Credit Risk Irb Approach2
Credit Risk Irb Approach2
Credit Risk Irb Approach2
Laurent Balthazar
1
Agenda
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 2
Agenda
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 3
1. Introduction to IRB: The 3 pillars
Basel Committee
4
1. Introduction to IRB: The 3 pillars
Scope of application
Holding => Basel 2 rules apply
6
1. Introduction to IRB: The 3 pillars
Deducted or
consolidated on a pro Risk Weighted
rata basis
7
1. Introduction to IRB: The 3 pillars
Deducted or other
method (national Risk Weighted
discretion)
8
1. Introduction to IRB: The 3 pillars
Majority-owned /
Controlled and Minority
investments
10
1. Introduction to IRB: Pillar 1
≥ 8%
11
1. Introduction to IRB: Credit risk
Credit risk
Standard approach
• Continuation of Basel 88
• New: internal ratings (+ asset class): 0%, 20%, 50%, 100%, 150%
12
1. Introduction to IRB: Why IRB ?
Standard approach
IRB
• If there should be only one reason:
13
1. Introduction to IRB: Why IRB ?
15
1. Introduction to IRB: Why IRB ?
16
1. Introduction to IRB: Why IRB ?
17
1. Introduction to IRB: From proposal to law
3-level structure
open
options
• EU : directives
open
options
18
1. Introduction to IRB: Exposure classes
• Sovereign:
- Countries
- Central banks
- Multilateral Development Banks
- Public Sector Entities (PSE) considered as sovereigns by regulators
(tax raising power)
- International organizations RW 0% in standard (e.g.: BIS, IMF …)
• Banks:
- Financial institutions
- PSE not considered as sovereigns
20
1. Introduction to IRB: Exposure classes
• Mortgages:
- Exposure covered by residential mortgage
- No size limit
- If rented must be limited
21
1. Introduction to IRB: Exposure classes
• Other retail:
- Loans on a group < 1Mio EUR
- Large number of exposures in portfolio
- Must be managed on a pool basis
- Individual analysis does not prevent mass treatment
23
1. Introduction to IRB: Terminology
IRB ingredients
Symbol Name Comments
It is the probability that the counterparty will not
PD Probability of Default
meet its financial obligations
It is the expected amount of loss that will be
LGD Loss Given Default incurred on the exposure in case the counterparty
defaults
It is the expected amount of exposure at the time
when a counterparty will default (the expected
EAD Exposure at Default
drawn-down amount for revolving lines or the off-
balance sheet exposure times its CCF)
M Maturity The average maturity of the exposure
A measure of association between the evolution of
ρ Asset correlation
assets returns of the various counterparties
The degree of confidence used to compute the
CI Confidence Interval
economic capital
24
1. Introduction to IRB: Terminology
25
1. Introduction to IRB: Terminology
1. Risk Components
3. Minimum
2. Risk weight function
Requirements
Capital requirements
26
1. Introduction to IRB: Roll out
Exceptions:
• Specialized Lending
• Securitisation
• Non-material exposures
P a ra lle l
ru n
R e la x e d : 2 ye a rs
R a tin g s ys te m : 3 y e a rs (*)
R e ta il: 5 y e a rs (*)
P D : 5 y e a rs (*)
L G D : 7 y e a rs
E A D : 7 y e a rs
T ra n s itio n p e rio d : 3 ye a rs
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
(*): re la xe d d u rin g tra n s itio n p e rio d
28
1. Introduction to IRB: Transition period
• 2009 … ?
29
Agenda
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 30
2. Credit risk in IRB: PD Minimum
Requirements
• Retail
Rating structure
• Corporate, Sovereign and Banks
• Retail
32
2. Credit risk in IRB: PD Minimum
Requirements
Rating criteria
33
2. Credit risk in IRB: PD Minimum
Requirements
Rating horizon
- PD on 1 year
- Ratings on longer horizon
- Rating: reimbursement capacity under adverse economic conditions
- PIT Vs. TTC issues (procyclicality…)
Business
Cycle ¾PIT rating
?
¾Average TTC
rating
¾Stress TTC
rating
Today Time
34
2. Credit risk in IRB: PD Minimum
Requirements
Use of models:
Current market practices
Weight of human expert in final rating
Expert
Bank and sovereign portfolio Models
Corporate portfolio
Weight of scoring model in final rating
SME Portfolio
Constrained
expert models
Retail portfolio
Statistical
Models 35
2. Credit risk in IRB: PD Minimum
Requirements
Use of models:
Requirements
36
2. Credit risk in IRB: PD Minimum
Requirements
Coverage
Integrity
38
2. Credit risk in IRB: PD Minimum
Requirements
Overrides
Data maintenance
- The bank must collect and historize all the data necessary for re-rating
- Corporate, Sovereign, Banks: Store rater name, rating data, Default Rates,
Migrations
- Retail: Store data that allows to classify in pools, observed DR
39
2. Credit risk in IRB: PD Minimum
Requirements
- The board must validate main aspects of the rating system and must
understand it
- There must be regular discussion about the rating system quality between
management and risk control.
- Ratings included in reporting to senior management
- The bank must have an independent risk control unit: model development,
monitoring, policies …
- Independent review at least once a year by audit or other similar
independent function
40
2. Credit risk in IRB: PD Minimum
Requirements
- Credit Approval
- Risk Management
- Economic Capital Allocation
- Corporate Governance
Overall requirements
- PD must be long term average of 1 year default rates (except for retail)
- Pooling authorized but data must be comparable
- Estimations must be grounded in experience
- Economic climate at the time of collection must be taken into account
- Usually a margin of error is included => conservatism
⇒ Often not enough data internally: conservative bias that might not be
used for internal risk management
42
2. Credit risk in IRB: PD Minimum
Requirements
Default definition
90 days delay
Unlikely to pay Calibration issue !
