The Computation of Convergence, or: How to Chase a Black Cat in a Dark Room
Journal of European Public Policy, Vol. 16, No. 7, pp. 990-1011, October 2009.
31 Pages Posted: 12 Jan 2007 Last revised: 24 Oct 2011
Date Written: January 8, 2007
Abstract
This paper seeks to bridge the gap between theories of convergence and its empirical tests. We start by assessing the variance approach, which is the dominant way of testing convergence theories. Based on analyzing various artificially generated convergence processes, we find that neither the standard deviation approach nor the coefficient of variation is capable to detect convergence when it is conditional. Based on our findings, we provide guidance to researchers who aim at developing and testing their theories of convergence. First, we recommend a set of questions which scholars should address within their theoretical frameworks. Secondly, we recommend estimating rather than measuring convergence. Estimating convergence bears the crucial advantage that researchers may a) test the causal relationship, b) control for structural constraints, c) control for the existence of convergence clubs, and d) test convergence to an equilibrium level of a policy. In addition, estimation eliminates noise such as unsystematic measurement error. Following both recommendations should help to close the existing gab between theoretical accounts and empirical analyses of convergence.
Keywords: convergence, clubs, variance approach, regression approach
Suggested Citation: Suggested Citation