While the recession has shaken the banking industry to its core, one of the biggest shocks for banks has been the realization that current capital reserves are insufficient to protect them during a crisis. Bankers and their regulators are asking some hard questions: Are the models we use to estimate capital requirements faulty? Are the inputs to those models of inferior quality? Did our risk governance processes fail? Did management respond too slowly?
Over the past decade, the increased complexity of banking instruments and the heightened need for accuracy have made models an indispensible part of portfolio risk estimation. Value- at-risk (VaR) models have gained acceptance as a credible approach to estimating overall portfolio risk. In the current environment, however, the performance threshold for these models has risen – and banks are keen on making more informed choices on their modeling approach.
With this interest on the part of banks in mind, we recently interviewed the risk management teams from 11 of the world’s leading banks to understand the landscape of credit economic capital modeling techniques. For most of these institutions, corporate assets make up more than 50 percent of their portfolio. In addition the more recent risk-management interviews, we also conducted a survey in 2007 of 22 banks, half of which are focused on corporate customers, the other half on retail. The findings summarized in this report are based on both the interviews and our earlier survey. In both cases, the participating institutions represented a diverse geographic mix of banks based in North America, Europe, and Asia-Pacific.
Through our research, we sought to answer the key questions faced by banks that either run or are setting up a credit economic portfolio system:
What characterizes an ideal credit portfolio model?
What is the role of stress testing and the best practices for managing it?
What lessons can be learned from the current economic crisis?
Our findings indicate that banks consider economic capital a valuable concept, but they have yet to nail down the ideal approach to estimating it. Our interviewees agree that a VaR model with a complementary stress-testing framework is necessary. Many banks are developing proprietary models because none of the existing third-party VaR models accurately measure the risk in portfolios constituted of both retail and corporate assets, with vanilla and exotic products.
We found from our discussions that the existing VaR models differ in three important ways:
Approach to correlation. An asset-based correlation approach is suitable for a corporate-focused portfolio, whereas an approach based on default intensity is suitable for a retail-focused portfolio.