Lessons in Econometric Methodology: The Axiom of Correct Specification

Leamer first pointed out that a regression model is valid only if all of the assumptions under which it is constructed are valid. In particular, this means that all of relevant regressors which are determinants must be included. In practice, applied econometricians assume that whatever model they put is valid, when there can only be one valid model. If any relevant regressor is omitted, then the equation is mis-specified and conclusions drawn from the regression can be seriously misleading. Even though this mis-specification analysis is included in textbooks, it is routinely ignored in applications, where researcher interpret their equations as if they have correctly specified all regressors, without testing to see if this may be the cases. Hendry's encompassing methodology provides a remedy for this problem, but seems to be unfamiliar to many. The purpose of this pedagogical note is to provide an introduction to an elementary but important aspect of this methodology.