Structural Decomposition Analyses with Dependent Determinants

Abstract

Structural decomposition techniques are used to break down the changes in one variable into the changes in its determinants. Typically, these determinants are assumed to be independent. Using the decomposition of value added growth as a prototype example, this paper examines the phenomenon that several of the determinants are not independent. The determinants are termed fully dependent if changes in one determinant cannot occur without corresponding changes in another determinant. In most empirical cases, full dependence exists between groups of determinants, not between separate determinants. It is indicated that dependencies may cause a bias in the results of decomposition analyses. An alternative to overcome this problem is proposed and the findings are illustrated by an empirical study for The Netherlands 1972-1986.