- Default indicators:
Non accrued status
Provisions
Selling at material loss
Restructuration with lower NPV
Filled for bankruptcy
43
2. Credit risk in IRB: PD Minimum
Requirements
Re-ageing policy
Regulators do not want the bank to play with rules:
There must be a clear policy for past due credits:
Overdrafts
If new limit, must be communicated to the client ! (no internal game)
44
2. Credit risk in IRB: PD Minimum
Requirements
Retail
45
2. Credit risk in IRB: PD Minimum
Requirements
- The bank must be able to verify that the seller/ servicer respect the contract
covenants
46
2. Credit risk in IRB: PD Minimum
Requirements
- The bank must have clear policies and pre-defined triggers to determine
when the deviation between expectations and observed values should lead
to a review of the parameters
47
2. Credit risk in IRB: PD operational
implementation
Operational development
of a rating framework
48
2. Credit risk in IRB: PD operational
implementation
Automated Traditional
Scoring Model Expert analysis
1) Regression techniques
- Multivariate Discriminant Analysis
- Ordinary Least Square
- Logistic regression
e.g. Logistic regression
1.2
default (1) / not default (0)
0.8
0.6
0.4
0.2
0
0 2 4 6 8 10 12 14 16 18
Financial ratios 51
2. Credit risk in IRB: PD operational
implementation
Current
asset Potential
value value of
Current assets in
value of the future
debts
1 year
52
2. Credit risk in IRB: PD operational
implementation
53
2. Credit risk in IRB: PD operational
implementation
Inputs
M
T O
R D
I I
A F
L Y
Wi
Results 54
2. Credit risk in IRB: PD operational
implementation
Statistical Inductive learning
Criteria Structural models
techniques techniques
- (Limited to listed
Applicability + +
companies)
Empirical validation
(out of the sample
+ + +
and out of the time
tests)
n.a. (Parameters must
not be statistically
- (No weights that
tested as they are
Comparison Statistical validation + can be statistically
derived from an
tested)
underlying financial
theory)
+ (We can see if the
weights of the + (The impact of the ++ (Structural
Economical various ratios ratios can be models are the only
validation corresponds to the estimated using derived from a
weight expected by sensitivity analysis) financial theory)
specialists)
+ (No model directly
based on neural
++ (Riskcalc of
networks to the
Moodys, Fitch IBCA
extend of our
scoring models,
Market reference knowledge, but a + KMV model
various models used
model of S&P is
by central banks of
based on Support
France, Italy, UK…)
Vector Machines that 55
is derived from NN)
2. Credit risk in IRB: PD operational
implementation
• Default data
- Most objective
- But often scarce data
- Basel 2 definition ?
• External ratings
- Especially for certain portfolios (Corp, Sovereign, Banks)
- Indirect way to modelize defaults
• Internal ratings
- When no other data available 56
2. Credit risk in IRB: PD operational
implementation
Regression techniques
• Default data
- Binary issue (default=1, not default=0)
- e.g. Binary logistic regression
• Ratings
- Multi-class issue
- e.g. Ordered logistic regression
- Rating coding: AAA=1, AA+=2, AA=3, … B-=16, CCC=17
57
2. Credit risk in IRB: PD operational
implementation
• Should integrate
- Number of type of counterparties (retail, Corp, banks…)
- Model by sector, region … => Data availability issue !
- Model use outside its development population (Generic Vs
Specific model)
• 2003 Study
- US banks: 5 non retail and 3 retail on average
- European banks
58
2. Credit risk in IRB: PD operational
implementation
Sector issue ?
Modelization steps
1. Data collection and cleaning
Regular
2. Univaried analysis
Feed-Back
3. Ratio transformation
4. Regression analysis
5. Validation
6. Calibration to Credit
60
8. Production
2. Credit risk in IRB: PD operational
implementation
61
2. Credit risk in IRB: PD operational
implementation
14 14
12 12
10 10
frequency
frequency
8 8
6 6
4 4
2 2
0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
rating rating
Which distribution ?
=> Bell shaped ?
=> Flat ?
=> keeping concentration ?
