Can risk modeling work?

Does the 2007/08 market crisis herald the end of risk modeling and the empirical method? This paper supports the hypothesis that recent risk modeling problems were caused by the use of inappropriate risk models which are fixable rather than fundamentally flawed. An extensive analysis including the sub-prime crisis shows that GARCH-based risk measures offer a potential solution to these problems. The paper also explores some risk modeling issues that arose during the crisis such as the appropriate choice of sample size and how to incorporate dynamic feedback effects into a risk model used for stress-testing. I illustrate a stress-testing method that applies the GARCH approach but results in relatively stable capital requirements preferred by practitioners. This method appears to address some of the concerns raised by regulators with respect to stress-testing practices during the market turbulence and would result in more conservative gearing levels for financial institutions.