“…The Hamilton model is, however, relatively uninteresting from a regime prediction viewpoint because it assumes the regime-switching probabilities are constant over time. Although Filardo (1994) has generalised the approach to permit the regime-switching probabilities to be functions of one or more leading indicators, Birchenhall, Jessen, Osborn and Simpson (1999), hitherto referred to as BJOS, find that US business cycle regimes are better predicted when the leading indicator information is used in the context of a logistic regression model. Thus, while modelling the regime implies a loss of information in comparison with the modelling of observations on GDP growth, it appears that the former yields better predictions when the regime itself is the focus of interest.…”