Evaluating Density Forecast Of China Output Growth And Inflation

The prediction of macroeconomic variables has always been an important research direction of the central bank and other monetary policy-making institutions.Among all the macroeconomic variables,output growth and inflation are two important indicators of a country's economic development.Better prediction of macroeconomic variables,including output growth and inflation,can guide the central bank to better formulate policies and help investment institutions make better investment decisions.We evaluate conditional predictive densities for China's output growth and inflation using a number of commonly used forecasting models that rely on large numbers of macroeconomic predictors.To put it in detail,we evaluate how well the conditional predictive densities based on commonly used normality assumption fit realizations out-of-sample to check the feasibility and efficiency of the normality assumption.Predictive density forecast,which is different from point forecast that is limited to "deterministic equivalence" and symmetric loss function,is the estimation of conditional predictive density based on existing information.The main contents of this paper are as follows:Firstly,we summarize the research findings at home and abroad of the influencing factors of output growth and inflation,the forecasting theory of different models,the forecasting directions and the evaluation methods of predictive densities.Secondly,we constructed the theoretical density forecast models of output growth and inflation and the corresponding probabilistic integral transformations of each predictive density.Thirdly,we construct testing methods to evaluate the consistency,the independence and the stability of the predictive density.Finally,we construct an economic data set with multiple macro variables based on the quarterly data from January 1999 to September 2018 and empirically evaluate the consistency,independence and stability of predictive density of output growth and inflation.We find that the normality is rejected for most models in some dimension according to at least one of the tests we use.However,the fitting effect of the combination forecasting models based on the normality assumption is relatively good:the simple average model when predicting output growth and the BMA-OLS model when predicting inflation.