Research On Long Memory And Fractional Cointegration In Financial Time Series

Lots of evidences indicate: traditional econometrics don't solve financianl time series,which is non-balanced and short-memory.This paper generalize long-memory concept, modeling, checking and evaluation on the basis of documents both here and abroad.The concept of cointegration describes the long-run linear equilibrium relation among vector time series.The traditional statistical method and dynamic model sepecialized theory is subtly mixed in cointegration modeling theory and the economy construction model describing unknowable data generation process is built,so regularity in financial time series can be illustrated more objectively.At present,research on vector time series cointegration mainly focus on the series whose variable's difference is integral and stationary,whereas the non-stationary property of time series can't be fully depicted by integration dimension.Long memory property is widely occurred in the time series of actual system and the difference of long memory time series is sometimes fractional dimension.In that,this paper cointegration of long memory vector time series modeling is present .This model enlarges the connotation and module of cointegration and illustrates the concept of long memory vector time series cointegration.In the meanwhile,compare relation of cointegration and persistence,and combining cointegration and vector FIGARCH, derive from second -order long term equilibrium.At last,the writer tests the model and method discussed above with date of Shanghai and Shenzhen stock markets to show the effectivenes.