自适应多元货币模型与人民币汇率预测 (Adaptive Multiple Monetary Model With Application in Forecasting the RMB Exchange Rates)

Abstract

Chinese Abstract: 在人民币国际化不断推进、人民币汇率双向波动加强的背景下，构建具有优良预测能力的人民币汇率预测模型愈发重要。参数模型对汇率预测的能力不仅取决于模型设定是否正确，同时还取决于能否迅速探测模型参数的结构性变化以使用最佳信息估计模型参数。本文构建了自适应多元货币模型，并对2011年1月以后的美元、欧元、英镑和日元兑人民币汇率月度数据进行样本外预测，发现在中长期(3至24个月)的人民币汇率预测中，自适应多元货币模型能显著优于随机游走模型、购买力平价模型、弹性货币模型、利率平价模型、泰勒规则模型与偏移型泰勒规则模型这六种汇率预测主流模型。该模型能准确捕捉模型参数的时变特征，探测汇率动态的结构性变化以实时检测参数同质区间，并寻找最佳参数估计以提高人民币汇率预测精度。 English Abstract: To develop an outstanding Renminbi exchange rates forecasting model based on the background of Renminbi internationalization and the increased two-way volatility becomes much more important. By developing an adaptive multivariable monetary model, we do the out-of-sample forecasting for the Renminbi exchange rates after 01/2011 and compare the forecasting accuracy with the other 6 different widely used competitive models. We find that the forecasting ability of a parameter model depends not only on whether it is correctly specified but also on the efficiency of whether it can detect the structure changes and using the effective observations to estimate the parameters. Our developed model can manage of capturing the parameters’ time-varying properties and significantly outperform the Random Walk, Purchasing Power Parity model, Flexible Price Monetary Model, Interest Rate Parity Model, Taylor Rule Model, and the Taylor Rule Differential Model in the middle and the long run (3- to 24- months ahead). In this paper, we have developed an adaptive multiple monetary regression model which can not only detect the structure changes automatically at every time point but can also detect the parameter homogenous subintervals and identify the longest homogeneous subintervals which are used as the best subinterval to estimate the parameters.