Research On China Macroeconomic Early Warning Index

Business cycle is one of the two core issues in macroeconomics research.In the New Normal economy,Chinese business cycle wave slowly and slightly.This new characteristic increases the difficulty of monitoring and forecasting the economic climate.How to scientifically and rationally select and compose a set of leading indicators which can predict the macroeconomic development and how to predict accurately the fluctuation of the business cycle are crucial to the monitoring of the business cycle fluctuation and have a prospective guiding significance for both the business decision-makers and the government policy makers.The traditional nonparametric NBER method is used to construct the leading composite index through the change rate of each of leading indicators,it is simple and easy.It does not need to estimate the uncertainty.However,the macroeconomic leading index constructed by the NBER method is not based on the strict statistical model,and it is more simple and subjective in the determination of the weights.Therefore,the macroeconomic leading index may be not the optimal.The Logit model is a method based on a strict statistical model,which predicts the macroeconomic development by using leading indicators,has a strict statistical basis and economic significance.But the Logit model is based on the strict parameter hypothesis,and the validity of the model can not be guaranteed once the process of the generating the real data violates these strict assumptions.The macroeconomic early warning index based on the Vine Copula model relaxes the conditions of the parameter hypothesis.Actually,This macroeconomic early warning index is a functional expression of the joint density function of all the leading economic indicators under the condition of the predicted event and not happening.This model allows the leading indicators to be biased,fat tail and non-normal and allows the nonlinear dynamic dependent structure between the leading economic indicators.In the meantime,The forecasting ability of the macroeconomic early warning index can be optimally measured by the Receiver Operating Characteristic Curve(short for ROC Curve).ROC Curve is a curve formed on the unit square,which is based on a series of different thresholds,with the false rate as the horizontal axis and the hit rate as the ordinate.Compared with the traditional methods of Mean Absolute Error(short for MAE),Mean Percent Error(short for MPE)and Root Mean Squared Error(short for RMSE),the ROC Curve is a trade off between the type I error and type Ⅱ error,and is not affected by the marginal distribution of the binary variables.Therefore,Using the Vine Copula model to construct the macroeconomic early warning index,Then measuring the forecasting ability of the macroeconomic early warning index constructed by the three method mentioned above by using the ROC Curve.The main work and innovations in this paper can be summarized as follows:1.About the selection and determination the leading period of the leading indicators.This paper enrich and supplement the method of selecting and determining the leading period of the leading indicators.When selecting the leading indicators and determining the leading period of the leading indicators,this paper utilizes the time difference correlation analysis,K-L information and ROC curve comprehensively.The results demonstrate that the ROC Curve has the ability to select the leading indicators and determine the leading period of the leading indicators.Compared with the time difference correlation analysis and K-L information,the ROC Curve not only has the statistics significance,also has the economic significance.2.About the empirical analysis based on Vine Copula Model.This is the new approach used to analyze the Chinese economic climate.Using the coincident composite index constructed by the NBER method as the benchmarked index.And constructing the macroeconomic early warning index is used to predict the recessions of the economy activity,by NBER method,Logit model and Vine Copula model respectively.The empirical result shows that the macroeconomic early warning index constructed by Vine Copula model has a higher forecasting ability than the methods of NBER and Logit method.3.About the effect of the leading period of the leading indicators.In the process of constructing the macroeconomic early warning index based on the Vine copula model.In this paper,we not only consider the same horizons of the leading indicators,but also consider the different horizons of the leading indicators.About the horizons of the leading indicators,the following results are obtained:(a)About the forecasting length,the former has a longer forecasting period;(b)On the prediction accuracy,the latter is slightly more precise than the former.