Yield curve risk management: adjusting principal component analysis for model errors

Nicola Carcano

Abstract

ABSTRACT

Previous research has shown that the improved capture of the dynamics of the yield curve does not necessarily lead to better hedging. Here we claim that these observations can be explained by the level of exposure to model errors and tested a model of principal component analysis (PCA)-hedging that controls this exposure. The results confirm our claim. Controlling the exposure to model errors leads to an average reduction in hedging errors of 35%. Also, this adjustment leads three-component PCA to outperform twocomponent PCA, as theory would suggest. Finally, controlling the exposure to model errors leads to a substantial reduction in the transaction fees.