| 摘要: |
| 传统的宏观经济变量预测使用相同频率的数据，将高频数据统一成低频，但这种做法可能会人为消除一部分高频数据所带来的信息，影响预测的准确性和及时性。本文使用拐点匹配和格兰杰因果检验找出合适的微观产业数据作为解释变量，使用混频数据抽样回归（MIDAS）将不同频率的数据进行回归并检验其预测效果。实证研究表明：使用MIDAS回归对PPI、M1同比增速、M2同比增速的预测效果较好。因此，在进行产业和货币政策制定时，采用MIDAS回归可以获得较好的预测精度。 |
| 关键词: MIDAS；EMD；格兰杰检验；宏观预测 |
| DOI： |
| 分类号:F831 |
| 基金项目: |
|
| An Analysis of Forecasting Macro Data Using Frequency-mixing Micro Industrial Data |
|
WANG Kai，CAO Zhe-zheng，HU Xiao-lin1,2,3
|
|
1.BOC International (China) CO.，LTD.;2.Fudan University;3.Shanghai University of Finance and Economics
|
| Abstract: |
| The traditional method of forecasting macro data is to use data of the same frequency or transfer high-frequency data to low-frequency one.But this method may artificially eliminate the information brought by high-frequency data，reducing accuracy and effectiveness.In this paper，we use turning point identification and Granger causality test to choose the appropriate micro industrial data as explanatory variables.Using mixed data sampling regression (MIDAS)，we test the forecasting efficiency of mixed-frequency data.The empirical result shows that MIDAS regression works well for PPI，M1 growth and M2 growth，indicating that the forecasting results of MIDAS regression are relatively accurate and helpful when making industrial and monetary policy. |
| Key words: MIDAS；EMD；Granger Test；Macro Forecasting |