62
2. Credit risk in IRB: PD operational
implementation
- Fraction issue
1600
1400
1200
Frequency
1000
800
600
400
200
0
84 5
26 6
1, 858
2, 899
2, 940
2, 981
3, 022
3, 063
4, 105
4, 146
5, 187
8
42 4
22
73
77
81
0,
0,
0,
0,
0,
0,
1,
1,
1,
1,
1,
1,
68
10
52
94
36
78
20
62
04
1,
- Studying outliers
65
2. Credit risk in IRB: PD operational
implementation
- Analysts feed-back
66
2. Credit risk in IRB: PD operational
implementation
2. Univaried analysis
- Study of the relationship between each variable and the average PD / the
average rating
50% 14
12
40%
Rating
10
30% 8
20% 6
4
10%
2
0% 0
-20% -10% 0% 10% 20% 30% -15.0% -10.0% -5.0% 0.0% 5.0% 10.0% 15.0% 20.0% 25.0%
ROA ROA
67
2. Credit risk in IRB: PD operational
implementation
16
14
e.g. Liquidity ratio 12
on large corporate 10
Rating
8
rating dataset
6
4
2
0
0.0% 50.0% 100.0% 150.0% 200.0% 250.0%
Cash/ ST debts 68
2. Credit risk in IRB: PD operational
implementation
- Analysts feed-back
69
2. Credit risk in IRB: PD operational
implementation
3. Ratio transformation
70
2. Credit risk in IRB: PD operational
implementation
- Analysts feed-back
71
2. Credit risk in IRB: PD operational
implementation
4. Regression analysis
1
π=
1 + exp[− (b1 x1 + b2 x 2 + ... + c )]
- Divide sample in construction and backtesting sample
- Selection process:
9 Forward
9 Backward
9 Stepwise
9 Manual
- Analysts feed-back
73
2. Credit risk in IRB: PD operational
implementation
5. Validation
- Performance measures:
74
2. Credit risk in IRB: PD operational
implementation
- Main tools
% of defaults AR
CND
100%
90%
Perfect model
80%
70%
60% Naive model
50%
40%
Tested model
30%
20% 0%
10%
0%
0 1 2 3 Default rate of Score
the sample
75
2. Credit risk in IRB: PD operational
implementation
- Benchmarking performances:
76
2. Credit risk in IRB: PD operational
implementation
- Analysts feed-back
=> weight of the different variables (e.g.: size for Corp at Dexia)
77
2. Credit risk in IRB: PD operational
implementation
6. Calibration
- If rating dataset:
78
2. Credit risk in IRB: PD operational
implementation
7. Test phase
8. Production
79
2. Credit risk in IRB: PD operational
implementation
Qualitative assessment
- Scoring tool delivers a first rating, but only based on financial assessment
Flexibility Subjectivity
=> in analyst freedom ? + -
=> or scorecard approach ? - +
80
2. Credit risk in IRB: PD operational
implementation
Qualitative scorecard
Qualitative score
Impact in steps Very good Good Neutral Bad Very Bad
AAA 0 0 0 -6 -7
Financial score
AA 1 0 0 -5 -6
A 2 1 0 -4 -5
BBB 3 2 0 -3 -4
BB 4 3 0 -2 -3
B 5 4 0 -1 -2
81
2. Credit risk in IRB: PD operational
implementation
82
2. Credit risk in IRB: PD operational
implementation
Combination of both
=> Model rating
Potential second
Overruling of the analyst
rating after
=> Final rating before sov ceiling
sovereign ceiling
83
2. Credit risk in IRB: PD operational
implementation
- Shareholders
- New credit
- Exceptional year
-…
84
2. Credit risk in IRB: PD operational
implementation
Final
Data collection validation
Model IT tools
and cleaning and first
Construction definition
backtesting
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
85
2. Credit risk in IRB: PD operational
implementation
Organizational issue
- We have a model to systematize the rating process
- But there is still freedom for the analyst
- Statistical development is not complicated, but will the model be used
correctly ?
Quality control
86
2. Credit risk in IRB: PD operational
implementation
- Main roles:
87
2. Credit risk in IRB: PD operational
implementation
Credit Analysts
Quality control:
Fill ratios and Give the final functioning
qualitative rating to the No agreement
scorecards borrower on overruling:
discussion
(rating
committee)
Difference is
within accepted
Both analysts and model
degree of freedom Independent review (senior analyst)
agrees on the rating – no
(e.g. 1 step)
additional control needed
Agreement on
overruling:
Regular joint report to
back-testing of the developers
model
Model developers
88
2. Credit risk in IRB: PD operational
implementation
Organization at Dexia
Credit analysts
Validation
Model developer department
Rating system
Rating Committee
Quality Control
Group of rating
Audit
follow up
- Support from senior management => resistance to change (micro view ><
macro view)
90
Agenda
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 91
2. Credit risk in IRB: LGD Minimum
requirements
Retail
Operational measurement
of LGD
93
2. Credit risk in IRB: LGD measures
Introduction
But …
Challenge…
Implied LGD
Workout LGD
Implied LGD
Workout LGD
Workout LGD
- Economical loss => including costs and discount effects
97
2. Credit risk in IRB: LGD measures
Example
− 1000 200.000
+ 2
LGD = 1 − 1 .05 1 . 05 = 64%
500.000
98
2. Credit risk in IRB: LGD measures
1. Costs
2. Null LGD
3. Default duration
4. Discount rate
99
2. Credit risk in IRB: LGD measures
1. Costs
=> Not important at portfolio level, but might be important at facility level
(pricing …)
100
2. Credit risk in IRB: LGD measures
2. Null LGD
⇒ B preferred by regulators
⇒ C coherent but changes Basel 2 default definition
⇒A…
101
2. Credit risk in IRB: LGD measures
3. Default duration
4. Discount rate
‘Firms should use the same rate as that used for an asset of similar risk. They
should not use the risk free rate or the firm’s hurdle rate…’ (CP 189, FSA)
‘A bank must establish a discount rate that reflects the time value of money
and the opportunity cost of funds to apply to recoveries and costs. The
discount rate must be no less than the contract interest rate on new
originations of a type similar to the transaction in question, for the lowest-
quality grade in which a bank originates such transactions. Where possible,
the rate should reflect the fixed rate on newly originated exposures with term
corresponding to the average resolution period of defaulting assets’ (pub
8/4/03, Federal Reserve).
103
2. Credit risk in IRB: LGD measures
• Historical rates or implied future ones (Rf 70’s 10%, 90’s 4%)?
104
2. Credit risk in IRB: LGD measures
105
2. Credit risk in IRB: LGD measures
Main findings
50
Frequency
50
40 40
30 30
20 20
10 10
0 0
0%
0%
0%
%
%
0
10
20
30
40
50
60
70
80
90
10
20
30
40
50
60
70
80
90
10
10
LGD LGD
107
2. Credit risk in IRB: LGD measures
LGD drivers
108
2. Credit risk in IRB: LGD measures
=> Study of Altman: fixed LGD against moderate correlation with PD.
