A Comparison About Jump Tests Based On High-frequency Intraday Returns

From the returns on assets to the volatility to the jump behavior,our research has been more and more detailed,it all depends on the convenience of the high frequency data's acquisition.As an important part of volatility and volatility is the basis of all kinds of pricing model and the index of market risk management,the jump behavior research is becoming more and more important.Economists at home and abroad have put forward and applied a variety of jump test methods.These methods separate jump variance from intraday volatility which could be used in volatility forecast model.HAR-RV-CJ model as a kind of separate jumps and volatility of forecasting model makes a great contribution to the research.For the emerging stock market of China,its significant heterogeneity and irrational features make the jumps happens more frequently and disorderly,also makes the study of jump behaviors more meaningful.For China's stock market,however,exactly which kind of jump tests looks more suitable has been a problem without conclusion.This article aims to comparing various jump tests used in China's stock market from the perspective of the empirical study.In addition to comparing the number of jumps,jump characteristics,this paper also focuses on comparing the fitting results within the sample and the prediction results out of sample of HAR-RV-CJ model which separates jump variance and continuous path of variance.At the same time,the prediction period is divided into 22 parts and comparing the results with SPA test saperately.Through the study of the csi 300 index data,we found that there isn't big difference between the results of fitting,coefficient of significant degree level as well as the goodness of fit are within the acceptable range.But,according to the prediction results from the sample,ZLog-MedRV has obvious advantages considering both the total number of rejected and refused to each cycle times.At the same time,the advantage is more and more small with the extension of forecast period,that is to say,when making long-term predictions,jump test methods have on difference.