Stretched margins: growing pains for Simm

The standard initial margin model could become more strained as its use expands in 2017

Tall order: the model will soon face a new set of challenges

So far, those behind the standard initial margin model (Simm) have plenty to be proud of. There were certainly jitters in the weeks and days preceding the launch of initial margin rules for non-cleared derivatives in the US, Canada and Japan on September 1, but the Simm's performance in the past three months with the 21 banks caught in the first phase of the rules has gone relatively smoothly.

To its supporters, the model designed by a forum organised by the International Swaps and Derivatives Association is a demonstration of how the industry can come out with efficient ways to make markets safer through consensus. The addition of more risk factors requested by regulators since September has provided a further test of the Simm governance mechanism. Indications so far are that it passed the test overall: four additional risk factors are being formulated and are slated to go online in March. Only one will miss a US regulator-imposed deadline of January, and so could be subject to the more punitive standard approach schedule for calculating initial margin (IM) for just a couple of months.

However, the model will soon face a new set of challenges. Regulatory demands are increasing and top-tier European banks will begin complying with the non-cleared margin rules from January 2017. From September, IM requirements will begin to extend to a much wider range of market participants. This will create a dilemma for Simm governance: sticking to the goal of staying transparent and harmonised on the one hand, while on the other being stretched to cover a much larger group of more diverse firms beyond the model's core governance group.

"Everything we are talking about and saying is such a tall order. We are talking about basically cobbling together extra risk factors, and then having massive backtests, which are just revealing that no matter what you do, people will breach. And you and I will breach for different reasons. And you are going to have one kind of net-to-gross-ratio factor and I'm going to add in a different one. I don't think that one can arrive at a homogenous solution. And banks are going to have to dedicate teams to these disputes, which impinges on front-office resources. So maybe it's worth it to just use the schedule – it may cost you twice as much in terms of capital, but the friction for your business is drastically reduced," says one senior quantitative expert at a firm caught in the first phase of the non-cleared margin rules.

We are talking about basically cobbling together extra risk factors, and then having massive backtests, which are just revealing that no matter what you do, people will breach
Senior quantitative expert

In many ways, the future challenges for the Simm lie in its very nature as a crowd-sourced project. Getting harmonised results across banks is a major trial and coming up with solutions can be tricky, as consensus has to be baked into every aspect of the Simm. Smaller firms will be phased in over the next four years. The Simm provides a margin amount that is less punitive than the standardised fall-back grid devised by the Basel Committee and International Organization of Securities Commissions (Iosco). Some say the grid can yield results as much as 15 times higher, although others say they have calculated it will only double amounts.

Click here to enlarge, or tap image on app

Additions and changes to the Simm are made via its Governance Committee, which consists of the forum, including all firms subject to margin rules, and the executive committee, which makes final decisions and features firms on Isda's board. Those working on the model are very aware of how it will be stretched in the months to come.

Indeed, the Simm was not designed to be static. It has built into it the capacity for additional risk factors, annual recalibrations and changes to methodology, based on factors such as reported shortfall and reconciliation issues, and changes in financial markets and technology.

"On the one hand we are trying to meet regulator demands, but we are also trying to keep it as simple as possible so we don't discriminate against smaller firms. Simm is trying to walk a balance between a complex model and something simple everyone can use," says one senior counterparty risk manager at another phase-one bank, who sits on the Isda Simm Governance Committee.

We're hoping the European regulators try to strike a balance with us
Senior counterparty risk manager

The trouble is that minor hiccups could become more serious as the model is expanded and extended to more users, while also becoming more difficult to solve as the group of governing firms grows. In addition, the framework for dealing with disagreements remains somewhat fuzzy, which could become more disruptive as the number of disagreements tick up when the overall number of trades captured by rules increases.

To top all of that, the European Commission's report on November 23, setting out the likely contents of a review of the European Market Infrastructure Regulation (Emir) next year, suggests Europe may follow the US in requiring regulatory approval for initial margin models. This could mean another draft of hoops to jump through, and the potential for differences between the European Union and US, adding a whole new layer of complexity.

"We're hoping the European regulators try to strike a balance with us. For the approval process, we don't know how that is going to play out and we're hoping we don't get inconsistent standards – number one, because that would be very problematic for us – and that European regulators try to work with us to get the balance right to keep the Simm usable for small firms and clients as well," says the senior counterparty risk manager.