30% of difference for capital at 99.9%
Statistic Fixed LGD Volatile LGD Correlated LGD Diff fix/ correl
99% VAR 435 437 564 29.6%
99.9% VAR 809 814 1053 30.1%
109
2. Credit risk in IRB: LGD measures
Stressed LGD
‘A bank must estimate an LGD for each facility that aims to reflect economic
downturn conditions where necessary to capture the relevant risks…In
addition, a bank must take into account the potential for the LGD of the
facility to be higher than the default-weighted average during a period
when credit losses are substantially higher than average... For this
purpose, banks may use averages of loss severities observed during
periods of high credit losses, forecasts based on appropriately
conservative assumptions, or other similar methods.’
(see article 468 of ICCMCS)
110
2. Credit risk in IRB: LGD measures
111
2. Credit risk in IRB: LGD measures
Model complexity
Complex
: LGD Basel II
: (costs
LGD Basel II
add-ons)
Basic LGD Recovery (transitions)
113
2. Credit risk in IRB: LGD measures
Model complexity
Country dimension
Complex
: ...
P
: UK
L
Basic NL
B
115
2. Credit risk in IRB: LGD measures
- Most banks will start with “recovery LGD” as a prudent first guess
116
Agenda
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 117
2. Credit risk in IRB: EAD requirements
and examples
- Conservative bias
Retail
2 types of EAD
- The one linked to the behavior of the client: will he draw its line, will it use
its guarantee
- The one linked to derivatives instruments, they are function of the evolution
of risk parameters (interest rates, exchange rates,…)
119
2. Credit risk in IRB: EAD requirements
and examples
120
2. Credit risk in IRB: EAD requirements
and examples
Line
Exposure
⇒Is the CCF 100% ? Bank penalized for a good management of risky
clients ?
121
2. Credit risk in IRB: EAD requirements
and examples
EAD linked to market parameters
OTC Derivatives Credit Risk is a function of two major drivers :
• The Current Credit Risk: equal to current Mark-to-Market
• The Potential Future Credit Risk or Add-On: due to potential future evolution of
Mark-to-Market
Total Credit Risk = Current Credit Risk + Add-On
Credit Risk
Current Mark-to-Market
Maturity
The current value of a transaction: assessed by calculating the value this transaction
would have if it were immediately hedged in the financial markets.
Default
Positive Mark-to-Market
Default
Negative Mark-to-Market
Forward Buy
Nov 2002
Positive MtM: The value of the transaction is positive. The transaction is ‘in the money’.
Fortis Bank has a credit exposure on the counterparty.
Negative MtM: The value of the transaction is negative. The transaction is ‘out of the
money’. The counterparty has a credit exposure on Fortis Bank.
123
2. Credit risk in IRB: EAD requirements
and examples
Reflect the evolution of the OTC Derivatives Credit Risk in the future
Add-On
?
Maturity
Regulatory add-on
For instance, if a bank has concluded a 3 years interest rate swap with
another bank, on a notional amount of 1000 EUR whose market value is
currently 10 EUR, the credit-equivalent would be
126
2. Credit risk in IRB: EAD requirements
and examples
127
2. Credit risk in IRB: EAD requirements
and examples
0.4+0.6×NGR
with NGR being the ratio of the netted MTM value (set to zero if negative) to
the gross positive MTM values.
128
2. Credit risk in IRB: EAD requirements
and examples
Example
NGR = 30 EUR (netted MTM of 100 – 30 – 40) / 100 EUR (sum of positive
MTM)=0.3
PFE (corrected for netting) = (0.4 + 0.6 × 0.3) × 290 EUR = 168.2 EUR
9.2%
7.0%
4.9%
3.5% Maturity
One month Add-on due to collaterisation
130
1m 6m 1Y 2Y 3Y 4Y
2. Credit risk in IRB: EAD requirements
and examples
Maturity
131
Agenda
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 132
2. Credit risk in IRB: Maturity
requirements and examples
Maturity calculation
- In IRBA calculated as
∑ tCF
∑ CF
- Min 1 year, Max 5 years
- National discretion: break of the 1 year min (e.g.: Repo, Forex, Securities
Lending…)
