Metodologia de estimação de matrizes de correlação de ativos e seu impacto no uso em estimativas de capital econômico para risco de crédito

Abstract

The estimation of losses distribution in a credit portfolio and related statistical measures, such as VaR, Economic Capital and Expected Shortfall, depends on several parameters, according to the model adopted. In Merton-based models, correlations play a key role. In credit risk management, one can define correlation as a measure of connection between events. The premise of independence (or lack of correlation) is not accepted in the literature of the subject, nor is it verifiable in practice. It is necessary to estimate measures of correlation with the systemic factor and between counterparts, which leads to an asset correlation matrix. This article presents a methodology for the estimation of asset correlation matrices, compatible with the Merton-based one factor model, and analyses the effect on statistical credit risk metrics when compared to the independence hypothesis. The theoretical model uses the probability of default and its volatilities to estimate the correlations between assets.