Although the spread has been widely used as a leading indicator of economic activity, a few studies have discuss the term spread-output growth relationship may exist unstable phenomenon cross time and nearly be nonlinearity. This paper using quarterly data for G-7 countries over the period 1970:1-2007-4, and applied linear model as well as nonlinear model to examine the strength and stability of the spread-output relationship. Our empirical evidence confirms that the link between spread and real activity exhibits asymmetries that present for different predictive power of the spread when past spread value were above or below some threshold value. It means under the threshold value the relation is significant and almost disappears and become negative when above the threshold value. Besides, we then compare the forecast performance of the linear and nonlinear model using RMSE and Theil’s U test, and shows the result that the nonlinear model perform better than linear model.
