A Capital Accord for Emerging Economies?

A Capital Accord for Emerging
Economies?
Andrew Powell
Universidad Torcuato Di Tella and
Visiting Research Fellow, The World Bank
(Financial Sector Strategy and Policy - FSP)
July, 2001
Motivation
• BCBS has published a proposal to change the
1988 Capital Accord
• The 1988 Accord was perhaps the most successful
of all ‘financial standards’
• Will the ‘New Accord’ be so successful ?
• Implications for emerging economies:
– Cost of capital
– Issues regarding implementing the New Accord
• Acknowledgements, but views strictly my own
Part 1:
On the 1988 Accord and the
Proposals
On the 1988 Accord...
• Designed by G10 supervisors and aimed at
their ‘internationally active banks’.
• But, very successful - applied in at least
100 countries and in many for all banks
(Barth, Caprio and Levine 2001)
• Easy to criticize such a simple yardstick as
‘assets at risk’ but the simple standard:
– Created a type of yardstick competition
– Was adapted to local conditions
– Was easy for legislators to understand
The ‘New Accord’: Overview
•
•
•
•
•
The New Accord
Pillar 1: Requirements
Pillar 2: Supervisory Review
Pillar 3: Market Discipline
Supporting Documents on the above and on
operational risk, interest rate risk, asset
securitization etc.
 this presentation focuses on Pillar 1 and within that
on capital requirements for credit risk.
Pillar 1: Requirements
• Capital Requirements
– Standardized
– Internal Ratings
• Sovereign, Corporate, Bank, (Retail)
• Enhanced rules for credit risk mitigation
techniques
The Standardised Approach
Table 1: Risk Weights Given External Credit Ratings
Basel Committee on Banking Supervision
January 2001
Sovereign
Bank
Alternative 1
Bank
Alternative 2a
Corporate
AAA to
AA-
A+ to
A-
BBB+ to
BBB-
BB+ to
B-
Below
B-
Unrated
0%
20%
50%
100%
150%
100%
20%
50%
100%
100%
150%
100%
20%
20%
20%
50%
150%
20%
AAA to
AA-
A+ to
A-
BBB+ to
BB-
Below
BB-
Unrated
20%
50%
100%
150%
100%
a: refers to rating of bank and not to rating of sovereign, for claims of less than 3 months
lent and funded in local currency.
The Standardized Approach
for Sovereigns
• Survey evidence says banks rely on rating
agencies.
– Banks have no superior information or are they just
equally at a loss ?
• There has been a focus on pro-cyclicality
– But up-front provisions already pro-cyclical
– Are banks more pro-cyclical than the agencies?
• But, rating agencies may just get it wrong!
– Very few opinions, assessments subjective but of
systemic impact (for country).
• And there may be circularity.
The IRB Approach
• Calibrated on experience with corporate claims
and applied to Sovereign, Bank and Retail.
• Method draws on advances in risk
management in large banks.
• Banks must slot claims into Borrower Grades
(minimum of 6-9 for performing and 2 for nonperforming)
• Grades ‘mapped’ into Default Probabilities
On Model Calibration
• Uses Merton risky debt model,
Creditmetricstm probability transition matrix
and G10 corporate default experience.
• Calibrated to cover expected and unanticipated
losses to 99.5% tolerance value.
• We are told that a 3 year loan with a PD=0.7%,
LGD=50% and average asset correlation of 20%
gives a risk weight of 100% (i.e.: an 8% capital
requirement).
Some Maths…
(1  PD) 

