Analysis And Comparison Of Time-varying Autoregressive Economic Modeling Methodology

As a emerging research areas, TVAR( Time-varying autoregressive model) in macroeconomic econometric analysis is in the stage of "Methodology forming", and it’s in the key link that enhancing from a methods to a general methodology. According to different structures of the models, the models can be divided into three types: fixed structure,time-varying structure and non-structural(semi) structural model. In this paper, on the basis of the relevant time-varying model review, modeling features, fields of application,advantages and disadvantages associated application cases in every model are introduced,and finally, introducing the law of its developments by studying their comprehensive comparison. This paper is both the review of time-varying modeling methods and its applications, and providing guidance and reference by improving promotion of methods and applications. It is helpful to enhance the relevance of TVAR, and it has important practical significance on perfecting TVAR methods and promoting their economic practice.The main innovation of this paper: First, put three types of models into a "time-varying structure(parameter) model" framework to the study, studying the modeling features and analyzing the modeling principles; Second, compare the time-varying parameter model in terms of history, the features of data, methods, uses, advantages and disadvantages in a comprehensive way; Third, take "time-varying structure(parameter) model" as a whole,summarizes the law of development of its modeling methodology.Paper is structured as follows:The first chapter is an overview of the study, sorting out the analysis based on existing work, analyzing aspects to be improved in the existing research, and defining the object of study.The second, third and fourth chapters are respective introductions of types of time-varying structural models. In these three chapters, time-varying models are divided into three types: structural break model(Breakpoint regression model and the threshold regression model), regime switching regression model and stochastic structural model,focusing on the modeling principles of each model; and introducing different applications of each model through the cases.The fifth chapter compares the time-varying parameter model in terms of history, the features of data, methods, uses, advantages and disadvantages in a comprehensive way, and summarizes the law of development of its modeling methodology based on the second, third and fourth chapters.Through the comparisons of the theoretical and applied cases of the three types of time-varying models, this paper argues, in the stage of setting macroeconomic models,time-varying structure model is weaker than the fixed structure model, but stronger than non(semi) structural model in assumptions and independent of economic theories; In the stage of parameter estimation and hypothesis testing, time-varying structural modeling approach is to develop and improve the fixed structure of the model, but basically built on the traditional method of estimation and testing basics, is much mature with respect to the non-parametric methods; In the stage of model applications, the time-varying structure model can be more in-depth characterization of the time-varying characteristics of the relationship between variables, parameters by the nonlinear relationship between the linear sequence of time, and also avoid the disadvantage that the model theory, structural analysis or extrapolation prediction can’t be verified in the non-structural model theory, so it is not only has a much broader range of applications than the fixed model structure and non-structural model, but also avoid the spurious regression which is often occurred in time-series analysis.