Risk Models after the Credit Crisis

Abstract

The 2008 credit crisis came as a complete surprise for almost all financial analysts and practitioners. The usefulness of ALM models has been questioned as the crisis was highly unlikely according to these models, and its impact on volatilities and correlations was not foreseen. This paper proposes a novel scenario generator model for ALM to partly remedy these flaws. A possible start of a crisis is modeled by including a low-probability jump process, that leads to a significant drop in the stock market, lower risk-free interest rates, and a severe increase in credit spreads. The risk characteristics of the crisis are captured by allowing for time-varying volatilities and correlations. The time-varying correlation between stock and bond returns is driven by two sources: monetary shocks leading to a positive correlation, and risk-aversion (or ‘flight-to-safety’) shocks leading to negative correlation. The time variation in correlations is due to the changing importance of these sources. The model stays within the essentially ane class, thereby allowing for closed-form solutions for arbitrage-free nominal and real bond prices of all maturities. The model can reproduce the time variation observed in both the conditional variances of yields and expected returns to bonds. Moreover, excess returns of both stocks and bonds can be explained by the same pricing model.