Since the start of Basel II, the dominant philosophy of rating models was that there are two options: a simple, standardized approach for credit risk (SA); or a more advanced, Internal Ratings Based Approach (IRB). For the adoption of IRB, the bank needs approval from the supervisor, but once this is granted, the bank could operate at a lower capital requirement, and put the risk models to good use also for secondary goals such as pricing and portfolio management.
The IRB models are complex, but also more risk sensitive. A risky SME portfolio will lead to a higher capital requirement than a low-risk SME portfolio. And whenever modeling starts to become extremely complex, one can buy off the model uncertainty with the help of uncertainty multipliers, penalizing factors, that increase the risk parameters and thereby also the capital requirement, generally known as margin of conservatism “MoC”.
For standard-setting bodies such as the Basel Committee on Banking Supervision and the European Banking Authority (EBA), the apparent underlying design principles for the overall IRB system are that the models have to be high-quality, risk-sensitive, well-calibrated, stable in time, sufficiently granular, and ultimately, conservative.
Gradually, with Basel III and now the finalization of Basel III (also known as Basel IV), a new design principle pops up: simplicity.
Marco Folpmers
Both supervisors and banks have called for simpler models, less rules and a smaller jump from SA to IRB. This new simplification criterion is explicitly addressed in the most recent EBA consultation paper on simplifying the IRB approach. Simplification has now become an extra, and important, additional design criterion.
This is not a standalone development at the high end of the complexity spectrum. There is movement at the SA side as well: These become more risk-sensitive and, hence, more complex. So it is fair to say that standardized becomes more advanced, whereas advanced becomes more standardized.
Let’s look at some examples.
A case in point is residential mortgages. Under Capital Requirements Regulation CRR 3, the risk weight depends on the exposuretovalue ratio (ETV). For the portion of the exposure up to 55% ETV, the applicable risk weight is 20% (CRR 3, art. 125), whereas for the portion above 55%, the exposure is treated at the unsecured retail risk weight 75% (CRR 3, art. 123).
In simple language, this means that for a retail mortgage exposure that extends only up to 55% of the value of the property, this is regarded as very low risk, and the risk weight is simply a low, fixed percentage of 20% of the exposure value. It is only in the extreme case – the value of the property declining by 45% – that the bank is looking at the beginning of a loss after default. If the exposure starts to get larger than 55% of the value of the property, the risk weight starts to increase.
For the entire mortgage loan, this produces an ETVdependent effective risk weight (illustrated in the graph below). For example, a 60% ETV is the weighted average of 11 5% buckets at 20% and one 5% bucket at 75%, which gives an effective risk weight of approximately 25% ((11 times 20% plus 75%)/12).
This is a move towards a more risksensitive treatment compared with the previous standard approach that applied a moderate 35% risk weight, until a much larger part of the ETV (up to 80%).
Figure 1: For retail mortgages, the risk weight depends now more directly on the Exposure-to-Value.
From the graph it appears that retail mortgages with an ETV above 55% are now strongly penalized and face around stiff increase in risk weight. In principle, that means there is a better business case for IRB for high-ETV retail mortgages, especially if they have low probability of default (PD). However, further analysis is needed since the business case also depends on complicating matters such as input floors and the transitional arrangements for the output floor.
At the same time we see the advanced approach becoming more standardized. In its recent discussion paper, around which the consultation process was launched, the EBA assesses the potential simplification of the IRB approach.
The paper presents a number of “first thoughts” on how the IRB calculations can be simplified. Whereas these proposals are yet far from certain to be implemented, they illustrate the supervisor’s way of thinking. And then it becomes clear that this way of thinking is already very concrete.
Let’s have a look at the EBA’s proposals for the risk parameters.
For loss given default (LGD), the EBA proposes a simplified approach for direct and indirect costs. These cash flows are indeed often in practice difficult to source for LGD modeling. For the defaulted portfolio, one needs access to the transaction costs that were incurred when seizing and selling the collateral.
The same is true for the indirect costs. These are typically the costs of the restructuring and recovery unit. These costs are allocated to the LGD per exposure. The EBA proposes to use, as a fallback option, a fixed (percentage) add-on to either the realized or the estimated LGD.
Also for the downturn estimation, the EBA proposes a simplified calculation, one that gives more prominence to the reference LGD value, i.e. the average value of the LGDs for the two worst loss years in the data history. Currently, LGD downturn modeling needs to be based on macroeconomic analysis and the reference value is only non-binding. This fits in a broader trend that tends to give more importance to the reference value, as we have discussed last year.
Also for the margin of conservatism, simple procedures are introduced. The MoC is one of the most complex parameters of the IRB calculation, meant to adjust for three different sources of uncertainty (in short, data for MoC A, update of lending standards for MoC B, and MoC C for the general estimation error).
For these MoC values, EBA is contemplating optional fallback options that banks may use if the modeling effort is too large and inefficient. Depending on the values to which these standardized MoCs may be set, some regulatory arbitrage may be possible here, but that depends on the exact specifications to be issued by EBA.
Whereas Basel II introduced the standardized and IRB approaches as distinct and bipolar, after the finalization of Basel III (also called Basel IV), the situation is no longer so clear-cut. Modeling complexity has become a continuous line, with more complex and risk-sensitive standard approaches, whereas the IRB approach moves towards standardization for important ingredients such as CCF (credit conversion factor), LGD and MoC.
The overall “simplification” umbrella under which the fallback options are presented for IRB models also have a strategic implication. The IRB approval no longer guarantees protection against new competitors that are also keen on capital-efficient IRB treatment. It is less a barrier to entry if these models can be set up more efficiently with fallback options that will considerably reduce complexity and modeling resources needed.
Dr. Marco Folpmers (FRM) is a partner for Financial Risk Management at Deloitte the Netherlands.