| Since Sims proposed the Vector Autoregression model(VAR),VAR model and related theories have been widely applied in many fields such as economy and finance,and have a good performance in time series data property research,prediction and macroeconomic policy analysis.However,due to the insufficient understanding of some scholars on the theories related to VAR model,there are a lot of abuse and misuse in the practical application.For example,the reasons for the selection of structured methods and variable order are not discussed in the impulse response analysis process,and the Vector Error Correction model(VECM)is blindly selected without combining the sequence properties in the Johansen cointegration test process,which may lead to wrong conclusions.Thus cannot reveal the correct law.Therefore,this paper starts from the theoretical level,combined with the problems in the application of in-depth research.In terms of theory,this paper explores the generation background of simultaneous equation model(including Keynesian large-scale macro model),VAR model and Structural Vector Autoregression model(SVAR)as well as the relations and differences among models,and discusses the structure.Deepen the understanding of econometrics macro models and methods.In terms of application,through literature review,this paper finds that there are a lot of common errors in the application of impulse response analysis problem and Johansen model selection problem.Therefore,based on correct theoretical knowledge and Monte Carlo simulation experiment,the two application problems are respectively explored.In view of impulse response analysis,this paper summarizes two common errors in application:(i)The structure of VAR model is not identified and decomposed,and the impulse is mistakenly understood as the impact of VAR model residual directly on variables.(ⅱ)When Cholesky decomposition was performed on the VAR model,the order of variables selected was inconsistent with the economic theory.Aiming at these two errors,this paper designs Monte Carlo simulation experiments respectively to analyze impulse response and explore the consequences of errors.It is found that the wrong understanding and operation in impulse response will make the result deviate from the correct value seriously,and the deviation is irregular.Aiming at the model selection problem of Johansen cointegration test,this paper introduces the solution method of high-order difference equation and solves five Johansen VECM models respectively.By discussing the possible stationary process,random trend,different deterministic trend and intercept term in the solution,the model and the sequence properties are closely related and distinguished.Most importantly,it reveals the deeprooted reasons for the differences between different models with the same sequence properties,which fills the research gaps in textbooks and literature in this part,and provides more powerful guidance and reference value for model selection.To sum up,this paper summarizes and studies the problems existing in the application of VAR model and related theories,and clarifies the application errors and explores the wrong results through the method of combining theory and experiment.This will enable subsequent researchers to further deepen their understanding of VAR model and related theories and increase their attention,so as to make the application of empirical analysis more mature and reduce the occurrence of errors. |
