Parametric Versus Semi Parametric VaR Models: A Comparative Approach

In this article, we compare the operation of parametric and semiparametric multivariate GARCH models portfolio including TEDPIX, DJIA, and Nikkei 225 in order to estimate the value at risk. To surmount the problem of parametric models which ignore the information existing in standard errors, we use semiparametric models. Therefore, after estimation of conditional covariance matrix using parametric multivariate GARCH models, we calculate standardized errors by kernel regression estimator nonparametrically. Semiparametric covariance matrix is obtained by combination of parametric covariance matrix and nonparametric residual matrix, then with parametric and semiparametric covariance matrix estimated the value at risk and evaluated the result. The result represents the excellence of semiparametric to parametric in value at risk estimation.