ASYMMETRY IN THE LINK BETWEEN THE YIELD SPREAD AND INDUSTRIAL PRODUCTION. THRESHOLD EFFECTS AND FORECASTING

We analyze the nonlinear behavior of the information content in the spread for future real economic activity. The spread linearly predicts one year ahead real growth in nine industrial production sectors of the US and four of the UK over the last forty years. However, recent investigations on the spread-real activity relation have questioned both its linear nature and its time-invariant framework. Our, in-sample, empirical evidence suggests that the spread real activity relationship exhibits asymmetries that allow for different predictive power of the spread when past spread values were above or below some threshold value. We then measure the out-of-sample forecast performance of the nonlinear model using predictive accuracy tests. The results show that significant improvement in forecasting accuracy, at least for one-step ahead forecasts, can be obtained over the linear model.