The Empirical Analysis Of Chinese Phillips Curve With SVR Inflation Expectations

The characterization of agent's learning behavior and the analysis of inflation expectations formation process through learning behavior have become the front of macro-finance,especially monetary finance research,at the same time,the development of computer technology and artificial intelligence has also provided a new and more effective research tool for the academic research-machine learning technology.Machine learning techniques are specialized in how computers simulate or implement human learning behavior,which makes machine learning a natural choice for learning expectations or public learning behavior in macroeconomics.Support vector machine(SVM)and support vector regression(SVR)have become one of the most widely used machine learning algorithm based on their relatively complete theoretical basis and many unique advantages in solving small sample,non-linear and high-dimensional pattern recognition.The SVR inflation expectation in this paper draws on the learning mechanism of adaptive learning expectation that characterized by incorporating new information in each period,and classifying the information according to whether the information is lagged.Then use the support vector regression(SVR)algorithm to create Inflation expectation,and use the GMM method to five different expected forms of Phillips curve for empirical analysis.The empirical analysis results show that:(1)The trade-off mechanism of output gap and inflation in the Phillips curve is invalid and the coefficients of different inflation expectations are very significant,so the central bank should focus on the independence of monetary policy and the management of expectation.The output gap coefficient is negative or not significant,which indicates that China's monetary policy transmission mechanism based on Phillips curve is invalid,that is,the central bank can not control the inflation rate by changing the output gap.And the coefficients of the different inflation expectations are very significant,which indicates that our Phillips curve has the characteristics of expectation or hybrid expectations.(2)SVR inflation expectation is expected to be a more "advanced"learning approach than adaptive learning.Compared with rational expectation and adaptive expectation,SVR inflation expectation is lagging and advancing.The mean,median,standard deviation and skewness of SVR inflation expectation are all minimal,so using SVR inflation expectation to describe learning behavior of inflation is more reasonable,and SVR inflation expectation is more suitable for inflation target of the central bank.(3)Chinese Phillips curve is characteristics of hybrid learning with SVR inflation expectation and rational expectation,and SVR inflation expectation is more significant than rational expectation.Hybrid learning expectations show that Chinese inflation expectations are not completely forward,but limited rational,and rather carrying out fully forward-looking monetary policy,the central bank should fine-tune the economy with continuous learning.The significant SVR inflation expectation shows that the ability of obtaining information in forming inflation expectations is strong and the dimension of information is high.Therefore,on the one hand,the central bank should guide the learningThe innovation of this paper lies in:First,the SVR inflation expectation is proposed to describe the formation mechanism of the inflation expectation when the agents face the high-dimensional information samples with lagging effect,and the SVR inflation expectation is estimated by the support vector regression(SVR)algorithm.In the process of estimation,this paper innovatively divides the variables into variables with and without lagging effect.Furthermore,this paper compares the SVR inflation expectation with the rational inflation expectation and the adaptive inflation expectation.By analyzing the timing chart and statistical characteristics of the three inflation expectations,it is considered that the SVR inflation expectation can more reasonably describe the public's learning behavior on inflation expectation and can be a more suitable choice of monetary policy inflation target,which provides the basis for the empirical analysis of the Phillips curve.Second,based on the GMM method,this paper analyzes the Phillips curve of five different expectation forms in our country,and compares the Phillips curve of the different expectation forms according to the empirical results.It is concluded that the hybrid Phillips curve of SVR inflation expectation and rational inflation expectation can better describe the transmission mechanism of China's monetary policy.And because of SVR inflation expectations can better describe the public's learning behavior,though empirical analysis of SVR expectation coefficient and rational expectation coefficient of hybrid Phillips curve,this paper further analyzed the relatively importance of learning behavior to rational expectations behavior of hybrid expectation of Phillips curve of the.