Combining Forecasts with Nonparametric Kernel Regressions

We introduce a flexible nonparametric technique that can be used to select weights in a forecast-combining regression. We perform a Monte Carlo study that evaluates the performance of the proposed technique along with other linear and nonlinear forecast-combining procedures. The simulation results show that when forecast errors are correlated across models, the nonparametric weighting scheme dominates. As a general rule, our simulation results suggest that the practice of combining forecasts, no matter the technique employed in selecting the combination weights, can yield lower forecast errors on average. An application to inflation forecasting is also presented to demonstrate the use of all forecast-combining techniques.