Frank Pierschel, Head of the Banking Supervision Division of the Department for International Policy, Financial Stability and Regulation, represents BaFin in different bodies of the Basel Committee which deal with this topic: he is a member of the Policy Development Group (PDG) and the Coherence and Calibration Task Force (CCTF) as well as Co-Chair of the Task Force on Standardised Approaches (TFSA). In this interview with the BaFinJournal, he explains and assesses the proposal.

Mr Pierschel, internal model approaches were introduced into the framework of regulatory capital requirements through Basel II and they have already been revised in Basel II.5 and III. Why are they being revised yet again, nearly ten years after the onset of the financial crisis?

First of all, I would like to draw your attention to the title of the document: "Reducing variation in credit risk-weighted assets – constraints on the use of internal model approaches". This encompasses two components which the Basel Committee has been dealing with for a long time. First of all, we supervisors regarded as critical the degree to which credit institutions were theoretically able to push down their risk-weighted assets for the calculation of regulatory capital requirements. This is why a lower limit was already formulated in Basel II, the so-called Basel I floor. We are retaining this concept in an altered form.

Second, the Basel Committee examined the implementation of Basel II, II.5 and III and determined that the standards rarely result in the same capital requirements for all credit institutions. Initially, this was to be expected, as the internal model approaches are indeed supposed to reflect the risks for institutions individually. Individual business strategies, the resulting adjusted portfolios and – not to forget – the exercising of different options in different countries provided good reasons for these deviations. What surprised us, however, was their scale. In parts, the capital requirements of banks using internal models deviated drastically from those of the institutions which use the standardised approach. It was therefore necessary to review the model approaches as part of the revision of the Basel I floor.

In 2013, the BCBS therefore established a high-level working group which was tasked with making specific proposals with the objectives of simplicity, comparability and risk sensitivity. The revision of the floor and basing it on the standardised approach were one of the central components; principles for the use of models for calculating regulatory capital requirements were the other.

And how does the proposal look now in detail?

Models for portfolios are only to be still permitted if they fulfil three criteria. First of all, the data available for the portfolio in question has to be sufficient in terms of both quantity and quality. Second, the institutions have to use their own information and experience for modelling; the data cannot come from publicly accessible sources alone. And third, the banks must use robust and generally accepted modelling techniques which are capable of validation.

Using these criteria, the Basel Committee examined the portfolios for credit risk, as it had already done for the internal model approach for operational risk. In general, it is planned to remove the internal model approaches for exposures to banks and other financial institutions and large corporates with total assets of more than €50bn and for equities. For all other portfolios, the proposal provides for constraints on the use of internal models. For exposures to corporates belonging to consolidated groups with annual revenues greater than €200m and total assets of up to €50bn, only the foundation internal ratings-based approach (F-IRBA) can still be used. For specialised lending, the standardised approach is to be used generally, although modelling is possible within the slotting approach which is anchored within the framework. The internal models for counterparty credit risk are to be limited by an output floor, i.e. a minimum capital floor based on the new standardised approach for counterparty credit risk. By contrast, internal models are no longer to be permitted for the calculation of capital requirements for credit valuation adjustment (CVA) risks. Internal model approaches for sovereign exposures are to be initially exempted from this regulation, but this will be reviewed during work on sovereign risks.

On the level of parameters, input floors will constrain the use of own estimations for the remaining portfolios, on an aggregated level output floors are specified. In the case of the input floors, the probability of default (PD) and the loss given default (LGD) will be subject to supervisory restrictions. For the consultation process, the BCBS initially set the PD floors at five basis points and ten basis points for revolving credits. The LGD floors vary depending on the existence and level of collateral between 0 and 25 per cent. In addition, the collateral values are to be limited in the calculation of the LGD floor through supervisory haircuts. For its part, the output floor will serve as a percentage brake which is to be based on the standardised approach and below which the capital requirement may not fall. The consultative document specifies a range of 60 to 90 per cent for this.

What does that mean for the calibration, do the proposals not result in a higher capital requirement? And does that not contradict the press release of the Governors and Heads of Supervisors, the Basel Committee’s oversight body, from January of this year?