133
Agenda
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 134
3. Understanding RWA function
135
3. Understanding RWA function
LGD
IRBA
EAD
F(x) = Capital requirements
Maturity
Correlation
Regulators
2. What is a risk ?
- Formula delivers capital to cover credit risk
138
3. Understanding RWA function
• Expected loss
• Anticipated average annual loss rate
• Foreseeable cost of doing business
• Differentiated cost of risk recovered through pricing
• Unexpected loss
• Unforeseeable
• Unanticipated Losses
• Requires balance sheet cushion of capital
• Differentiated capital sustained with appropriate return
139
3. Understanding RWA function
3. Correlation issue
4 .5 0 %
4 .0 0 %
3 .5 0 %
Year 1
Year 2
3 .0 0 % Year 3
default rate
Year 4
2 .5 0 %
Year 5
2 .0 0 % Year 6
Year 7
1 .5 0 % Year 8
Year 9
1 .0 0 %
Year 10
0 .5 0 %
0 .0 0 %
100 500 1 ,0 0 0 1 0 ,0 0 0 5 0 ,0 0 0
140
3. Understanding RWA function
2.5%
Default rate
2.0%
1.5% Average
1.0%
0.5%
0.0%
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
Year
141
3. Understanding RWA function
12%
10%
8%
6%
4%
2%
0% 100.0
120.0
140.0
160.0
180.0
200.0
20.0
40.0
60.0
80.0
0.0
Systemic Idiosyncratic
Correlated risk risk
143
3. Understanding RWA function
Probability of Confidence
default interval
With some ⎛ φ −1 ( PD ) + ρ φ −1 (CI ) ⎞
statistical φ⎜ ⎟
developments … ⎜ − ρ ⎟
⎝ 1 ⎠
Asset
(Inverse)Normal correlation
distribution
144
3. Understanding RWA function
200.00 100.00
180.00 90.00
160.00 80.00
Frequency
Frequency
140.00 70.00
120.00 60.00
100.00 50.00
80.00 40.00
60.00 30.00
40.00 20.00
20.00 10.00
0.00 0.00
0.00%
0.03%
0.05%
0.08%
0.11%
0.14%
0.17%
0.21%
0.26%
0.31%
0.36%
0.42%
0.50%
0.58%
0.68%
0.79%
0.93%
1.10%
1.32%
1.60%
1.99%
2.58%
3.69%
7.81%
0.15%
0.33%
0.40%
0.46%
0.51%
0.56%
0.61%
0.66%
0.70%
0.75%
0.80%
0.85%
0.90%
0.96%
1.02%
1.08%
1.15%
1.23%
1.32%
1.43%
1.57%
1.76%
2.06%
2.93%
Default rate Default rate
Corp: PD 1%, Correl 20% Qual. Rev. Expo: PD 1%, Correl 4%
=> Basel 2 capital= 14.6% => Basel 2 capital= 4.1%
145
3. Understanding RWA function
- The different risk classes are associated to different risk weighting function,
the only difference is correlation
Retail
SME
UL
UL Large Corporates
UL
146
3. Understanding RWA function
⎛ φ −1 ( PD ) + ρ φ −1 (CI ) ⎞
φ⎜ ⎟ - Signification: not enough capital =>
⎜ 1− ρ ⎟
⎝ ⎠ Accepted Bankruptcy
- Calibration: Desired PD ?
147
3. Understanding RWA function
Risk= Unexpected
⎡ ⎛ φ −1 ( PD ) + ρ φ −1 ( 0 .999 ) ⎞ ⎤
⎜
⎢φ ⎜ ⎟ − PD ⎥ xLGD
⎢⎣ ⎝ 1− ρ ⎟ ⎥⎦
⎠
149
3. Understanding RWA function
• Likelihood that losses will exceed the sum of Expected Loss (EL) and Unexpected
Loss (UL)
• 100% minus this likelihood = confidence level
• Capital set according to the gap between EL and VaR, and EL is covered by
provisions or revenues, likelihood that the bank will remain solvent over a one-
year horizon = confidence level.
Under Basel II, capital is set to maintain a supervisory fixed confidence level
150
3. Understanding RWA function
We can work on 1 year horizon because we suppose that we can sell all the
bank assets …
Assets Liabilities
100 EUR credit BBB 5 EUR capital
(e.g. 20Bp EL Enough ?
5% at 99.9%
95 EUR debt
5 years)
151
3. Understanding RWA function
Assets Liabilities
Stressed year ! 100 EUR credit 5 EUR capital
BBB (20 Bp EL
5% at 99.9%
95 EUR debt
- 5 EUR loss because of default => covered by capital 5 years)
- 95 EUR of credit still at risk and no more capital => we sell them
- But … New rating % of
year end portfolio Spread New value
AAA 0% 0.05% 0.0
AA 5% 0.10% 4.8
A 10% 0.20% 9.5
BBB BBB 50% 0.30% 47.5
Spread BB 20% 1.00% 18.5
0.30% B 10% 2.00% 8.9
Default 5%
Intitial Value Final Value
95 89
Delta
-6 152
3. Understanding RWA function
25.0%
Regulatory capital
20.0%
0.1% PD
0.5% PD
15.0%
1.0% PD
1.5% PD
10.0%
5.0%
0.0%
1 2 3 4 5
Maturity 153
3. Understanding RWA function
- The formula:
⎡ ⎛ φ −1 ( PD) + ρ φ −1 (0.999) ⎞ ⎤ 1
⎢φ ⎜⎜ ⎟ − PD ⎥ xLGDx
⎟
x(1 + ( M − 2.5) xb)
⎢⎣ ⎝ 1− ρ ⎠ ⎥⎦ (1 − 1 .5 xb )
154
3. Understanding RWA function
⎡ ⎛ φ −1 ( PD) + ρ φ −1 (0.999) ⎞ ⎤ 1
⎢φ ⎜⎜ ⎟ − PD ⎥ xLGDx
⎟
x(1 + ( M − 2.5) xb) x1.06
⎢⎣ ⎝ 1− ρ ⎠ ⎥⎦ (1 − 1.5 xb)
May be
reviewed based
With b = (0.11852 − 0.05478 x ln( PD))² on QIS 5 results
155
3. Understanding RWA function
SUMMARY
Asset
Proba correl
(Merton) CI for BBB+ /
Default
A- rating
⎡ ⎛ φ −1 ( PD) + ρ φ −1 (0.999) ⎞ ⎤ 1
⎢φ ⎜⎜ ⎟ − PD ⎥ xLGDx
⎟
x(1 + ( M − 2.5) xb) x1.06
⎢⎣ ⎝ 1− ρ ⎠ ⎥⎦ (1 − 1.5 xb)
156
3. Understanding RWA function
• Fixed LGD
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 158
3. Detailed issues of Risk quantification
1. Calibration of RW functions
2. Correlation level
3. Example of RWA
4. Specialized lending
5. Equity exposures
6. Purchased receivables
159
3. Detailed issues of Risk quantification
1. Calibration of RW functions
0
Standardised
Standardised
Current
Accord
Approach
Foundation
Current
Accord
Approach
IRB
IRB
Source: CBF
160
3. Detailed issues of Risk quantification
Standardised
Further
Foundation
Standardised
Foundation
Current
Accord
Approach
review of the
Accord
Current
Approach
IRB
IRB
IRB
calibration
Source: CBF
161
3. Detailed issues of Risk quantification
SME
UL
Corp, Sovereign, 12%-24% UL
UL
Large Corporates
162
3. Detailed issues of Risk quantification
18.0%
26%
16.0%
24%
14.0%
22%
12.0%
Correlation
Asset correl
20%
10.0%
18%
8.0%
16%
6.0%
14% 4.0%
12% 2.0%
10% 0.0%
0.00% 1.00% 2.00% 3.00% 4.00% 5.00% 6.00% 7.00% 0.00% 2.00% 4.00% 6.00% 8.00% 10.00% 12.00% 14.00% 16.00%
PD PD
163
3. Detailed issues of Risk quantification
SME
25.0%
20.0%
Correl 15.0%
10.0%
5.0% 45
0.01%
0.80%
25 Turnover Mios
1.70%
2.60%
3.50%
4.40%
5.30%
EUR
6.20%
5
7.10%
8.00%
8.90%
PD 164
3. Detailed issues of Risk quantification
Business Line
Shared
Fixed Costs Cost BL
?
