impairment parameters look well below the surface joaquin gutierrez the world bank finsac - vienne...

59
Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Upload: hubert-todd

Post on 29-Dec-2015

222 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Impairment ParametersLook well below the surface

Joaquin GutierrezThe World Bank

FinSac - Vienne -- October 22, 2014

Page 2: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Approaches for provisioning• Risks from derogating national accounting • (a few) Key issues in implementing IAS 39 • Take-away lessons• Appendixes (further reading)

Outline of Session

Note: This presentation provides a short summary of issues relevant for supervising under IAS 39. The presentation does not expand in discussing the basic standards proposed by IAS 39 and IFRS 9 neither provides a comparison of IAS 39 versus Basel II compliance differences.

2

Page 3: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Traditional rules-based matrix (our national GAAP)∙ Simple, easy, consistent, less room for ‘optimization”.

• Principle-based: Current IAS 39 – Next IFRS 9∙ Divergent practices, Basel II effects ∙ More discretion/room for method/model “optimization”

• Must discover issues regarding the “key estimates”∙ IT/Data: length, scope, adjustments to historic estimates ∙ Future cash flows (FCFs) - assumption for projecting ∙ Modelling key parameters for collective provisioning∙ Depth of empirical loss evidence (loss crystallization)

• Mid-Path: centralized incurred loss model

Approaches (pros, cons, trade-offs)

3

Page 4: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Weak reporting by EU banks (boiler plate wording disclosures).

• Challenges in clarity and universality.• Limited comparability of asset quality from obscure

definitions. • Lack of clear accounting definition of impaired loans

in IFRS. • Overlapping indicators of credit quality.• Refinancing obscures harmonized delinquency

triggers (90+)• Lack of comparable qualitative classification triggers• “Grey” forbearance practices – EBA raises the bar

(by Dec. 2014) ∙ Interpretation difficulties - Comparability problems ∙ Vague qualitative classification triggers∙ Little disclosure on model risk inputs (PD, LGD,

Cures, etc.).

Substantial divergence in EU practiceFitchRatings Special Report (13 May 2014)

4

Page 5: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Class/Criteria Payment

Experience

Industry Trend Enterprise Position

Financial Condition

Management Governan

ceViability Outlook

 1-3 -- PASSGeneric

Provision = 1.5% -5 %

Factors (a) concentration, (b) underwriting, (c)

credit risk management; (d)

quality of information; (e)

economic outlook.

PunctualHigh Account Turnover

AcceptableAdequate demandAdequate profitsLiberalized industryMinimal competition

Above sector averageStrong competitive positionGood products and market

Profitable (ER)Liquid (LQ)Sufficient cash-Flow (CF)Low leverage (LV)Two repayment sourcesWorking Capital (WC) loans clearly supported

Capable/qualifiedNo doubt of integrityClear strategic visionVery professionalGood control/MISGood External Audit

No significant Risks

 

4- SPECIAL MENTION

(potential problems)

Provision = 5% - 15%

Delays < 90 DaysOccasional overdraftsHigh average balancesMedium turnoverMinor contract breachNew Loans supported 

Some questionsIncome decreasingCompetition increasingPrice competition upOperating costs up 

Within sector’s averageSome competitive Weakness

ProfitableAcceptable liquidityModerate leverageTwo repayment sourcesCF does not cover all operating costs and replacement of assets

Capable/qualifiedNo doubt of integritySome strategic problemsImproving control/MISCommitted owners and managersAcceptable Audit

Will survive problemsHas strength to copeOwners can supportNew capital available if neededNo major labor issues

 (5) SUBSTANDAR

D Provision =

15-50%Interest

suspension + reversal 

Past due > 90 daysRecurrent overdraftsLow account turnoverContract breach > 90dRenewals conceal financial problemsNo seasonal clean-upsWeak Documentation

VolatileWeak Co. under pressureIncome downDemand downLiberalization riskRaw materials riskDevaluation riskAdministered prices

Under sector’s averageDefined competition problemsTechnological weaknesses

Income low to zeroLow liquidityHigh leverageOne repayment sourceWeak cash flowCF < Principal + InterestIncrease in WC masks problems

Weak, low capacityLow experienceIntegrity in questionNo strategic visionWeak controls/MISGovernance conflictsWeak External Audit

Reliant on financingOwner support = ??Requires new marketingLatent future risksMinor labor excessesBasic problem = financialProduct and markets can recover

 

