摘要
本研究利用中国宏观经济指标构建了基于贝叶斯估计的混合频率向量自回归模型(MF-BVAR),并对该模型预测中国宏观经济运行情况的效果进行了检验。本文模型在允许多变量、不同频数据共存的条件下提高了模型估计的自由度,从而实现高精度预测。实证结果显示,在对宏观经济管理部门所关注的核心经济变量CPI、RPI和GDP等进行预测时,MF-BVAR模型相对于目前广泛应用的同频向量自回归模型和MIDAS模型,预测精度都有显著改善。本文亦发现房地产投资对于模型预测能力的重要作用,从样本外预测的角度佐证了房地产部门对于中国宏观经济的重要影响。本文也验证了中国股票市场表现不能对预测宏观经济运行提供额外贡献。
This paper constructs a Bayesian mixed frequency VAR( MF-BVAR) model to study the dynamics of China's macro-economy and to forecast the key macro variables. The MF-BVAR model can nest high frequency macro information( e. g. capital market price) without compromising to the low frequency information( e. g GDP,investment) in economic projection. Empirical evidence shows that the MF-BVAR model dominates other classic models on forecasting key macro indicators such as CPI,RPI and GDP growth. The study further demonstrates that the real estate investment plays a significant role in forecasting China 's economic dynamics,while the stock market is insignificant in macro projection.
作者
张劲帆
刚健华
钱宗鑫
张龄琰
ZHANG Jinfan;GANG Jianhua;QIAN Zongxin;ZHANG Lingyan(School of Management and Economics,Chinese University of Hong Kong(Shenzhen)/ Shenzhen Finance Institute/ International Monetary Institute,Renmin University of Chin;China Financial Policy Research Center,School of Finance,Renmin University of Chin;Department of Industrial Engineering and Operations Research,Columbia Universit)
出处
《金融研究》
CSSCI
北大核心
2018年第7期34-48,共15页
Journal of Financial Research
基金
国家自然科学基金项目(批准文号:71733004、71503257)资助