Red Herrings and Revelations: The Destabilizing and Stabilizing Effects of Economic Theory

This paper assumes that output depends on a large number of variables, but that economic theory has only revealed a subset of these variables. The variables that determine output include exogenous shocks, as well as endogenous variables that depend on both the exogenous variables and agents’ expectations. This paper examines a policy maker’s attempts to forecast output under two different types of expectations formation. I first analyze the case where the policy maker knows the structural coefficients associated with all known variables, and uses this knowledge to form expectations. Under these structural coefficients expectations, if economic theory reveals a new variable, then welfare will improve with probability between one-half and one. Under adaptive learning, however, the revelation of a new variable may worsen welfare more often than it improves welfare. This scenario occurs when the policy maker already knows most of the variables that determine output and when the model includes significant endogeneity. Furthermore, the policy maker is typically better off choosing to abandon structural coefficients expectations in favor of adaptive learning, regardless of how many variables theory has revealed. JEL Classification: E27, E61, D81, D83.