A Comparative Study On VaR Implied By GARCH-JUMP Models

VaR (Value at Risk), is the measurement of the biggest loss for a risk position in a given probability for a certain period.In China,this tool is still at the stage of research and discussion.We use a set of GARCH-JUMP models to estimate parameters of time series.The financial time series we use are SHANGHAI SHENZHEN300index and share price index futures over the same period.These parameters are used to calculate VaR series. Comparing these VaR series,we know which model is best for VaR estimation. Broadly speaking the empirical part of this article has the following three steps:fitting time series to obtain parameters, calculating VaR sequences using the relevant parameters, analysising the performance of different VaR sequences.The innovation of this paper is to put the new algorithm of GARCH-Trend model in VaR calculation. Another innovation is to put forward a new set of evaluation indicators. In the empirical research, we also use VaR algorithm provided by Matlab.Besides the ARJI-R2model algorithm which deducts to an abnormal sequence,we compare the following four algorithms:GARCH-JUMP model algorithm, GARCH-ARJI model algorithm, ARJI-Trend model algorithm and Matlab algorithm.The first one is the best.Without considering skewness of the series,the Matlab algorithm performents worst.