Empirical research of two-factor Vasicek model based on copulas

Abstract

It analyzes the non-linear term structure of interest rates of Treasury in Shanghai Stock Exchange, using two-factor Vasicek model and Kalman filter in the state-space framework. The observation errors of one-year and 20-year interest rates are extracted and estimated applying the Archimedean copulas and mixture copula by maximum likelihood estimation. It has shown that the Gumbel copula and mixture copula could be more proper to capture their dependence structure. It is also found that Gumbel copula is superior over Gaussian copula, Frank copula, Clayton copula and mixture copula to test the VaR of Treasury portfolio by Monte Carlo simulation, for the latter four categories of copulas would underestimate the risks of portfolio.