EU cohesion policy and “conditional” effectiveness: What do cross-section regressions tell us?

About one third of total EU budgetary resources are spent on implementing cohesion policy. Therefore, it is understandable that the European Commission and especially donor states (acting for the taxpayers) need to be reassured that their contributions are spent wisely and are being used effectively in achieving their stated goal of promoting growth and thereby reducing welfare differences throughout the Union. Different evaluation methods have been proposed to look at the likely impact of Structural Funds interventions ranging from macroeconometric models to case studies. Recently, evaluation results based on enhanced growth rate regressions with panel data have received wide interest. Ederveen et al. (2006, 2002) are two widely cited works that address the evaluation of the effectiveness of cohesion policy using the single equation, panel dataset approach. The results support a serious critique of cohesion policy, asserting that its effectiveness is conditional on country characteristics that may be in short supply in many poorer member states (e.g., the quality of public institutions), and that cohesion policies should not be implemented in the new member states unless the institutional capacities are installed. This paper takes a closer look at the Ederveen et al. results, mainly from three directions. Firstly, we discuss some issues concerning the general set-up of the database and the time period that was used, secondly show that their preferred regression seems mis-specified and instable concerning the countries included and the time period used and thirdly discuss in more general terms that the use of this methodology in the whole area of policy evaluation has been shown to be deeply flawed and to tell us nothing about the effectiveness of public policy. Our analysis of the methodology and results of Ederveen et al. drive us to the conclusions that the policy recommendations derived from this work are unsound, unwise and without merit. In particular, the recommendations concerning the new EU member states should not be based on an appeal to the cross-section regressions that are presented in their 2006 paper. In contrast, we propose two other approaches – the macroeconometric modelling approach and the microeconomic approach - which, if developed together, hold out the possibility of more robust and insightful analysis and conclusions. Only by looking deeper into the manner in which EU Cohesion Policy is actually designed and implemented, the manner in which national governments operate parallel regional policies with no reference to Brussels, and by making use of more searching and holistic models is it likely to be possible to deliver verdicts on whether or not the EU has a role in this important area of integration, and if the answer is “yes”, how that policy can be modified in light of the recent enlargements. Dogmatic conclusions reached in the literature, mainly negative, but the point also applies to supportive conclusions, are premature and almost certainly wrong.