Regime switching and the forecasting bias of asymmetric error correction models

Abstract

The potential forecasting bias of asymmetric error correction models is examined via an application of the Granger and Lee and Escribano and Pfann specifications to the earlier data of Davidson et al. Cook's results concerning sample space regularities are extended, with partitioning of the error correction term shown to induce regularities in the forecasting space in the form of forecasting bias.