Evaluating market risk in a portfolio with heavy-tailed risk factors using Monte Carlo Methods

In this master thesis we study and implement a model for market risk in a portfolio consisting of both stock and bond indices. It is based on an article by Glasserman, Heidelberger and Shahabuddin from year 2002. The model uses the assumption that the joint distribution of the losses follows a multivariate t-distribution. The model also uses Monte Carlo simulations with importance sampling in order to improve the performance of the simulations, which is necessary to achieve statistical certainty when working with high percentiles of the losses. We focus mainly on Value at Risk, but we will also mention Expected Shortfall. We test the certainty of our model in numerous different ways