Reducing the variation of credit risk-weighted assets

Contents

Abbreviations

1 Introduction

2 IRB approaches for credit risk

3 Reasons for variation in credit RWA

4 Proposals of the BCBS to the IRB approaches for credit risk
4.1 Applicability of internal modelling
4.2 Parameter floors
4.3 Parameter estimation practices and fixed parameters
4.4 Output floors

5 Assessment of the consultation items

6 Conclusion

References

Abstract

On March 24, 2016, the Basel Committee on Banking Supervision published a consultation document which contains several proposals regarding a revision of the internal ratings-based approach. The proposals are a response to the markets participants’ loss of confidence in the regulatory capital ratios. The objective of this paper is to present the Basel Committee’s proposals and to investigate if the measures are sufficient to reach the aspirations of the Basel Committee.

The proposed measures aim at reducing complexity as well as improving comparability of the Basel framework and, in particular, reducing the excessive variability in credit risk-weighted assets. It is intended to curtail the use of internal models for certain exposure categories and to introduce model-parameter floors, greater specification of parameter estimation practices and output floors based on the standardised approach.

The result of this study incicates that the measures are hardly suitable to restore confidence in capital ratio. A major reason for this is that most measures will reduce the risk-sensitivity of the Basel framework and, thereby, the relation of banks’ capital resources and banks’ risk taking will not be reflected adequately.

Abbreviations

Abbildung in dieser Leseprobe nicht enthalten

1 Introduction

“Addressing the issue of excessive variability in risk-weighted assets is fundamental to restoring market confidence in risk-based capital ratios”¹, said Stefan Ingves, chairman of Basel Committee on Banking Supervision (BCBS). In 2013 the BCBS conducted a study to find out the main drivers for differences in risk-weighted assets (RWA) in the banking book calculated by using the internal ratings-based (IRB) approach.² The IRB approach allows banks to use internal models as inputs for calculating their regulatory capital requirements for credit risk. Building on the result of this study, in March 2016, the BCBS published the consultation document BCBS 362 which contains several proposals of changes to the IRB approach.³ The proposed changes are part of a package of measures of the BCBS designed to balance risk sensitivity, simplicity and comparability of the Basel framework.⁴ The BCBS 362 includes four main proposals which aim at reducing the variation in credit RWA calculated by using the IRB approach. The BCBS proposed to curtail the use of internal models for certain exposures categories and intends to introduce internal model-parameter floors, greater specification of parameter estimation practices and output floors based on the standardised approach (SA).⁵

The aim of this paper is to present the different proposals of the BCBS on the IRB approach for credit risk and to find out about the effects that such measures may have on the financial sector as well as assessing if the measures are sufficient to reach the aspirations of the BCBS.

The remainder of this paper is organized as follows. The next section briefly describes the IRB approach under the Basel II accord and its current design. Section 3 discusses the reasons for the excessive variability in the regulatory capital requirements for credit risk. The focus of this paper is placed on section 4. It presents the different proposals concerning the IRB approach of the BCBS consolation document. Section 5 investigates the effects of the proposed measures of the BCBS. The final section summarizes the main conclusions.

2 IRB approaches for credit risk

In 1999 the BCBS disclosed the “Revised international capital framework”, the so- called Basel II accord. Basel II proposed to permit banks to choose between two different methodologies for calculating their regulatory capital requirements for credit risk: the simple standardised approach and the more complex internal ratings-based approach.⁶ The IRB approach allowed banks to use internal models to calculate their credit risk requirements for the first time.

The BCBS associates two key objectives with the IRB approach. On the one hand, the banks’ internal estimates may be more sensitive to credit risk and therefore, the IRB approach may increase the risk sensitivity of the Basel framework. On the other hand, it is incentive compatibility because it may encourage banks to continuously improve their internal risk management technics.⁷

In order to receive authorization to adopt the IRB approach and to be allowed to use it continuously, a bank must fulfil certain minimum requirements that are reviewed by the banking supervisors.⁸

A bank has to take two main steps to use the IRB approach. First, it has to categorise its banking book exposures into six different asset classes: corporates, sovereigns, banks, retail and project finance. Second, to calculate the credit risk, the bank has to estimate the following risk parameters that are inputs to risk-weight functions which map the risk parameters to RWA: probability of default (PD), loss given default (LGD), exposure at default (EAD) and maturity. The exact type of risk parameters a bank has to estimate, depends on the specific type of IRB approach a bank uses. The Basel framework provides two broad approaches: the foundation IRB (F-IRB) approach and the advanced IRB (A-IRB) approach. Under the F-IRB approach banks are allowed to estimate its own PD parameters for non-retail portfolios. The other risk parameters which are required to calculate RWA are provided by the banks’ supervisor. For retail portfolios, a bank is required to use the A-IRB approach. The A-IRB approach can be also used to calculate the RWA of other asset classes, however, under the A-IRB approach banks have to estimate all risk parameters by themselves.⁹

3 Reasons for variation in credit RWA

In recent years, several studies raised concerns about whether the implementation of the IRB approach to credit risk might be unequal or not. These studies show huge differences in banks’ RWA.

