【摘要】 考虑到不同类经济指标之间具有非线性的结构信息,对128个美国宏观经济指标建立分层动态因子模型。按实际经济意义将指标分为收入产出、劳动力、消费和投资、价格、货币信用,以及利率汇率6部分,分别使用bottom-up和top-down两种方法提取公共因子和块级因子,并对其经济意义进行解释。最后对INDPRO,CPIAUCSL,PAYEMS和FEDFUND四个指标建立关于公共因子和块级因子的两种预测模型,和时间序列模型的预测结果进行对比,结果表明:分层动态因子模型可以有效提取不同类型指标的结构信息,使得因子的可解释性大大提高,基于因子的预测模型预测效果也有显著提升。更多还原

【Abstract】 By taking into account the nonlineal structural information between different types of economic indicators, established in this paper was a hierarchical dynamic factor model for 128 U.S.macroeconomic indicators. According to the actual economic significance of the indicators being divided into six, namely, income output, labor, consumption and investment, price, monetary credit, and interest rate exchange rate six blocks, respectively, using bottom-up and top-down two methods to extract public factors and block-level factors, with its economic significance explained. Finally, two prediction models of common factors and block-level factors are established in terms of DPRO, CPIAUCSL, PAYEMS and FEDFUND, with the prediction results of time series models compared. Results show that the hierarchical dynamic factor model can effectively extract the structural information of different types of indicators, which greatly improves the explanatoryness of factors, and the predictive effect of factor-based predictive models is also significantly improved.更多还原