FX Fixes: Forget the Weights, Check the Length

All dealer-to-client trading in the FX market comes with a degree of tension around the outcomes for each party, and always has, but perhaps nowhere is it more pronounced than in trading at benchmark fixes – especially the London 4pm Fix. Here, large trades are executed routinely, with a surge in activity at the month-end, with, historically, better outcomes for dealers than clients.

If a client insists on a large order being executed within, for example, the five-minute window, the market impact is outsized – as witnessed by such an event in USD/SEK a few years back when the market was moved more than 10 big figures – if the order is hedged by the dealer ahead of the window, the client does not receive a “true” reflection of the transaction cost in the benchmark rate.

Previous literature by Johannes Muhle-Karbe and Roel Oomen has highlighted the broader conflict between the two sides, and the pair have now published another paper, A Comparison of FX Fixing Methodologies, which demonstrates that the length of the calculation window is the primary driver of dealer and client outcomes, rather than the various methodologies deployed by Bloomberg, Siren FX and WM. “A fixing window that is too narrow leads to excessive client transaction costs while a window that is too wide makes the hedging of fixing risk commercially unviable for the dealer,” the paper states. “This trade-off should be balanced by the fix administrators when they settle on their preferred methodology.”

The paper studies the methodologies of the B-Fix, Siren and WM benchmarks, noting that innovation in this area has grown in recent years, however it observes that many of the changes that have occurred – Bloomberg quietly changed its methodology earlier this year, which raised a few eyebrows among market participants who were unaware – are the result of a compromise between the views of clients and dealer, rather than based upon broad empirical research. To fill this gap, it presents a framework to evaluate the various methodologies and their impact on client transaction costs and dealer risk/reward.

The framework uses the model proposed by the authors, along with Benjamin Weber, earlier this year, for optimal hedging of fixing flows by dealers, thus enabling the authors to distinguish between the impact of a different methodology and window length. It finds that a wider fixing window reduces client costs, but notes “this needs to be balanced against an increased risk for the dealer in managing fix exposures and expected losses for smaller transactions.”

Notably, the paper highlights how the WM design that works well for smaller trades, is sub-optimal for larger tickets or less liquid markets. “It is therefore on the client to make an informed choice whether a fix is the right trading mechanism for their purposes and on the dealer to make an expert judgment about how to best manage their fixing risk given the size of exposure in relation to prevailing liquidity conditions and fixing methodology,” the authors state, very much reflecting the guidance in The FX Global Code and Financial Stability Board literature on using a benchmark fix.

Source: Oomen, Roel C.A. and Muhle-Karbe, Johannes, A comparison of FX fixing methodologies

In the section that empirically compares the methodologies, the authors indirectly highlight a growing challenge around the WM Fix, namely that it is very hard to replicate because it is a median, rather than mean, calculation (for more on this see The Last Look…16 June 2024). This has resulted in what some dealers have described as “strange” prints from WM, not least last month-end in the NZD. Notwithstanding this, in the absence of highlight volatile price action, the authors suggest their framework provides an “accurate approximation” of the median.

The paper presents different outcomes using various window lengths and hedging styles (i.e. in the window only, or ahead and during the window). It notes that hedging in the pre-fix window on WM can reduce transaction costs (as measured by arrival price) by 1bp, but also that this can be detrimental to the client, especially if 20 minutes is sufficient to execute the entire order. It also, interestingly, measures the impact of each methodology across one, five and 20-minute windows – hence the finding that the methodology is of secondary importance to window length.

That said, the paper highlights that with a weighting towards the end of the window, as used by Siren and, before it changed earlier this year, B-Fix, the benefit is skewed towards higher dealer profits and higher client execution costs. “This is intuitive,” the authors observe. “With concentrated weights, the dealer’s optimal hedging trajectory is more concentrated as well, and the impact they generate feeds into the fixing calculation with higher weights.”

The paper finds that a longer window length is harmful to the dealer on smaller trades, it also observes that an excessively long window – it does not suggest a time horizon – would lead dealers to exit the business. Alongside the finding that a short window would increase client costs, the authors observe “Both of these scenarios are harmful to the client.”

The paper also investigates what would happen if, for example, the Siren methodology was deployed, namely the different arrival price would reduce the projected 6.1bp saving from using Siren, to 4bps. The authors note, however, that if dealers are hedging ahead of the WM fix (as they often are), then the empirical data may be more representative of what one could expected with a longer Siren fix. The paper uses the model to measure an execution starting 15 minutes before the WM window and finds that the projected savings are 3bp, versus a projected 3.6bp.

Ultimately, this research highlights an argument that has been made in these pages and those of Profit & Loss Squawkbox, for some time – the length of the fixing window is important. As academics (albeit with “real world” experience), the paper has to consider all aspects of the topic, however FX fixings are largely associated with large trades. The major dealers all deploy netting processes to reduce the trading requirements at the Fix, which means the resulting trades – especially at month-end – are predominantly large.

If this is the case, and with the empirical evidence growing that larger trades need a longer window, then surely it is past the time when users of FX benchmark fixes reconsider their options and finally ask – and answer – the pressing question; ‘Is the benchmark I am using really fit for purpose?’

The full paper can be accessed here