Friday, September 10, 2010

What I'm Writing

Roger Koppl and I are currently working on a book chapter titled "Animal Spirits and Cognition in Macroeconomics." I'll post the full draft shortly. Here's a snippet on rational expectations which (at least I hope) takes aim at the typical presentation of the assumption (you know: the Phillips curve assumes expectations are unchanged; then Friedman and Phelps said that was stupid; and Lucas, building on Muth, discovered rational expectations as the logical conclusion to the argument of Friedman and Phelps--as if no one had thought of it before):
It is historically inaccurate (and, to the extent one knows otherwise, intellectually dishonest) to claim the rational expectations assumption was discovered in the 1960s with Muth. The theoretical power behind the assumption of full and complete expectations had in fact been known for some time. Among others, Hicks (1936, p. 241) had expressed the basic idea in admitting it “unrealistic to assume that an important change in data—say the introduction or extension of a public works policy—will leave expectations unchanged, even immediately.” Rational expectations was not discovered, but rather rediscovered in the 1960s because earlier theorists had explicitly rejected the assumption. In the case of Hicks, it was assumed that “there is a psychological unknown, affecting the magnitude of the impact effect” and, as such, “[w]e must not expect the most elaborate economic analysis to enable us to see very far ahead” (p. 241). Earlier economists were aware of rational expectations and, as Meacci (2009, p. 1) describes, saw the assumption primarily as “a device […] to conceal the link between the disappointment of expectations and the theory of fluctuations.” It was not a lack of knowledge that had left the assumption largely unemployed, but the feeling that it was wholly inappropriate to use in addressing a topic so intimately linked to the process of time.
The comments are open. I'd love to hear what you think.


Eli said...

I think of RE as mean-zero errors. Did the early RE types think of it in those terms?

Also, note that RE is consistent with clusters of errors (as in, for example, *Risk and Business Cycles*). I think there is not any good reason to believe that RE + clusters is inappropriate, "process of time" or no.

Will Luther said...

Mean-zero errors is how Muth articulated it. I am not sure if anyone explicitly stated it like that prior to Muth.