Only 55.25 percent of closed credit-risk measures fully remedied
Full remediation occurred in 55.25 percent of 8,106 credit-risk supervisory measures issued and closed during 2016–2025, European Central Bank (ECB) researchers find. Data, reporting and IT deficiencies proved harder to resolve.
Closure does not mean resolution
Alessandra Agati, Adele Fontana, João Matos Leite and Claudia Schwarz classify 43.24 percent of measures as partially effective and 1.51 percent as ineffective.
Multi-topic findings and more severe measures are associated with lower full-remediation rates and longer remediation periods.
On-site inspection measures achieve better outcomes and faster closure than horizontal reviews, ongoing monitoring and the Supervisory Review and Evaluation Process.
Recommendations achieve full remediation more often than binding obligations, but legal status does not affect speed.
Stronger governance and profitability correlate with better outcomes, though bank-level effects weaken with supervisory-team controls.
The findings represent the authors' views, not necessarily the ECB's.
AI reads the closure record
GPT-4.1 classifies confidential measure descriptions and closure narratives across ten runs, using majority voting and expert review.
Agreement with o4-mini reaches approximately 90 percent.
Logistic models assess full remediation; duration models cover only fully remediated measures.
Each additional topic has a Cox closure hazard ratio of 0.84. However, proportional-hazards tests fail globally.
A Weibull alternative confirms the direction but estimates only 2 percent longer duration per additional topic and 4 percent longer for review-process measures versus inspections.
Outcomes, not administrative completion
Formal closure is not evidence that a bank has resolved its deficiencies.
The paper measures documented remediation, not subsequent reductions in credit risk.
Its observational design and exclusion of ongoing measures prevent causal claims and limit comparisons involving recent years.