Machine Learning Applications in Agricultural Economics

Samenvatting (Engelstalig)

The European Union has ambitious environmental goals, but evaluating whether current policies effectively promote sustainable farming remains challenging. This dissertation investigates how machine learning—computational methods that identify complex patterns in data—can improve agricultural policy evaluation. The research develops novel frameworks combining machine learning's flexibility with traditional econometrics' causal inference capabilities. Applying these methods to evaluate EU Common Agricultural Policy initiatives (2014-2020) reveals important findings: many agri-environmental programmes show windfall effects, where farmers receive payments without changing practices. Organic farming conversion analysis uncovers strategic behavior, with farmers reducing costs during transition periods before adjusting management after certification. The work demonstrates that machine learning enhances policy evaluation by handling complex relationships and revealing varied effects across farming systems. These findings suggest current sustainability policies may not achieve intended behavioral changes, highlighting the need for refined policy design to support European agriculture's environmental transition.