However, the pandemic has been a reminder of the continuing importance of human discretion in FX trading, with many on the buy-side relying on their relationships with clients at a time of unprecedented volatility and uncertainty.
The goal of this article is to assess and highlight a number of themes, such as changes to average daily volume (ADV) over the years along the lines of geography, FX instruments, and electronic versus voice trading. Some of the topics that have dominated the industry at the peak of the pandemic will be discussed, such as the growing realization of automation in FX markets, algo adoption, transaction cost analysis (TCA), and the enduring value of human discretion as well as the skill set of the modern trader.
State of the market
The U.K. has recorded the highest ADV by a clear margin each year from 2009 to 2020. In October 2020, the U.K. reported ADV of US$2.6 trillion, down 10% from the same month in 2019 while many of the other reporting centers remained constant without drastic change (Figure 1). From a growth perspective, Hong Kong, Canada, and Singapore experienced the most exceptional development, with a growth rate of over 100% over the past decade.
In the last couple of years, Singapore has gained traction with an influx of banks, nonbank liquidity providers, and trading venues setting up pricing engines in Singapore’s data centers, fueled by the partnership with the Monetary Authority of Singapore (MAS) to expand their business in the financial hub of Asia—a trend that will drive up the trading activity in the region.
Since 2009, FX swaps ADV has grown by 90%, outright forwards by 85%, and spot by 23%. As an essential instrument for emerging market currencies, nondeliverable forward (NDF) trading continues evolving as market players strive to improve e-trading infrastructure and regulatory support. Further electronification and growing ADV necessitate automated trading of NDFs, as executable streaming is becoming mainstream and banks are offering algos to gain competitive edge. Trading at swap execution facilities (SEFs) experienced steady growth in the past three years, and off-SEF trading started to pick up activity levels with new trading venues joining the market.
Speaking of the broad consensus of further electronification across FX markets, Figure 3 demonstrates the percentage of trading volume in the U.K. traded via various electronic methods of transacting, including electronic brokering systems, single-dealer trading systems, and multidealer trading systems. Overall, 46% of total transacted volume was electronically traded in October 2020. While e-trading has grown as a percentage of total trading, the growth is slow and inconsistent for non-spot products. Nonetheless, 69% of spot trading was transacted electronically in October 2020, a new high for spot over the past decade.
From a technology perspective, the industry coped particularly well with the unprecedented trading environment brought about by the COVID-19 pandemic. Buy-side firms are constantly searching for the next innovative trading solution, while streaming and algos offered by liquidity providers enable the sophisticated traders to achieve efficiency. The need for transparency and best execution requirements will drive more buy-side adoption of TCA and other trading analytics tools.
Algos prove their value
FX trading volume rose dramatically in March and April 2020, and many buy-side traders relied on banks’ algos and outsourced easy tasks to banks so they could focus on more complicated trades. The debate as to whether it is best to buy or build when it comes to algos has subsided, as many buy-side funds have reported to be using more bank algos—partly due to the falling costs. Algo trading has proved to offer efficiency and performance optimization to traders while creating a revenue stream for the banks. The wider spreads have let algos to reject more orders, and the demand for reliable bank algos increased. As the market became more jittery in March 2020, the algos began to get more passive by resisting orders. There is a place for more aggressive algos, but some buy-side traders would prefer to use smarter, more passive algos. Firms that were already using algos expanded their usage this year. The number of firms using algorithmic trading strategies has increased and stayed high, representing a high proportion of trades.
The range of users has broadened as well, now ranging from hedge funds and asset managers to corporations, smaller managers, and even pension funds. Algo usage has also played a role in optimizing liquidity for the growing trading volume in emerging market currencies beyond the G10. Even in less liquid NDF markets, algo trading is becoming more mainstream. There has been a surge in liquidity providers wanting to trade NDFs via algos, and buy-side traders can see the cost savings by using execution algos for nondeliverable currencies comparing to trading against traditional risk transfer prices. Smaller fund managers will benefit more from this development as they outsource execution to banks’ algos and then leave it for a day without sourcing new technology internally. However, the NDF market is less fragmented than the G-10 deliverable currencies; hence, accessing multiple liquidity pools can be difficult. The challenges of NDF algos are the wide range of liquidity levels for each currency product and various liquidity concentrations for certain dates.