Daily margin calls

So far, the main issues on daily margin calls have involved some differentiation in the calculation of risk sensitivities, but more substantial disagreements on trade populations and product mapping – as was predicted when rules were first launched. For example, the senior quantitative analyst at a phase-one bank has seen a reconciliation report showing a difference of more than 30% on the trade count with one counterparty. He says the list of breaks such as this was getting longer every week as trading picks up.

Sources say confusion about trade populations can be caused by trades being completed at the end of the day, where one firm catches them in the daily margin call and another doesn't until the next day. More seriously, differences can be caused by disagreements over what the margin rules cover. The Blazer platform provided by AcadiaSoft finds mismatches in a reconciliation process before margin calls are made to minimise disputes, while their project partner, TriOptima, isolates the mismatch. However, these mismatches still require human intervention to get the trades matched correctly.

The problem is likely to get worse when European rules go online next year, because the regulations themselves differ. For example, commodity forwards are not in scope in the US, while certain commodity derivatives based on spot prices are in Europe.

"If you take that example, an EU bank not under US regulation will call for margin on a portfolio of trades, including certain commodity forwards, and it will be trying to reconcile its call from a counterparty in the US on a portfolio that does not include any commodity forwards. So they will do two different calculations, I expect, once EU rules come into force in January, and then other rules from other jurisdictions will have slightly different definitions of what a derivative is, and so this issue will become more obvious and more problematic," says one risk management adviser.

New risk factors

Banks are also grappling with variability in how risk sensitivities are calculated. So far, the differences between banks have been limited, sources say, but as more counterparties and a wider range of risk factors are added, this could also get worse. As firms beyond the major global banks are phased in, the data, systems and models for calculating sensitivities will vary. Also, new risk factors requested by regulators involve data that, historically, differs significantly between banks.

"It's not difficult for us, technically, to make these additions to the model – it's just more of the same. But I think it is ropier stuff, because it is begging for more disputes," says the senior quantitative expert at the phase-one bank.

US regulators ensured Simm models would be approved in time for the September 1 deadline, but on the basis that certain conditions would be met, including on model validation and governance, as well as the addition of certain risk factors. One such factor, dividend swaps, was to be in place by January 2017, followed by credit correlation, inflation and cross-currency basis risk factors by March. Requests for changes were made directly to the Simm Governance Committee by regulators, as well as in conversations with firms, which were then fed up to the committee.

"We discussed in the committee what the priorities for different firms were, in terms of the enhancements requested by the US regulators, and concluded which ones to focus on, as well as the timeline for their introduction," says Panagiotis Koumantanos, director in the risk and capital group at Isda.

"In adding the new products to the Isda Simm, we need to gather the relevant data to calibrate the risk factors and then there's a testing phase to validate the design. If it passes, we go ahead. If not, we tweak the initial design. So we have defined the enhancements and we do have initial design proposals, and it is now up to us to calibrate and test," he adds.

According to Isda, the industry is on track to have all four in place by March, meaning that dividend swaps could be subject to the standard grid for a limited amount of time. Regulators have also asked that the Simm methodology and documentation be modified to allow for add-ons and multipliers, which some say means dividend swaps could be subject to a less punitive add-on rather than the grid. Isda says everything is expected to be ready to start the testing and calibration phase in mid-January, after which documentation can be prepared for submittal to regulators by the end of that month.

Scott O'Malia

"US prudential regulators have requested product-specific enhancements to the methodology and set tight deadlines for achieving this. These additions, along with future enhancements and clarifications, are being worked on by the Isda governance forum," says Isda chief executive Scott O'Malia.

However, the addition of cross-currency swaps is an oft-cited example of a risk factor for which sensitivities may vary widely between banks. Differences may be difficult to resolve on a bilateral basis because of the highly sensitive nature of the information used in calculations.

"The Simm model is simple to apply once you have the sensitivities, but the sensitivities depend on the way they are calculated," says Claudio Albanese, founder of software firm Global Valuations, which works in partnership with another firm, Riskcare, on IMEX, an interbank utility that standardises Simm sensitivity inputs. Albanese co-wrote an article for Risk.net earlier this year, supporting a move to standardised risk factors.