(6) DOUBTFUL Provision = 50% - 75%

Interest suspension and

reversal 

Past due > 180 daysPermanent OverdraftsContract breach >180dRenewals capitalize interestPoor legal documentation (loan or claim to collateral)

PoorEarnings = or < zeroAcute price competitionHigh risk of liberalizationPrices downOperation restructuring requiredPolitical prices

Well < sector averageSerious competition problemAcute technology problemUrgent need to modernizeLosing marketsProduct problemsOver-extended

Operational lossesIlliquidSelling assets to surviveCF < interest serviceExcessive leverageInadequate payment sourceIncreased WC hides operational losses

Poor -- against the wallIncompetent -- hidingUn-cooperative, hostileDoubts on integrityLack of control/MISOwnership problemNo source of new capitalPoor External Audit

Operational problemsMajor labor excessRequires debt reliefDeep product restructuringDeep process restructuringNo full cost recovery

 

7 - LOSTProvision = 75% - 100%

Interest suspension and reversal

Past due > 360 daysNew loans to finance operational lossesClearly lack of evidence of loan or ability to liquidate collateral at predictable value

DyingStructural weaknessesAnachronisticLiberalization = extinction

Lower quartileCan’t competeObsolete technologyWeak productCountry riskMarginal role

High lossesSelling assets at lossesAcute CF & LV problemsCF < production costsNo identifiable repayment sources (except liquidation)

Very poorCan’t be trustedIncompetent and desperateChance of fraudNon-existent governance

Extremely questionableShould be liquidatedShould be fragmentedBase value on liquidationMinimal buyers

The rules based

(matrix

)

IBRD used since

late 80s

5

Page 6: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• EU members and accessing (Art.24 EU 575/2013)∙ Rely on IFRS by “derogation” of national accounting

• Consequences from derogation (new risks):∙ Lose the driver’s seat (now

pulled-by-bankers/auditors)∙ Mitigate discretion (conservative, cautious, justifiable)∙ Deploy capacity and expertise to assess estimates:

Stocktaking of local practice (survey – visits – compilation) Toolkit to benchmark banks’ assumptions and model inputs Repository of loss experience (require empirical evidence)

• Must offset potential surplus (Pillar 2 and overlays)

• Need to specify prudential standards/expectations

Repeal of the rule based approach

6

Page 7: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

1. Precision of triggers for objective evidence of impairment (OIE) 2. Forbearance (treatment of refinancing, re-aging, restructuring,

other)3. Cur-off point: individually significant vs insignificant (collective)4. Estimation of cash flows (debt service, collateral, guarantees)5. Segmentation into pools of similar risk features (diverse attributes)6. Parameters for estimating collective provisions (PD, LGL, cures…)7. Modelling approaches for loan provisioning (≠ credit risk capital)8. Validation/backtesting of risk inputs (model results vs actual losses)9. Collateral eligibility, tests of effectiveness, valuation, discount10. Length of the emerging period (EP) for the IBNR provision11.Interest recognition through the unwind discount (mortal

sin) 12. Disclosure (see previous slide on the Fitch IBCA report)

Areas of interest (IAS 39 issues)Beyond the incurred loss model principles

7

Page 8: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Bankers’ strategies to “model” reported earnings• Extend and pretend (roll-over, renew, refinance)

∙ EBA’s ITS implementing Art. 99(4) of EU 575/2013

• Modify terms without work-out (debt overhang)• Hold and hope (delay foreclosing the collateral)• Dual FCF: service + collateral, but no re-default

∙ Projection of FCFs: (conservative, cautious, justifiable?)∙ Ability and willingness to foreclose (test and trust)

• Optimistic optimization of assumptions and inputs • Lack of data/IT-MIS (obligor/transaction attributes)

The devil is in many of the details

8

Page 9: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Three different (related) processes

ODR and PDs• ¶ 424-433 - Rating operations

16.0% • ¶ 446 - Bank to estimate a PD for each internal borrower grade15.0% • ¶ 447 - PD a long-run average of 1yr defaulr rate for that grade14.0% • ¶ 461-467 Requirements specific to PD Adverse scenario13.0% • Restructuring ¶ 453.4 and Reaging ¶ 45812.0% • Process to tag and track cures, re-aging, refinancing, restructuring.11.0% • In most cases: there is not actual empirical loss experience (no foreclosures)10.0% • Basel II ¶ 468 uses downturn/discounted/fully costed9.0% • IAS 39 uses Point-in-time LGD8.0% Base scenario7.0%6.0%5.0%4.0%3.0%2.0% • Process of transformation of ODR into PIT-PDs

