论文部分内容阅读
2008年全球金融危机之后,行业周期循环受到外部冲击,由于钢铁行业受投资拉动较大,与宏观经济相关性较强,原有先行、一致指标体系出现不稳定。本文根据K-L信息量和时差相关分析建立了一套新的先行、一致指标体系。在此基础上首次将广义动态因子模型(FHLR)和混合频度数据预测(MIDAS)方法引入到行业景气预测领域,广义动态因子解决了因子模型中的随机扰动项同期非正交性,刻画了先行、一致指标周期循环之间的交互影响;混合频度数据预测方法可将新的高频数据信息加入季度景气预测中,提高了信息利用率,扩展了先行、一致指标的数据频度范围。两类方法在7个季度的预测中均表现较好,准确的预测出了景气趋势和谷底出现的时点,其中FHLR方法对行业景气变化趋势更为敏感,而MIDAS方法预测误差更小。
After the global financial crisis in 2008, the cycle of the industry was subject to external shocks. Since the steel industry was heavily driven by investment and highly correlated with the macroeconomy, the original and consistent index system was unstable. Based on K-L information and time difference correlation analysis, this paper establishes a set of new leading and consistent index system. On this basis, for the first time, the general dynamic factor model (FHLR) and the mixed frequency data prediction (MIDAS) method are introduced into the field of industry climate prediction. The generalized dynamic factor solves the simultaneous non-orthogonality of stochastic perturbation terms in the factor model, First, and consistent index cycle cycles. The mixed frequency data prediction method can add new high frequency data information to the quarterly climate prediction, improve the information utilization rate, and extend the data range of the leading and consistent indicators. Both methods performed well in seven seasons, and predicted the trend and the trough at the bottom accurately. The FHLR method is more sensitive to the trend of industry sentiment, while the prediction error of MIDAS method is smaller.