My view, with both all due respect and all due derision, is that the Robert Lucas types are like the early Copernicans here. There’s something admirable in their insistence that it ought to all work out to an easily modeled system grounded in compelling theoretically considerations. The New Keynesian model is a mess, like late-Ptolemaic astronomy, thrown together to account for observed reality. But you don’t fly to a moon with an elegant model that delivers mistaken predictions about where the moon’s going to be. And what we actually need is a Kepler to give us an elegant model that actually predicts the phenomena, and then a Newton who can explain what that model means.
Not sure I agree with the main point, but worth a read.