We investigate subnational disparities in preliminary reference activity by locating national courts on maps of the EU territory. Spatial visualization reveals that involvement in the preliminary ruling procedure tends to be concentrated in a relatively small subset of regions within member states. Using a machine learning approach, we explore a wide range of possible predictors and the relations among them. Our data-driven analysis shows that regions that are the seat of a peak court and have a large cargo port are associated with higher referral rates. So too are regions that are the seat of the country's capital and regions exhibiting greater economic dynamism. Our ndings directly inform the theoretical discussion and suggest ways to reconcile varying strands of research on trade, courts, litigation and institutional change in the EU context.