Application Of FAVAR And Its Time-varying Models In The Macroeconomy In China

Since Lucus criticism, simultaneous equations for macro economy are not popular. Sims come up with VAR model instead of it, and then study about VAR models is fashionable. VAR models quite succeed at many fields such as forecasting for macro economy, and structure inference. However, VAR models also have disadvantages. For example, excessive parametric models, and not processing time varying structure for economy. The shortcoming of excessive parametric models is that estimated parameters are not stable, and then the results we want are also not stable. Ordinary VAR models cannot capture the nonlinear characters for macro economy, and then models specification of ordinary VAR models about macro economy is wrong. So, we hope that we can develop other models which can solve the questions. The models should inherit the advantage for VAR models, and overcome the shortcomings for VAR models.The purpose for writing the dissertation is that we want to solve the questions. The first, we state the theory structures for VAR models, especially the recent development about structural VAR models. Their emerging make structure inference for macro economy believable. The secondly, in view of models with many variables being excessive parametric, but excluding of the variables from the models would make the informations loss, so, to achieve a effective tradeoff between the excessive parametric and information loss, I introduce Fator Augemented VAR model. The models can extract effective information from a lot of variables in macro economy, and let the refined information form a VAR model. The same time, I state the Bayesian models construction, and the models can reduce the dimensions for VAR models by adding prior information in models. The last, to solve the question that ordinary VAR models cannot capture the changing for economic structures, we state the time varying VAR models in Bayesian perspective. The disstertation has two demonstrations about Chinese economy. The first, we measure the effectiveness about Chinese monetary policy by making use of FAVAR models, and find output effect from the changing of required reserve is not big, just 55. However, the price effect is very big, and it is 54/3%. Moreover, the time-lag for changing of required reserve is short, and it is generally 2 month. The same time, I find that collectively-owned enterprise is more sensitive to changing of monetary policy than state-owned enterprise, and the living standard for countryside is more sensitive to changing of monetary policy than the one for city. The secondly, I use the time varying coefficients model to fit the data from Gao Xu(2008), and I find that phenomenon from the slope puzzle in Gao Xu(2008) disappeared. It indicates that Chinese macro economy has nonlinear characters, and ordinary VAR models cannot capture this character.Contributions of the dissertation have some aspects. I state it as follow:the first is the innovation of econometric models. In china, ordinary VAR models are used generally, and Bayesian VAR or time varying coefficients VAR seldom is used. The dissertation states the models construction in detail, and has some contributions for improving the econometric level in domestic demonstration study. The second is innovation of research methods. In domestic demonstration study, ordinary VAR or nonlinear VAR that has thresholds are used generally, and the dissertation firstly models the GDP and inflation rate in China using Bayesian estimated methods. The third is the conclusion of the dissertation about effectiveness measurement for monetary policy has important reference meaning for monetary policy makers. The aspects of the changing of required reserve to output and price is asymmetric, and the same to countryside and city. The time lag of required reserve is 2 months, and it may provide policy makers with important reference.