Algorithmic Approaches to Stress Testing

Algorithmic Approaches to Stress Testing

	One drawback of historical and hypothetical stress scenarios is that the risk manager is never fully confident whether the entire risk of the portfolio has been exposed.

	An alternative is to design a systematic approach. An algorithm can be created that will search for the worst possible outcome for the portfolio within some feasible set of risk factor values.

	Three uses need to be addressed under this approach:
		To find out the relevancy of the relationships (like correlation and other measures of interdependence) used in the optimization for identifying worst-case scenarios.
		To measure the authenticity of the algorithm with respect to identification of worst-case outcomes globally within a given feasible set.
		To examine whether the feasible set includes plausible outcomes and again whether those outcomes are represented disproportionately in the optimal result.