The Analysis Of Threshold And Fractional Cointegration And Its Applications In Economics

In nearly twenty years, it has been focuses of Time series analysis to test and model nonstationary time series. Theory of cointegration provides a new approach to model nonstationary time series. Cointegration means variables of system can keep long-term equilibrium, which can be reached by adjustment of variables of system themselves. The long-term equilibrium is considered to be short-memory and the adjustment is considered to be symmetric by classical cointegration theory. Research suggests that, however, the long-term equilibrium is long-memory, and the adjustment is asymmetric. So, this paper introduces threshold cointegration and fractional cointegration. The former is about asymmetric and the latter concerns long-memory. The key of cointegration are testing and estimating of it. In this paper I discuss testing of threshold cointegration and fractional Cointegration. The following five parts are included.Firstly, background of the theory of cointegration is recalled. And the theoretical basis for the existence of spurious regression is inspected. We also elaborate the Basic Content, basic form, the popular test and estimation methods of the cointegration.Secondly, threshold cointegration is introduced for the asymmetric of the long-term equilibrium between system variables. We also discuss the model, testing and parameter estimation of the threshold cointegration. It is a creation that we bring up the threshold cointegration testing based on the GLS method. It can be prove that this kind of testing can greatly improve the power of the threshold cointegration testing.Thirdly, in this paper cointegration theory is extended to more generally occasions for the existence of the long-memory of long-term equilibrium and fractional cointegration is introduced. I analyze the model, testing and estimation of fractional cointegration as well. Another creation is FDF testing of fractional cointegration in this paper. Monte Carlo simulation_establish that this test improve the testing power greatly.Fourthly, Bayesian methods are used to study fractional cointegration. This further extends the fractional Cointegration study method. We can identify the cointegration rank of a model by posterior probability of cointegration rank; Thus Cointegration relation can be identified. Monte Carlo simulation proves that the identification rate is rather high.Fifthly, there are applications of threshold cointegration and fractional cointegration in this paper. Consider the Non-Linear Effects of Fiscal Policy. I use threshold cointegration testing to test the sustainability of budget deficit. I also calculate the dynamic critical value of budget deficit Consider the fractional cointegration between the spot prices and futures prices of Stock index futures; we use fractional Vector Error Correction Model (VECM) to estimate the optimal hedge ratios of stock index futures.