Nonparametric inference of value at risk for dependent financial returns

Abstract

This thesis considers the estimation of a popular financial risk measure, Value at Risk (VaR), for dependent financial return series. Two nonparametric estimators for VaR and their properties are examined. The involvement of the dependence when considering the estimation for the VaR makes the inference more interesting and affects the variability of the estimates. The estimation of the standard errors of the proposed nonparametric estimators is studied based on the kernel smoothing of the spectral density of a derived series. The performance of the estimates has been evaluated by simulations of the commonly used financial models of returns and their applications to real financial time series.