人民币汇率的可预测性与预测因子选择  

Predictability of the RMB exchange rate and predictor selection

作　　者：戴志锋 康杰 王雄[2] DAI Zhifeng;KANG Jie;WANG Xiong(College of Mathematics and Statistics,Changsha University of Science and Technology,Changsha 410114,China;College of Business,Central South University,Changsha 410083,China)

机构地区：[1]长沙理工大学数学与统计学院,长沙410114 [2]中南大学商学院,长沙410083

出　　处：《系统工程理论与实践》2021年第11期2822-2836,共15页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(71771030,11301041);湖南省教育厅资助科研项目(20B035)。

摘　　要：随着我国汇率市场化改革深入发展,人民币的国际地位不断提高,人民币汇率波动也明显加强.在此背景下,设计有效的汇率预测方法具有重要意义.当数据充足时,预测问题的解决思路之一是寻找有效的预测因子.本文基于众多经济与技术预测因子,分别根据样本外预测能力和样本内拟合能力的惯性假设,构造了两种动量因子选择方法,并在美元和英镑兑人民币月度汇率预测问题中将其与多种主流模型进行对比.实证结果表明,动量因子选择方法在多数情况下显著强于随机游走基准模型,且相比于其他竞争模型,动量因子选择方法使用强预测因子的频率更高,能实现更小的预测误差以及更准确地判断汇率变化方向.此外,从预测因子分类和样本外预测时期划分角度考察模型的预测效果,动量因子选择方法均稳健地优于竞争模型.With further development of market-oriented exchange rate reform in China,RMB has played a more important role on the world stage,and the volatility of RMB exchange rate is also more drastic.In this context,it is of great significance to design an effective exchange rate forecasting method.When the data is sufficient,one of the solutions to the forecasting problem is to find powerful predictors.Based on a number of economic and technical variables,we construct two momentum predictor selection methods according to the momentum assumptions of out-of-sample forecasting ability and in-sample fitting ability,respectively,and compare the monthly forecasting ability of simple model selection methods for the exchange rate of USD and GBP against RMB with those of various other prevalent models.Our empirical results show that the simple model selection methods are significantly stronger than the random walk benchmark model in most cases,and compared with other competitors,the simple model selection methods use powerful predictors more frequently,and can achieve smaller forecasting error and more accurate prediction of the direction of exchange rate change on the whole.In addition,from the perspective of predictor classification and out-of-sample period division,the simple model selection methods deliver more robust performance in contrast with the rival models.

关 键 词：人民币汇率 样本外预测 动量因子选择 统计降维 

分 类 号：F0G4.1[经济管理—政治经济学]