Estimating market risk metrics using gaussian mixtures [Estimación de métricas de riesgo de mercado usando mixturas gaussianas]

The most commonly used financial models for the estimation of market risk either assume that asset returns follow a Normal distribution or are based on the empirical distribution. More often than not, the normality assumption is taken for granted. However, it is not realistic due to skewness and excess kurtosis observed in the actual behavior of asset returns. In this work we show evidence that finite Gaussian mixtures are an efficient model for the distribution of asset returns. We study the model and obtain expressions to estimate the usual market risk metrics. We illustrate its application by estimating risk figures for a portfolio of Mexican assets using the proposed model and comparing them against values produced with the most widely used models. © 2015 Universidad Nacional Autónoma de México, Facultad de Contaduría y Administración.