1.0%

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

PD for Loan Loss Forecast

Same parameters but different dimension: (1) Point-in-time; (2) Forecasted; (3) Trough-the-Cycle/Downturn

(1) Loan Provisioning (AQR) - (2) Loss Forecast (ST) - (3) Capital Requirements

Observed Default Rates (ODR) affected by cures/reaging

Long term average PD for RCAP

Level of s

TodayPIT

Tomorrow

May be

PD/Rating model: Point-in-time vs through-the-cycle 9

Page 10: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Many moving parts (key parameters)Two different standards (Basel II and IAS 39)

Elements Methodologies Key Parameters Assumptions• OEI Event triggers • Rating Systems ¶ 394-437 • Breach of contract (days-past-due) • Historical vs. PIT market conditions• Historical loss experience • Score models • Refinancing/reaging ¶ 452 • Emerging period• PD adapted to PIT cycle conditions • Roll-Rate models • Cure rates from work-out • Observation length• LGD adapted to PIT cycle conditions • Transition matrix • Trouble loan restructuring • Re-default rates• Exposures @ Default • Markov Chain Models • Client/business assumptions • Future Cash Flows (single/dual)• Collaterals (elligibility and tests) • Vintage Models • Significant difficulties • Collateral valuation (yields & rents)• Foreclosure/liquidation elements • Logistic models • Significant vs. Non-significat cut-off • Time-to-liquidation/Fire-sale

• Macro-economic scenarios (factors) • Econometric models • Leverage/debt-service/employment • PD conditional to macro-scenario• Categorical variables • Expected loss models • LTV and peak-to-trough price ∆ • LGD conditional to macro-scenario• Pre-provision profits • Financial projections • Income revenue from NPLs+ • Payment principal only

• Long Run average PD ¶ 447 • Logistic Regression Models • Factors driving PD risk drivers • State-of-economy TTC adjustments• Downturn LGD ¶ 468 • Transition Matrix • Factors driving LGD risk drivers • One year time horizon ¶ 285

• Credit Conversion Factors ¶ • Left blank intentionally • Left blank intentionally • Downturn conditions ¶ 468

PD LGD EAD Cure Rates

Basel II • Long run average (not Trough-Cycle) • Fixed (45/75%) or downturn • Drawn + CCF (undrawn) • ¶ 458 reaging definitionIFRS • Point-in-time (PIT) • Point-in-time (PIT) • Outstanding balances • Not mentioned

Basel II • 100% (¶ 452 default definition) • Downturn, discounted, full-costed • Outstanding balances • Not mentionedIFRS • 100% only if/when impaired • PIT • Outstanding balances • Not mentioned

NPLs vs Defaults vs Impaired

Inconsistent Standards Diversity of Concenpts - Parameters - Approaches

Prov

ision

sRe

gulatory

Ca

pital

Loan

Loss

Fore

cast

Performing loans

10

Page 11: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

The loss cliff effect (keep it in mind)Oliver Wyman – Sep.2012 – Spain AQR and Stress

IFRSProvisionToday PIT

LossForecast

Next 3yrs

Level of stress 3.0x = PD2012/PD2011

Pro

visi

ons

vs L

oss

fore

cast

11

Page 12: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

AQR: Elements (Ireland & Spain based)

Typical AQR Workstream and Process

Follow-up Plan

}AQR Adjusted

CET1 % as Input for

Stress Test

~ Quality Assurance

and Project Management

u Processes,

Policies and Accounting

Rewiew| Level 3 Fair Value Exposures Review

y Collateral and Real Estate Valuation

z Extrapolation of Findings from the Credit File Review

{ Collective Provision Analysis

vLoan Tape w Sampling x Credit

File Review

|•u Revaluation

of non-derivative

Level 3 securities

|•v Review of Trading

Book Processes

|•w Review of Derivative

Pricing Models

Stress test exercise

Adjusted Balance

Sheet and CET1%

RemediationPlan

• Pro-forma financial statements and financial projection (base/stress)• Capacity of business to reach economic break-even• Level of recurrent pre-provision profits with suspension

12

Page 13: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Typical bank methodology per segmentMultiplicity of compliant approaches (Basel II-IAS 39)

Quantitative/Objective Qualitative/Subjective

Rating

Consumer/CardsMortgages

SME

CREs

Local authorities

Multinationals

Country risk

Corporates

Scoring

• Ordinal stage: classification by means of discriminant techniques• Cardinal stage: calibration of a default probability (based on ODR)• Adjustment of PD to cycle-neutral (state-of-economy)• Validate and back-test all the above

Banks use a suite of methods and modeling techniques (see appendix).

