Comment of the Day: Charles Steindel: Economics, DSGE, and Reality:
The immediate response to these incredibly sensible criticisms is typically either...
..."Prove we are wrong" (aka, observational equivalence, or, it's all in our error bands), or the "Well, it takes a model to beat a model. Show me yours." Somehow, there doesn't seem to be any way to move academic macro out of this rut.
One of the things on my "bucket list" would be to go back and look at those legendary large errors of the Keynesian models from the early 1970s. My suspicion is that after 40 years of other models making errors they don't look all that large. Unfortunately, that project would involve a fair amount of RA work grubbing through paper documents, and wouldn't move anybody's views. The professoriate would say that the size of the errors, if now seen to be not all that large, solely reflected the "discretion" of the forecaster, not the model's underlying properties (as if the choice of model isn't discretionary!), while those who still take a very old-fashioned mechanical Keynesian view would think there was never any reason to advance from those.