Predicting Delays in Cohesion Infrastructure Projects

Public investment in infrastructure is essential for economic growth, but delays in project implementation can undermine its benefits. This paper examines the determinants of such delays using data from cohesion projects in Italy. We predict which projects are likely to experience delays and identify the key contributing factors by means of machine learning (ML) techniques. To avoid endogeneity, we use only (lagged) features observed at the start of the project as predictors. Our findings show that socioeconomic factors and institutional weaknesses in various regions play a significant role in these delays. The discipline imposed by rules and strict implementation timing on EU funds seems to work, lending credibility to the hypothesis of the benefit of an external commitment. Results underscore the potential of ML in designing appropriate implementation policies, enhancing project management, and improving the outcomes of public investments.