The Manipulation of Basel Risk-Weights
Mike Mariathasan
University of Oxford
Ouarda Merrouche
Graduate Institute, Geneva
CONSOB-BOCCONI Conference on Banks, Markets and Financial Innovation;presented by: Charlotte Werger, EUI
May 22, 2013
Introduction
Bank capital regulation
� core of (microprudential) bank regulation� global standards continuously revised: Basel I, Basel II, Basel III, ...
� dual objective:� enhances shareholder monitoring ex ante� protects depositors (tax payer) ex post
� trade-off:� stability� inefficient funding structure
� central element: minimum capital requirement x , ERWA ≥ x
� risk-weighted assets (RWA) =�
i Aiωi
Introduction
Basel risk-weights� idea: more risky assets require more monitoring/ larger buffer
� Basel I:� regulator decides riskiness of the asset & assigns risk-weight� few & simple risk categories (simpler than banks’ risk models)
� Basel II:� more complex risk categories� external credit ratings (Standard Approach, SA)� internal models (Internal Ratings-Based Approach, IRB)
� F-IRB: banks calculate PD, (M)� A-IRB: banks calculate PD, LGD, EAD, M
� Basel III:� “more of the same, but better” (Haldane, 2011)� internal models still used
� literature: limited risk-sensitivity of risk-weights under IRB
Introduction
Basel risk-weights� idea: more risky assets require more monitoring/ larger buffer
� Basel I:� regulator decides riskiness of the asset & assigns risk-weight� few & simple risk categories (simpler than banks’ risk models)
� Basel II:� more complex risk categories� external credit ratings (Standard Approach, SA)� internal models (Internal Ratings-Based Approach, IRB)
� F-IRB: banks calculate PD, (M)� A-IRB: banks calculate PD, LGD, EAD, M
� Basel III:� “more of the same, but better” (Haldane, 2011)� internal models still used
� literature: on-going search for determinants
Introduction
Basel risk-weights� idea: more risky assets require more monitoring/ larger buffer
� Basel I:� regulator decides riskiness of the asset & assigns risk-weight� few & simple risk categories (simpler than banks’ risk models)
� Basel II:� more complex risk categories� external credit ratings (Standard Approach, SA)� internal models (Internal Ratings-Based Approach, IRB)
� F-IRB: banks calculate PD, (M)� A-IRB: banks calculate PD, LGD, EAD, M
� Basel III:� “more of the same, but better” (Haldane, 2011)� internal models still used
� literature: incentives for banks to misreport risk (Blum, 2008)
Contribution
This paper
� does the data support Blum (2008)’s hypotheses?
� average reported riskiness (RWATA ) declined upon IRB adoption
� specifically for weakly capitalised banks
Contribution
This paper
� does the data support Blum (2008)’s hypotheses?
� average reported riskiness (RWATA ) declined upon IRB adoption
� specifically for weakly capitalised banks
� competing explanations� IRB induced re-allocation of resources towards safer assets� Risk-weights implied by Basel I were fundamentally too high� IRB weights are too low (by accident)� IRB weights are too low (intentionally)
Contribution
This paper
� does the data support Blum (2008)’s hypotheses?
� average reported riskiness (RWATA ) declined upon IRB adoption
� specifically for weakly capitalised banks
� competing explanations� IRB induced re-allocation of resources towards safer assets� Risk-weights implied by Basel I were fundamentally too high� IRB weights are too low (by accident)� IRB weights are too low (intentionally)
� corresponding empirical predictions
Contribution
This paper
� does the data support Blum (2008)’s hypotheses?
� average reported riskiness (RWATA ) declined upon IRB adoption
� specifically for weakly capitalised banks
� competing explanations� IRB induced re-allocation of resources towards safer assets� Risk-weights implied by Basel I were fundamentally too high� IRB weights are too low (by accident)� IRB weights are too low (intentionally)
� corresponding empirical predictions� no effect when controlling for loan categories
Contribution
This paper
� does the data support Blum (2008)’s hypotheses?
� average reported riskiness (RWATA ) declined upon IRB adoption
� specifically for weakly capitalised banks
� competing explanations� IRB induced re-allocation of resources towards safer assets� Risk-weights implied by Basel I were fundamentally too high� IRB weights are too low (by accident)� IRB weights are too low (intentionally)
� corresponding empirical predictions� effect should disappear when controlling for loan categories� banks should be able to reduce capital and remain stable
Contribution
This paper
� does the data support Blum (2008)’s hypotheses?