Costs
Marginal Costs
Product
Capital
Crédits
(Unexpected Loss)
Provision YES
(Expected Loss) Transfer price
Options:
ALM
• Reservation
• Max. Price
• Early Term.
NO Market price
Funding
165
3. Detailed issues of Risk quantification
Corporate Retail
20% €1.60
Basel II 50% €4.00
Standardised €100 X X 8% = €100 X 75% X 8% = €6.00
100% €8.00
Approach
150% €12.00
€5m €50m
Basel II *11.3% - 14.4% €0.90 to €1.16 4.45% €0.36
IRB Approach X
€100 X 8% = €100 X X 8% =
(PD 0.03% to 20%;
LGD 45%)
X
188% to 238% €15.07 to €19.06 100.3% €8.02
*Shows Firm –size adjustment effect
From Basel II Annex 3
for Sales of €5m and €50m
166
3. Detailed issues of Risk quantification
3. Example of RWA
Original calibration !
PD (%) BRW PD (%) BRW
0.03 14 1 125
0.05 19 2 192
0.1 29 3 246
0.2 45 5 331
0.4 70 10 482
0.5 81 15 588
0.7 100 20 625
167
3. Detailed issues of Risk quantification
300,00%
150,00%
Risk Weight Standard
ECAI RATED
100,00%
Risk Weight Standard
ECAI UNRATED
50,00%
0,00%
%
0%
00
05
14
33
75
30
00
00
00
,0
0,
0,
0,
0,
0,
1,
2,
3,
5,
10
168
3. Detailed issues of Risk quantification
THE 4 CURVES
Maturity : 2,5y Sovereign Retail Other Retail
LGD : 45% Corporate Residential Retail Qualifiying
PD Bank Mortgage Revolving
0.03% 14.44% 4.15% 45.00% 0.98% RISK WEIGHTED ASSET
0.05% 19.65% 6.23% 6.63% 1.51%
0.10% 29.65% 10.69% 11.16% 2.71%
DERIVATION for
0.25% 49.47% 21.30% 21.15% 5.76% Corporate, Sovereign,
0.40% 62.72% 29.94% 28.42% 8.41%
0.50% 69.71% 35.08% 32.36% 10.04%
Bank & Retail Exposures
0.75% 82.78% 46.46% 40.10% 13.80%
1.00% 92.32% 56.40% 45.77% 17.22% June 2004
1.30% 100.95% 67.00% 50.80% 21.02% Without 1.06 conservative
1.50% 105.59% 73.45% 53.37% 23.40%
factor
2.00% 114.86% 97.94% 57.99% 28.92%
2.50% 122.16% 100.64% 60.90% 33.98%
3.00% 128.44% 111.99% 62.79% 38.66%
4.00% 139.58% 131.63% 65.01% 47.16%
5.00% 149.86% 148.22% 66.42% 54.75%
6.00% 159.61% 162.52% 67.43% 61.61%
10.00% 193.09% 204.41% 75.54% 83.89%
15.00% 221.54% 235.72% 88.60% 103.89%
20.00% 238.23% 253.12% 100.28% 117.99%
169
3. Detailed issues of Risk quantification
4. Specialized lending
170
3. Detailed issues of Risk quantification
5. Equity exposures
Approaches:
• a methodology based on market risk and stress testing (target: equities held
mainly for capital gains purposes)
• Simple risk weight
• Internal models (VAR)
171
3. Detailed issues of Risk quantification
172
3. Detailed issues of Risk quantification
6. Purchased receivables
Corporate:
• If such receivables are purchased, banks are expected to estimate the risk
of their default (“bottom-up” approach).
• The alternative "top-down“ approach is possible as long as:
• receivables are purchased from third parties,
• must not be subject to netting agreements,
• if not secured by collateral, their maturity must not exceed 1 year,
• concentration limits must be kept.
• Within the IRB approach bank is expected to calculated not only the capital
requirement for default risk (like for any other exposures) but also capital
requirement for so called „dilution risk“
• For the purpose of calculating the RW for such risk, into the corporate risk
function is set LGD = 100 % and the portion of expected loss from the exposure
will be equal to PD.
