How Common is Identification in Parametric Models?
Posted: 14 Oct 2008
Date Written: 1992
This paper examines whether identification is a 'common' or 'rare' phenomenon in nonlinear parametric models. For several broad classes of models, it is shown that there is an open and dense subset of identified models, and that consequently the set of nonidentified models is nowhere dense. The results of this paper suggest that the assumption of linearity in functional relationships to ease the conceptual and computational development of a theory can drastically limit the estimability of the model. An interesting feature of this paper is that global identification is established in the context of nonlinear models.
Keywords: Identification, nonlinear parametric models
JEL Classification: C1
Suggested Citation: Suggested Citation