Diego Zuluaga // 28.04.2016

The Institute for International Finance has just released an interesting briefing paper on the trends in risk-weighted assets (RWA) among major banks before and after the financial crisis. RWA is the main measure of risk mandated by the Basel Committee, the forum of central bank supervisors charged with global standard-setting to achieve financial stability. The first such regulations, known as Basel I, were issued in 1988. They were subsequently updated in 2004 (Basel II) and, following the crisis, in 2009 (Basel III).

The IIF paper’s main objective is to show that banks’ internal risk models, which some in the Basel Committee now blame for the risk-management failures during the financial crisis, cannot actually have been to blame because most banks only began using them in 2008. The authors note that, since the implementation of this Internal Ratings-Based (IRB) approach, average RWA for major global banks has gone up, suggesting more caution and better risk management. The implication is that banks should continue to be allowed to use internal models when complying with capital requirements.

The IIF is responding to Basel’s plans to partly centralise credit risk measurement, which seek to prevent banks from gaming the system. The Committee presumably fears that, under the IRB approach, banks have a certain amount of leeway to tweak their models so as to increase returns without any increase in measured risk. The crucial point is that risk would in fact have increased but would not be reflected in the banks’ internal risk models, therefore placing them in a more vulnerable position were a crisis to hit. This in turn could have systemic implications if one bank’s weakness was perceived as reflecting wider problems across the system, as was the case in 2008.

There are a number of problems with using an increase in RWA as indicative that internal models are working. Firstly, the figures are averages and mask potential discrepancies between institutions and across jurisdictions. Secondly, in the absence of an effective way to judge whether a certain RWA ratio is sufficient, an absolute increase in the average ratio has little meaning. Additionally, it is important to remember that the weights assigned to various types of assets are determined by regulators, and that they have failed before to capture the true risk – and especially risk correlations between similar instruments – of bank balance sheets.

That doesn’t mean, however, that the authors lack a point. For the most part, banks know their balance sheets better than external observers, including regulators, so in principle the idea that internal models pre-approved by public authorities should be used for risk assessment is not at all imprudent. However, that assumes incentives for regulatory arbitrage do not compromise the objectivity of the model. It also assumes a reasonable predictive ability and information on the part of whoever is doing the modelling. Because criteria are determined globally by the Basel Committee, any mistakes made will be magnified on an enormous scale. Mortgage-backed securities in the run-up to the last crisis are a poignant example.

Furthermore, in any risk management – whether public or private – there is arguably a trade-off between simplicity – a set of rules which are clear, easy to understand and hard to game – safety – using estimates which will be robust even to hard-to-predict fat-tail events – and accuracy – measures which reflect the true risk contained in a balance sheet at a point in time. This is not to say that one cannot have the three, but that – for instance – aiming for simple and safe measures will often result in overly conservative estimates which do not accurately reflect risk – rather, they overestimate it – thus hampering the flow of capital. Similarly, one could try to develop criteria which are both safe and accurate, but the resulting rules would probably be numerous and complex, making risk management by private firms costlier and more difficult, and gaming harder to supervise by regulators.

As mentioned before, these problems are hugely magnified in systems of centralised regulation where most criteria are determined by statute on a global scale. In a competitive system where banks were able to develop their own ways to manage risk, there would be as many attempts to balance simplicity, safety and accuracy as there were banks, which over time would be more likely to deliver better solutions. In contrast, the Basel Committee’s approach reduces flexibility, magnifies the costs of regulatory error, and fails to adapt to changing circumstances in time to react to new problems. Moreover, the criteria determined at Basel are far from a merely technocratic exercise, but are subject to intense political bargaining by countries. This further weakens the Committee’s effectiveness.

Of course, in a world of state-backed deposit insurance, implicit bailout guarantees and many statutory interventions into financial markets it is difficult to imagine a competitive system of private risk management. But that doesn’t mean that the alternative – publicly set criteria, implemented (modelled) either by the regulators or the regulated – is satisfactory.

Given the incentives for gaming and regulatory arbitrage that come with complex rules, and the experience of risk weights in the previous crisis, it is at least warranted to consider other, simpler and more robust, indicators of banks’ risk exposure, such as plain leverage ratios of assets over equity. These are more straightforward than RWA, can be set at conservative levels and – according to the BoE’s Andy Haldane – are no worse predictors of risk than risk-based measures. It may well be that simple heuristics – rules of thumb – are better for risk management than complex, pseudo-scientific formulas.
The conclusion would seem to be that, rather than spend its energy debating the propriety of internal models, Basel should aim to move away from risk-based measures altogether.