三步回归滤波建模的拓展及其在汇率预测中的应用研究

Extension of Three-pass Regression Filter Model and Its Application to Exchange Rate Forecasting

对关键经济变量的准确预测不仅能为当局采取合适的政策以实现有效调控提供启示,还能为公众(包括企业)制定投资决策提供参考,因而具有非常重要的理论、现实与政策意义。为避免统计数据信息缺失并提高分析问题的时效性,文章对传统三步回归滤波建模进行拓展,将混频数据抽样模型(MIDAS)建模思想纳入其中以形成混频三步回归滤波模型(MF-3PRF)。基于大样本性质的理论模拟发现,MF-3PRF能够实现参数估计的有效性与一致性,且具有较好的理论预测能效。进一步地,文章基于MF-3PRF模型与备择模型(DMA/DMS、TVP-VAR、BMA)并选取无抛补利率平价、购买力平价、弹性/粘性价格货币模型、巴拉萨-萨缪尔森模型、资产组合模型、泰勒规则等理论模型对人民币汇率进行预测,结果表明MF-3PRF并结合泰勒规则能够较好地捕捉人民币汇率的动态变化特征,因而能为公众在新的汇率形成机制条件下深入认识人民币汇率动态变化与波动成因提供启示。

Accurate prediction of key economic variables not only enlightens the authorities to adopt appropriate policies that effectively realize effective regulation,but also provides a reference for the public( including enterprises) to make the right investment decisions. Thus,this has very important theoretical,practical and policy implications. In order to avoid the loss of statistical information and improve the timeliness of analysis,this paper extends the traditional three-pass regression filtering model with the idea of a mixed-frequency data sampling( MIDAS) model to form the mixed-frequency three-pass regression filter model( MF-3 PRF). Theoretical simulations based on the nature of large samples indicates that MF-3 PRF can achieve validity and consistency in parameter estimation,and has a good theoretical prediction of energy efficiency. Further,based on the MF-3 PRF model and alternative models( DMA/DMS,TVP-VAR,BMA) with theoretical models such as uncovered interest parity,purchasing power parity( PPP),elastic/sticky-price monetary model,Balassa-Samuelson model,portfolio model,the Taylor rule,etc,the article conducts a prediction for the RMB exchange rate. The results show that MF-3 PRF combined with Taylor rule can better capture the dynamic characteristics of the RMB exchange rate,thus providing insights for the public to understand the causes of the dynamic changes and fluctuations of the RMB exchange rate under the new exchange rate formation mechanism.