� average reported riskiness (RWATA ) declined upon IRB adoption
� specifically for weakly capitalised banks
� competing explanations� IRB induced re-allocation of resources towards safer assets� Risk-weights implied by Basel I were fundamentally too high� IRB weights are too low (by accident)� IRB weights are too low (intentionally)
� corresponding empirical predictions� effect should disappear when controlling for loan categories� banks should be able to reduce capital and remain stable� very hard to distinguish
Contribution
This paper
� does the data support Blum (2008)’s hypotheses?
� average reported riskiness (RWATA ) declined upon IRB adoption
� specifically for weakly capitalised banks
� competing explanations� IRB induced re-allocation of resources towards safer assets� Risk-weights implied by Basel I were fundamentally too high� IRB weights are too low (by accident)� IRB weights are too low (intentionally)
� corresponding empirical predictions� effect should disappear when controlling for loan categories� banks should be able to reduce capital and remain stable� supervisory scrutiny should matter more in case of intention
Contribution
This paper
� does the data support Blum (2008)’s hypotheses?
� average reported riskiness (RWATA ) declined upon IRB adoption
� specifically for weakly capitalised banks
� competing explanations� IRB induced re-allocation of resources towards safer assets� Risk-weights implied by Basel I were fundamentally too high� IRB weights are too low (by accident)� IRB weights are too low (intentionally)
� corresponding empirical predictions� effect should disappear when controlling for loan categories� banks should be able to reduce capital and remain stable� supervisory scrutiny should matter more in case of intention� would expect faulty models to imply more impaired loans
MotivationRWATA (%) before & after IRB adoption
Full implementation of the advanced approach
Resolved
Not resolved
5055
6065
7075
-10 -5 0 5 10Quarters before and after implementation
� strong decline and leveling with non-resolved banks afterwards
� harder to disentangle stable from fragile banks
Setup
Sample
� 115 banks that have been approved for IRB adoption� A-IRB & F-IRB� mostly for corporate & retail loans� 77% of assets covered on average
� annual balance sheet data, 2004-10 (Bankscope)
� 21 OECD countries
Setup
Model�
RWATA
�
i ,t= α0 + α1 · 1IRB
i ,t + β�Xi ,t + ui ,t
� 1IRBi,t : IRB adoption dummy (= 1, for t ≥ implementation date)
� Xi,t : control variables� Ln(TA), GDP growth, year dummies, bank FE� Bank level: gross loans, corporate loans, residential loans, liquid assets� Country level: short-term rate, public debt/ GDP
� interaction effects� variations of the LHS variable
� cross-sectional panel regressions (fixed effects)� clustered standard errors (country level)
Results
� IRB adoption associated with lower reported riskiness� magnitude: 0.0293 ≈ 0.64 bp change in the CAR
� A-IRB vs. F-IRB does not seem to matter (low coverage for A-IRB?)
� degree of coverage appears to be secondary (little variation?)
Results
� consistent with re-allocation of resources (column 1)
� weakly capitalised banks report (relatively) lower riskiness (Blum, 2008)
Results
� well capitalised banks increase capital after IRB adoption
� not weakly capitalised banks
� consistent with justified adjustment, but ...
Results
� consistent with misconduct
� but: closer supervision could also enhance model quality
� additional evidence: prudence more generally & loan quality
Results
� credit risk not systematically underestimated by weakly capitalised banks
� weakly capitalised banks behave less prudently
Results
� credit risk not systematically underestimated by weakly capitalised banks
� weakly capitalised banks behave less prudently
Results
� credit risk not systematically underestimated by weakly capitalised banks
� weakly capitalised banks behave less prudently
Results
� credit risk not systematically underestimated by weakly capitalised banks
� weakly capitalised banks behave less prudently
Conclusions
Internal risk-models under Basel II� have lead to a reduction in reported riskiness� less if banks are better capitalised & more if supervision is weak
Consistent with� Blum (2008)’s model� “DIY capital” hypothesis
Lessons for regulation� value of simple & transparent rules� tight auditing rules could complement regulation� leverage-dependent scaling factor?� Basel III goes in the right direction (leverage ratio constraint)