Reducing asset weights' volatility using importance sampling in a stochastic credit portfolio

Abstract

The aim of this dissertation is to improve portfolio risk management with the aid of variance reduction techniques. Especially in rare events that generally entail high losses, proper portfolio risk modelling can benefit from reducing great losses. In order to achieve lower credit risk, optimal asset weights, Value-at-Risk (VaR) and Conditional Value-at-Risk (CVaR) measures were optimized by a method based on simulation scenarios. Simulation for this particular study was done through two models. The first model utilizes original crude stochastic optimization, whereas the second model makes use of the variance reduction technique Importance Sampling (IS). In fact, when comparing the relevant results obtained by these two models, one would immediately note an overall positive variance reduction. This implies that augmenting the traditional stochastic optimization procedure by the variance reduction technique, a more efficient risk model was achieved due to lower resulting variance in all considered measures. All modelling is done by use of Matlab.