“…Hamilton (1996) set out general methods in the tradition of the first category, defining a set of extended models and then using the scores of these extended models as the basis of specification tests. But there have been only a few applications of his methods; exceptions include Hamilton and Perez‐Quiros (1996), Asea and Blomberg (1998), Driffill and Sola (1998) and Layton and Smith (2000). One reason for this low level of application might be the fact that, in practice, MS models fitted to data are often more complex than those specifically dealt with by Hamilton (1996) 1 .…”