The current enthusiasm for simpler measures of banks’ capital solvency is understandable but dangerous. Abandoning risk-based measures in their favour could make the financial system less safe while simultaneously impeding the flow of credit to the real economy.
Take the so-called “leverage ratio”, whose main selling point is its simplicity. There is no risk-weighting of assets. It measures a bank’s capital against a straightforward accounting definition of its assets. Yet this means a dollar of very low-risk assets backed by collateral is treated exactly the same as a dollar lent to a risky borrower on an unsecured basis. So as a true measure of solvency, the leverage ratio fails since it tells you nothing about the nature of the assets.
Moreover, treating risky and safe assets the same way when deciding how much capital banks should hold will have predictable consequences. Banks will shed safe assets with low returns (since they will have to put too much capital against them) and focus on riskier products and segments with higher returns.
I guarantee that a regulatory framework that uses the leverage ratio as a primary measure will make banks and the banking system as a whole more prone to crisis. And using the leverage ratio in this way will, in addition, constrain the provision of mortgages and trade finance. This will damage economic growth and job creation.
The leverage ratio can play a useful role as a backstop and cross-check for regulators. It would only occasionally bite directly, but its existence would focus attention on banks’ relative leverage and on changes in leverage over time.
When I make these points to proponents of tighter leverage ratios, the argument quickly switches to the flaws in risk-weighting. How can people trust banks using their own models when they are so opaque and produce such different results? This concern underpins the enthusiasm for the leverage ratio and why some advocate a return to the so-called 'standardised approach'.
Under this approach, there are no sophisticated models, simply standard risk-weightings for different categories of asset that all banks have to apply. This gives the appearance of comparability, but it is entirely illusory. Massive differences in risk profile are simply smoothed out by very crude assumptions. Like the leverage ratio, it assumes lending to a start-up and an established multinational are equivalent risks, and takes no account of collateral. Indeed, imposing the standardised approach amounts to imposing a very poor model we know is wrong.
The standardised approach and leverage ratio share two characteristics. First, they simplify a complex reality. But the allure of simplicity should be resisted if the simplification so dangerously distorts and obscures the real picture. Second, they narrow the difference in regulatory approach between risky and safe assets, creating perverse and powerful incentives for banks to run higher risk portfolios.
Instead we should focus on working to make risk-weighting function better. The fact that different institutions give apparently similar assets different risk weightings is a real problem, and it has created a serious credibility gap.
Some of the differences between model results are good, reflecting real variations in intrinsic risks and in the effectiveness of different banks’ risk-management approaches. But some differences are bad. They are the outcome of unwarranted differences in methodology, data sets or technical definitions. There are also variations driven by different parameters imposed by different regulators – some sensible, some fairly arbitrary. We can fix these: it is possible for banks and regulators to iron out most of the bad differences; the enhanced disclosure the industry is implementing will enable investors and regulators to scrutinise the residual ones.
We should also take the opportunity to address some longstanding weaknesses in risk-weighting. The structure of these internal models and many important parameters are imposed by regulators. Indeed, every model, and every material change is approved by regulators. But some crucial aspects of the framework are more than 20 years old. The models do not capture tail risks well, deal effectively with low data portfolios or incorporate diversification benefits. The underlying mathematics is dated. While no model that predicts the future will be perfect, we can certainly make them much better.
I am convinced that improving risk-weighting must be at the heart of a robust approach to bank solvency, and that we can remedy its most significant flaws. Yet I would never argue that it should be the sole tool. The leverage ratio provides a good backstop. Stress tests can also reveal potential blind spots or correlated risks. Probing supervisors and challenging boards can play a crucial role in ensuring we do not become too model driven. While a robust, risk-based measure of capital solvency is crucial, we should not be too reliant on any one tool.
Peter Sands is chief executive of Standard Chartered.
Copyright the Financial Times 2013.