"Banks don't share sensitivities with each other. If I share the sensitivity for cross-currency swaps with another bank, I am disclosing my basis and how I price the cross-currency swap, and so the other bank can undercut me. That is considered to be highly private information. Banks are sharing through the AcadiaSoft hub their IM on an aggregate basis, so you cannot reverse-engineer the basis swap, but they are not going to share the sensitivities themselves," he adds.

Cross-currency swaps are also presented as an example of how Simm can get complicated and costly, but where a single move by regulators could wipe out months of expensive work. Currently, margin rules exclude principal exchanges and catch only the cross-currency basis. Implementing this exemption has been a highly complicated process, according to numerous companies, and some firms fear the exemption could eventually be removed because it makes bilateral swaps less expensive than cleared swaps, which do not benefit from the exemption.

"It's theoretically unsound, it clutters up your code – which is something you don't want to think about, but which matters when you want to maintain a business – and, frankly, it could disappear in the end because it is so suspect," says the senior quantitative expert at the first bank.

No definition of dispute

If, as some predict, disagreements increase, then attention will soon turn to the fact there is neither a formal definition of a daily IM dispute nor a well-defined set of guidelines about what to do in response to a dispute.

"The problem is the industry hasn't created a buffer around margin to say ‘this is a dispute, this is not a dispute'. Technically, it's probably wrong for you to pay $95 if I asked for $100, even if I don't care, because one could, as a lawyer, argue that because you only paid $95, it's a dispute. So I think there's some concern over that – not the difference between $100 and $95, but that the industry hasn't defined a threshold," says the risk management adviser.

He says there had been some industry discussion about formally defining disputes as those where the difference is both more than $20 million and over 20%, but there was no industry appetite to go forward with the proposal.

Guidance given by the Isda Simm governance document says users are expected to inform Isda promptly of reconciliation issues that result in a dispute of at least 10 days in duration. It also defines a dispute as a "genuine margin reconciliation problem" where the full called-for margin is not being paid, and which is not due to a difference arising from mismatching trade portfolios or incomparable sensitivities.

Without a formal definition of dispute, dealing with any differences, large or small, is tricky, as no-one knows what should be formally addressed and what should just be smoothed over in favour of easier trading. The same problem arises in the world of backtesting the Simm, say sources, as no one has agreed on what constitutes backtesting and how to address differences in the outcomes between banks.

Backtesting

According to a third phase-one firm, many banks were expecting US regulators to scrutinise the Simm when providing model approval and were surprised when a lot more attention was paid to the backtesting process instead. Even so, this has not meant that guidance on how to go about backtesting has been particularly clear. US regulators have diverged in how they ask backtesting to be carried out, so there is no set way of going about it; neither is there guidance on what to do with the results or in the case of disputes.

Margin rules are applied to firms by their relevant regulator, with the CFTC covering swap dealers and major swaps participants not captured under the prudential regulators. The National Futures Association (NFA) is carrying out model approval on behalf of the CFTC, and multiple sources say the NFA had singularly asked banks to backtest the Simm by comparing their outcomes against 10 days of profit and loss (P&L). By contrast, prudential regulators have not provided specific guidelines on how ongoing testing should be carried out. However, others say the NFA is now working with the industry to come out with a less stringent approach.

"NFA recently completed its initial margin model reviews of CFTC-covered SDs (swap dealers) subject to the initial September 1, 2016 compliance date. As part of NFA's review process, these firms submitted policies and procedures for backtesting, backtesting results and their procedures for ongoing monitoring. NFA is now working with the CFTC and prudential regulators to develop its monitoring procedures, and during this process, NFA is receptive to understanding the industry's concerns regarding backtesting," says Kristen Scaletta, a spokesperson for the NFA in Chicago.

Regulators say firms have to conduct backtesting, but they don't say how… nowhere does it say what happens if you fail backtesting
Another senior quantitative expert

Beyond that, firms do not have any standard for model validation and backtesting, and so have gone about these processes in different ways.

"Regulators say firms have to conduct backtesting, but they don't say how. They might just say ongoing monitoring, but nowhere does it say what happens if you fail backtesting – they just say you need to take appropriate action, but it doesn't say what that is, and how and when you are supposed to take that appropriate action," says another senior quantitative expert at a fourth phase-one firm.

