An Empirical Study Of Inflation Forecast In China Based On Dynamic Factor Model

With the rapid development of the economy today,economic data is becoming more and more complex,and economic indicators are becoming wider and wider.On the one hand,the traditional inflation prediction model is difficult to process data sets containing hundreds of variables.On the other hand,econometricians The discovery factor model can process high-dimensional data efficiently and easily.Inflation is no longer determined by a single economic variable,such as the unemployment rate,but the influencing factors are more complex.The study of inflation plays a key role in the formulation and implementation of policies.For this reason,this paper selects a large number of macroeconomic data for the forecast of inflation.Firstly,the classical factor model is selected to select common factors for the data set,and the DI-AR model is used for prediction.Then,in order to improve the prediction accuracy,the paper improves the prediction from two aspects.First,the hard thresholding method is used to select the variable set in the factor selection,and the common factor is selected in the variable set.Second,the model averaging is introduced in the prediction model,and multiple single models are combined by weights.The model averaging method has become a hot issue for metrologists with its many advantages such as good robustness and loss of useful information.The key to model averaging is how to choose weights.This paper uses the minimum Mallows criterion to get the weight of the combination.The conclusions drawn from empirical research include: First,the impact of macroeconomic variables in the data set on China’s inflation is obviously different.It can be seen from the factor component table in empirical analysis that currency and investment have a very significant impact on China’s inflation forecast.Second,based on the dynamic factor model,the prediction of China’s inflation is better than the single variable model.Factor model is an important tool for processing high-dimensional time series.It can easily and efficiently screen out afew factors from thousands of data sets,which not only preserves the prediction information of China’s inflation,but also reduces the data dimension,making the prediction and analysis more convenient and accurate.Third,in order to make the prediction accuracy further improved,this paper improves the prediction of China’s inflation based on the classical factor model in two aspects.The prediction accuracy of China’s inflation combined with the target predictive factor and the model average dynamic factor model is significantly better than the single prediction model.Fourth,whether it is based on the classical factor model to predict China’s inflation,or combined with the target predictive factor and the model’s average factor model to predict China’s inflation,it is found that the effect is not ideal when forecasting one year.This may be due to insufficient macroeconomic variables in the data set and the short data length.