基于TVP-FAVAR模型的中国金融状况指数的构建和预测

Construction and Forecast of China’s Dynamic Financial Condition Index Based on TVP-FAVAR Model

一个良好的金融状况指数既能准确反映金融市场的状况,又具有较好的宏观经济预测能力,对政策制定者有重要的指导意义。在时变参数因子增广向量自回归模型(TVP-FAVAR)的基础上,采用动态模型选择与平均(DMS和DMA)的技术,构建了一个能准确反映中国通胀和经济状况未来走势的金融状况指数(FCI)。在评估FCI的质量时全面地覆盖了多种评价指标;增加了短期利率和房地产价格方面的指标,范围广,信息充分;考虑了潜在金融市场风险对金融市场整体形势的影响;不仅在TVP-FAVAR模型中提取FCI,也在TVP-VAR模型中对通胀和宏观经济景气一致指数两个变量进行预测,捕捉了FCI与两个宏观经济变量之间的动态互动关系,同时进行了模型选优。研究结果表明:第一,DMS和DMA方法构造的FCI不仅与中国通货膨胀率和宏观经济景气一致指数的走势大体一致,而且分别领先于通货膨胀率和宏观经济景气一致指数1年和3~6个月;第二,在DMS中,构成变量在每个样本点对目标宏观经济变量的影响并非一成不变,它们随着经济结构或政策的变化进入或退出模型;第三,新构建的FCI对中国通货膨胀率和宏观经济景气一致指数有良好的预测能力,在向前1~4个月的预测中,相对于常系数VAR、TVP-VAR以及FAVAR,TVP-FAVAR-DMA模型的预测效果是最优的,即DMA方法测度的FCI的宏观预测能力强于其他方法。

A well-performed financial condition index can not only accurately reflect the situation of the financial market,but also have good macroeconomic forecasting ability,which has important guiding significance for policy makers.Based on the time-varying parameter factor augmented vector autoregression(TVP-FAVAR),a new financial condition index(FCI)that can accurately reflect the future trend of China’s inflation and economic conditions,which is constructed by using the dynamic model selection and averaging(DMS&DMA)techniques.Compared with previous studies,a variety of evaluation indexes are comprehensively inclued when evaluating the quality of FCI;The indicators of short-term interest rate and real estate price have been added,with wide range and sufficient information;Considering the impact of potential financial market risks on the overall situation of the financial market;Not only the FCI is extracted from the TVP-FAVAR model,but also the inflation and macroeconomic consensus index are predicted in the TVP-VAR model to capture the dynamic interaction between the FCI and the two macroeconomic variables.At the same time,the model is optimized.The results show that:The FCI constructed by DMS and DMA methods is not only generally consistent with the trend of China’s inflation rate and Macro-prosperity Consistency Index,but also ahead of inflation rate and Macro-prosperity Consistency Index for 1 year and 3-6 months respectively;In DMS,the influence of constituent variables on target macroeconomic variables at each sample point is not fixed but enter or exit the model with the Change of economic structure or policy;The reconstructed FCI has good prediction ability for China’s inflation rate and Macro-prosperity Consistency Index.In the previous 1-4 months,compared with constant coefficient VAR,TVP-VAR and FAVAR,the prediction effect of TVP-FAVAR-DMA model is the best,that is,the macro prediction ability of FCI measured by DMA method is stronger than other methods.