Back-testing the VaR risk measure: an empirical study

Abstract

This thesis verifies the worst case losses (Value-at-Risk) of financial returns over a specified time period with a certain level of confidence. The measurement of VaR hinges on the distribution of investment returns. In order to test whether or not the VaR model accurately represents reality, back-testing is carried out for one day horizon for a yearly rolling window. The standard VaR parametric model which is based on normal distribution of returns is tested on real data. Findings are that this model is better for historical VaR estimation for bigger exceedance probabilities such as 5%, 1%, 2% etc, while the Student’s t-distribution seems to be better for smaller exceedance probabilities such as 0.5%, 0.1% etc.