Severe model defects drive majority of supervisory capital add-ons
Severe model defects, rather than the total number of flaws, drive supervisory capital add-ons imposed on euro area banks using internal ratings-based models. European Central Bank research shows most high-severity findings stem directly from non-compliance with data governance rules.
The heavy cost of severe defects
Analyzing 267 Internal Model Inspections across Single Supervisory Mechanism banks between 2014 and 2020, researchers found that supervisory limitations required banks to hold double-digit billion euro amounts in additional Common Equity Tier 1 capital.
This impact vastly exceeds the EUR 44.3 million in total supervisory sanctions imposed by the European Central Bank over the same period.
Econometric results demonstrate that the overall number of identified model deficiencies is not a statistically significant driver of risk-weighted asset increases.
Instead, capital add-ons are almost exclusively explained by the presence of top-tier severity findings, classified as F4 deficiencies.
Data governance as the main channel
High-severity model deficiencies trace directly to non-compliance with Capital Requirements Regulation articles governing bank data management and data governance.
Poor data inputs systematically compromise model output reliability, creating an early warning indicator for supervisory model risk.
The econometric findings remain robust when controlling for bank size, profitability, capital headroom, and cost of risk, as well as accounting for the impact of the Targeted Review of Internal Models conducted across major euro area institutions.
Quality over mechanical checklists
The paper reveals a clear inefficiency in supervisory checks that obsess over minor technical flaws.
Prioritizing data governance over administrative tick-boxing would streamline supervision without eroding resilience.
Supervisors must urgently abandon mechanical checklists in favor of targeted risk-based enforcement.