China's Unemployment Forecas

This paper forecasts the employment index and uses the correlation between the employment index and the urban survey unemployment rate to try to obtain an accurate prediction of the monthly unemployment rate.Based on the deformed Minnesota prior distribution and the bayesian shrinkage method to reduce the risk of over-fitting caused by too many parameters,a high-dimensional Bayesian vector autoregressive model with 152 variables is established which applies the relative mean square error and the overall tightness to test the forecasting result.Furthermore,this paper adopts the method with the model comparison to analyse the forecasting result through VAR、BVAR and LBVAR models and the forecasing result illustrates that the LBVAR model with 152 variables has the best forecasting performance by comparing the monthly prediction results of the three core indicators which include the consumer price index,the Value-added of industry,and the PMI employment index.Acoording to the the different-dimensional LBVAR models with 11 20 and 152 variables,it is found that the overall relative mean square error of LBVAR model with 152 variables is smaller and its short-term forecasting result is the best among these models.What’s more,this paper also explores the one-step forecasting result by selecting two different models-FAVAR and LBVAR to find the optimal model.In order to analyse the international factors whether or not affect the explanasory ability of LBVAR model,this paper also incorporates some variables which mainly have CPI,the unemployment rate and industry production index in several countries to form models with 166 variables to carry out the robustness test.It is the first attempt to use a large bayesian vector auto regressive model to predict the employment index in the domestic,and the aim is to establish a correlation between the employment index and the surveyed urban unemployment rate.This model can provide with helpful forecast references in promoting employment by including more macroeconomic variables.