The Manipulation of Basel Risk-Weights

Transcription

1 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

2 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, risk-weighted assets (RWA) = i A i ω i E RWA x

3 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

4 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

5 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)

6 Contribution This paper does the data support Blum (2008) s hypotheses?

7 Contribution This paper does the data support Blum (2008) s hypotheses? average reported riskiness ( RWA ) declined upon IRB adoption TA specifically for weakly capitalised banks

8 Contribution This paper does the data support Blum (2008) s hypotheses? average reported riskiness ( RWA ) declined upon IRB adoption TA 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)

9 Contribution This paper does the data support Blum (2008) s hypotheses? average reported riskiness ( RWA ) declined upon IRB adoption TA 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

10 Contribution This paper does the data support Blum (2008) s hypotheses? average reported riskiness ( RWA ) declined upon IRB adoption TA 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

11 Contribution This paper does the data support Blum (2008) s hypotheses? average reported riskiness ( RWA ) declined upon IRB adoption TA 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

12 Contribution This paper does the data support Blum (2008) s hypotheses? average reported riskiness ( RWA ) declined upon IRB adoption TA 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

13 Contribution This paper does the data support Blum (2008) s hypotheses? average reported riskiness ( RWA ) declined upon IRB adoption TA 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

14 Contribution This paper does the data support Blum (2008) s hypotheses? average reported riskiness ( RWA ) declined upon IRB adoption TA 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

15 Motivation RWA TA (%) before & after IRB adoption Resolved Not resolved Full implementation of the advanced approach Quarters before and after implementation strong decline and leveling with non-resolved banks afterwards harder to disentangle stable from fragile banks

16 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, (Bankscope) 21 OECD countries

17 Setup Model RWA = α 0 + α 1 1 IRB i,t + β X i,t + u i,t TA i,t 1 IRB i,t : IRB adoption dummy (= 1, for t implementation date) X i,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)

18 Results IRB adoption associated with lower reported riskiness magnitude: 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?)

19 Results Explanation 1: IRB induced re-allocation of resources towards safer assets

20 Results consistent with re-allocation of resources (column 1) weakly capitalised banks report (relatively) lower riskiness (Blum, 2008)

21 Results Explanation 2: Risk-weights implied by Basel I were fundamentally too high

22 Results well capitalised banks increase capital after IRB adoption not weakly capitalised banks consistent with justified adjustment, but...

23 Results ex post capital was fundamentally too low

24 Results Explanations 3 & 4: Models implied too little capital. By accident, or intentionally?

25 Results more supervisory scrutiny associated with higher reported riskiness

26 Results more supervisory scrutiny associated with higher reported riskiness

27 Results results robust at the country level

28 Results consistent with misconduct

29 Results consistent with misconduct but:

30 Results consistent with misconduct but: closer supervision could also enhance model quality

31 Results consistent with misconduct but: closer supervision could also enhance model quality additional evidence: prudence more generally & loan quality

32 Results credit risk not systematically underestimated by weakly capitalised banks weakly capitalised banks behave less prudently

33 Results credit risk not systematically underestimated by weakly capitalised banks weakly capitalised banks behave less prudently

34 Results credit risk not systematically underestimated by weakly capitalised banks weakly capitalised banks behave less prudently

35 Results credit risk not systematically underestimated by weakly capitalised banks weakly capitalised banks behave less prudently

36 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)