Here is a Stiglitz talk at the Lindau meetings. He gets going, and you think it is just going to be Krugman ideas - kibitzing about modern macro in general. He makes the common error of finding fault with models that did not predict the crisis. If you build a good model of a financial crisis, the people living in that artificial environment know that a crisis can happen, but they won't be able to predict it. And you won't be able to use the model to predict a crisis in the real world either.
However, the talk then gets more interesting. It becomes clear that, when he's finding fault with models that are in common use, he has actually talked to people and tried to figure this out. The models he finds fault with are apparently New Keynesian models. He thinks that wage rigidity is unimportant, that the distortions that New Keynesians worry about don't matter much, and that we should be more concerned with financial frictions and credit. Excellent! A central banker tells him that the model used in his or her central bank has no banks in it. Stiglitz is flabbergasted. Stiglitz is now my hero.