Simple banks better left to use the old matrix.

12 parameters x 7 classes x 3 orbits ≈ 252 moving parts

13

Page 14: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• IAS 39.59: significant financial difficulty of obligor

• IFRS 9 (5.5.3): significant increase of credit risk

• CRR 575 A178.1: unlikely to pay w/o recourse• Basel II (¶ 452 & ¶ 453): unlikely to pay

(+90pd)• Must increase the precision of above

principles.• The Risk Grade Concept: adopt a “Migration”

code∙ The concept of credit migration is valid under IAS∙ ∆ in the risk of default since recognition (IFRS 9)∙ Bring full focus onto credit rating systems (Basel 2)

Key issue: Precision on impairment triggersHow difficult must be the borrowers’ difficulties?

14

Page 15: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• +90 dpd without re-aging / refinancing• -90 dpd re-aged/refinanced without full

interest paid• Restructured (modified due to borrowers’

problems) • Decrease in rating to ‘default’ or by ‘two+’

grades• Breach of contract (worsened debt service

prospect) ∙ drop of CF < debt service + margin for

reinvestment ∙ Increase in leverage

• Decrease in collateral value (∆ LTV)• Legal changes (bankruptcy and related) ….

Triggers of loss event

15

Page 16: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

“ Loans in risk grades below the highest quality grade qualify as having experienced an event that results in a reduction in estimated future cash flows. The concept is applied to loans for the collective impairment assessment. The theory is based on the assumption that an "event" occurred that led to a loan being down graded from the highest quality grade (e.g., investment ≥ Baa3) to a lesser grade.”

• Renewed focus on the internal credit rating systems:∙ Diversity of attributes in the MIS to tag and track conditions∙ Meaning of the explanatory variables used in the ordinal model ∙ A decrease in the rating to default line or by two or more grades

Adopt a strong Risk Grade Concept: Applicable under IAS 39 and IFRS 9

16

Page 17: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• What is significant (cut-off point > ½ of 1%)?• Linkages with Basel II implementation• Supervisory needs and practice• Application to normal situations and AQR• Implications for calculating provisions

∙ From where come the estimated CFs?∙ When do dual CFs make sense (debt se

Key issue: Significant loans with OEILarge borrowers (only – or all OEI borrowers?)

T

t t t

eff t

r

CF 0

) ( . ) 1 (

Provision = Book less Recoveries

17

Page 18: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• In any extension, modification or alteration:∙ Client pays from its own funds flow all interest due without

new financing granted for such purpose by the bank or its affiliate to the client or its associates.

• Otherwise (EBA is getting there by Dec. 2014):∙ Above extension is ‘problem loan refinancing’; ∙ does not suspend aging or arrears (migration continues) ∙ provisions increase with age (neutrality)∙ suspend accruals and reverse of unpaid amounts;∙ apply cash payments to principal (neutrality)∙ graduate based on explicit performance tests

Key issue: Rolling-on Bad Loans Citi’s golden rule to control re-aging risk

18

Page 19: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Provide standard practice not available in IAS 39:

• Move from contractual to probable (re-default weighted• Allow for debt-service FCF subject to evidenced projection• Zero debt service FCF if collateral dependent (define cut-off)• Exclusion of debt FCF if 180+ DPD• Direct method to estimate OCF (sales - ΔA/R – Δstocks etc.)• Sensitivity, quality and quantity of projected OCF (FX stressed)• Cautious and conservative assumptions to project OCF• Consequences if projection is not conservative: no debt

service• Use of re-default rates (RDR) if restructured or refinanced• De-minimis RDR (50%) if lack of evidence in calculating FCF• Must have paid all arrears, including if capitalized• Add other relevant issues to local practice.

Key Issue: FCFs from debt service

19

Page 20: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Define collateral dependent loan (only one cash flow)• Filter bad types of collateral (illiquid – debtor essential)• Up-to-date appraisals (spot market or lowest)• Enrich/reinforce collateral appraisal standards • Realistic time to foreclosure (TFT) towards 5

years• Conservative fire sale discount (FSD) towards

30% • Factor in administration and legal costs• Re-appraisal versus revaluation based on index• Land underdeveloped and associated issues• CRE yields and income (see ECB collateral criteria) • Add other relevant issues to local practice

Key Issue: FCF from collateral

20

Page 21: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Operational losses (new loans to finance ∆ working capital)∙ Economic evidence of viable working capital ∆ (stocks, A/Rs)

