Even if this upshot is accepted, the appropriate steps to be taken are less obvious. An interesting example of how to proceed in this situation is the paper by Fuhrer (1997). Though with more elementary methods, he similarly states that “expectations of future prices are empirically unimportant in explaining price and inﬂation behavior” (Fuhrer 1997, p. 349). Noting that the data cannot reject the hypothesis with any conﬁdence that expectations are rather purely backwardlooking (ibid, p. 344), he then goes on to examine the dynamic implications of a backward-looking, accelerationist Phillips curve with those of a mixed forwardlooking and backward-looking speciﬁcation. In a disinﬂation experiment he ﬁnds that the former “implies implausibly long and vigorous responses to events many years ago” (ibid, p. 347), whereas the time path of the model with some forwardlooking price behavior conforms considerably better to the conventional wisdom that monetary policy does not have pronounced long-run effects. Thus, for policy simulations, a combination of forward-looking and backward-looking elements may yield more reasonable long-run behavior than the accelerationist Phillips curve, without sacriﬁcing too much on empirical performance (Fuhrer 1997, p. 349).