Correlations under Stress

Empirically, correlations between equities increase in bear markets. However, in the industry-standard, (one-factor) Gaussian credit-risk model for a single sector, the tail correlations between equities converge to zero in a market downturn. This introduces significant model risk in portfolio credit-risk management, due to the resulting underestimation of losses from correlated defaults or downgrades in severe market-downturn scenarios. An alternative model is proposed, in which the systemic factor undergoes exponential, downward jumps and for which the conditional correlations increase to a certain level under stress. An example is given using Dow Jones US sector index returns.