The eternally simmering blog debate over "microfoundations" has reached a sort of balance, with Simon Wren-Lewis and Paul Krugman on the skeptical side, and Chris House, Steve Williamson, and Tony Yates in support of the dominant paradigm. But "balance" doesn't mean "boring", so I encourage you to read the latest round, which is about the history of New Keynesian macro. Wren-Lewis and Krugman say that New Keynesians, by embracing the microfounded approach, gave up an important type of modeling tool unnecessarily; House and Williamson say that the thing that was given up was not useful at all, so it deserved to go.My comment:
Balance ! We don't need no stinking balance ! Look I'm not in their league or anything but I am very vigorously participating in the discussion and I don't find my view there at all. I disagree with Krugman and Wren-Lewis because I think the approach for which new-Keynesians abandoned the old approach is worthless.
Now I guess my contribution to the debate hasn't gotten all that much attention, but I haven't gotten a one tenth of the way convincing answer to the question "what have microfoundations ever done for us." I am particularly pleased that you lumped Wren-Lewis and Krugman together because it gives me an excuse to embed this link (it really shows that their views are different -- it's here because it proves Krugman typed "Robert Waldmann" but it's relevant).
One would think that being accused of offering "very thin gruel" by Krugman would stimulate some further effort. I do think that. I am confident that Wren-Lewis has devoted considerable effort to the search for payoffs from the micro-foundations approach. I think he hasn't found any, because there aren't any to find.
update: comment after reading
You wrote "And it seems to be rapidly erasing the "freshwater/saltwater" divide." You read Krugman. You know that the divide seemed to be rapidly vanishing way back in 2000-05. Then it reappeared with freshwater economists saying saltwater economists must be fools knaves or both because Calvo pricing was fine until models with it were used to justify Obama's fiscal policy and then it was known to be nonsense as it had been known for decades. If a macro consensus is patched up again, it will hold exactly until the next time there is a major partisan debate about macro policy.
you wrote "the stagflation of the 70s (which seemed to fit with RBC models) and the Volcker Recessions of the early 80s (which seemed to fit with New Keynesian models)."
Ah you are so young. I wish I was so young. The original RBC models didn't include a price level. They had nothing to do with inflation let alone stagflation. The model which seemed to fit the stagflation was, in fact, the Phillips curve as discussed by Samuelson and Solow in 1960. It was perceived to be the critique of a strawman made by Friedman and Phelps (who added nothing and subtracted hysteresis). Later it was thought that the stagflation era and earlier data might fit the Lucas supply function which was Huge for a while. Later it was noted that those data reject the Lucas supply function which was a clearly silly idea to begin with. As noted by Krugman, RBC emerged when freshwater economists had to come up with something new or admit that saltwater economists were right.
New Keynesian economics does not fit data from the Volcker deflation except to the extent that it is identical to old Keynesian economics (see above). Also over here (in Europe) the last war was the huge unemployment problem of the late70s 80s and 90s which is radically inconsistent with new Keynesian models (as note by, among other, leading new Keynesian O.J. Blanchard).
Macroeconomists are totally unprepared to fight the last war or the war before that.
Your idea that macroeconomics always changes so that it wouldn't be defeated in the same way again assumes that old macroeconomics was defeated in battles which newere macroeconomics would win. My view is that this is totally false and that not only have macroeconomists not gained the ability to predict the future but they have lost ability to fit past data.