Back testing multi asset value at risk : Norwegian data

Abstract

This paper attempts to e stimate Value At Risk (VaR) for a multi asset Norwegian portfolio, using some of the most popular estimation methods , Variance Covariance Method, Historical Simulation and Monte Carlo Simulation . The Variance Covariance Method is applied with both time varying and constant volatility . Each VaR estimation method ’ s accurac y is tested , using Kupiec’s univariate test ing framework , for multiple single points in the left tail of the portfolio’s return distribution, and Pérignon and Smith ’s multivariate framework for a larger subset of the left tail. It compares each method ’s ov erall results for the Norwegian portfolio with those found by Wu et al. (2012) on a similar Taiwanese portfolio . And finally , based on the empirical testing , it attempts to draw a conclusion on which method is best suited for Norwegian data.