• CF diversion: from earning assets to non-earning assets.• Material decrease in turnover/loss of a major customer.• Default or breach of contract: overdraft rules, # times rolled• Projected debt service capacity: OCF < P+I

+Reinvestment• Financial performance:

∙ Related by policy to key financial indicators (KFIs) modelled∙ Cut-off points (split) used in internal rating models for KFIs∙ At which point migration leads to “significant financial problems”

FCFs: Other key considerations

21

Page 22: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Incurred versus expected loss issues

𝑬𝑳=𝑬𝑨𝑫∗𝑷𝑫∗𝑳𝑮𝑫𝑰𝑳=𝑬𝑨𝑫∗𝑷𝑫∗𝑳𝑮𝑫∗𝑬𝑷

• Banks implementing FIRB or AIRB per Basel II can move from an expected loss (EL) to an incurred loss (IL) to calculate portfolio based impairments

• Additional requirement is to estimate the Emerging Period (EP) factor with empirical historical data.

• Portfolio segmentation and the other parameters should be in place: subject to in-depth review and challenging from us and validation/backtesting.

Simplified algorithms

22

Page 23: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Must invest more in IT, data, systems, expert staff.

• Enact a supervisory methodology requiring:∙ Collection of historical data ∙ MIS “field-attributes” per client/transaction∙ Segmentation into homogeneous segments∙ Standard calculation method based on historical

data∙ Consideration of future needs for IFRS 9 EL model

• Decide among alternatives (joint or not)∙ Central algorith/methodology (e.g, Mexico,

Colombia)∙ Bank specific implementation of portfolio based

method

Standardized (no IRB) local banks

23

Page 24: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Adopt-adapt local practice to an IFRS compliant ILM

𝑰𝑳𝑷=(𝑬𝑨𝑫−𝑪∗𝑫𝑭 (𝑪𝑻 ,𝑳𝑻 ))∗𝑷𝑫 ¿• Discount collateral ‘C’ with a “DF” as per the

next slide• Include ‘transitory’ benchmarks (FSD, TTL)• Differentiate collateral type (“CT”) and loan

type (“LT”)• Calibrate a central PD based on loan type

considering the Financial State (“FS”) of the client and the Loan Performance status (“LP”) – Risk Grade Concept

• Move banks to invest in IT architecture and systems

• Consider relevance of a regional level (pool effort and data)

24

Page 25: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Example (Mortgages – Mexico)

cCovbovrecAdditionalaVTCL

TR

.

ˆ1

%10,%)201()1( TRMaxSP  Region A Region B Region C

With legal contract or trust in guarantee

a=0.57b=0.7

7c=0.93 a=0.45

b=0.65

c=0.89 a=0.36b=0.5

5c=0.84

Without legal contract or trust

a=0.47b=0.7

1c=0.91 a=0.37

b=0.60

c=0.86 a=0.29b=0.5

1c=0.82

)ˆ...ˆˆ(11101

1),...,(ˆ

nn xxne

xx

Severity:

Number of delinquenciesMax delinquencies last 4 paymentsAverage paid last 7 monthsCredit Loan to value (CLTV)Seller’s participation

25

Page 26: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Key issue: Collective provisioning

Discount factor, DF = 1 / (1 +

effective interest rate) TTL

 

 

PD = 1 for NPL exposures TTL = time to liquidation after foreclosure WOC = work-out costs

EAD =Exposure-at-defaultEP =Emerging period HPI =Collateral-Price-Index PTT =Peak-to-trough

LGL =Loss-given-liquidationFSD =Fire-sale-discountC =Collateral

•Generic formulation: LGD

Anolli, Mario, Becalli, Elena: “Retail Credit Risk Management”, 201326

Page 27: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Lack of standard concept/practice (IFRS and BCBS)

• EBA would like to guide on this element of alchemy

• Rate at which loans in arrears become performing

• Time span to consider a work-out “effective”• Effective: Re-default rate ≈ equivalent

marginal PD• Obscure alchemy with material impact in

provisions.• Key to quantify the “severity”: LGD =LGL*(1-

CR)• Key to filter “false cures” from the ODR and PD• Observed or ‘modelled” and where is the

evidence?

Key issue: from where the Cure rates?See more hints in the appendix

27

Page 28: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Months performing after modification as per new terms• Measures to ensure that once a loan is modified the

performance tests remain prudent and traceable.• These tests preclude the release of a provision and the

conditions for such release.• Determine if banks retain provisions for a longer period,

or at least in sufficient amount to cover capitalized interest.

