IRB models lower bank risk density and shift credit to large firms
A Bank of Italy study finds that internal credit risk models lower risk-weighted asset density without opportunistic risk underestimation by weak banks. Analyzing 124 euro area banks from 2015 to 2023, researchers show that models prompt banks to shift credit toward large corporations.
Supervisory oversight limits risk gaming
Researchers Maria Alessia Aiello, Salvatore Cardillo, and Caterina Ciancaglioni analyzed supervisory data for 124 significant euro area banks operating under Single Supervisory Mechanism (SSM) oversight between March 2015 and December 2023.
The empirical findings confirm that adopting internal ratings-based (IRB) models leads to a measurable decline in risk-weighted asset (RWA) density, measured as RWA divided by exposure at default.
However, contrary to earlier studies from the pre-SSM era, weakly capitalized or fragile banks did not cut risk density more aggressively than well-capitalized peers.
This indicates that centralized, harmonized supervision through initiatives like the Targeted Review of Internal Models effectively restrained opportunistic capital optimization.
Reallocating capital to corporate borrowers
Beyond measuring risk density variations, the paper evaluates how internal models reshape bank balance sheets.
Adoption of IRB models incentivizes banks to reallocate portfolio exposures away from low-yielding, zero-risk-weight assets—such as central bank reserves and sovereign debt—toward more profitable non-financial corporations.
This credit expansion is concentrated in large corporate exposures.
Lending to small and medium enterprises remained unaffected, as SMEs had already received capital relief under the EU SME supporting factor framework enacted in 2014.
Disciplined models, skewed incentives
Harmonized SSM oversight has successfully constrained opportunistic risk-weighting by European banks.
Yet, credit expansion concentrated in large corporates reveals how risk-sensitive rules reshape real-economy lending.
Regulators must ensure model optimization does not obscure underlying risks in stress periods.