6.3 Modelling Requirements

6.3.1	The models employed to compute the metrics above should follow the principles articulated in the MMS. In particular all IRRBB models should follow the steps in the model life-cycle. The assumptions and modelling choices surrounding IRRBB models should not be the sole responsibility of the ALM function nor the market risk function. Rather, these assumptions should be presented to and discussed at the appropriate governance forum reporting to the Model Oversight Committee.
6.3.2	The modelling sophistication of the EVE should depend upon the size and complexity of institutions. For that purpose, different requirements are defined in function of their systemically important nature. The modelling requirements presented hereby should be regarded as minimum standards. To remain coherent with Basel principles, higher standards apply to large and/or sophisticated institutions (“LSI”). However, the other institutions may choose to implement models with higher standards than the one prescribed for them. This proportionality is an exception to the MMG due to the prescriptive nature of the Basel methodology surrounding IRRBB.
6.3.3	The requirements below refer to the methodology articulated in section IV (“The standardised framework”) of the Basel Standards on IRRBB. All institutions are requirements to fully understand this framework.
	Table 10: Components of IRRBB models

Component	LSIs	Other institutions
Computation granularity	Facility level or facility type if groups of facilities are homogeneous	Summation of facilities within buckets, according to the Basel Standards
Time buckets	Granular bucketing depending on the composition of the books	Standardised bucketing according to the Basel Standards on IRRBB
Option risk	Included in both EVE and NII	Included in EVE Optional from NII
Commercial margins	Optional from EVE Included in NII	Optional from EVE Included in NII
Basis risk	Included	Optional
Currency	Estimation for each material currency	Estimation for each material currency
Scenarios	Standard plus other scenarios defined by institutions	Standard six scenarios
IT-system	Dedicated system	Spreadsheets can be used if the model and its implementation are independently validated

