Linking the problems of estimating and allocating unconditional capital

Abstract

This paper addresses two problems related to determining the unconditional capital required by a credit portfolio: Estimating it using Monte Carlo simulation and allocating it among the different risk units that form the portfolio. By elaborating on a tractable analytical framework, we propose a new simulation algorithm and a new allocation
method. Both contributions rely on the conditional loss distributions and share the same core idea. We discuss their optimality, consistence and practical advantages. In an empirical study based on American data, we show the remarkable gains in efficiency achieved by the former and the improvement in the standard variance-covariance allocation provided by the latter.