Yes, if you compare the capital requirements of the present regulation and the new proposal with each other, a higher capital requirement can be expected from the parameters currently specified in the consultative document. However, I would also point out that the floors are an essential component in the calculation of the overall capital requirement. For credit risk, this means that the new standardised approach can provide the gauge for measurement. This is also currently being recalibrated.¹⁾ The overall calibration is subject to a very broad quantitative impact study. If the calibration here turns out to be too high for banks using the internal model approach, it will become necessary to take countermeasures by means of the output floor, i.e. the scaling factor, which is currently 1.06, or other readjustments. I do not believe that the Basel Committee would present a proposal which ignores this important requirement to its oversight body, a requirement which in any case merited a press release from the GHOS.

Would it not have been better not to permit internal model approaches at all to begin with if they are now causing such problems?

In general, I view the use of models positively. They assist complex credit institutions in particular to better assess their risk situation, make their decisions accordingly and take precautions against potential risks. However, there are of course model risks. Nobody knows with certainty whether the parameter estimates will actually materialise as expected. The estimated default rates have to be validated against the actual rates in use tests and, where necessary, adjusted in an iterative process. Here in Germany, we therefore rely heavily on the use test with which risk indicators (which are determined using internal models) are also used for risk management processes within the banks and the model assumptions are subjected to a financial litmus test on an ongoing basis. In addition, with the leverage ratio, an instrument which is intended to act as a buffer for zero-risk- and low-risk-weighted assets has also been introduced.

In Germany, models are to be permitted if they meet the model approval conditions. If we have doubts, we demand further documents or formulate requirements and monitor their fulfilment. A good model results in capital backing which is more risk-based and should therefore be retained. By contrast, there are portfolios which may be modelled according to the framework but, after an extreme optimisation process by banks, their supervisory advantage can no longer be perceived. However, we should not forget that standardised approaches also result in their own problems. Nonetheless, the overall advantages of internal models outweigh the disadvantages for me.

So what will happen next?

The proposals in the consultative document are already a compromise between those who distrust all internal model approaches and those who would prefer to change nothing at all in the framework. Market participants should take this into consideration when they are commenting on it. Their reactions are very important to the Committee in the assessment of the consequences of the regulatory changes. I can therefore only encourage them to state reasons for their criticism – and there will certainly be plenty of it – and to substantiate it with numbers. I would like to remind anyone who thinks that it is all over and done with about the reintroduction of external ratings in the second consultative document on the standardised approach for credit risk which was based on the consultation. It is also clear that the technical details in particular will be dealt with here. The portfolios which are no longer eligible for use of the internal model approaches and the use of input and output floors are not a matter for debate in the consultation process any more.

I believe that we can expect the publication of the finalised version of the Basel III regulatory package at the start of 2017. In addition to the new regulation of the internal model approach for credit risk, this will include the standardised approaches for credit, market and operational risk and the final calibration of the leverage ratio.

Footnote:

¹⁾ The second consultative document was published in December 2015. See also BaFinJournal April 2015 for background information.

Further steps

Comments on the consultative document can be submitted to the BCBS until 24 June 2016. Parallel to the consultation process, the quantitative impact study (QIS) is also starting. The results of the consultation process and the QIS will be taken into account in the final design of the framework.

Internal model approaches according to Basel

• Internal model: model specific to an individual institution for the calculation of capital requirements; possible alternative to the standardised approach
• Advanced IRBA: advanced internal ratings-based approach; all parameters are estimated by the credit institution
• Foundation IRBA: foundation internal ratings-based approach; only the probability of default is estimated internally
• Slotting approach: approach with parameters which change risk weights and are to be determined by the credit institution
• Basel I floor: limitation of the capital requirement for IRBA approved pursuant to Basel II, calculated on the basis of the reported Basel I capital requirement
• Input floor: minimum parameter which is to be used in the calculation of the capital requirements and which limits the model parameters
• Output floor: minimum capital requirement for portfolios eligible for use of internal model approaches as a percentage of the capital requirement determined according to the standardised approach
• CVA risk: credit valuation adjustment risk; the risk of a change in value of transactions with over-the-counter derivatives due to changes in credit quality of the counterparty
• PD: probability of default
• LGD: loss given default
• Use test: practical test of internal models prior to approval
• Leverage ratio: capital requirement orientated towards leverage for the coverage of model risks with zero-risk- or low-risk-weighted assets