STR Models And Nonlinear Studies Of Chinese Monetary Policy Transmission

Before 1980s, classical linear time series econometrics had the dominant place.However, in practical application, classical linear time series models, such as standard linear time series models (such as AR(p) / ARMA(p,q)) ignore structure change of the model, which leads to failure of forecasting 1970s, oil crisis and can not explain the phenomenon of"Black Monday"happened in 1987.And until now, economists can not find out significant information variation which led to breakdown of the stock markets. For above reasons, traditional linear time series modeling is faced to more and more challenges. At the same time, nonlinear science developed from 1960s is changing traditional view of people towards the world and more importance is attached to it by science circles. With the development of computer technology, more and more nonlinear theoretical models can be chosen to depict nonlinear relationship between variables studied by economists. Unfortunately, theoretical modeling often does not translate easily into a testable time series econometric model. Smooth Transition Regression or STR models have the following advantages which make Smooth Transition Regression or STR models a good first choice for time series modeling: Firstly, STR models possess a set of operable procedures of estimation and test. Secondly, STR models can be applied to different types of data. Finally, STR models possess rich economic sense. Especially, more and more importance is attached to STR models when studying the problem of monetary policy transmission mechanism.When applied to study the problem of monetary policy transmission mechanism, traditional linear time series models are on the assumption which they implicit that micro individuals have the same reaction to monetary policy. For example, in different economic phases, such as in a high inflation phase and in a low inflation phase, there is the same linear relationship between the variables. However, the fact is, in different phases, different economic behavior of individuals may lead to nonlinear relationship between the economical variables. For example, in a high inflation phase, micro individuals always have strong desire to adjust their cash in hand, but in a low inflation phase, micro individuals may be insensitive to prophase inflation because of adjust cost. When carrying out monetary policy we also find out the contractive monetary policy can effectively keep down economic overheat, but expansionary monetary policy has no effect on controlling economic declining. However, traditional linear time series models (such as AR (p) / ARMA (p, q)) can not describe the properties of diversity and asymmetry of above economic behavior. But most studies about Chinese monetary policy transmission mechanism are made by applying linear time series models, or Markov Switching Regime model or MSR model is also adopted in few nonlinear time series studies on Chinese monetary policy transmission mechanism. In a MSR model, the transition variable is not observable and much information is needed when we decide in which phase the variable we study is, and wrong information may give wrong conclusions. Further more, generally a MSR model can not describe the type of nonlinear transition, and only can infer the probability of transition between different regimes, which limits the application of MSR models. But the transition variable in a STR model is observable, and STR models can mimic smooth transition between different behaviors. For above properties of STR models, they will be widely applied to the study on nonlinear effect of monetary policy transmission.However, it is very scarce of applying STR models, especially extended STR models to the nonlinear studies of Chinese monetary policy transmission mechanism. This dissertation investigates specification,estimation and other aspects of different kinds of STR models, and has an empirical study on nonlinear effect of Chinese monetary policy transmission. Compared to existing literature, the innovative points and significance of this dissertation are as follows: Firstly, this dissertation has innovative angle of view of study. this dissertation firstly applies micro panel data to the studies of Chinese monetary policy transimisson.The existing literature study Chinese monetary policy transmission from macro angle of view by applying macro aggregate data.However,models which apply macro aggregate data have many disadvantages, such as obscuring micro mechanism of monetary policy transmission , and problems of scarcity of sample data and identification. But models which apply micro data can overcome above disadvantages. This dissertation shows the effectiveness of interest policy and credit policy from micro angle of view.Secondly, this dissertation has innovative methods of studies. For considering that monetary policy variables probably are endogenous variables, vector smooth transition autogressive regression model is firstly applied to the study of Chinese monetary policy transimissoin.Furthermore, the technique of generalized impulse response function analysis is firstly used and the results show that, with cyclical variation of credit, the effect of Chinese monetary policies has significant asymmetry. For example, in the short-run the change of product has more sensitivity to the change of interest rate as economy is situated in credit contraction. For considering that different micro individuals may have different reaction to monetary policy, threshold panel model and nonlinear smooth transition panel model are firstly used in this dissertation and the results show that the effect of monetary policy is not identical between state-owned-holding and non-state-owned-holding listed companies, and between listed companies which possess different payment capacity for debt and different profit rate. For example, asset price channel of monetary policy transmission is ineffective for state-owned-holding listed companies but shows nonlinear effect for non-state-owned-holding listed companies, and the effect of monetary policy transmission through credit channel gets weaker with the increase of payment capacity for debt of listed companies, and the effect of interest policy is stronger for high-profit and red companies but the effect of credit policy is stronger for ordinary-profit companies.Thirdly,on the estimation of the models ,the technique of simulated annealing is widely used in this dissertation. Compared to the techniques existing literature used, such as the principle of Gauss-Newton iteration, simulated annealing can search more closely and can avoid the problem of local optimal solution, so the results this dissertation concludes are more credible. Finally this dissertation has richer economic and policy sense. This dissertation finds out that the effect of monetary policy is different when economy is in different credit phases. For example, in the short-run, the positive shocks of change of money have a positve cumulative impact on output as economy is situated in credit expansion,but the positive shocks of change of money have a negative cumulative impact on output as economy is situated in credit contraction, and to control overinvestment problem of some high-profit industries, such as real estate industry, we can take interest policy as dominant policy, and credit policy as assistant policy.Obviously,above findings are derived from nonlinear STR models. Otherwise we can not get those findings. From this point, this dissertation has much significance of knowledge and application.