Ken Monahan

Greenwich survey shows impact of transparency issues on FX algo adoption

June 2023 in Algo Tech

A paper published by Greenwich Associates before the Covid-19 volatility event indicated very low adoption rates of FX algos among the market participants surveyed. Even though the crisis may have permanently changed participant behaviours, the market structure issues which had held back many from algo execution may still persist.

In the paper, Digitization Delayed: Why Algos Aren’t More Popular in FX, Greenwich Associates found that prior to the crisis, just 37% of market participants were using FX algos and even then were only using algos to execute 22% of their trading volumes. This is despite virtually all FX market participants now trading electronically, according to the findings. FX algo execution was less than half that of the uptake in DMA, smart order routing or algorithmic execution by the equities market, while well over half of the firms surveyed did not use algorithmic execution at all.

However, the current crisis and its impact on FX volatility, combined with many participants now having to trade from home, led to record numbers turning to algos to execute their trades in March and April.

Ken Monahan, Vice President, Market Structure and Technology at Greenwich Associates, says that since writing the research paper, the crisis may have already accelerated changes to the way participants use FX algos and related TCA toolsets. “In talking with industry insiders about this paper people tell me, ‘I feel kind of bad for you. That was a really insightful paper, you made a lot of great points. But I can tell you that since you published it, our algo volumes have exploded,’” he explains. “I think that people have had a hard time getting trades done, or getting them done at a level they like, so they’ve been using algos far more often.

Keep in mind also that a lot of algos have their own benchmarks, so there are often embedded metrics which can help you judge performance in a way that is hard to do with a human trader. I think the demands for transparency will always be there and it will always be a struggle to meet them, but the crisis pushed people to try a lot of new things. Many of them will not go back to the old ways.”

Based on 97 respondents. Source: Greenwich Associates 2019 Market Structure and Trading Technology Study

Measuring success

Monahan explains that the genesis of the paper was a conversation he had with several FX practitioners from the buy side and the sell side. In the course of that conversation, he says he was surprised by how few said that they had used algos at all and even from among the users, what a low proportion of their flow they executed algorithmically. He adds: “Since Greenwich is a pan-asset class research house, I knew how different these patterns were from other asset classes and, given how electronic FX is, I wanted to understand more about it. So in that sense, the top level findings were not surprising. I had just concretized what I had heard anecdotally and was able to see that effect much more clearly with a robust dataset.

With regard to the reasons for this, I knew the data issue was present but I did not expect to see what I saw with regard to transaction cost analysis (TCA) and benchmarking.”
The issue with TCA, as highlighted by the paper, is the low rate of adoption in its use among FX participants with less than half (47%) saying that they used TCA at all. As highlighted by the report, this is important because one of factors driving algo adoption in equities and other asset classes is a virtuous cycle between TCA and algo development. According to Monahan, one of the spurs to the adoption of new technology is the ability to measure its success. In addition, the absence of a reporting requirement in FX has hampered the uptake of TCA.
The report also found around 40% of FX flow executed by algos used TWAP algos, compared to a fairly even split between liquidity seeking algos, at 24%, and passive algos, at 22%, while VWAP algos accounted for only 9% of flow. Monahan explains that the lack of a reporting requirement in FX means that it is extremely difficult, if not impossible, to calculate a VWAP, since the volume traded at any given price cannot be known with certainty. As the popularity of TWAP vs. VWAP demonstrates, the scarcity of data is also a significant hinderance, he adds. Price, however, is available, making TWAP the closest many can get to a benchmark in FX.


Overcoming the data hurdle

In fact, algos with self-referential benchmarks are nearly twice as popular in FX as in equities. This is because benchmarking is a significant challenge in FX, and not only because of a lack of data, with over half the respondents (57%) having said they have no specific benchmark to compare trading performance to. Monahan explains: “The fact that many firms had no benchmarks at all really surprised me, as did the relative reliance on the dealers and platforms themselves for TCA. What was even more interesting was that this varied significantly by sub-segment of the survey participants. Asset managers were more likely to have a benchmark and a TCA process than hedge funds. This may come from a philosophical difference, where hedge funds think of the execution as part of the alpha, and the asset managers view it as a cost to be minimized.” Improvements in TCA offerings and the analytical tools available will be extremely significant if they can gain traction, but it’s a bit like squaring the circle, warns Monahan. From the conversations he has had with the industry, participants raise the issue of access to data constantly but, at the same time, he believes there is also a lot of resistance to making their data available. “Still, there is a ton of demand for TCA, both from a Best Ex perspective and also from a trading optimization perspective,” he adds. “As the sophistication and the results continue to improve outcomes in other asset classes the demand will continue to rise. To meet that demand, firms are going to be creative about how they get FX data and they’ll start creating better economics for those who own it.”
The report also noted that in the pre-crisis markets, the most popular use of algos was for working large orders over time, but interestingly, they were also being frequently used for large orders that need to be executed quickly. “Given the relatively fragmented nature of FX markets, this makes a lot of sense. Managing a large order over time across multiple venues is absolutely something you would want a machine to do,” Monahan says.

Traders were also marginally more comfortable using them in stable conditions rather than volatile ones, but notably the difference between the two is very small, a finding that is significant in light of the surge in FX algo use reported in March’s volatile trading period. However, the report concluded on an optimistic note with the split between participants who believed their FX algo use will increase over time (44%), compared to just 5% who envisaged their use would decrease while 51% expected their current usage levels to remain the same.  Monahan also observed that the reason for the plateau in e-trading adoption is that the last redoubt of voice execution continues to be large orders, which he notes is precisely the order set that algos dominate.