Increased dispersion of Risk Weighted Assets (RWA) troubles regulators as potentially undermining prudential supervision. We study the determinants of RWA/EAD (Exposure-At-Default) on data painstakingly compiled from Basel Pillar-Three for 239 European banks over 2007–2013. We improve on most previous studies, which consider instead RWA/TA (Total Assets). Indeed, Internal-Rating-Based (IRB) models allow lawful capital-saving Roll-Out effects which RWA/TA analyses disregard and likely misidentify as regulatory arbitrage. Instead, encapsulating Roll-Out effects, RWA/EAD avoids false positive identification. We find that regulatory arbitrage: (i) was present; (ii) likely materialized via risk weights manipulation with IRB models; (iii) was stronger at Advanced-IRB vs Foundation-IRB banks.