Current Forecast Of Service Sector Economy Based On Factor Mixed Frequency Data Model

Since 2012,China’s economic growth has slowed down,and China’s economy has entered a new normal stage characterized by deceleration and shift of gears,structural adjustment and shift of driving forces,showing new features different from those in the past.Under the background of the new normal of China’s economy,it is of great practical significance to strengthen the monitoring of the service industry’s prosperity.On the one hand,it is helpful to reflect the business situation and development trend of the service industry timely and accurately.On the other hand,China is in service industry and industrial parallel development stage,due to the current macroeconomic prosperity monitoring research framework widely used industrial added value as a benchmark,so the prosperity monitoring of services can so as to further perfect China’s macroeconomic monitoring framework and help the government a more comprehensive and in-depth grasp the macroeconomic operation condition.When traditional econometric models are used to study the service industry and other issues,only regression is conducted for variables with the same frequency and length.Due to the limitation of sample data structure,it is usually unable to include enough impact factors in the same model.In order to overcome the limitations of the same frequency data model with high requirements on sample structure and insufficient utilization of high frequency data,domestic and foreign scholars continue to study and improve a variety of mixed frequency data models.Method for mixed frequency according to this study model facing the challenge of trying to use dynamic factor model(DFM)to high frequency data dimension reduction,with the low frequency data into the mixing Bayesian vector autoregressive model(MF-BVAR),build factors mixed frequency(FA-MF-BVAR),according to the model to forecast the macroeconomic indicators,which is a breakthrough than traditional direct use of original data using MF-BVAR forecast macroeconomic indicators.In this paper,FA-MF-BVAR is applied to the current forecast of the service industry,and 8 monthly indicators related to the service industry and macro economy and 1 most representative comprehensive quarterly indicator are selected.The monthly data were introduced into DFM for dimensionality reduction,and the monthly data and quarterly data were introduced into MF-BVAR to construct FA-MF-BVAR.The model is based on the steady-state prior,inverse Wishart prior,constant error covariance matrix posteriori,common random fluctuations posteriori and other estimation methods,to get the estimation of the VAR model and state variables,so as to forecast the current situation of the service industry in China.The following conclusions are drawn:1)In the steady-state priors with and without stratified contraction,the posteriori intervals are similar,and the value-added of service industry expressed by grade specifications is narrower than that shown in the figure.Due to the similarity between the models,the model evaluation result is that the first-order difference of the year-on-year added value of the service industry is in a relatively stable state,indicating that the year-on-year added value of the service industry is stable,and the service industry is expected to rise slowly and return to a stable state in the long run.The FA-MF-BVAR model can achieve a better prediction effect in terms of whether the actual added value of the service industry is rising,declining or maintaining the normal year on year.2)in the use of a constant or variable distribution of forecast error covariance of the steady state prior to generate the random fluctuations of the current forecast narrow distribution,from the use of stochastic volatility to the similarity between the two models for prediction of distribution is greater than any model with stochastic volatility and the similarity between the model with constant volatility.Therefore,the choice of random volatility may be less important than whether or not it is included.3)As can be seen from the standard deviation chart of the error term in the growth equation of service industry,the standard deviation of the error term in the added value of service industry has a great change.In 2020,a period of particularly high volatility is evident,and models with factor stochastic volatility and models with common stochastic volatility have similar estimated volatility,both of which capture the same peak.The standard deviation lines for models with constant volatility are almost always below the baseline from which time-varying volatility is now somewhat off.Therefore,the prediction distribution generated by the steady-state priors with constant or time-varying error covariance is narrower under the random fluctuation specification.In a word,DFM and MF-BVAR are applied together,and good results are obtained in the current forecast of service industry.