• When the receivable is secured by a guarantee, the procedure is the same as for
IRB approach for corporate exposures, regardless whether the guarantee covers
the risk of default, dilution risk or both.
175
Agenda
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 176
4. Credit risk mitigation
1. General issues
3. Funded
4. Unfunded
a. Eligible instruments
b. Calculation
177
4. Credit risk mitigation
1. General issues
• Institutions using both the standardized approach, or foundation IRB, may take
into consideration the effect of Credit Risk Mitigation (CRM) during RW
calculation.
• All possible credit risk mitigation methods must be legally enforceable in all
relevant countries.
• In addition, institutions must ensure the efficiency of the entire process and
address risks connected therewith.
• Institutions must always ensure compliance with minimum requirements for
CRM procedures.
• If all defined requirements are met, the institution may decrease the value of
RW or EL in compliance with CAD.
• No exposure with a CRM instrument may have a RW higher than an identical,
but unprotected exposure.
• In cases when CRM has been incorporated in the calculation of RWA, the
provision on CRM is not further applied. 178
4. Credit risk mitigation
Eligibility
• Low correlation between the creditworthiness of the debtor and the value of the
collateral.
• In addition, under the simple approach with respect to financial collateral it is required
that residual maturity of the collateral is not shorter than the residual maturity of the
exposure.
181
4. Credit risk mitigation
• In order to mitigate the capital requirement, banks may use the simple approach
replacing the risk weight of the counterparty by the risk weight of the collateral.
However simple approach is NOT allowed within IRB.
Comprehensive approach
184
4. Credit risk mitigation
Haircuts adjustment
- Those haircuts are for a 10 day holding period and should be adjusted for other
effective holding period with
185
4. Credit risk mitigation
• To transform the supervisory haircuts for the ten days holding period to haircuts
adapted for the transaction-holding period, banks have to use the square root of
time formula.
186
4. Credit risk mitigation
Example :a bank has a three years BBB bond as collateral to cover a three years
secured lending operation. The bond is marked to market every week. The bond
issuer is a corporate and the face value is 100 EUR. The haircut is calculated as
follow:
• The supervisory haircut for a three years BBB bond issued by a corporate is 6%
• The minimum holding period for secured lending is 20 business days
• As the bond is not revaluated daily but weekly (every five business days), the
minimum holding period must be adapted to 24 (as there are five days instead of
one between revaluations)
• Then the supervisory haircut that is based on a ten days holding period is scaled
up using the square root of time formula
24
Haircut = 6 .0 % x = 9 .3 %
10
• The value of the bond is then 100 EUR x (1-9.3%)= 90.7 EUR. 187
4. Credit risk mitigation
- Internal VAR models may be used to estimate haircuts but subject to following
requirements
188
4. Credit risk mitigation
At national discretion, some collateral can receive zero haircuts when used in
REPO-style transactions
189
4. Credit risk mitigation
Netting agreements
Hs = haircut appropriate to Es
Efx = absolute value of the net position in a currency different from the settlement
currency
Financial receivables
• Claims <1y, repayment will occur through the commercial or financial flows
related to the underlying assets of the borrower.
• Legal certainty: the bank can take control of it
• Risk management requirements (collection cost, concentration risk, …)
191
4. Credit risk mitigation
Other collateral
• Liquid markets
• Public prices
Banks requirements:
• First claim
• Loan agreements should describe valuation procedure
• For inventories: physical inspection
• Collateral accepted, valuation methods should be described in banks
procedures
192
4. Credit risk mitigation
IRBF
193
4. Credit risk mitigation
⇒The LGD on the 100 EUR exposure would then be 45% (assuming senior
corporate exposure) on 71.4 EUR and 35 % on 28.6 EUR.
194
4. Credit risk mitigation
IRBA
• the rules are less strict as any kind of collateral can be recognized
• as long as the bank has historical data to support its valuation (at least 7
years of data on average recovery value on the various types of collaterals it
plans to use).
195
4. Credit risk mitigation
Maturity mismatch
Maturity of exposure is defined as the longest possible residual period until the
debtor meets its obligations (maximum 5 years).
Maturity of protection is defined as the shortest possible delay when the credit
protection can be cancelled.
Calculation:
t − 0,25
• In general, the value of CRM instrument is multiplied by: ,
T − 0,25
• where:
• t … is maturity of protection (in years)
• T … is maturity of exposure (in years)
196
4. Credit risk mitigation
IRBA:
• No limit
197
4. Credit risk mitigation
198
4. Credit risk mitigation
Requirements: Guarantees
- Credit events should cover: default, bankruptcy, restructuration with lower NPV
(or partial recognition)
- The one who determines default cannot be protection seller alone
- If a reference asset is used: should be at least pari passu, same issuer
199
4. Credit risk mitigation
4. Unfunded b) Calculation
G* = G × (1 − HFX )
200
4. Credit risk mitigation
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 202
5. Securitization
203
5. Securitization
Traditional securitisation
• A structure when the cash flow from the underlying pool of exposures relates at least to
two different exposures or tranches with different degrees of credit risk.
• The payment to the investor depends on the development of the underlying exposure.
• Categorized structure (in tranches) characterizing securitisation differs from regular debt
instruments in that the “junior” tranches may absorb losses without affecting more senior
tranches.
Synthetic securitisations
• A structure of at least two categorized (stratified) risk positions or tranches taking into
consideration a different degree of credit risk while the credit risk of the underlying pool
is transferred through guarantees or credit derivatives (funded credit linked notes or
unfunded credit default swap).
