The Modeling And Application Of Panel Smooth Transition Regression Models

With the development of econometric modeling technology, more and more people are aware that the linear model faces many limitations but the application of nonlinear model sometimes can better solve these problems. The nonlinear model is mainly composed of chaos models and switching regime models. Moreover, the latter models contain Markov swithing regime model (MSR), transition autoregressive model (TAR), and smooth transition regression model (STR). This paper will only focuses on the panel smooth transition regression model (PSTR). PSTR model can be dividend into static PSTR and dynamic PSTR, which is based on whether PSTR model includes an endogenous variable. Because the two models have different estimation methods, this paper discussed them apart.About the static PSTR, this paper first introduces the modeling procedure of Gonzalez, Terasvirta and van Dijk (2005), and in the empirical analysis, the paper discusses Chinese "demographic dividend" with the model. The result shows that the relationship between Chinese population structure and incurred economic growth is nonlinear. When the labor population burden coefficient is less (?)han the " Lewis turning point "(0.3276), Chinese economic development has the " demographic dividend ", on the other hand, especially when the labor population burden coefficient is greater than0.4, the "demographic dividend " is very limited.The current estimations of dynamic panel smooth transition regression model are all GMM method. This paper puts forward a kind of indirect inference algorithm and discussed the model respectively based on whether the transition variable is an endogenous variable and whether the model contains exogenous cont-ol variables. Secondly, in order to verify the algorithm’s feasibility, I apply the Monte Carlo method to simulate and the results show that, the algorithm will improve the accuracy of the estimates with the sample size increases, particularly to increase the number of panel period, what’s more, adding to the exogenous variables in the model will also improve the precision of estimates. Finally, this paper re-examines the Phillips curve in China with the dynamic PSTR. The result shows that Chinese Phillips curve is also nonlinear and the parameters of the linear Phillips curve model all fall in the parameter scope of nonlinear Philips curve. In addition, with increasing of the transition variable output gap, the inflation rate will increase, and the influence of output gap on the inflation rate also increases, but the effect of supply shocks is from reducing inflation to increase inflation, at the same time, inflation expectations will also change.