• Until building a sufficient buffer of debtor’s equity (e.g., reach a LTV of 80 percent, of OCF > trigger level)

• Or, align to EBA (2 years) or BE (2 years/20% repayment)

• Traceable in internal MIS with tagged attributes

Key issues: Performance periods (test)

28

Page 29: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• EP refers to the detection risk (from re-aging risk)

• Time span between the occurrence of a loss event and the date that loss event is identified.

• Usually applied to non-impaired as IBNR provision

• IBNR = EAD• [ 1 – (1 – PD) EP • LGL• (1- Cure rate)]

• Practice from 3m (mortgages) to 12m (corporate)

• Best practice: take a sample (say 30 randomly loans by pool/product) that went bad in each category and examine the files to see how long the emergence period for each really was (to the point of performing the impairment test).

Key Issue: Emerging period (EP)

29

Page 30: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• LGD/LGL continue to be the less well know risk inputs• Few actual charge-offs result into limited loss experience.• No empirical loss data from holding “loss loans” without foreclosure:• Lack of rules to govern charge-offs and clean up the back-bad-books • Protracted legal process for recovery: adds uncertainty/widens

spreads• Collective provision reported mixed with the specific provision• Set in place process to compile data for gaps and use proxies• Consider using transitory proxies (until there is empirical evidence)

∙ e.g., no cures (or minimal cures) absent of systems to tag and track cures∙ 50% re-default rate absent of systems to quantify re-default∙ Longest Time-to-Foreclosure-Liquidation ∙ Highest Fire-Sale-Discount∙ Forced write-off time limit∙ Phase-out the eligibility of PV(collateral)

Key issue: Limited loss evidence

30

Page 31: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Concept used by (1) policy (2) practice and (3) regulation

• (1) model input risk parameters (2) assumptions & projections.

• Time (1) to foreclosure vs. (2) to divestiture (via OREO)

• Management assumptions (practice) versus empirical evidence

• “Hold and hope” strategies: Mark in lieu-of-repossession

• Fair Value determination: (1) appraisal versus (2) internal valuation

• Pricing: (1) House/Commercial price index (HPI) (2) peak-to-through (PTT) to LTV

• LTV used: (1) historical versus (2) rebased LTV (procedures used)

• Collateral related costs: (1) repair (2) renovation (3) administration (4) maintenance (5) sale

• Fire sale discounts and administration costs

Key issue: Collateral dependency

31

Page 32: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Time to liquidation (TTL)

Segmenting by type of collateral, location, condition, age, etc.

Must compile and track appropriate information

32

Page 33: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Key Issue: Recovery (net of all costs)

Must compile and track appropriate information

33

Page 34: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Control ‘your’ migration risk (many more moving parts)• Implement a migration strategy (see annexes), or:• (A) Keep old matrix-system for small/non-complex, or• (B) Decide for a centralized solution/algorithm• Require IT/Data investment for credit risk MIS• Compile bankers’ practices, risk inputs and assumptions• Revamp means, skills, people, processes, tools• Develop a benchmark to measure gaps and distances• Set regulatory parameters when no clear empirical data• Consider a risk grade concept based on internal ratings

Lessons (take away)

34

Page 35: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Acronyms• Templates to compile

practice/parameters• Cure rates (might distort PD and LGD)• Quick model/methods overview• References

Appendixes

35

Page 36: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Acronyms

• DPD= Days past-Due• DCF = Discounted cash flows• EAD= Exposure at Default• EP = Emergence Period• FCF = Future Cash Flows• FSD= Fire-Sale Discount• IBNR = Incurred But Not Reported• LGL = Loss given liquidation• LGD = Loss given default• MIS = Management information

systems• OCF= Operational Cash Flow• ODR = Observed Default Rates

• OEI = Objective evidence of impairment

• PIT= Point-in-Time• PTT = Peak to Trough (real estate price)• PD = Probability of Default • TTL = Time to Liquidation• TTC = Trough-the-Cycle

36

Page 37: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Appendix: Planning the transition • Need to learn IAS39-IFRS9

• Have a migration strategy• Aware on IAS 39 key issues• Keep IFRS 9 in mind• Prepare dedicated people• Learn credit risk modelling• Warehouse stronger data• Develop benchmarks/tools

• Meaning, concepts, practices, issues, alternatives• Incurred vs Expected vs. Cycle adapted • Point-in-time (PIT) vs. Through-the-cycle (TTC)• Accounting provisioning vs. Loan loss forecasts vs. Basel II capital • Base normal conditions vs. adverse/downturn conditions • Broken standards advise the use of regulatory accounting practices• See presentation of Mike Moore (use regulatory reporting) 37

