The Modeling Research Of Financial Assets Cojumps Based On High-frequency Data

Under the increasingly complex financial market environment, investors pay more attention to the risk assessment and management while seeking the maximize compensation for purpose,so the risk management become one of the important financial subjects. With the nonparametric estimation being widely used of the volatility estimation,scholars pay more attention to the daily rich information, which is important to the risk management Jump behavior research based on high-frequency data in the reality can well explain the impaction of the intraday abnormal events to the price and the risk. As jumps play an important role in accurately forecasting variance,cojumps make an significant rule in accurately covariance prediction in the same way, thereby it’s useful for the reasonable portfolio allocation and the optimization of asset management.Therefore,based on available literature and the high-frequency data,this paper will make some modeling research about the cojumps between different financial assets from following aspects:1)Systematically introduces the recent research on financial assets cojumps,and theoretically analyse jump-diffusion process, providing theoretical basis for the research belowed.2)Formulate the common different intraday jump test statistics, and carries on the comparison of these intraday jumping test statistics from empirical angle.By respectively using simulated data and empirical data,this section finds that the ABFN statistics has certain advantages in testing the intraday jumps.3)After the jump variance and jump covariance adjustment by using the ABFN statistics,build the MHAR-RV-CJ model to forecast variance and covariance, and compared the prediction accuracy with MHAR-RV model,this section approves that the cojumps is important of predicting covariance,and considering the cojumps will make more sence to predict the covariance.4)Based on the MHAR-RV-CJ model and MHAR-RV model,this section constructs a equally-weighted portfolio,measure the VaR and ES value of the equally-weighted portfolio,it results that the MHAR-RV-CJ model including jumps and cojumps perform better than the MHAR-RV model which doesn’t include jumps and cojumps.The ES measurement is more accurate compared to the VaR measument,which is the better risk measurement.The research is sponsored by National Natural Science Foundation of China:Research on Jump Behavior of Financial Assets Based on Realized Measurement Methods (NO.71171056).