A Bayesian Perspective on Inference from Macroeconomic Data.

Bayesian methods are logical, admissible, linguistically natural, helpful, conceptually simple, and honest. But they are also preposterous and very labor intensive. Generalized Bayesian methods allow data to be confusing and inconclusive. A Bayesian viewpoint allows one to get a handle on the conundrums of exploratory data analysis which are prevalent when the theory is vague and experiments are expensive. These ideas are illustrated with a Bayesian analysis of the determinants of the U.S. unemployment rate using four different priors as inputs. Copyright 1991 by The editors of the Scandinavian Journal of Economics.