Study on Portfolio La-VaR Analysis Based on Copula-Kernel Model

Abstract

Traditional VaR method has many defects in measuring portfolio risk, this paper modifies BDSS model and gets revised BDSS model - La-VaR model based on relative price. For fitting the sequences of the rate of return and relative price, this paper adopts Gaussian-kernel function with good smoothness and Copula-kernel model to portray marginal distribution and correlation structure. Afterwards sequence of empirical distribution is produced through Monte Carlo simulation. The empirical results show that Copula-kernel model has a high accuracy in fitting sequences of the rate of return and relative price. Liquidity risk in La-VaR model is getting significant as the decrease of confidence c, but back testing shows that both VaR and La-VaR model overestimate the risk.