COVID-19 Market Volatility and SIMM Model Governance – The Implications for phase 5 of UMR
The market volatility of recent weeks in response to the COVID 19 virus becoming a pandemic in every metric but name, will bring with it many significant knock-on effects and challenges for risk managers. These will include counterparty risk issues, operational problems as systems struggle to cope with increased volumes and product control pressure as liquidity issues and stressed markets make establishing accurate marks more challenging. This piece is focused on one specific one, which, like COVID 19, is novel for risk managers in stressed markets; SIMM model governance.
For those using the SIMM model, including phase 5 entities who will start in the SIMM regime come September 2020, the COVID 19 market volatility will likely generate significant model governance issues for each SIMM calculation, in particular for portfolios with exposure to asset classes that have experienced the greatest volatility. It's possible that come September, a phase 5 SIMM IM amount at “go-live” will actually be set by a backtesting / benchmarking model and not the current calibration of the SIMM model.
Why is this the case?
The regulatory wording in almost all of the major jurisdictions, enshrines the requirement to maintain Initial Margin (IM) levels at a 99th percent confidence level for a 10 day close out. To ensure the SIMM model is performing at least as indicated above, dealers need a comprehensive SIMM model governance framework to assess the IM level calculated for each portfolio. However, SIMM levels are only as good as the freshness of its parameters and this is where the problems lie. SIMM’s re-calibration process only occurs on an annual basis, and even then the refreshed parameters are only released into production several months afterwards. The recalibration of SIMM parameters that may improve the model’s performance currently is a long way off. Not to mention, with such a significant time lag between assessments of appropriate parameters versus their final release, even a re-calibration may not solve the problem. Compare this with the London Clearing House, which in some instances uses self-correcting models (filtered historical simulation), reviews risk parameters quarterly and can release parameter changes immediately, or in the case of extreme volatility, can review parameters on an ad-hoc basis, changing them on the fly.
Backtesting and governance
This is where the IM model governance kicks in – if SIMM levels are less than the 99th percentile requirement, regulation requires that dealers monitor for these cases, identifying problems and correcting them. This identification occurs through routine backtesting and benchmarking with this output also used to establish appropriate strategies to mitigate the shortfall. Possible mitigating strategies have been outlined by ISDA and include a range of options from ‘brute force’ portfolio multipliers, to posting block amounts as permanent extra margin, or to more refined and best case ‘risk factor’ level intervention.
If you are sensitive to how much IM you might post to maintain your risk positions, you need a stake in any conversation regarding SIMM shortfalls. That stake can only be provided by well specified and transparent backtesting and benchmarking output. How will you know if there is a more efficient solution to your SIMM shortfall problem if you have no detailed benchmarking output to demonstrate it? If dealers are faced with a large number of SIMM portfolio failures in a highly volatile market, will they have the risk resources to put in the extra effort to find the most efficient shortfall mitigation strategy by the timelines laid out in the regulation? Or will they just apply a brute force multiplier, with all the excess (and unnecessary) IM that may bring? While solutions have to be bilaterally agreed, for the best funded dealers, they may let counterparty risk requirements and their own limited risk resources dictate a process that then rides roughshod over funding cost considerations.
Either way, come mid-September, as portfolios build up at go-live, some phase 5s may be drawn very quickly into unanticipated conversations with their phase 1 - 4 dealers, with expectations on when thresholds would be breached proving to be wholly incorrect as backtesting data flashes red for phase 5 SIMM IM levels.
Andy Shaw, Links Risk Ltd
The author has validated every risk model in use at the LCH, has validated numerous SIMM portfolios for various fintechs and tier 1 banking clients and was formerly an Expert Witness on risk models for JP Morgan.