Assessing the Finalised Basel III Banking Regulation Regime

Abstract

The authors summarise the final version of Basel III, reducing thousands of pages to a necessary few. They argue that supervisors appear to believe the problem all along was just the need for greater granularity and more model-based sensitivity, rather than fundamental flaws in the framework itself (its one-size-fits-all and portfolio invariance assumptions; and contaminating bank risk models by linking them to regulatory capital charges). They opine that banks have defended their risky business models very well and thus the reform process is not over: separating investment banking and imposing a sufficient binding leverage ratio (LR) remain on the table. They suggest that such a pre-emptive approach is needed to deal with the underpricing of risk and are dismissive of mechanical modelling of stress scenarios to estimate capital requirements.

Appendix to Chapter 8: Technical Details of Revisions to Securitisation , Market Risk and Counterparty Credit Risk Capital Charge Frameworks

Revision to the Securitisation Framework

Securitisation (discussed in Chapter 2) was perhaps one of the most important drivers of the crisis. Securitisation involves many players that ultimately generate investable instruments based on underlying pools of assets (that can be held on balance sheet or structured as liabilities of SPVs in senior, mezzanine and equity tranches with external credit ratings). The role of credit rating agencies in this process was very problematic (see Chapter 4). Banks can be originators of securitisations and/or investors in the liabilities. Regulatory treatment of off-balance sheet vehicles depends on the degree of control over the SPV . Where there is some control relationship, the revised rules try to reduce the role of external ratings—which is only partially achieved.

The new framework introduces a hierarchy of three choices: (i) if approved for the IRB modelling approach in the capital framework, then use it (and internal assessments can be used for asset-backed commercial paper); (ii) if not IRB approved but approved for external ratings, then use those; and (iii) where neither of these is possible, use the SA. The last of these becomes necessary in many cases as the new framework has more granular tranches (and also distinguishes the so-called thickness of tranches) that need to be rated. The SA applies typical Basel formulas to calculate the securitisation capital charge, based on inputs that start with the 8% capital charge that would apply had the exposure not been securitised (the underlying pool). It then makes adjustments for delinquencies and the attachment and detachment points of the tranches (see Chapter 4). If none of the above three approaches can be used, then a risk weight of 1250% is applied to the exposure.

There are likely to be mixed pools in this process—for example, where the IRB approach applies and but where the standardised approach is needed for parts that can’t be modelled. Further complicating the picture is the use of maximum risk weight caps for senior tranches based on good quality underlying pools (which might have risk weights as low as 15%). Furthermore, when all the capital charges are added up, the bank can apply a maximum capital charge of no more than what it would have been for the underlying (not securitised) pool of assets. The reason for these caps is that policy makers want to encourage the rejuvenation of the securitisation markets to help the economy.

The amount of discretion available to banks in this framework still allows plenty of scope for bank optimising their capital requirements.

Market Risk Reform

MR relates to risk of losses on- and off-balance sheet arising from changes in market prices of the so-called trading book assets. Recall from Chapter 2 that the capital to be required to cover for MR is included in the RWA formula scaled up by 12.5²⁸:

$${\text{RWA}} = 12.5({\text{OR}} + {\text{MR}}) + {\text{SUM}}[{{w}({i}){A}({i})}]$$

Prior to the fundamental review of the trading book and the resulting final reforms, MR was based on default risk and an incremental charge related to rating migration risk (rating changes that trigger mark-to-market losses) and an additional value-at-risk (VaR) capital charge based on a stressed scenario. Securitisation was excluded (treated separately from MR in the banking book). Otherwise, the boundary between what should be in the trading book and in the banking book was left unchanged.

The fundamental review is expected to significantly raise capital charges for MR .²⁹ It first clarifies the boundary between banking and trading books: there is now a clear list of presumptive inclusions in each based on intent-to-trade versus hold-to-maturity. Second, a more risk-sensitive approach is adopted for the standardised and internal models approaches to MR in the final version.

Standard Approach to MR

The SA to MR is the sum of three elements: (i) a sensitivity-based method (SBM) risk charge; (ii) a default risk charge (DRC); and (iii) a residual risk add-on (RRAO).

• For the SBM: there are seven risk classes: general interest rate risk (GIRR); credit spread risk (CSR) for non-securitisation; CSR for securitisations not correlated to the trading portfolio; CSR for securitisations that are correlated; equity risk; commodity risk; and foreign exchange risk. Risk factors are mapped into these risk classes: that is, variables in the pricing function used by the bank trading desk for given instruments for reporting profit and loss positions to management.³⁰ These risk factors are shocked for a stress scenario, and sensitivity risk positions are calculated. These are then multiplied by prescribed risk weights and risk positions with common characteristics (‘buckets’) are aggregated. Diversification benefits are allowed within each risk class but not between them.

• The DRC: is to capture stress events in the tail of the default distribution not captured by credit spread shocks in mark-to-market risk. A jump-to-default (JTD) is calculated, which is a function of the notional value of each position (separately), its market value and the prescribed LGD. Some offsetting diversification benefits are allowed so that net JTD can be calculated. These are allocated to risk buckets, and the appropriate risk weight is applied. Total DRC adds up these elements.

• The RRAO: is for instruments where model sensitivities cannot be calculated.³¹

The three risk positions are then added to obtain trading book MR (which multiplied by 12.5 gives the RWA ).

Internal Models Approach to MR

For the internal models approach (for approved sophisticated banks), the main innovation compared to previous Basel attempts is to move away from VaR method towards the concept of ‘expected shortfall’ (ES) under a stress scenario, which is somewhat tougher. VaR asked what the maximum 1-day loss in dollar terms would not be exceeded at the 99% confidence level over a given trading period. ES asks, conditional on a tail event actually occurring (i.e. being in the tail beyond the 97.5 confidence level), what could be the potential loss? This change requires risk factors for modelling losses to use more complex metrics from back-testing over a stressed period, using the finance concepts of delta, vega and curvature risk (see Glossary).

All banks must calculate the SA charge (even if they use the internal model approach). Much stricter conditions apply for banks being able to use internal models for nominated trading book portfolios. In particular, they have to be based on a database of verifiable market prices. Both the SBA and the DRC are treated as being able to be modelled for trading book exposures, but the RRAO is not.

Counterparty Credit Risk

CCR for derivatives (risk of bilateral counterparty default) is calculated either by using the standardised approach (SA-CCR) or by using internal models.³² CCR exposures are calculated as a multiple of the replacement cost of the exposure (for an immediate close out) plus the potential future exposure. The replacement cost of un-margined derivatives is the maximum of: (i) zero or (ii) the sum of mark-to-market exposures within a netting set, minus any related collateral held. For margined derivatives, it is the maximum within a netting set of: (i) zero; (ii) the sum of values in the netting set less associated collateral ; or (iii) the largest exposure that would not trigger a margin call net of any independent collateral held.³³ Potential future exposure is a multiplier (to allow for over-collateralisation) times an aggregate add-on for each netting set within an asset class (interest rate, foreign exchange, credit equity and commodity derivatives ) based on the notional amount, its maturity and its delta-adjustment to the underlying price.