Expectations Robust Policy Evaluation is Impossible

Macroeconomists have a problem. We know that agents' expectations are very important, but we don't understand how people form expectations. This is a problem for psychologists and sociologists. They haven't solved it and we can't wait decades (or centuries or millennia) until they do.

One proposal was to assume that people have rational (that is model consistent or Nash equilibrium) expectations. This requires either a unique Nash equilibrium or the abandonment of methodological individualism. The assumption that the world is in Nash equilibrium and it is unique rules out bubbles. So we can either advise policy makers on how to deal with the housing bubble and the dot com bubble and their aftermaths by assuming that bubbles are impossible, or try something else.

Hansen and Sargent (extremely eminent Nobel prize winning fresh water economists who were once enthusiasts for rational expectations) have thoughts about what to do. These thoughts are expressed in papers which I haven't read. I will debate Straw Hansen and Sargent (SHS)

SHS says we should seek policy rules which yield OK outcomes for any conceivable expectations formation mechanism. SHS doesn't understand that this quest is utterly futile, because SHS has limited imagination and can conceive only of expectations mechanisms which have been explored by SHS.

In fact, there can be no policy which gives OK outcomes for any conceivable expectations mechanism. The idea that knowing that people form expectations somehow has policy implications is like the idea that assuming that people maximize some utility function has policy implications. Any illusion of implications can be based only on a failure of imagination.

So consider economy X and proposed policy A. I may make any assumption about the expectations of agents in the economy I please. I assume that if the policy maker implements any policy other than A then agents have adaptive expectations. I also assume that if the policy maker implements A then agents believe that there will very soon be an instantly spreading plague which will paralyze all people and make us die slow horrible deaths (without the ability to put ourselves out of our misery.

Thus the agents all maximize their utility by killing themselves (altruistic agents kill others first).

I don't think this is an OK outcome.

Now the expectations mechanism which implies that just because policy A is implemented everyone will instantly believe something crazy is not plausible.

But at this point, all hope of having anything useful to say is based on the belief that we can have some confidence that human psychology (and sociology) fall in some plausible subset of the conceivable. That's not robust in any well defined way.