Wading into the WMR Debate

The latest column in our Voice of Experience series sees Simon Wilson-Taylor, senior partner at FastFin, takes a look at one of his previous responsibilities, the WM Fix, and discuss potential shortcomings in the arguments for change, as well as solutions as to how to improve the process

After the recent paper on Benchmark fixings in FX (A comparison of FX fixing methodologies: Muhle-Karbe/Oomen) and commentary from Colin Lambert at The Full FX, I finally decided I should weigh in on the topic.

What follows is not a defence of the WM Fix by any means, but as WMR was one of the businesses that reported to me during my time at State Street I have a somewhat holistic view of some of the challenges. While the ‘academic’ analysis somewhat acknowledges some of the issues I raise here they don’t propose solutions, which I will attempt to remedy.

Colin’s commentary essentially concluded that improvements to the current benchmark should include much higher frequency data collection (from 1 per second to 200 per second) and longer trading windows customised to eliminate market impact for each currency separately.

There are four things I would point out in response:

1. The WM rate is calculated as a median, not an average, therefore, the only value in WMR collecting higher frequency data is if there is a data point that is outside the range of all data points captured during the window that is otherwise being missed. I am not certain, but WM used to capture the high/low for each second as well, which, if used in the calculation, actually captures any outliers anyway. This negates the ‘academic’ research and Colin’s suggestion to use higher frequency data collection (and saves WM a lot of money upgrading its architecture!).
2. Extending the trading window could have unintended consequences and might actually add significant risk due to increased aggressive trading, this is also a function of the calculation being a median – because as soon as you have observed one more than half of the data points in the window you know that the final rate must be within the range of rates you have collected. Mathematically it cannot be outside the range of the first half of the window. This characteristic has been exploited for a long time, with more time to play it’s possible it could get worse with a longer window?
3. The ‘smart money’ trading the WM (banks and hedge funds) also has a pretty good idea of the direction of demand, especially during periods of re-balancing or large movements in the stock and bond markets. Simply tracking the stock market moves in the MSCI Global Index gives you a pretty good start in knowing what the index managers need to do to hedge and re-balance. Add this information to any directional moves that occur during the trading window, exploiting the median characteristics in the point above, and you have significantly better than 50% chance of making money on the trade, especially on ‘big’ days. Longer calculation windows will definitely not stem this behaviour.
4. The need for ‘“internal consistency” in the data is a problem that Siren (and most others) conveniently ignore. This means that EUR/JPY must exactly equal USD/JPY X EUR/USD.  This is required so that the math works for all the index managers, custodians, actuaries etc. This has to work across 170 currencies with at least three base currencies (and for Colin’s benefit, must even work for PLN/ILS). Changing the methodology to having windows varying in some way due to the anticipated trading volumes for each currency would completely mess this up, particularly if also changing from a median to an average.

The WMR benchmark was created by a Scottish firm of actuaries to provide a reasonable industry standard reference for currency valuations, and was never intended to be used for trading purposes. That came as an unintended consequence though, once MSCI adopted the WMR rate for its valuations, due to the need for index fund managers (in particular) to track precisely to the index. It is this preoccupation with minimising tracking error that has created so many issues for the FX trading world.

If we think radically about how to change behaviour, the big things we could consider doing include:

• Get MSCI to change provider. I’m not sure that anyone has actually tried, but they can’t even consider changing provider until they get total coverage – all 170+ currencies (whether tradeable or not) in a methodology that is “better” for trading. No-one has done that yet – they are not even trying – and the problem is now much bigger than MSCI. While that remains very influential, the WMR has now been adopted by almost every index provider, actuary and custodian out there, and all would have to change provider to make this work.
• Educate all retail investors about the issues relating to FX and get them to accept the risk of a different benchmark. Not happening…
• Convert the FX market to an exchange, and have a formal end of day close which can be used for valuations, as in the equities world. Not happening (although I have heard more than one very senior buyside head of trading seriously propose this…repeatedly)

So, we are kind of stuck – or are we?

I have come to the conclusion that there are two feasible ways to move forward. As an industry we might need to do both at once and have the ideas converge:

1. Encourage a major Index manager to establish an index fund that does NOT use WMR – promoted as an “Index Plus” fund using Siren (or similar) for the major currencies only. If one manager successfully broke the mold and was able to show positive tracking gains against WMR it would essentially set a new standard that others would need to follow. Unfortunately, index-only managers do not normally have the advanced FX trading skills required to get this right, but we are inching closer to the time when this may be possible.
2. Build something new in the forwards market. I’m in danger of being accused of talking my own book here, as FastFin are working on a couple of projects in this space (ping me directly to hear more) but here goes. All the discussion around WMR is relating to spot, but almost all the trading activity of the buy-side is non-spot…and a lot is just rolling hedges. WMR’s swaps point data is probably even less fit for use in trading than its spot data, and the increased focus on the FX swaps market is another major accident (or scandal) waiting to happen. Creating a new FX swaps benchmark now, and doing it right, could set the standard for how a benchmark used for trading should be created in FX. Combining that with a new and better electronic trading environment for FX swaps would be a win for all.
3. I said two, why now three? This is where the bridge back to solving the spot market problem comes in. An FX swap benchmark has to have a spot basis underlying it. Build the swap benchmark correctly, and you have the infrastructure to back into solving the spot problem.

Gordian knot – finally untied…

All those with an interest feel free to message me directly…together we can Fix this…

Simon Wilson-Taylor

Rowayton CT USA

simon@wilson-taylor.net