BRW C ( PD)  976.5 * N (1.118 * G ( PD)  1.288) * 1  0.0470 *
0.44 
PD


 LGD

RWc  Min 
* BRW c ( PD),12.5 * LGD
 50

BRW = Benchmark risk weight, N(.) is cumulative normal,
G(.) is the inverse cumulative normal, PD is probability of
default and LGD is Loss Given Default. Subsequent
adjustments may then be made for maturity and ‘granularity’.
IRB Approach for Sovereigns
• AAA’s – A’s get PD=0
• No rules on how many sovereigns in each
Borrower Grade.
• The remainder of the calibration follows
that for corporates.
Part 2
On the Cost of Capital for
Developing Economies
Ratings and Capital Requirements
(from Powell 2001, 'A Capital Accord for Emerging Countries?)
50
15.0%
45
% Capital Requirements
40
35
6.6%
30
25
20
15
1.4%
10
5
0.2%
0
0.03%
AAA
0.03%
AA
0.03%
A
BBB
BB
B
Ratings
Current Requirements
Standardized
Internal Ratings
CCC
Implications of the Proposals
• For IRB approach, Required Capital much more
sensitive to PD at higher levels of PD.
• This “convexity” of IRB approach, a consequence
of credit risk mathematics (not clear how the
standardized approach is ‘calibrated’).
• Perverse incentives for banks ?
• Estimates of effects on spreads…
Estimated Effects on Capital Requirements
For BIS Reporting Banks Lending to Developing Countries
Rating
AAA
AA
A
BBB
BB
B
CCC
Number of
Countries
BIS Consolidated
Liabilities
$bn
Change in Cap. Requirements
Standardized
IRB
1
1
11
16
14
17
1
100.1
28.5
193.6
186.3
178.6
228.2
1.6
-79.1%
-97.2%
-62.6%
-18.7%
6.8%
10.5%
59.3%
-85.2%
-98.0%
-79.9%
3.7%
112.5%
380.7%
524.5%
Total Excl Mexico and Korea
Below BBB
Total Incl Mexico and Korea
Below BBB
33
1028.4
408.4
1150.9
530.9
-17.6%
9.1%
-5.6%
36.7%
90.5%
261.8%
110.1%
305.4%
Source: author’s calculations based on S&P ratings and BIS consolidated claims on developing economies
and some strong assumptions!
Effect on Average Cost of Funds for Developing Countries
(Assuming Cost of Equity of BIS Reporting Banks=18% and Cost of Funds 5%)
700
600
500
Basis Points
400
300
200
100
0
BBB
BB
B
CCC
-100
Standardized
IRB
Below BBB Excl
Mexico
Below BBB Incl
Mexico
An alternative IRB Approach for
Sovereigns
• Banks must assign an ‘internal’ rating on a
Moody’s, S&P or Fitch type scale
• Capital Requirements are then given by the
relevant bucket in the standardized
approach
 Given the uncertainty of applying the corporate-calibrated,
IRB approach to sovereigns, this would create a less
controversial and less convex scale.
Part 3:
How Emerging Economies
may implement the New Accord
Implementing the New Accord :
Two important issues
• Level of Consolidation
– consolidation required to one level above
internationally active bank
– but many emerging economies have not yet
implemented consolidated supervision
• Related Lending
– material exposures of more than 15%
(individual) and 60% (aggregate) deducted
from capital
Standardised versus IRB
approach?
• The standardised Approach would lead to little change due
to the limited universe of rated institutions (eg: Argentina
has about 150 rated corporates but there are 25,000
corporates in the BCRA ‘credit bureau’)
• Or, very sharp rise in demand for ratings and potential
‘race to the bottom’ in terms of rating quality.
• The Internal Ratings approaches will be very difficult to
implement
 Likely result: will ‘implement’ the standardised approach and
little will change (execpt enhanced rules on collateral and other
supporting documents – Pillars 2 and 3 etc.)
Provisions and Capital
• In Latin America:
– some regulators have more legal autonomy to
determine provisions, capital requirements are
determined in laws
– provisions do not only reflect ex post
accounting losses
– level of provisioning often higher than in G10
– in some countries, provisions set through highly
developed credit bureau policies in attempt to
reflect expected losses
Probability
Graph 1: The credit loss distribution:anticipated vs unanticipated
losses, provisions vs capital
0
2500
5000
7500
Provisions =
Expected Loss
10000
12500
15000
17500
Capital
Requirements =
Unexpected
Loss
20000
22500
25000
Credit Loss ($)
27500
More general points regarding
model calibration...
• Obvious doubts:
– As models are non-linear, difficult to recalibrate to local conditions (not clear what
11.5% and not 8% means), needs full recalibration
– Not clear whether the correlation of 0.2 nor the
‘granularity’ adjustment are appropriate in
emerging economies with less diversifiable
risk.
On Minimum Requirements for
Internal Rating Authorisation
• The focus is on a bank developing an internal rating
system, for that system to satisfy a set of minimum
requirements and to be integral to the bank’s
operations.
• Would give a degree of autonomy in setting
regulatory capital that many emerging economy
supervisors may find unacceptable.
• Minimum laid down requirements for the rating
process may not ‘fit’ emerging economy data.
Credit Bureau Policies
• At the same time, some emerging countries have
developed ‘credit bureau’ policies. Miller (2000)
“Credit Reporting Systems around the Globe”,
provides a review.
• In 21 of the 30 countries reviewed, there is a
rating/grade (normally 5 or 6) and in most cases this is
set by the bank but scrutinised by the regulator.
• The objectives have been to improve the information
in the financial system but in many cases the ratings
feed directly into provisioning requirements.
The Credit Bureau Policy of the
Central Bank of Argentina
• Started in 1992, gradually increasing in coverage,
now all loans > $50, 8 million entries per month.
• 5/6 grades: 1 and 2 performing, 3-5/6 nonperforming (does not satisfy Basel II).
• Database available for individual inquiries free
through the internet, all information >$200k and
‘good’ information >$200k ‘sold’ on a cd
• Corporate >200k grades forward looking (ex ante),
retail grade is function of arrears (ex post).
• Feeds directly into provisioning requirements
Use of Credit Bureau Data to
Assess Capital Requirements
• Anticipated and unanticipated losses: two statistics
from the same distribution
• Can use non-parametric models (eg: Carey 2000
and Falkenheim and Powell 2000) or parametric
models (eg: Creditmetricstm) or study the
probability of default from a credit scoring model
with a parametric model (eg: Creditrisk+tm) to
determine anticipated and unanticipated losses.
• Central Bank of Argentina has Creditrisk+tm up
and running with a preliminary credit scoring
model.
Emerging Country Supervisors
face a difficult choice
• Use model results to re-calibrate internal
ratings approach to local conditions.
• Adapt ‘credit bureau’ policies to ‘fit’
minimum laid down standards.
• Use existing credit bureau policies as a
basis for PD estimates.
Another Alternative is to
Develop a Separate Standard
For Emerging Economies
Autonomy
Flexibility
Simplicity
Complexity
Standardization
Control
Conclusions
• 1988 Capital Accord was a tremendous success
• The ‘new Accord’:
– May imply significant increases in the cost of funds
for lower rated emerging economies.
– Regarding implementation, may be largely
irrelevant (standardised approach) or difficult
(IRB) approach.
• The more the New Accord ‘fits’ risk management
policies of large international banks, the less it
may fit most banks in emerging economies
• The time has come to think about a new
Accord for emerging economies