An initial margin model is the calculation method used to determine the amount of initial margin that a party must post on a portfolio of derivatives or other cleared trades. Initial margin is collateral held to cover the loss that could arise between a counterparty's default and the closing out or replacement of its positions. The model estimates that potential loss to a chosen level of confidence over a chosen period.
Models are used in two settings. A central counterparty applies its own model to the positions of its clearing members. For contracts that are not cleared, the two counterparties each calculate the margin the other must post, using either a regulatory schedule or an approved model. Initial margin differs from variation margin, which settles losses that have already occurred as prices move, and the variation margin calculation requires no model beyond valuation.
What a model has to estimate
Every model answers the same question: how much could this portfolio lose over the time needed to deal with a default, in conditions worse than nearly all those expected? Three choices define the answer.
- The margin period of risk. This is the assumed time from the last margin payment to the completion of close-out. It is shorter for liquid exchange-traded contracts and longer for over-the-counter and uncleared trades.
- The confidence level. Margin is set to cover losses up to a high percentile, typically 99 per cent or more.
- The data. The price history used must include a period of market stress, so that margin does not fall too low in calm times.
A model must also decide how far to recognise offsets between positions. Giving credit for well-established correlations lowers margin; relying on correlations that break down in a crisis leaves the holder under-protected.
Models used by clearing houses
The oldest widely used method is SPAN, introduced by the Chicago Mercantile Exchange in 1988. It revalues a portfolio under a fixed set of scenarios for price and volatility changes and takes the worst loss, with adjustments for spreads between contract months and between related products. Many exchanges adopted it for futures and options.
For cleared derivatives such as interest rate swaps, clearing houses generally use portfolio models based on value at risk or expected shortfall. Historical simulation applies several years of past market moves to the current portfolio; refinements scale those moves to current volatility. Add-ons are charged for concentrated or illiquid positions, for credit quality and for wrong-way risk. European rules on clearing houses set minimum confidence levels and liquidation periods, with higher standards for over-the-counter derivatives than for listed products. The margin a member posts is the first resource used if it defaults, ahead of the default fund described under central counterparty clearing.
Models for uncleared derivatives
After the 2008 crisis, the Basel Committee and the International Organization of Securities Commissions agreed a framework requiring financial firms and large non-financial firms to exchange margin on OTC derivatives that are not centrally cleared. Initial margin must be calculated to a 99 per cent confidence level over a ten-day horizon, using data that includes stress. It is exchanged gross by both sides and held in segregated accounts. The requirement was phased in between 2016 and 2022, starting with the largest dealers.
Firms may use a standardised schedule, which applies fixed percentages to notional amounts by asset class and maturity, or a model approved by supervisors. The schedule recognises little netting and gives high figures, so most large firms use a model. To avoid disputes arising from each party running a different model, the industry adopted a common one: the Standard Initial Margin Model, known as SIMM, published by the International Swaps and Derivatives Association and first used in 2016. SIMM is based on sensitivities. Each firm computes how the value of its trades responds to changes in defined risk factors for interest rates, credit, equity, commodities and foreign exchange, and the model applies prescribed risk weights and correlations to those sensitivities to produce the margin.
The amounts are called and settled under collateral documents such as a credit support annex, and differences between the parties' calculations are handled through the collateral dispute process.
Validation and governance
Models are tested by backtesting: actual portfolio losses over the margin period are compared with the margin held, and the number of breaches is checked against the number implied by the confidence level. Clearing houses also run sensitivity analysis and stress tests, and significant model changes need approval from their risk committees and regulators.
Criticism
The main concern is procyclicality. A model driven by recent volatility calls for more margin just when markets are under strain, forcing participants to find cash or sell assets and adding to the stress. Sharp increases in margin calls in March 2020 and in energy markets in 2022 prompted international reviews. Regulators require clearing houses to use tools that dampen the effect, such as a buffer that can be released, a floor based on a long look-back period, or a weighting towards stressed observations, and have called for greater transparency so that members can anticipate calls. Margin also has a cost: collateral tied up in this way cannot be used elsewhere, a factor in decisions about collateral optimisation and counterparty risk management.