Data-Driven Supervision in a Decentralised Supervisory Environment

This working paper argues that data-driven supervision (DDS) provides a principled and practical resolution to the tension between supervisory convergence and institutional decentralisation at the heart of the current EU debate on supervisory architecture. Rather than forcing a binary choice between centralised EU-level supervision and the status quo, DDS enables a collegial centralisation of risk identification, metric definition, and supervisory expectations while preserving operational decentralisation of day-to-day supervision. The paper formalises supervision as a problem of statistical inference under measurement error and resource constraints, and proposes a five-step supervisory cycle through which a decentralised network of National Competent Authorities can achieve common risk identification, shared metrics, harmonised data standards, consistent triage, and continuous collegial calibration. The SPMU (Severity, Persistence, Materiality, Uncertainty) composite scoring architecture provides the analytical core. The paper analyses the economics of DDS - demonstrating scale economies, multi-mission amortisation, and early-intervention savings - and develops the governance, model-risk, and institutional-sustainability requirements for sustainable deployment. The central conclusion is that, for market segments characterised by large dispersed populations under harmonised EU legislation, a decentralised DDS architecture delivers effective, convergent, and proportionate supervision at lower institutional cost and with fewer transition risks than full centralisation of supervisory powers.