believe the results, but then took too much time in making the critical decisions that, if made sooner, could have averted the crisis.
In response to the weaknesses exposed in their risk management systems, our interviewees say they are taking the following steps:
Adopting a comprehensive stress-testing approach. In line with the best practices described above, banks are currently defining their governance structure for stress testing, creating a set of complementary and complete stress tests, and developing a process for incorporating results into decision making.
Refining model assumptions. The crisis has illuminated new drivers of risk and linkages across drivers. Banks are therefore refining their models to capture these effects.
Limiting exposures to complex instruments. The crisis also has revealed a lack of understanding about how complex instruments behave under stress. In the current environment, banks prefer to be risk-averse and limit their exposures to such instruments.
As some market participants have observed, the fact that many banks use the same models led to herding behavior: their models all rang an alarm at the same time, causing everyone to “run to the door.” This effect seems to imply that using an alternative approach – one that might trigger an earlier warning in some cases – could create significant value. It also suggests that regulators aiming to ensure functioning credit markets during times of distress should try to promote adoption of more diverse models.
The current economic crisis is one of the worst to hit the banking system since its inception. There is at the same time, comfort in knowing that credit portfolio risk models have, by and large, performed well during the crisis, and that the crisis in no way brought an end to VaR models.
Given the rapid evolution of the markets, banks need to take a much more systematic and vigilant approach to managing and updating their risk models – regularly identifying the risks underlying their portfolio, developing more sophisticated methods to estimate the portfolio impact of those risks, and designing contingency plans for plausible scenarios. Likewise, they need to refine the assumptions and methodologies in VaR models to suit complex portfolio exposures and evolving dependencies across business units. In addition, banks need to complement these models with stress-testing frameworks that work in data-poor environments of heightened uncertainty from new sources of risk. Last but not least, bank leaders should continuously challenge the effectiveness of contingency plans and risk management measures.
Tobias Baer is an associate principal in McKinsey’s Taipei office, Venkata Krishna Kishore is a consultant in McKinsey’s risk practice, and Akbar N. Sheriff is a consultant in the San Fransciso office.