The BCBS also conducted an analysis within its “Regulatory Consistency Assessment Programme” to find out about the reasons for variation in credit RWA in the banking book. As credit risk is the largest component of RWA, it accounts for 77% of the BCBS’ observed variation in RWA.¹⁰ Up to three quarters of this 77% are explained by differences in underlying risk arising from the composition of banks’ assets. This kind of RWA differences are consistent with the risk sensitivity intended by the Basel accords. In addition to the risk-based factors, the BCBS also identifies practice-based factors which contribute to RWA variability. These practice-based factors are inter alia deviation in national implementation of the Basel standards, banks’ choices under the IRB framework and conservative adjustments to IRB risk parameter estimates, differences in banks’ modelling choices as well as differences in interpretation of the Basel framework.

In the BCBS study, banks were given a hypothetical portfolio with largely low-default exposures. For these exposures, they should estimate the PD and LGD. The results showed that banks’ capital ratios vary by up to 20% if one translates the implied risk weight estimations into potential impact on banks’ capital ratios. The BCBS stressed the less-advanced stage of development of LGD modelling and the lack of appropriate data for low-default portfolios as a significant driver for RWA variability.¹¹

Besides studies from regulatory organisations, which often focuse on model calibration as a reason for RWA variation, there are different studies from banks and banking associations emphasising the differences in the application of regulatory principles. For instance, the results of a survey of the European Banking Federation (EBF) highlighted that RWA comparability is affected by the inconsistency of approaches to model add-ons and layers of conservatism related to model uncertainty, model risk and data issues.¹² Moreover, the EBF argued that different definitions of default significicantly impact banks’ PD and LGD estimations.¹³

4 Proposals of the BCBS to the IRB approaches for credit risk

The excessive variability in the capital requirements for credit risk in the IRB framework has compromised comparability. Some market participants have therefore questioned the reliability of the IRB approach.¹⁴ To restore the trust in IRB models the BCBS published the consultation document BCBS 362 which proposes several measures to reduce the variation in credit RWA calculated by the IRB approach. The key proposals are described below.

4.1 Applicability of internal modelling

One proposal to reduce the variation in credit RWA is to restrict the scope of use of internal models. The BCBS presents three different criteria which must be met before an asset class is suitable for internal modelling. First, appropriate data must be available for the respective risk, second, the banks have to use individual data which are unavailable for the broader market so that they are an information advantage for the individual bank and third, robust and universally accepted modelling techniques must be available for the validation of the exposures.¹⁵

The use of IRB approaches should be completely removed for the category banks, which also contains other financial institutions, because in the view of the BCBS it does not fulfil the criteria for modellability. Furthermore, banks should no longer be allowed to use the IRB approaches for exposures to corporates and corporates belonging to consolidated groups with total assets greater than €50bn. The BCBS argues that for banks, other financial institutions and large coporates there are no sufficient data available to produce reliable PD estimations. The mentioned asset classes are usually considered to be low-default exposures which require a lot of observations of defaults for reliable PD estimations. However, defaults are too rarely observed due to their rare appereance. In addition, banks usually do not have any information advantage with regard to exposures to banks, other financial institutions and large corporates which are usually the subject of many market analysis and rated by rating agencies. Therefore, banks’ internal models create no added value compared to the SA.

Moreover, the BCBS proposes to remove the A-IRB approach for exposures to corporates and corporates belonging to a group with annual revenues exceeding €200m as well as to remove both IRB approaches for specialised lending due to insufficient data availability.¹⁶

Since banks use the same public data to calculate the risk-weights for equity exposures they should be subjected to the SA. What’s more, equity exposures are just a small part of financial institutions’ balance sheets, and so from the BCBS’s point of view the application of the SA for equity exposures would significantly increase the simplicity and comparability of the Basel framework and just slightly reduce its risk sensivity.¹⁷

In the BCBS review of the credit valuation adjustment (CVA) risk, which is an adjustment of the price of a derivative instrument due to counterparty credit risk, the BCBS cast doubts whether CVA can be reliably captured with banks’ internal models which are designed to capture market risks in the trading book.¹⁸ That is why the BCBS decided to prohibit the use of the internal models approach (IMA- CVA) for the calculation of CVA risk.¹⁹

