A forecasting property of the unrestricted, restricted, and partially restricted reduced-form coefficients

Abstract

In this paper, we show that the unrestricted, restricted, and partially restricted reduced-form (URF, RRF, and PRRF) coefficients display a remarkable property in the context of forecasting from a linear simultaneous equation model. If the values assumed by the exogenous variables in the forecast period are equal to their sample means, the RRF and PRRF estimates derived from the k-class estimates yield exactly the same forecasts of the endogenous variables regardless of the k value we use in the k-class estimates. The common forecasts are equal to the forecasts from the URF estimates, i.e., the sample means of the endogenous variables.