Page 38: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

SurveyBank

practices

DevelopBenchma

rk

IssueGuidelin

es for IAS 39

Calibrate

Pillar 2Overla

y

Plan for

IFRS 9

now

Appendix: Migration strategy

Develop templateCompile bank practiceDo targeted visits

Time to liquidation (TTL)Fire sale discounts (FSD)Cures net of refinancingPD rates PIT (vs. actual ODR)Emerging Periods 12/24m

Define good/bad practiceSet expectations & detailsIndicate processes and roles

Overlays to close gaps Use standard benchmarksPreserve regulatory reporting

38

Page 39: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Unite forces to facilitate transition (not alone). • Form a regional committee and a working

group.• Agree to upload/share anonym information.• Obtain expert support for a medium term

project.• Develop templates to compile regional

practice.• Suggestions included in the appendixes

• Prepare people with specialized skills (learn).• Discover and benchmark techniques,

estimates.• Understand materiality of differences.

Appendix: Way Forward

39

Page 40: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Ap

pen

dix

: I

np

uts

sp

ecifi

c p

rovis

ion

ing

Tem

pla

te t

o c

om

pile p

racti

ces

By segment (COR, CRE, SME, RSM, CON)All pools PNF PFB NPFB NPNFB Other Notes

Cut-off point usedType of rating model

Methodology(ies)In-house or vendorNumber of gradesPD calibration methodData length & attributesValidation/backtest policy

Discount rate usedEIROther rate

Debt service flows computedContractualProjectedStressedWeighted bySuspended at daysRe-default usedOther

Collateral flowsTime-to-liquidationFire-sale discountLegal expensesAdministration +Spot priceOther priceIndex-rebasedCollateral dependentOther

Old regulation metricsCriteria (a)Criteria (b)Criteria (c)

Volume of EADNumber of exposuresProvisions per bank IFRSProvisions (old regulation)

Significant - Individual - Specific Model Risk Parameters

and Assumptions

Per

form

ing

not e

ver

forb

orne

Per

form

ing

forb

orne

Non

-Per

form

ing

forb

orne

Non

-Per

form

ing

nor f

oreb

orne

40

Page 41: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Ap

pen

dix

: In

pu

ts f

or

collecti

ve p

rovis

ion

ing

Tem

pla

te t

o c

om

pile p

racti

ce

By segment (COR, CRE, SME, RSM, CON)All pools PNF PFB NPFB NPNFB Other Notes

Cut-off point usedType of rating model

Methodology(ies)In-house or vendorNumber of gradesPD calibration methodData length & attributesValidation/backtest policy

Discount rate usedEIROther rateTime-to-liquidationCollateral discounts

Loss ratesHistoric averageHistoric adjustedRoll-ratesLGLLGDCure rate usedCure cleanedOther

Old regulation metricsCriteria (a)Criteria (b)Criteria (c)

Volume of EADNumber of exposuresProvisions IFRSProvisions (old regulation)Include specification of formula or algorithm used per segment

Collective (excluding IBNR) for NPLsModel Risk Parameters

and Assumptions

41

Page 42: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Ap

pen

dix

: In

pu

ts f

or

IBN

R p

rovis

ion

ing

Tem

pla

te t

o c

om

pile p

racti

ce By segment (COR, CRE, SME, RSM, CON)

All pools PNF PFB NPFB NPNFB Other Notes

Point-in time

Through-the-cycleMixed PDOther approachEmergence period

Loss ratesPD pitPD ttcHistoric averageHistoric adjustedRoll-ratesLGLLGDCure rate usedCure cleanedOther approach

Old regulation metricsCriteria (a)Criteria (b)Criteria (c)

Volume of EADNumber of exposuresProvisions IFRSProvisions (old regulation)Include specification of formula or algorithm used per segment

Model Risk Parameters and Assumptions

IBNR for Performing

PD

42

Page 43: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Appendix: Importance of Cure Rates

Cure rate LGD=0The rate at which loans in arrears become performing P(C|D)Actual effectiveness of work-out operations

P(D|P)

Danger rate LGD>0

P(CO|P)

Loss-Given-Default (LGD) = Loss-Given-Liquidation (LGL) * (1 - Cure Rate)Charge-off LGD frequently is a model output (regression) due to no empirical lossP(C|D)*charge-off LGDP(D|P)*doubtful LGD+P(CP|P)*charge-off LGD

