| The yield curve has been regarded as an important indicator of financial activities,which plots the interest rates of Treasury bonds according to their maturity.The different shapes of the yield curve indicates different economic meanings.Haubrich and Dombrosky(1996)explained that,normally,a yield curve is upward sloping which shows the yield for long-term bonds is higher than for short-term bonds;conversely,if interest rate for short-term bonds is higher than long-term bonds,it would be an invert shape,which indicates a sign of recession;a flat yield curve indicates there are close interest rates for both long-term and short-term bonds.Numerous researchers track the term structure of interest rate as a predictor of future economic activities.The purpose of this paper is to measure the predictive power of yield curve in USA in terms of different approaches and analyze factors that may affect the performance of yield curve models.Specifically,the study focus on the generally applied linear approach to model the yield curve.Besides,a nonlinear approach which modeled the term structure through the dynamic version of Nelson-Siegel(1987)model is applied as a further check of predictive power for term spread.Under the Nelson-Siegel model,the term structure can be modeled through three factors,the slope,level,and curvature of the yield curve(Abdymomunov,2013).The study found that the yield curve has predictive power on economic activity of GDP growth,especially under long-term forecast horizon.Furthermore,add variables of lagged GDP growth or short-term interest rates enhance the performance of yield curve in predicting economic activity.Besides,the study also found that the dynamic version of yield curve model outperform the linear model under out-of-sample tests in predicting real GDP growth. |
