Far from the Modelling Crowd

It’s been clear for some time that the enthusiasm for internal bank capital models has been waning within certain parts of the regulatory community. The latest signal of that decline is a recent Basel Committee on Banking Supervision proposal to restrict the use of models for the calculation of credit risk-weighted assets. Within that proposal is a very clear decree: the use of the internal model approach (IMA) for the calculation of credit valuation adjustment (CVA) capital is no longer allowed.

The timing of that announcement came as a surprise. One reason given for the decision is that the Basel Committee has doubts CVA can be effectively captured by an internal model. However, only one quantitative impact study (QIS) has so far been completed on proposed revisions to the CVA capital framework, and that was hindered by lack of completeness, an absence of clarity (particularly over the treatment of portfolio hedges) and time constraints (the QIS was run in conjunction with a QIS on the Fundamental Review of the Trading Book (FRTB), but with less preparation time to make the necessary changes to bank systems).

As a result, the Basel Committee launched a second, comprehensive QIS earlier this year. But the decision to eliminate internal models for CVA has been taken while that second QIS is still in progress, before the Basel Committee has even seen the data submissions from banks.

Regulators further justified their decision to eliminate IMA-CVA by noting that CVA capital will be significantly reduced anyway due to greater use of central clearing and the introduction of margining rules for non-cleared derivatives. That’s true, but certain counterparties – non-financial corporates and sovereigns, for instance – are exempt from mandatory clearing and margining. As a result, trades with these end users will be subject to a CVA charge calculated using a standardized or basic approach. An overly conservative methodology will therefore particularly affect those counterparties.

The decision to eliminate the IMA-CVA follows other, similar developments elsewhere: a requirement for all banks to model market risk capital using a standardized approach, and for those outputs to potentially act as a floor for internal models; a proposal to introduce capital floors more broadly; the emergence of non-risk-based backstops such as the leverage ratio; and the ditching of the advanced measurement approach for operational risk.

Some regulators have highlighted complexity and variation in risk-weighted assets (RWAs) as a rationale for wanting to restrict the use of internal models. ISDA understands these concerns, but believes there are ways to address trepidation about RWA variability without eliminating internal models – through greater consistency of model inputs or through ongoing testing procedures, for instance. Opting instead for a broad restriction in the use of internal models, or disallowing their use entirely, has several important implications.

For one thing, internal models are much more sensitive to risk and better align with how banks actually manage their business. In comparison, standardized models are relatively blunt, meaning the required capital charge for holding a particular asset might not adequately reflect its risk. This can lead to poor decision-making: a bank might choose to pull back from low-risk assets, counterparties or businesses where capital costs are relatively high. Conversely, they might opt to invest in higher-risk assets that appear attractive from a capital standpoint. These issues were what prompted the Basel Committee to create incentives for the use of risk-sensitive internal models in the first place via Basel II.

We believe, as a general point, that capital levels should reflect risk as closely as possible. A less risk-sensitive capital framework leads to the possibility of a misallocation of capital and an increase in systemic risk. Making decisions in a business that is intrinsically about taking and managing risk, based on a capital framework that is being made purposely less risk sensitive, creates its own hazards – as described in this recent article from Risk.

Another likely impact of this shift away from models is an increase in capital. That’s because standardized approaches tend to be more conservative. For example, an industry study on a draft of the FRTB rules, conducted by ISDA and other trade associations last year, revealed a move from internal models to the standardized approach would result in a jump in capital of between 2.1 and 4.6 times, depending on the trading desk. The Basel Committee has since published its final FRTB framework, and ISDA is involved in another study to determine whether these cliff effects still exist.

The Group-of-20 nations and the Basel Committee have both stated that further refinements to the capital framework should not result in significant increases in overall capital levels. We believe that is the right approach given the significant increases in capital that have occurred already as a result of Basel III. The challenge is that each individual measure tends to be considered in isolation. So while a single refinement might not meaningfully increase overall capital levels by itself, it might, when combined with all the other little tweaks, end up leading to higher capital levels in total.

Only a comprehensive impact study to determine the overall effect of all the changes together, including the changes to models, will provide the answer to this. And that should occur sooner rather than later.

Leave a Reply