Identifying farmers' response to changes in marginal and average subsidies using deep learning

Abstract

AbstractMuch of the developed world has adopted substantial, complex agricultural subsidy schemes in an attempt to produce desired rural livelihood and environmental outcomes. Understanding how farmers adjust their production activity in response to farm subsidies is crucial for setting optimal agricultural policy. Whereas standard economic theory suggests that farmers largely adjust production levels in response to prices and marginal subsidy rates, recent work in consumer behavior suggests that average (dis‐)incentives may play a relevant role. We use a unique panel covering all farms applying for subsidies in Norway and a flexible deep‐learning method to exploit kinks in the subsidy scheme to answer whether farmers respond more to average or marginal subsidies. In contrast to the standard economic theory of production, we find suggestive empirical evidence that farmers respond more to changes in average payments than to changes in marginal payments. We anticipate that our findings on the relevance of average payment levels for farmers' decision making may inspire further theoretical and empirical inquiries into agricultural policy effects. The study also highlights how novel deep‐learning tools can be applied for detailed policy analysis and what advantages and challenges come with it. We believe that this approach has substantial potential for analysts and policymakers to evaluate and predict the impacts of policy options.