The heterogeneous causal effects of the EU's Cohesion Fund

Abstract

This paper estimates the causal effect of EU cohesion policy on regional output and investment, focusing on the Cohesion Fund (CF), a comparatively understudied instrument. Departing from standard approaches such as regression discontinuity (RDD) and instrumental variables (IV), we use a recently developed causal inference method based on matrix completion within a factor model framework. This yields a new framework to evaluate the CF and to characterize the time-varying distribution of its causal effects across EU regions, along with distributional metrics relevant for policy assessment. Our results show that average treatment effects conceal substantial heterogeneity and may lead to misleading conclusions about policy effectiveness. The CF's impact is front-loaded, peaking within the first seven years after a region's initial inclusion. During this first seven-year funding cycle, the distribution of effects is right-skewed with relatively thick tails, indicating generally positive but uneven gains across regions. Effects are larger for regions that are relatively poorer at baseline, and we find a non-linear, diminishing-returns relationship: beyond a threshold, the impact declines as the ratio of CF receipts to regional gross value added (GVA) increases.