Comparison Of Statistical Models Based On A Number Of Macroeconomic Monitoring

Economic supervising and early warning carries on total, comprehensive, overall, systematic analysis and judgment to the whole economical development. It supervises a series of indexes, which reflect economic activity process and actuality, and put forward the alarm to the future possible economic transition and great change based on supervised result, which is the barometer and alarm of economic by utilizing a series of economic indicator.It is very important to develop an early warning system (EWS) of Chinese macro-economy. It can help analysis the macro-economy situation. For example, is it hot or cold, what problem exists in the economy and is the policy taking useful for solving the problem. It provide an instrument that helps policymakers identify and anticipate situations in which crises are more likely to happen, so that the symptoms can be detected sufficiently in advance to allow them to adopt preemptive measures.The traditional methods are mostly stepwise regression analysis, which is based on the linear assumption. Bur in reality, the relationship between indicators is often non-linear. Therefore, we introduce the hidden Markov models and Markov switching models for macro-economic monitoring and early warning, which is different from the VAR model. And then we compared the fitting accuracy and the prediction accuracy of each model. Empirical results show that hidden Markov model with order two has better performance compared with the traditional linear model.This article consists of four chapters. The first chapter is introduction section. In this section, the research background of the main macro-economic monitoring and warning system is described and the previous application of a variety of research methods are reviewed. In chapterⅡ,14 indicators are selected for the macro-economic monitoring and early warning indicator system. In chapterⅢ, the three research methods used in this paper are described, which include vector autoregressive model, hidden Markov models and Markov switching models. In chapterⅣ, the results of different models are compared. And chapterⅤis a summary of this paper.