The question of microfoundations raised by Brad DeLong, and discussed by Simon Wren-Lewis is more relevant to the current policy failures in monetary policy than it appears. Simon, who’s blog is a breath of fresh air, is far too generous to New Keynesian modelers. He suggests the focus on microfoundations is a methodological issue and economists don’t like methodology. He’s right about them not liking “methodology” – most are not sure what it means. But this is not a methodological issue. The Lucas critique – which kicked-off the obsession with “microfoundations” – is absolutely valid, in the narrow sense that we need to model people using forward-looking expectations, and model parameters clearly should change in response to policy shifts, because (at least some) people are aware of policy. If there is an approach to economics that can avoid those two issues, I am not aware of it (pace Steve Keen). That said, we can also do way better than rational expectations.
Brad similarly gives up too easily in assuming that in order to produce more accurate models we should abandon microfoundations – arguably you can only substitute one set for another. The challenge is to ensure one has the best microfoundations available: i.e. plausible representations of how people actually form beliefs and make decisions, without creating unhelpful model complexity. The problem with New Keynesian modelers is that they show little evidence of a desire to pursue accurate microfoundations – and a huge desire to increase the formal complexity of their models. As the brilliant work of Christopher Carroll shows, more accurate microfoundations of consumer decision-making are generally simpler than assumed in NK models. Incorporating them would greatly improve analysis of the modeling of policy-effects. I am convinced the same is true with the formation of expectations more generally. It would not surprise me if more realistic models were simpler, and more accurate. But that may not be a conclusion consistent with the preferences of modelers. If you incorporate realistic microfoundations, the treasured tenets of NK policy-making breakdown. For starters, lower real interest rates don’t raise demand. Try getting that published. Or proposing it as the basis for forecasts at a central bank.
 In fact, Simon’s productivity is such that he may have forgotten an earlier post of his on the subject (from 2013) which suggests as much (he is responding to this by Tony Yates. He recommends an “eclectic” approach, and he gives the example of modeling consumer behavior, “Suppose in the real world some consumers are credit constrained, while others are infinitely lived intertemporal optimisers. A microfoundation modeler assumes that all consumers are the latter. An eclectic modeler, on finding that consumption shows excess sensitivity to changes in current income, adds a term in current income into their aggregate consumption function, which otherwise follows the microfoundations specification.” I’m entirely in agreement with the “eclectic”, only I would argue that his approach has better microfoundations. Empirical microeconomics shows clearly that many consumers are credit constrained, and NONE are infinitely-lived optimisers! Since when did “microfoundation” mean “empirically falsified” … or is that too methodological?