Twenty years ago, Paul Krugman warned that the liquidity trap was not just an issue in the economic history of the 30s. He noted there was every sign that Japan was in the liquidity trap in the 90s, then argued that a liquidity trap was theoretically possible. I guess one lesson is that economists even including Paul Krugman had more respect for theory back then.

Now he, frankly, boasts about being decades ahead of the rest of the profession. (pdf warning) and I mean frankly “This paper is an exercise in self-indulgence and self-aggrandizement.” It is also extremely excellent and very much worth reading. I will mainly discuss two sections near the end (which are the only weak points I perceive — Krugman's post is brilliant clear and concise)

First, there was a theoretical proof that economies with flexible couldn't be in a liquidity trap. The claim was that if the price level fell enough, the real value of money holdings (real balances) would be a significant part of wealth and this would cause high consumption (this is called the Pigou effect). The argument is neither fish nor fowl. There is no Pigou effect in the IS-LM model. It is also a relephant in optimizing models with Ricardian equivalence because currency is not net wealth. Krugman notes that formal math says if real balances are huge because of a liquidity trap, people know that their apparent wealth will be consumed by inflation tax (or by taxes needed to retire money if the monetary authority doesn't accept inflation).

This is Krugman's number 1 example of why formal models are good (and also a case in which the reduced form treat correlations and causation accidental theory works just as well). It also shows the power of two (also self-promotion and I thank Krugman for critiquing “a commenter” without naming me).

“My tendency to do New Keynesian models that say they have infinite horizons, but whose analysis always seems to boil down to just two periods, “now” and “forever after”.” I call this “Two Keynesian economics”.

It made it clear to Krugman that one can't believe both in Ricardian equivalence and the Pigou effect. I think the power of two is one of Krugman's most useful and important insights (which is saying a lot). Almost all of the differences between New Keynesian and old Keynesian models can be explained with two periods. The difficult models without closed-form solutions are needed. I add that, I think, the (limited) influence of academic macro on policy is based on claims about the long run.