# FRTB: The ascent of the standardized models

- The post crisis period has seen the emergence of several new standardized risk models
- FRTB-SA and SIMM are the more sophisticated methodologies and share the most similarities, but also have crucial differences
- This blog explores these similarities and differences, and shows how they will bear on the capital markets landscape in years to come

The post crisis period has seen the emergence of new standardized risk models. It began in 2014 with **SA-CCR** for derivatives counterparty credit risk exposure, followed in 2016 by **FRTB-SA** for market risk and most recently in late 2016 with **SIMM** for initial margin calculation. These standardized models share several characteristics, such as multi-level aggregation by risk class, allocation of exposures to risk class buckets and risk correlation matrices. Worryingly for banks, these models are substantially more complex than those that preceded them.

FRTB-SA and SIMM are the more sophisticated methodologies and share the most similarities. Clearly the SIMM model developed by ISDA was heavily influenced by the FRTB-SA Sensitivity Based Approach (SBA). Both are driven by delta, vega and curvature sensitivities computed in the front-office, and broadly follow the popular Delta-Normal VaR methodology.

However, there are substantial differences too. The SIMM model, because of its role in calculating initial margin, incorporates liquidity through a liquidity concentration threshold parameter. Under this approach large exposures in illiquid markets carry a heavy risk charge that depends on the size of the position. The SIMM model is also an evolving framework, incorporating back-testing that drives the annual recalibration of the risk weights, correlations and concentration thresholds. By contrast FRTB-SA has been calibrated on a period of stress and no recalibration is anticipated.

Critically, the aggregation for SIMM is computed at the product class level (RatesFX, Equity, Commodity, Credit). Every trade is assigned to a product class that is most representative of the underlying risks. Trades that bear exposure to multiple risk classes (anything from a structured product to a simple equity swap) are not offset by a natural hedge falling in a different product class.

SIMM is also calibrated using a VaR metric, rather than the Expected Shortfall introduced with FRTB in order to better align to the backtesting procedure. Due to the annual calibration of parameters, SIMM has also opted to steer away from the correlation regimes used in FRTB-SA.

While FRTB is a mandatory regulation for financial institutions, ISDA SIMM is a proprietary model developed by ISDA to meet regulatory requirements around Uncleared Margin Rules (UMR) which has largely been adopted by the market to become the de-facto standard.

### Front to risk alignment

In capital markets, banks typically control risk by minimising the portfolio sensitivity to changes in underlying market prices. However, under the standardized market risk under Basel 2, the approach taken was more reductive. Trades get assigned to time bands, on the basis of either the maturity date or duration before a series of disallowances are applied to account for maturity mismatches, basis risk, coupon discrepancy etc. The accuracy of the methodology is sacrificed for the ease of implementation.

A consequence of this approach is a decoupling of the risk capital and front-office systems. Market risk capital charge is derived from static properties of the trade such as notional, maturity date^{1}, currency, sector, bond issue etc.

Both FRTB-SA and SIMM are based on trade sensitivities. A perfectly hedged portfolio will not incur a risk charge under FRTB-SA (assuming no residual risk add-on which cannot be netted) or SIMM (assuming the trades fall under the same product class). This contrasts with the lack of risk sensitive metrics under Basel 2 which meant that disallowance is applied indiscriminately.

One area where SIMM and FRTB-SA methodologies diverge is in the requirement for front to risk alignment under FRTB-SA, which mandates that the models used to calculate sensitivities are the same ones used for P&L computation. However, given the similarity of the sensitivity requirements it makes sense to use the same engine for both SIMM and FRTB-SA. With interest rate risk for example, the risk metric (PV01) is the same and the term structure of risk factors is almost identical, differing only in that SIMM has more granularity at the front end of the curve, which lends itself to a common computational framework.

### Risk charge calculation

In the trivial case of a trade with a single delta sensitivity measure the capital charge for both FRTB-SA and SA-CCR is the product of the risk sensitivity and the risk weight^{2}. Using the PV01 sensitivity computed in the front office we can compare the interest rate risk capital for different maturity trades. The FRTB-SA charge is consistently higher, as the risk weights have been calibrated on a stress period, and while the SIMM risk weights ramp up on the longest tenors, the risk charge across the maturity profile broadly follows the same pattern. SA-CCR takes a different approach and the notional is scaled by the supervisory duration, a function of term to maturity and margin period of risk, which is capped at one year.

When the portfolio has sensitivity to multiple risk factors comparisons across the different methodologies are more intractable. FRTB-SA uses a parametric approach for interest rate correlations, and innovatively applies correlation regimes where the capital charge is based on the maximum under low, medium and high correlation regimes.

Curvature is another area where the two methodologies diverge. Under FRTB-SA curvature is computed using a large shift in the risk factor. Negative curvature is capitalized and is computed by comparing the change in value under a large shift, for interest rates a 1.7% change in rates, with the linear change as extrapolated from the 1bp delta sensitivity. SIMM by contrast adopts a different approach based on the vega sensitivity, the implied volatility, and a scaling function to link vega and gamma.

SIMM also works with a reduced number of interest curves. If a portfolio is sensitive to any other curves these must be mapped on to the reduced set.

Together, these standardized methodologies will have a substantial bearing on the capital markets financial landscape in the coming years.

**FRTB: Our new report**

For more details about how FRTB will affect the roles and responsibilities of heads of trading desks, including guidance on the changes and how management can address the new challenges, see our new report, “Changing Roles of the Head Trader under FRTB”.

**How Finastra can help**

Finastra is the leading provider of capital markets solutions with the largest global install base with its Fusion Treasury, Fusion Markets and Fusion Risk products. Finastra delivers out of the box FRTB-SA and works with banks to deliver a FRTB-IMA solution. Click here to learn more.

_{1 The market risk capital charge also allows a duration-based approach, which is less commonly used.}_{2 The sensitivity measure for FRTB-SA is normalized hence the risk weights are an order of magnitude smaller than those used for SIMM.}