How well can machine learning predict export credit insurance claims, and which technique works best?
Using Berne Union data, the most comprehensive collection of export credit insurance data, the study compares decision trees, random forests, neural networks and probabilistic neural networks against a simple benchmark in predicting claims. All techniques predicted reasonably well whether claims would occur and, with limits, their order of magnitude, but none predicted actual claim ratios satisfactorily. Random forests performed best across all tasks.
Can free trade agreements backed by export credit agencies give the EU a resilient answer to protectionism, conflict and trade policy uncertainty?
How do European export credit agencies finance digital infrastructure and services abroad, and what reforms would make them enablers of partners’ digital transformation?
How do multilateral and plurilateral trade agreements shape the room African countries have to design their own trade policy?
