Article: The Swiss franc blip: time FX leverage limits moved from flat percentage to risk-based?

06 March 2015 | Jon Skinner

The dust has settled from the initial fall out from retail brokerages going out of business as a result of the Swiss National Bank's abrupt removal of the CHF vs EUR peg several weeks ago.  

So I think it's time to ask a deeper question - does it make sense to use risk-based margins in retail FX proprietary trading?   After a quick review I can't see a good reason why not.

The initial response

For a short period it was common to see reports of retail brokerages going out of business or suddenly changing margin levels.  FXCM in the US, Alpari in the UK and Global Brokers NZ of New Zealand are just a few of the names in severe difficulties or out of business altogether.   Some brokers (including Lukascopy and FXPro are already increasing minimum margins to 10% or 10:1 leverage instead of the prior 2% or 50:1 leverage minimum (according to this article from Leaprate).   In particular broker IG is causing a stir by increasing it's margins on pegged or controlled currencies and stop loss positions according to this Reuters UK article - indicatively up to the 30% in some cases.

So far regulators are monitoring the situation and the NFA has temporarily increased the US minimum FX margin from 2% to 5% with CFTC and FCA among others keeping a watching brief.  As background there was a prior effort in 2010 to increase minimum margin to 10% which didn't not go through and recently there was even some noise that 2% was over the top (see e.g. this WSJ article (subs required)).  

Why did the brokers go out of business?  The speed of the move compared to client margin levels

Whilst minimum margins are agreed at the discretion of the client and the broker and this could in theory encompass a risk-based assessment of both the client and the market, in practice they can be as low as 2% and it may be that this has become the norm in a "race to the bottom" of brokers.  

If the SNB had "guided" the market on the move in advance there may have been more days and more margin calls or position close outs to absorb it.  As it was they did not and the one-day appreciation of CHF vs EUR and USD of 18+% (source: oanda.com) any participant with significant CHF exposure as a proportion of the total FX exposure in their account would have burned quickly through their margin.  For the good of the market, brokers are on the hook to make up the difference to the extent additional intraday margin calls to clients were not met.   (So FX brokers are a bit like FCMs in guaranteeing the performance of their clients to the market).   

Free market Idealogues will no doubt seize the moment to trumpet the futility of central bank intervention.  Putting that on one side what is clearly empirically evident is that it doesn't always work and that central bank resources are ultimately limited and cannot be fully relied upon.   

Was this a systemic risk problem or a historic market move?  No and no
On systemic risk, only 4% of the FX market is retail.  If a few end-users and a few retail brokers unaffiliated to systemically important institutions lose their shirt this is very bad news for those involved but likely not a systemic event unless a systemically important institution has been compromised.  Perhaps there'll be further news on the impact on major banks and on CCPs over the coming days?  

Whilst the speed and size of the move was surprising it is relatively common to see significant shifts in FX values.  In fact according to OANDA.COM historic FX rates, 9 times in the past 5 years there has been a 1-day CHF vs EUR move greater than 2%.   And 26 times in the past 5 years there has been a 2-day CHF vs EUR move greater than 2%.

Given no tangible value how a flat percentage work well?

No doubt some are comparing 2% FX minimum margins to the US reg T minimum of 50% but that misses the point that - unlike securities and commodities which are underpinned by tangible assets / defined cash flows - FX is determined only by supply and demand effect.  Therefore there is no tangible or par value on which to base pricing - only the current price itself.   This implies that a flat percentage margin rate is on shaky ground (just as a % of notional approach for OTC derivatives or futures is shaky ground).

A first quantitative indication can be gained from the historic market data.  9 out of 5-years is 0.72% or 99.28% coverage.  26 out of 5 years is 2.08% or 97.92% coverage.   So if there is signfiicant risk it takes more than one day to close out it looks like 2% is not enough.  And this doesn't even consider the coverage of stress.
Is risk-based margin relevant?  Possibly
Risk-based margin techniques are used in OTC derivatives IM methods partly because percentages of notionals are a shaky foundation.  It seems the same is true of FX.
The first step might be to thoroughly review the existing 2% / 5% minimum levels using backtesting techniques common in the derivatives industry.  I hope NFA and CFTC are already looking at this.  FX historic market data is readily available.
The next step might be to design a new risk-based margin methodology to substitute for crude flat minimum approaches
  • Design it: a bit of a "dark art" but a well versed quant should be able to rustle up a starter-for-10 hVaR method, stress method and a liquidity add-on method fairly rapidly  
  • Backtest it: play the margin results against a sample of FX portfolios to demonstrate exceedances are within the bounds of knowledge
  • Commercialize it: either have each broker develop competitively or have a small number of vendors build and license their methods
  • Build it: harness suitable IT staff or vendors.
  • Approve it: regulators would want to approve and set minimums of a different kind e.g. confidence intervals, margin period of risk (MPOR) etc.
This is some significant work but it's a walk in the park compared with other regulatory changes that are going through.
What should be the FX close out period / MPOR?   2 days at least
FX is probably the most liquid market of all.  However, even here a 1-day MPOR looks aggressive.  In the recent CHF blip, were all participants able to continue to trade CHF FX pairs in any size or did liquidity get disturbed by the market move?   Even if not, liquidity disturbance seems more likely at least for less liquid / emerging market currencies.  Could the clients of the brokers who got into difficulties close out their positions on the first day?  I don't know but it would seem prudent to allow another day on the holding period.  (This is essentially the same reason why EU regs have even futures MPOR as 2-days).   A longer MPOR also makes the losses covered less sensitive to the speed of the move or whether there is market guidance or not from the central bank.
What else is needed?   Full margin model analysis, design and implementation - ideally by a third party
A deeper analysis would be required to create and maintain a fully fledged margin model.  Examples would include historic market data analysis across all currencys, design of the approach to stress, clarifying the approach to risk correlations between currencies and backtesting the methodology to demonstrate upfront and periodically it's adequacy.
Aren't retail clients and brokers going to struggle with the mathematics?  Not if there are handy tools provided
They need to be able to press a button and calculate the margin and see a graph of margin percentages for typical trades not understand how the margin is calculated.  Some handy web-based tools might be enough to overcome this one.  Frankly most people in the brokers / banks will also struggle with the mathematics too.  For many therefore it would be useful if they could outsource the creation and implementation of a more sophisticated risk-based FX margin model to a third-party provider.  If the market settled on a single or small number of margin model / calculation providers, this would yield considerable transparency and economies of scale for the industry.
Why would risk-based be better?  Self-calibration
Setting a flat rate even with some back testing at a point in time to show that it is adequate is naturally a hostage to fortune.  Only through unusual / crisis events like the Swiss franc blip do we discover problems.  A risk-based margin method in skilled provider's hands along with periodic re-calibration based on historic market data in an explicit and transparent manner to regulators and participants would offer the hope of more naturally adjusting margins to current volatility trends whilst maintaining stress coverage at all times.
Conclusion: I say "why not?"   
Comments below please.
Bill: Risk also covered the same topic considering how VaR performed for many firms when this event occurred, story here (subs required).