TCA comes into its own
TCA for FX has already started to gain traction as a means of assessing performance and ensuring compliance. Increasingly, TCA tools are being used to mine the information prior to the trade. These pre-trade analytics give an indication as to what kind of spread to expect, and this is becoming more central to FX trading. TCA gives a clear picture of what works best. Some firms utilize TCA to conduct venue analysis and remove venues with which they do not want to interact. TCA is also being used to evaluate algos in FX, especially with algos being much harder to read and predict during times of turbulence.
The unprecedented volatility since March 2020 has intensified buy-side usage of TCA tools, as 10 years’ worth of data from an unchanged market regime may be less useful than three months’ worth of data from a different market regime. Spot spreads are around five to 10 times bigger than swaps, and buy-side firms are unlikely to find this data from a decade ago. Undoubtedly, TCA is a critical component for the overall workflow as it becomes part of the pre-trade analytics traders use to decide on their strategy. Consequently, TCA vendors have ramped up their development to deliver the next generation of tools to cope with the higher demand.
Automation has enabled the new normal
The changes brought about by the COVID-19 pandemic have seen FX markets follow the example of other markets, such as equities, that have for a long time now incorporated automation into their processes, as many firms see it as way to achieve a more efficient risk transfer. The widespread adoption of either cloud-based or vendor-hosted solutions has been particularly important for traders’ ability to ensure a high degree of business continuity as the industry traded in the office for the kitchen table or a home study at short notice.
With many firms in major financial centers keeping their staff at home for the duration of 2020 and likely much of 2021, a more dispersed trading floor opens opportunities for technology vendors. While there were concerns that spending would slow as projects were put on hold, investment to enable automation has increased, with solutions designed with remote working in mind at the top of the agenda. Automation has also positioned firms with staff working remotely to manage issues around compliance and audit trails better than the more traditional methods, such as voice trading, would have permitted. Many firms have made it clear they would much prefer their buy-side traders to deal electronically rather than over the phone.
Wide spreads have posed technological challenges, and they were particularly pronounced in emerging markets. Liquidity providers refreshed their prices much more often than they usually would to protect themselves, and the display of the latest prices on execution management systems over traders’ home internet caused issues. To cope with challenges like these, buy-side traders have reported building proprietary exception-rule-based logic to automate trades. This means that from the moment an order comes in through the point at which it is executed, the order must pass through a series of hoops by satisfying the safety criteria. Checks are likely to consider market conditions, spreads, volume, price, and limits. Pre-validation tables specify the limits and sizes for certain currency pairs at certain points in the day. Most of the orders are rejected at the beginning because the logic is yet to be refined, but over time, the trade process takes shape.
Many trading systems reported strong performance, and the trading environment held solidly with deep liquidity during the peak of the market stress. This underlines how important automation is to the industry’s ability to weather unprecedented periods of volatility. Trading platforms have been rolling out new features and enhancements over the years to streamline the workflow. For instance, FXGO, FXall, and FX Connect all have been enhancing their auto-execution processes in recent years. The process is getting more granular and customizable, allowing traders to create different goals by componentizing every functionality. The key to well-designed automation is to provide flexibility, allowing clients to automate a piece of the flow and reduce a process from four mouse-clicks to one or two.
While rules-based automation for trade execution can increase efficiency, questions arise about its resilience within the context of the market stress and unprecedented trading environment.
Human intelligence in the era of automation
Many experts see the role of the FX trader in the future developing into something resembling a pilot in a cockpit. Even with a range of tools available, the individual is still very much in control of how those tools are deployed.
The pandemic, by creating a stress-test environment, has undoubtedly underlined the ability of modern technologies to realize business agility. However, while automation increases safety and efficiency, it is still best used as a machine in the hands of a trader rather than as an autonomous agent itself. The automation of braking in cars, for example, does not render the role of the human driver redundant. Automation is useful for certain segments of the trading process, but the industry is far from a time when the entire life cycle of a trade can be automated. There are always going to be elements that are best left to the trader’s judgment. For example, NDFs and other derivative markets and unusual currency pairs are examples of areas that will require significant human discretion for the foreseeable future.
It is true that the skill set required of traders is changing, with the demand for quantitative skills in coding, computer science, and mathematics rising. But it is important to remember the human side to trading, as much of the communication and relationships between counterparties cannot be meaningfully done by machines.