The key insight of rational expectations (which should be called consistent expectations) is that if you model the economy in a way where policy X produces result Y, you should not assume the the rest of the public believes policy X produces result Z. This is especially true of public policies. It is very unlikely that a policy regime will be effective if it is based on the assumption that the public will respond foolishly to your policy. They might behave foolishly, but you can’t count on it.
The rational expectations revolution also showed that:
1. Today’s AD will be heavily influenced by changes in tomorrow’s expected AD, and thus by changes in the expected future path of monetary policy.
2. Changes in the expected future path of policy show up immediately in the auction-style commodity, stock, and bond markets.
The idea is the world is classical in the long run (permanently doubling money, doubles prices), but prices are slow to adjust in the short run. Some prices adjust faster than others so today’s GDP can change in response to expected future changes in monetary policy. So even in a weak monetary trap (where short-term rates are zero) if the monetary authority can affect expectations, they can have an affect on GDP.
The predicate — “if the monetary authority can affect expectations” — is where the disagreements are. Krugman thinks normal open market operations are the only mechanism for the Fed to affect expectations. Most everyone else thinks otherwise. Read the rest of the Sumner article to get six reasons to think Krugman is wrong.