Kay's post tackles a puzzle I have been pondering for some time, since well before the financial crisis caused rational expectations economists to engage in all kinds of reality-denying, mental gymnastics to save their model. Why do very smart, creative, and obviously able economists become so attached to their models that, whenever some conflict arises between their models and events in the real world, the models must prevail? This is true, for example, of the many economists who insist on applying the "Coase Theorem" to problems in the real world, despite the fact that Coase himself told us, as soon as he created the model (purely for analytical and pedagogical purposes),* that its assumptions made it useless for understanding transacting in the real world.
Kay's analysis of the puzzle and its causes certainly is deeper and more illuminating than any I can provide. The only thing I can add is a bit of emphasis to a point Kay only hints at near the beginning of his analysis, where he notes that all science relies on simplifying models. The basic axioms and assumptions of neoclassical economic theory unquestionably take us a good way down the road of understanding how the social world works; they have substantial, but far from complete, explanatory and predictive value. Alternative models of economic behavior that excluded basic insights relating, for example, to price theory (as in at least some versions of the so-called "post-autistic economics") are so impoverished as to be useless. They would throw the baby out with the bathwater.
It is a long way, however, from recognizing the substantial utility of necessarily simplified models of the world to supposing that such models - even those as powerful as neoclassical theory - could ever provide a sufficient, let alone a complete, explanation of the social world. Crossing that gap requires a thoroughly unscientific leap of faith.
Finally, I will take issue with Kay on one point, which is his needless denigration of logical consistency as a goal in model-building. Without logical consistency, any model of the world (and any work of creation) would lack coherence; arguably, it would not even count as a model. Contrary to Kay, I believe logical consistency was a primary virtue for the likes of artists such as Mozart and Shakespeare; their creations complied (though not blindly) with basic rules of composition that impose a logical consistency. In any case, the minimal requirement of logical consistency does not require anyone to blindly follow unrealistic models.
*Coase would likely insist that he did not "create" the model, which George Stigler coined the "Coase theorem;" it was merely Coase's restatement of the neoclassical model that many economists since Alfred Marshall had contributed to building.