The Theoretical and Empirical Research of Threshold Cointegration Based on the Equilibrium-Deviation Term
|School||Huazhong University of Science and Technology|
|Keywords||Threshold cointegration Equalizer error term Threshold autoregressive model Grid search sup-Wald Block bootstrap based on residuals Term Structure of Interest Rates|
The theory of the Engles and Granger cointegration 1987 (cointegration) to help raise awareness of the long-run equilibrium relationship between the non-stationary time series, the theory assumes that the time series of the trend back to the deviation from the long-run equilibrium adjustment is continuous. With further research, a lot of evidence that adjustment costs or policy intervention due to the presence of many important economic and financial time series does not comply with the above assumptions, the long-run equilibrium trend back to adjust the non-continuous. Therefore, the use of traditional cointegration theory of such time series cointegration test will be larger deviation, or even cause erroneous statistical inference. The threshold cointegration theory of (threshold cointegration) It is in this context that the progressive development together. The theoretical threshold autoregressive (TAR) model accurately portrays the time series trend back to adjust the long-term equilibrium of the non-continuous. The theory of the research focused on the threshold estimation method and Threshold Cointegrated test methods. Depending on the research framework, existing-door limit cointegration mainly divided into two categories, a class-based the cointegration system of balanced error term, and the other based on the threshold vector error correction model (TVECM). Based Cointegrated Systems equalizer error thresholds is estimated and doors limit cointegration test methods were studied, and the use of the above estimation and testing methods empirical analysis of U.S. interest rate term structure. The first set several common door threshold regression model, and analyze their data characteristics; Second method of Monte Carlo experiments on two common door limit parameter estimation method (arranged from the regression method and grid search method) the estimation accuracy under different sample compared to the small sample grid search method is more reliable conclusions; This article focuses on establishing a two-step threshold cointegration test method, a residual sub-block (RBB ) bootstrap algorithm were constructed to achieve the first step of sup-Wald statistic nonlinear unit root test and the second step of the cointegration test; same time, the method of Monte Carlo experiments on the finite sample properties of the unit root test and ADF test contrast, proved to have higher test potential RBB bootstrap sup-Wald test than ADF test in the context of a threshold effect; empirical analysis of the term structure of interest rates in the United States, we get existence of threshold cointegration conclusion.