Abstract
This article characterizes the local effects of parametric behavioral model change on relationships between aggregate variables, and it presents consistent estimators of such effects using cross-section data. Two equivalent interpretations of model-change effects are given: an “average-marginal” formulation and a cross-section regression formulation. The relation between model-change effects and maximum likelihood estimation of the behavioral parameters is explained. Finally, the article addresses the question of whether R 2 [from a crosssection ordinary least squares (OLS) regression] is a general measure of the sensitivity of aggregate relationships to model-change effects.