Forecasting Quarterly GDP in China with Mixed-Frequency Data: A Model Comparison
Read
DownLoad |
| Title | Forecasting Quarterly GDP in China with Mixed-Frequency Data: A Model Comparison
|
| Author | Zheng Tingguo,Xia Kai and Dang Jue |
| Organization | XIAMEN UNIVERSITY |
| Email | zhengtg@gmail.com;xiakai1986@gmail.com；dj.alex@foxmail.com |
| Key Words | GDP forecast, Mixed-Frequency Data, VAR model, DFM models, MIDAS model |
| Abstract | Governments and individuals need to have accurate and timely forecasting of quarterly GDP to provide a reliable basis for decisions. This paper presents the main strands of mixed-frequency methods which aim to make proper use of high frequency economic information in forecasting or analysis, and compares their performance in forecasting quarterly GDP in China, using quarterly VAR model as benchmark. It turns out that, for both real-time data and final-time data, the mixed-frequency approach is more timely comparing to the benchmark model. And the State Space Mixed-Frequency VAR model and State Space Mixed-Frequency Dynamic Factor Model provide substantial improvements in forecasting accuracy within three quarters and eight quarters respectively, whereas Stacked Vector Mixed-Frequency VAR model improves the performance moderately for all horizons. However, MIDAS model only outperforms the benchmark model up to three months. Finally, by comparing State Space Mixed-Frequency VAR model’s forecasting with those published by IMF Outlook Database, it shows our results are more time efficient and more accurate, hence we conclude State Space Mixed-Frequency VAR model is a reliable method for short-run GDP forecast in China. |
| Serial Number | WP660 |
| Time | 2014-08-19 |
|