• The risk of investor again depends on the development of the underlying exposure.
204
5. Securitization
Investor:
• institution which is not the originator, sponsor or provider of services and
carries the economic risk of the securitisation exposure.
205
5. Securitization
Capital requirements:
- Banks that invest in exposures that they originate themselves that receive an
external rating below BBB- must deduct them form their capital base
- For off-balance sheet exposures, CCF are used (if they are externally rated, the
CCF is 100%).
206
5. Securitization
IRB Approach
The Rating Based Approach (RBA) that must be applied when the securitised
tranche has external or internal ratings.
The Supervisory Formula (SF) is used when there are no available ratings.
The Internal Assessment Approach (IAA) can also be used when there are
no available ratings but only for exposures extended to ABCP programmes.
207
5. Securitization
208
5. Securitization
Not senior
Senior
RW Rating tranches and N<6
tranche, N≥6
N≥6
AAA 7% 12% 20%
AA 8% 15% 25%
A+ 10% 18%
A 12% 20% 35%
A- 20% 35%
BBB+ 35% 50%
Long Term
ratings BBB 60% 75%
BBB- 100%
BB+ 250%
BB 425%
BB- 650%
Unrated and <
Deduction
BB-
A1/P-1 7% 12% 20%
A2/P-2 12% 20% 35%
Short Term
A3/P-3 60% 75% 75%
Ratings
Other and
Deduction
unrated
209
5. Securitization
• Only applies to ABCP programmes. Banks can use their internal ratings
if they meet some operational requirements, mainly:
- The ABCP must be externally rated (the underlying, not the securitised
tranche).
-is used when there is no external rating, no inferred internal rating and no
internal rating given to an ABCP programme.
212
5. Securitization
- Until the sum of the subordinated tranches and tranche for which the
capital is calculated is inferior to the regulatory capital had the exposures
not been securitised, the capital rate is 100%.
- Then it decreases sharply until the marginal capital rate becomes close
to zero, as illustrated on the following graphs (in this example the capital
had the exposures not been securitised would be 8.14 EUR, and the credit
enhancement equals 5 EUR)
Senior Tranche
Securitised (100 – 3 – X) EUR
assets: 100 EUR
Bank investment
Bank buys
X EUR
K irb= 8.14 EUR
First loss 5 EUR
213
5. Securitization
4
Capital
S iz e o f th e tr a n c h e 3 .1 4
3
C a p ita l 3 .1 4
0
0 5 10 15 20 25 30 35 40 45
1 2 0 %
S iz e o f th e tr a n c h e
1 0 0 % S iz e o f th e tr a n c h e 3 .1 4
C a p ita l r a te 1 0 0 %
Capital / tranche size
8 0 %
6 0 %
4 0 %
2 0 %
0 %
0 5 1 0 1 5 2 0 2 5 3 0 3 5 4 0 4 5
T r a n c h e s iz e 214
5. Securitization
1200%
1000%
800%
600%
400%
200%
0%
AAA AA A+ A A- BBB+ BBB BBB- BB+ BB BB- B+ B B-
1. Introduction to IRB
2. Risk parameters in IRB
• PD: Minimum requirements and operational implementation
• LGD: Minimum requirements and operational implementation
• EAD: Minimum requirements and examples
• Maturity: Minimum requirements and examples
3. Risk quantification
• Understanding the Risk weighting function
• Detailed issues of Risk quantification
4. Credit Risk Mitigation
5. Securitization
6. Application for supervisory approval 216
6. Application for supervisory approval
Nov: QIS2,5
1960 1970 1980 1990 2000 2001 2002 2003 2004 2005 2006 2007
June/1999: CP1
1996: Market risk
July: QIS3 results
published
May: Release of CP3
Home-host issue
Consolidated solvency reporting
Consolidated
Consolidating supervisor Supervisory review Bank
Framework
• Overall responsibilities (going concern & emergency situations)
• Assessment of compliance with the requirements for FIRBA, AIRBA
and AMA
• Coordination of gathering & dissemination of information
• Planning and coordination of supervisory activities, including Pillar 2
• Other specific provisions
• Permission to go FIRBA, AIRBA and AMA
• Reach a joint decision within six months
• In the absence of joint decision, make its own decision
• The decision of the consolidating supervisor will be recognised and
applied by the local supervisors
219
6. Application for supervisory approval
Calibration
Monitor
Support
Oversight and
Data management
control mechanisms
system
Quantification
process
220
6. Application for supervisory approval
Rating system
(PD and LGD)
Quantification
process
- Conservative bias ?
222
6. Application for supervisory approval
Oversight and
control mechanisms
- Follow up of overrulings ?
223
6. Application for supervisory approval
Rating Committee
Quality Control
Validation Committee
Model manager Credit analysts
Validation Department
Rating tool
224
6. Application for supervisory approval
Data management
system
225
6. Application for supervisory approval
Duration 3 months
Regulators 4 2 Quantitative Specialists
2 Process Specialists
Review documentation, interviews ( credit management,
analytics, chief risk officer, business)
Report CEO, CCO, Chief Auditor, External Auditors
Point of attention
Outliers detection
Coherency of default definition
Documentation of methodological choices (test)
Central Documentation Repository
Overruling process
227
Agenda
Conclusions
228
7. Conclusion
• A permanently strong Credit Risk framework is the aim and key driver,
Basel 2 gives the needed leverage
• A lot of work has been done but still a lot needs to be done : we know we
can rely on your dedication and professionalism to succeed
229
7. Conclusion
231
Consortium Partners
232