6.3.4	The estimation of EVE should be based upon the following principles: (a) it includes all banking book assets, liabilities and off-balance sheet exposures that are sensitive to interest rate movements, (b) it is based on the assumption that positions roll off, and (c) it excludes the institution’s own equity. The approach subsequently depends on the type of institutions.
	(i)	LSIs should compute EVE as the difference between discounted assets and liabilities at a granular level. Institutions should aim to perform this computation at a facility level. For practical reasons, some facilities could be aggregated, provided that they are homogeneous and share the same drivers and features. All inputs including, but not limited to, cash-flows, time buckets, risk-free rates, option risk and basis risk should also be estimated at a granular level.
	(ii)	It should be noted that the Gap risk and the Gap risk duration are not directly used to estimate EVE in the context of a granular full revaluation. However, the Gap risk and Gap risk duration should be estimated and reported in order to manage IRRBB.
	(iii)	Non-LSI can compute EVE at a higher level of granularity, according to the principles outlined in the Basel Standards on IRRBB and in particular according to article 132. The methodology is based upon the summation of discounted Gap risk across time buckets, rather than a granular NPV estimation at facility level. Institutions should pay particular attention to the cash flow allocation logic in each time bucket.
	(iv)	Irrespective of their size, all institutions should compute ?EVE as the difference between EVE estimated under interest rate scenarios and the EVE under the current risk-free rates. The final EVE loss and the standardised risk measure employed in Pillar II capital should be computed according to the method explain in the article 132 (point 4) of the Basel Standards on IRRBB, whereby EVE loss should be aggregated across currencies and scenarios in a conservative fashion.
6.3.5	The estimation of NII should be based upon the following principles: (a) it includes all assets and liabilities generating interest rate revenue or expenses, (b) it includes commercial margins and (c) no discounting should be used when summing NII across time buckets. The approach subsequently depends on the type of institutions.
	(i)	LSIs should compute NII at a granular level, both for facilities and maturity time steps. NII should be based on expected repricing dates upon institutions’ business plan of future volume and pricing. Therefore LSIs should estimate ∆NII as the difference in NII under the base and the stress scenarios. Such granular computation should include option risk and basis risk.
	(ii)	Non-LSIs can compute ?NII by allocating interest revenue and interest expenses in the standardised time buckets used for EVE. Non-LSI institutions can compute ?NII by estimating directly their earning at risk on each expected repricing date.
	(iii)	For the purpose of risk management, institutions are free to model NII based on static or dynamic balance sheet assumptions (although LSIs are recommended to employ the latter). Institutions can also choose the NII forecasting horizon. However, for Pillar II assessment as part of the ICAAP and for reporting to the CBAUE, the following, institutions should compute NII over 1 year; in addition LSIs should also compute NII over 3 years.
6.3.6	Institution’s own equity: For NII estimation, institutions should include interest-bearing equity instruments. For EVE, in the context of the MMG, institutions should compute two sets of metrics by first including and then excluding instruments related their own equity. These two types of EVE will be used for different purposes.
	(i)	CET1 instruments should be excluded at all times to avoid unnecessary discrepancies related to the choice of behavioural maturity associate to this component.
	(ii)	Some institutions have a large proportion of interest-sensitive instruments, in particular in the AT1 components. Consequently, these institutions should estimate and report a first set of EVE sensitivities by including these instruments. This type of EVE is useful for proactive management of IRRBB.
	(iii)	Conversely, one of the objectives of assessing IRRBB is to ensure that institutions hold enough capital to cover such risk, which is articulated through the ICAAP. Institutions should not use part of their capital to cover a risk that is itself generated from capital. Therefore, institutions should also compute and report EVE by excluding their own equity entirely. This type of EVE is useful to estimate the Pillar II capital charge arising from IRRBB.
6.3.7	Commercial margins: The treatment of commercial margins is different between NII and EVE. However, the recommendation is similar for both LSIs and non-LSIs.
	(i)	All institutions should include commercial margins in NII estimation. Margins should be adjusted based on business plans and expected customer behaviour in a given interest rate environment. For instance, it might be assumed that margins will be increased to retain depositors in a falling interest rate environment.
	(ii)	All institutions have the option to include or exclude commercial margins in EVE estimation. However, institutions should also aim to estimate the impact of commercial margins on EVE. For consistency, if margins are included in the cash flows (numerator), then discount factors should also reflect the corresponding credit spread of the obligors (denominator). Such estimation should be done at homogeneous pools obligors with similar credit risk profiles.
6.3.8	Basis risk: This risk arises when assets and liabilities with the same tenor are discounted with different ‘risk-free’ interest rates. Potential credit risk embedded in these rates makes them not entirely risk-free, hence the existence of bases. A typical example is an asset priced with the US LIBOR curve but funded by a liability priced with the US Overnight Index Swap (“OIS”) curve, thereby creating an LIBOR-OIS basis leading to different NPV and NII from both the asset and the liability. Another example is the recent introduction of USD Secured Overnight Financing Rate (“SOFR”) creating a LIBOR-SOFR basis. LSIs are required to fully identify and assess basis risk. They should employ the appropriate risk-free rate for each instrument type, thereby capturing basis risk in all the IRRBB metrics. While non-LSIs are not expected to fully quantify basis risk on a regular basis, they should perform an approximation of this risk to assess whether further detailed quantification is necessary.
6.3.9	Currency risk: The currencies of assets and liabilities have a material impact on the resulting IRRBB, therefore this dimension should be fully addressed by institutions’ modelling practice.
	(i)	All the IRRBB metrics should be estimated for each currency in which the institution has material exposures, i.e. when the gross exposure accounts for more than five percent (5%) of either the gross banking book assets or gross liabilities. For those, the interest rate shocks should be currency-specific.
	(ii)	For the estimation of the capital charge, the Basel Standards on IRRBB suggests to sum the maximum change in EVE across currencies without offsetting. While the CBUAE recognises that no offsetting is conservative for pegged currencies, (typically USD/AED), institutions should manage basis risk appropriately since material bases have been observed between USD rates and AED rates. Consequently, each institution has the option to offset ?EVE between pegged currencies, only if it can demonstrate that it does capture the basis risk between these currencies with dedicated stress scenarios.
6.3.10	Non-performing assets (“NPA”): Institutions should define clearly the treatment of non-performing assets in their modelling practice, according to the following principles.
	(i)	NPA (net of provisions) should be included in the estimation of EVE. In most default cases, LGD>0% therefore a recovery is estimated at some point in the future. The LGD is estimated by discounting expected recoveries with a discount rate generally based on the effective interest rate of the facility. In the context of IRRBB, a change in the interest rate environment should have an impact the present value of discounted recoveries and therefore on LGD. This effect could likely impact EVE. Finally, consideration should also be given to rescheduled facilities and/or forbearance with payment holidays where interests are accrued. The postponement could results in lower PV under scenarios with increasing rates.
	(ii)	The treatment of NPA (net of provisions) for NII computation is left to the discretion of banks. Under a static balance sheet assumption, non-performing assets will not generate cash inflows. A change in rates would have no impact the NII from such assets. However, under dynamic a balance sheet assumption, some NPA could return to a performing status and therefore impact NII.