Baltimore
In general how have your algos have been performing during this disruptive period?
TB: From a connectivity standpoint, our access to algos has been seamless, which is a big deal in the current work-from-home environment. Whether it is from Hong Kong, London, or Baltimore, our traders are able to access our algos via FX Connect (our current trading platform for algos) with a reliable lack of latency, even from home.
Regarding TCA, we did have a slight dip in overall performance versus both arrival mid and risk transfer price, but this was more or less true across the board. The biggest trend we noticed was an increase in the variance of performance (versus both arrival and risk transfer). Our results exhibited fatter tails given the market volatility.
What types of algos have you been using?
TB: As a firm, we continue to favor opportunistic and/or participation-rate-based strategies, although there has been some additional uptick in the usage of internal-only/non-displayed strategies of late. This is at the expense of more purely passive, pegging-based strategies, which we shied away from for the reasons mentioned above.
Algo usage gets significantly trickier in increased volatility
Have you experienced any difficulties in the use of your algos?
TB: One specific difficulty we had in the use of algos in this environment relates to their flexibility. Specifically, having the ability to partially fill an algo, mark the portion traded as complete, and partial off the balance would have been a great benefit to us in a few situations. This is expected to go live on FXConnect within the next couple of months, and we are looking forward to this enhancement.
What direct benefits has the use of FX algos delivered for you during this crisis?
BM: Some of the direct benefits of using algos included the ability to capture spread and achieve cost savings versus risk transfer, efficiency versus traditional voice execution with sell-side desks being dispersed, scalability (the ability to multi-task via the use of algos), and better/more transparent TCA. Most of these are the same as they’ve always been, in our view, but they’ve been perhaps magnified during these times.
What are the takeaways from the crisis in terms of shaping future FX algo development?
BM: At a high-level, pre-trade TCA (both on the bank side and our internal capabilities) needs to improve in terms of its reflection of market conditions. How volatile are the conditions? Is the market trending or is it range bound? How much volume is going through that the bank feels is accessible? Simply looking at a historical market impact versus risk curve for various strategies given the currency pair, trade size, and time-of-day does not cut it in this environment.
We’ve also noticed an increased interest amongst the buy-side for access to unique liquidity models, namely peer-to-peer. It remains to be seen whether any of these newer venue offerings will take off in any meaningful capacity in the spot FX space.
Avoiding pure passive strategies in trending conditions is a key takeaway
Do you anticipate using FX algos more or less in the future as a result of their performance during this crisis?
TB: During the course of the crisis, our algo usage definitely increased on an absolute basis, but on a relative basis, the increase in usage was much less significant. Given our increased experience and data capture around algorithmic trading in the last few months, we envision our usage to continue to increase organically. That said, algos cannot be a one-size-fits-all approach, and we will continue to use these as simply another tool in the toolset.
What lessons have been learned about when and how to deploy FX algos as a result of Covid-19?
BM: As alluded to previously, algo usage gets significantly trickier in increased volatility. Avoiding pure passive strategies in trending conditions is a key takeaway. Other strategies can have more success here, but having the ability to adapt on the fly is crucial. For example, if the market is moving against you with a low participation rate, you’ll likely need to speed up execution, but if you seem to be causing the impact, stepping back to allow liquidity to replenish may make the most sense. To this end, we’ve found that leveraging genuine real-time TCA is essential, particularly for orders with a longer time horizon. Providers who have invested heavily in these platforms and prioritized ease-of-access (e.g., via Bloomberg) have an edge here.
During the course of the crisis, our algo usage definitely increased on an absolute basis
Do new algos need to be developed in order to trade in more volatile times?
TB: One of the biggest takeaways we’ve had in terms of future algo strategies is the need for more adaptive strategies. Yes, we would like to capture spread and remain passive, for the most part, but if a trade is running away from us without our participation, we would like to become more aggressive in order to minimize slippage from arrival, and thus, risk transfer. The majority of our trades are benchmarked against arrival price and risk transfer price.
How easy / difficult is it to measure the transactions costs in more volatile times?
BM: Measuring transaction costs versus arrival price is fairly straightforward, even in volatile times. That said, we have had increased scrutiny around risk transfer prices, both from banks and our third-party TCA provider, in these conditions. If the risk transfer price is based upon a historical model, it can fail to adapt to what the true risk price would be if we took that route instead. Benchmarking is important, and both we and our sell-side and TCA partners have work to do here.
The office interior photos included in this story were taken prior to the Covid-19 pandemic and the implementation of T. Rowe Price’s work-from-home arrangements and social distancing guidelines.
