The concentration of trading around the five-minute fix window, and the widespread use of time-weighted average price (TWAP) algorithms, is creating strong momentum in rate changes throughout the window, often followed by a marked reversion, according to Pragma, a provider of algorithmic trading technology.
The effect is most pronounced at month- and quarter-end, when trading is heaviest.
Pragma research has found the pattern of trading during the five-minute window is usually set within the first minute.
Pragma first identified this trend last July, in research that identified new trading patterns that had arisen as a consequence of the widening of the fix window. Its latest report suggests the trend has been surprisingly persistent, and that the momentum and reversion pattern is markedly stronger at month- and quarter-end than on ordinary trading days.
The study states: “Conditioned on the direction of the rate move in the first minute of the fixing window, returns peak about 10 basis points higher, and revert about 6bp on quarter-ends. On month ends, the rate momentum after the first minute of the fixing window peaks at about 6bp, and reverts about 4bp.”
Guy Debelle, assistant governor of financial markets at the Reserve Bank of Australia and chairman of the Bank for International Settlements' (BIS) FX Working Group, acknowledges use of algorithms increased markedly as a result of the WM window being widened to five minutes in late-December 2014.
This helped banks track the fix price, and was also an easy way to implement the regulatory recommendation for execution separation, he tells Euromoney.
More generally, the share of algo trading has increased inside and outside of the window, says Debelle.
Banks are shifting away from handling fix orders as principal trades on the spot desk and moving toward agency-style execution, typically using TWAP algorithms that trade steadily during the five-minute fix window, according to Pragma.
The TWAP algorithm, also known as time-slicing, breaks an order into smaller trades that are executed at evenly spaced intervals, providing execution prices that are likely to be closer to the average rate during the trading period, and therefore the fix. The closer traders are to the fix, the less their tracking error against their benchmarks.
Although Pragma observed the momentum effect during the fixing window on ordinary days has weakened since the WMR fix methodology change last February, the reversion appears to be just as strong. This effect, too, is likely to decline over time, but for the trend to have persisted for as long as it has – at least a year – is unusual.
Traders typically adjust their behaviour to avoid or exploit these kinds of trends, which quickly causes them to change or break down. However, traders have less flexibility to do this when it comes to the fixing window.
David Mechner, CEO at Pragma, says: “The pattern around the fix is unusual, especially in a market as liquid as foreign exchange. The microstructure effects our research typically focuses on tend to be more granular and less robust. But the combination of factors driving people to trade around the fixing, such as regulation, bank policies and buy-side mandates, makes this pattern very strong and persistent.”
The research shows that while the market generally operates efficiently, there are issues the industry needs to work on to deliver further improvements, says Mechner.
“When we first observed the effect, we expected it to weaken more quickly than it has," he adds. "These kinds of inefficiencies don’t usually last, especially after research is published exposing them. But in this case, the constraints on behaviour are very hard to change. It is going to take time for the industry to resolve these issues.”
One possible resolution would be for the market to diversify away from the 4pm fix, benchmarking to fixes at other times of day. The pressure on firms to trade this window is felt across the industry, but some institutions, such as passive index funds, are particularly constrained in when they trade.
At the other end of the spectrum, the fix is less relevant for absolute returns investors such as hedge funds. However, banks and other funds that do trade the fix could change their behaviour, and if some do it might encourage others to follow.
Mechner says: “I think we will see a gradual move away from trading the 4pm fix, led by firms that are less constrained by tracking-error risk. I think some firms have an exaggerated perception of what the risk of trading outside the fixing window is, but some will increasingly weigh this risk against the increased execution costs."
Debelle notes the FXWG has already recommended traders think about whether they need to use the 4pm fix. However, there has been a limited migration away from it, partly because of the work involved for fund managers changing their benchmarks.
“In the end, a lot of them don’t care about price predictability or even best execution," he says. "All they care about is minimizing tracking error.”
|Guy Debelle, BIS|
Therefore, while those who do not need to trade in the window might be looking at alternatives, or might have already changed, for the large – and growing – funds industry, there is little prospect of a significant move away from the 4pm fix, says Debelle.
The findings will lead to questions about whether the widening of the fixing window has sufficiently resolved issues around the benchmark.
Mechner says: “Trading was far too disjointed in the one-minute window, so five minutes is certainly an improvement. If it were extended to 30 minutes, I would expect this pattern to completely disappear, so in that respect it would be more efficient.
"However, it would make it more difficult to get close to the benchmark, which would create its own problems for many market participants, so we need to be mindful of the knock-on effects of any changes, even if they are well-intended.”
Debelle says: “Algos have and will become more sophisticated over time because of the possibility of predictability. That was happening before and it will continue to happen.”
And Pragma itself is at the forefront of this trend, having designed an algorithm that uses the predictability of trading during the window to help traders minimize their tracking error. It might be more difficult to create algorithms to generate alpha from these patterns, because the effect is only significant intermittently, at month- and quarter-end.
Pragma's research analysed data on all normal, month-end and quarter-end days from February 15, 2015, through December 31, 2015.