Two Approaches in Reading Model Interdependencies

Abstract

In order to introduce the basic notions used in our analysis, let us consider the structural form of our econometric model given by the set of n structural equations h(y,z) = 0, where y,(y∈Rⁿ) and z,(z∈R^m) are the endogenous and exogenous variables respectively. To be regular, such a model has to verify the existence of a complete matching between the set of relations and the set of endogenous variables. This means that it has to be possible to assign a different left-hand variable to each equation, which enables us to write the model as a set of n explicit functions y_i = g_i(y;z). Then, if the model is written in this form, a non-zero derivative ∂g_i/∂y_jindicates the existence of a causal link going from variable y_j to variable y_i. The set of all these links defines what is called the causal structure of the model. The present study is based exclusively on information of this type.