4.2 Parameter floors

For the remaining asset classes, for which the BCBS still permits the use of IRB approaches, the BCBS proposes floors for the model parameters. Floors are proposed for the PD, LGD and the credit conversion factors (CCF) which are used to calculate the EAD of off-balance sheet items. Floors range from 5bps to 10 bps for PD and between 0% to 50% for LGD, the EAD floor is set at 50%.²⁰ As the introduction of input floors is associated with several trade-offs regarding the appropriate calibration of the floors, the BCBS will conduct a quantitative impact study to test also other parameter floors. One trade-off is for instance that parameter floors would narrow the range of model outcomes and therefore reduce the variability in credit RWA, however, it may incentivise financial institution to invest in riskier assets such that they are no longer affected by the parameter floors.²¹

From the market participants’ perspective portfolios with exposures to retail and medium-sized enterprises especially in jurisdictions with historically low actual LGD would probably be mostly affected by the new input floors. Most banks which use the A-IRB approach for non-mortgage retail portfolios and corporate exposures show lower LGD values than requested from the BCBS in its consultation document.²² Furthermore, the rating agency Fitch estimates that risk-weights for retail mortgages would increase by 50% if the proposed PD floor of 5 basis points for retail mortgages would be applied.²³

4.3 Parameter estimation practices and fixed parameters

In order to reduce the variation of practices which banks use to estimate the model parameters under the IRB approaches the BCBS published several proposals with regard to parameter estimation practices and fixed supervisory parameters.

The consultation document contains four proposals with regard to PD estimations. The BCBS would like to reinforce the stability of assigned ratings and associated PD by proposing “through-the-cycle” models instead of “point-in-time” models. Furthermore, the data used to estimate PD should include a mix of “good” and “bad” years with at least one out of ten years must be a downturn year. To ensure granularity of PD estimations, PD should be estimated for each rating grade. In addition, banks should include seasoning effects into the estimation of PD for retail exposures.²⁴

Referring to the F-IRB approach, the BCBS proposes a new calculation methodology to determine the downturn LGD for fully and partially secured exposures. The introduction of a single formula and the related elimination of the two currently existing approaches intends to simplify the framework. In addition, the hair-cuts applied to non-financial collateral should be increased on a flat rate of 50% because the BCBS has evidence that collateral values are regularly overpriced in stress situations. In return, supervisory collateral hair-cuts for LGD applied to fully secured exposures should be reduced and the minimum collateralisation requirement should be removed for all collaterals that are eligible under the IRB approaches.²⁵

The downturn LGD, applied to unsecured exposures under the A-IRB, were identified as a significant driver of RWA variability. To address this, future banks should calculate downturn LGD add-on for each exposure additionally to the longrun average LGD. Downturn add-ons must be separately modelled and the BCBS consider an additional floor for the add-ons. Moreover, the BCBS also thinks about the introduction of a floor for LGD estimations of fully and partially secured exposures. This floor should be calculated as the weighted average of the floors applied for unsecured and fully secured exposures.²⁶

[...]

¹ c. f. Binham, Caroline, 2016

² c. f. Basel Committee on Banking Supervision, 2013a

³ c. f. Basel Committee on Banking Supervision, 2016b

⁴ c. f. Basel Committee on Banking Supervision, 2013b

⁵ c. f. Basel Committee on Banking Supervision, 2016a, p.1

⁶ c. f. Basel Committee on Banking Supervision, 2004, pp. 1 - 2

⁷ c. f. Basel Committee on Banking Supervision, 2001, p. 1

⁸ c. f. Maxwell, Fiona, 2014, p. 42

⁹ c. f. Basel Committee on Banking Supervision, 2001, pp. 4-7, p. 13, p. 57

¹⁰ c. f. Basel Committee on Banking Supervision, 2013a, p. 6

¹¹ c. f. loc. cit., pp. 7 - 8

¹² c. f. European Banking Federation, 2012, p. 6

¹³ c. f. European Banking Federation, 2012, p.12

¹⁴ c. f. Busse, Johannes, 2016, p. 18

¹⁵ c. f. Basel Committee on Banking Supervision, 2016a, p. 3

¹⁶ c. f. Basel Committee on Banking Supervision, 2016a, p. 4

¹⁷ c. f. loc. cit., p. 4

¹⁸ c. f. Basel Committee on Banking Supervision, 2015a, pp. 1, 4

¹⁹ c. f. Basel Committee on Banking Supervision, 2016a, pp. 4 - 5

²⁰ c. f. loc. cit., p. 6

²¹ c. f. loc. cit., pp. 6 - 7

²² c. f. Alexander, Philip, 2016, p. 51

²³ c. f. Verma, Sid, 2016, p. 23

²⁴ c. f. Basel Committee on Banking Supervision, 2016a, p. 7

²⁵ c. f. loc. cit., pp. 8 - 9

²⁶ c. f. Basel Committee on Banking Supervision, 2016a, p. 10