ABSTRACT

This paper presents a new probabilistic definition of component value-at-risk (VaR). It is shown that, for well-behaved return distributions, this definition is equivalent to the standard definition of component VaR. However, this new approach is convenient because it has a graphical interpretation and facilitates rigorous analysis. Furthermore, it leads naturally to component VaR estimators for both parametric and scenario-based risk management systems. These estimators are applied to a sample portfolio and present a simulation to recommend which estimator should be used in practice.