Threshold Model And Its Application In China's Macroeconomic Research

Economics theory shows that many important macroeconomic time series have nonlinear character.When we analyze macroeconomic time series without thinking about this,we will get wrong conclusion.So if relax the nonlinear restriction it can improve our analysis.The paper was divided into 7 chapters.Chapter 1,which is the Foreword, introduces the writing background,research method,framework of the paper.Chapter 2 makes a retrospect for nonlinear models,the development of STAR model,tests of stationary in STAR framework,the union tests of stationary and linearity of STAR model.Chapter 3 we proposed a simple test to distinguish between linear unit root processes and stationary nonlinear LSTAR processes is proposed.New limiting nonstandard distribution results are provided.Chapter 4 considers the OLS and GLS detrending procedure for unit root tests against alternative hypothesis where the time series data under investigation follow either globally stationary LSTAR or ESTAR processes with deterministic components being present via Monte Carlo simulation. Chapter 5 we proposed a Wald test to test the union of linearity and nonstationary of STAR model.Chapter 6 is experience analysis including the Asymmetric Effects of Monetary in China and The Monetary Policy Uncertainty and It's Effects on Macroeconomic stability in China and Inflation,Inflation Uncertainty,and Monetary Policy uncertainty in China and Measuring Expected Inflation and the Ex-Ante Real Interest Rate in China.Chapter 7 makes a conclusion for the paper and point the future analysis.We focused on theoretical analysis in the paper;we relax the restrictions of the model and improve the size and power of the new test.First on basis of George Kapetanions,Yongcheol Shin and Andy Sneli(2003)'s analysis of unit root test in exponential smooth auto regression;we extend the test to logistic smooth auto regression with concept and trend.New limiting nonstandard distribution results are provided.We find the critical value for our test via Monte Carlo simulation exercises. We also find that under the alternative of a globally stationary LSTAR process,our proposed test has better power than the standard Dickey-Fuller test.The second we consider the OLS and GLS detrending procedure for unit root tests against alternative hypothesis where the time series data under investigation follow either globally stationary LSTAR or ESTAR processes with deterministic components being present via Monte Carlo simulation.It is found that the proposed testing procedures have considerable power gains against existing nonlinear unit root tests recently proposed by Kapetanios et al.(2003) and the chapter 3.If there is only concept,OLS is better than GLS in ESTAR.In LSTAR,when T＜50,OLS is better,but when T＞100,GLS is better.If there is concept and trend,when T＜50,GLS is better in LSTAR,but OLS is better in ESTAR.When T＞100,GLS is better in ESTAR,but both are very good in LSTAR.The third we consider the union test of the stationary and the linearity of STAR model.We found that Wald test may have size and power distortion.Experience analysis includes several aspects.The first the structural vector auto regression methodology is used to decompose the China nominal short-term interest rate into an expected inflation and an ex-ante real interest rate component.To identify the structural vector auto regression model,we restrict the long-run effect of an ex-ante real shock on the nominal rate to zero.Then we use HP filter to get the trend of the ex-ante real rate and cumulate the short-run shock in the structural vector auto regression to get the stochastic component of the ex-ante real rate.The ex-ante real rate may be a useful indicator of the monetary policy stance of China.The second we constructed a time series of monthly inflation and money supply in China from 1992 to 2007 using time-varying parameter and time-varying parameter markov switching models and investigated the link between inflation,inflation uncertainty,and monetary policy using Granger tests.Through the models the author gets long-term, short-term,and expected inflation uncertainty.Also through the models money supply uncertainty can be decomposed into two components:the policy uncertainty and the real economy shock uncertainty.The author found strong statistical support that inflation uncertainty is the Granger reason of inflation in high inflation term.The evidence on the effect of money supply uncertainty especially the policy uncertainty on inflation uncertainty is strong in high inflation term.The third we consider the LSTR modeling approach for the demand for money in China.We use Ter(a|¨)svirta (2004)' procedure to test the linearity of an error correction model against a smooth transition regression(STR) non-linear alternative model.It found there is a critical threshold figure for GDP growth rate affecting real money demand in China,at about 8.8%.The high and low GDP growth rates regimes are quiet different form each other, so that non-linearity in the model is strong.The fourth we graft asymmetry onto a aggregate supply-aggregate demand model considering a simple flexible price.The impulse response functions from the LSTVAR model are used to show the asymmetric effects of monetary policy.The empirical study shows the effect of monetary policy possesses asymmetry in policy direction in recession,but it doesn't pose in prosperity.The efficiency of the expansionary monetary policy has asymmetry in terms of the business circle,the inflationary monetary does not have. Both of the monetary has effective compact on output.The effect of monetary policy possesses asymmetry in policy direction in both in inflation and deflation.The efficiency of the expansionary monetary policy has asymmetry in terms of the business circle,the inflationary monetary does not have.Both of the monetary policy has not effective compact on price.These results show that output in China is below full employment and the aggregate supply curve is not kinked.