中文摘要：

由于利率期限结构中包含未来经济运行的信息,利用2006年4月到2014年12月中美两国利率期限结构的月度数据,通过动态Nelson-Siegel模型抽取两国利率期限结构的相对水平,斜率和凸度三因子,基于三个相对因子检验其对人民币/美元汇率的预测能力。实证研究表明：（1）相对因子模型对汇率在1到12月期具有可预测性,相对水平因子或相对斜率因子增加1%分别导致人民币升值1%和2%而相对凸度因子增加1%会导致人民币贬值1%;（2）基于CW检验统计量的滚动窗预测表明：在所考虑的各个滚动窗下,相对因子模型预测能力优于随机游走和非抛补利率平价模型。

英文摘要：

Since the term structure of interest rates embodies information about future economic activity, this paper uses dynamic Nelson-Siegel model to extract relative level, slope and curvature based on monthly data of interest rate of term structure of China and United States from April in 2006 to December in 2014 and analyses forecasting ability of relative factors on Renminbi/Dollar exchange rate. The empirical study shows that（1） Relative factors model can predict exchange rate changes 1 to 12 months ahead, and 1 percentage point increase in relative level or slope predicts 1% and 2% annualized appreciation of the Renminbi respectively, 1 percentage point increase in relative curvature predicts 1% annualized depreciation of the Renminbi;（2）Rolling window forecasting based on Clark-West statistics shows that relative factors model outperforms random walk and uncovered interest parity model.