"As a result, firms have tried to comply with the requirement for backtesting and monitoring, but not necessarily in a uniform way," he adds.

According to the Simm governance document, when exceptions occur, firms should conduct bilateral discussions to validate exceptions and agree on causes and solutions. Any issue that requires remediation has to be reported to the Isda Simm Governance Committee.

"We have a process where users of the Simm are expected to provide feedback to the committee on any shortfalls versus the 10-day 99% risk that should be covered. If we see any systematic non-coverage, then it will be discussed at this forum, where we will decide if we need to amend the Simm," says Isda's Koumantanos.

Firms can carry out backtesting using several techniques, according to Naresh Malhotra, a director at KPMG in New York, who says the most common approaches include: periodic backtesting, where Simm results are compared to historical and model-generated P&L; continuous backtesting, where daily P&L and scaled Simm results are compared; and non-linear or delta-neutral backtesting, where first-order sensitivities are removed to focus on the impact of non-linearities.

"These are the three backtesting methods that we have seen the most of. Banks have started or are planning on doing either a subset or all three of them," he says.

However, some fear these differences will lead to mounting disputes about whether risk is sufficiently covered or not. They are advocating a standard approach to backtesting, based on what banks are currently asked to do to validate the value-at-risk models used to determine capital requirements for market risk, for which banks have to compare daily P&L from all trading positions with the corresponding one-day total VAR model estimates. The proposal stops short of prescribing how to address exceptions, for which there are no set guidelines.

"I see little benefit in having a firm-specific approach to Simm backtesting for ongoing monitoring. It is not as if we are trying to create a barrier to entry; we want to ensure less disruption occurs in the practice of monitoring. The more different monitoring is between firms, the more disputes there are, which is not conducive to good market behaviour. We are recommending a more standardised approach to backtesting, but which falls short of actually prescribing the application of multipliers or defining any consequence of failing backtesting," says the fourth firm's senior quantitative expert.

Critics counter that a standard approach to backtesting would mean firms will have to invest in another complicated infrastructure, which could defeat the point of the Simm being applicable to all types of banks covered by the margin rules. Further, they warn no model is perfect and that standardising models could mean risks go unseen. The discussion is understood to currently be happening on a firm level not involving Isda.

"Yes, if we all agree on the same way of measuring it that shows there is no under-coverage, we don't have a disagreement. But that doesn't mean there really was not an under-coverage. You get a natural evolution of excellence by people doing different things," says the risk management adviser.

"The whole point of Simm was to come up with something we could all build, that would be standard and simple, and we could create our own inputs. People do all their VAR models differently and you can back-simulate the P&L in so many different ways. And you have to think about the risks not in Simm, so you are creating a model that is standard, but which is even more complicated than Simm. It's just not going to happen," he adds.

Standardisation

The debate of standardising versus keeping inputs to the Simm firm-specific extends not only to backtesting, but to all the factors outside of the model itself. While some argue there is little appetite among firms for standardising sensitivity inputs, and that differences haven't proven large enough to warrant it, software firms argue outsourcing the calculations will ensure smoother sailing.

Firms are lining up to provide solutions to standardise aspects of the Simm process. In addition to Albanese's work, clearing house LCH launched the SwapAgent platform in November. The utility, which is supported by about a dozen dealers, calculates variation margin calls as well as standardised risk factors, which can be fed into the AcadiaSoft utility to be run through the Simm.

"Given the sheer number of entities expected to post margin is in the 5,000-8,000 range, I don't think it will ever be feasible for them to synchronise all of their collateral calculations. Either one third-party emerges, like an exchange for instance, and computes margin for everyone, [as] is already done in the cleared markets, or it is going to be a total disaster. It is just not going to work: you will have a liquidity crisis," says Albanese.

Firms think those standardisation efforts could also be used for model validation; for example, for both benchmarking exercises and backtesting.

"We think there is a definite opportunity around that. Firms are using very similar models and going through very similar validation processes, so there is a substantial duplication of efforts across firms. Yes, they are responsible for individual submissions on model validation, but the work they do independently overlaps. So there is some potential for us to identify and package up aspects of validation that can be standardised across firms," says OpenGamma chief executive officer Peter Rippon in London.