Observed vs Modelled Cure Rates

Must filter reaged and refinanced cases from the real cures calculated

PERFORMING

DOUBTFUL

CHARGE-OFF

CURED to Performing

CHARGE-OFF

43

Page 44: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Find out the concept/practice of “cures” used by the industry• Clarify what un-friendly solutions are: repossession and

implications • Determine the rational for high level of cures in banks• Confirm that/if:

∙ CRs are conditioned by an early definition of default∙ The MIS tags & tracks defaults that return to performing status with no loss∙ The “cures” form part of the loss history on which LGD estimates are based∙ “Incomplete workouts” in the recovery process are also part of LGD history∙ Banks tag and track the above − Discuss implications to validate income

flow

• Ascertain in which segment cure rates are used and implications• Determine practice in estimating probabilities of cures:

∙ Empirical observed versus modelled cures (local or parent ‘global experience’)

Appendix: More on cure rates (1)

44

Page 45: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Determine proper use of “default count” basis• Ascertain if a high level of cures tend to fall hard during the downturn• Prove expected ∆ of CRs in a downturn (before including in LGD)• Determine the precise course of events that allow cures to take place • Discuss those events and how are filtered from standard refinancing• Clarify controls in place if cures are driven by firm’s own policies• Determine if longer realization periods and larger forced sale discounts

are applied to the exposures that do not cure• Considering evidence of higher default rates on the books as a whole• See if firms are accounting for links between cures & re-default rates• See if earlier cure definition (6M) result in higher level of re-defaults

Appendix: More on cure rates (2)

45

Page 46: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Discuss if auditors approve CR considering reliability of MIS

• Ascertain means to compensate the exclusion of CR restructures from PD/LGD data

• Determine work-arounds to compensate the exclusion of CR and incomplete workouts

• Means of modelling and approach: back-testing and validation

• Ascertain the variables and inputs for modelling cures = F(age, LTV, equity, other)

• MIS attribute handling: tagged procedures for extensions, re-aging, novation, refinancing

• Test the source of empirical data: frequency, number, amounts, tracing durability

• Filter false cures: temporary forbearance – instability of observed cures

Appendix: More on cure rates (3)

46

Page 47: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Consider how/which are:∙ the rating/scoring models built∙ the discriminant factors used

(quantitative/qualitative)∙ their correlations and how correlation dealt with∙ The key powerstat and fitness measures (ROC, etc)∙ re-ages, cures, refinancing… tagged, tracked,

factored

• Whether the rating model is fit to support a Risk Grade Concept (downgrade to below BBB equates to significant deterioration as an event of impairment) and the contribution and cut-off point of each factor used in the model.

Appendix: The ordinal rating model

47

Page 48: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Appendix: The calibration of the PDA PD assigned to each score/rating grade

• Different techniques (learn them)• PD12M = F(season, moment of the cycle) • Failed cases / (Failed cases + pass cases) by grade or score or

group_of_risk

• Calculation of calibration curves (functions) by groups of risk

• Adjustment to a central tendency of the portfolio• Cycle neutral calibration (as for capital’s State of the

Economy)• Validation and back-testing (policies and process)

−− Normal adjusted to CT−− Exponential adjusted to CT

Calibration of PD to Score/Rating Grades

48

Page 49: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Appendix: LGD - Rules vs Models

Time window/historical depth ( BII ¶ )� Sample to fit default definition and collection process Data availability� All available default history into development-test samples Default definition (coherence with PD models)� Collect atributes and macro-drivers to test Close/open defaults (per model type)�

Workout approach� Attributed to each facility with direct-indirect costs Discount factor definition� Selected target variable for LGD or for sevity models Discount of cash flows� Separate considerations for Basel and for impairment models Recovery process� Impairment models more adhrent to recent trends in recoveries

Both for segmentation and estimation purposes� Drivers most likely to imapct LGD Macroeconomic factors� 3. Key macroeconomic indicators 4. Vintage atributes Product-specific drivers� 1. Transaction/product features 2. Borrower attributes Based on explanatory power�

Estimated on defaulted contracts� 1. Conditional means2. Linear regression

Applied to all portfolio exposures� 3. Chain ladder4. Target variable

Measure model performance on current portfolio� Monitor and backtest model periodically � Model calibration/update as needed�

LGD Model Development Steps

RELEVANT PERIMETER

RECOVERY FLOWS

SELECTION OF POWERFUL DRIVERS

LGD ESTIMATION

VALIDATION - MONITORING -BACKTEST

• LGD/LGL continues to be the less well known 49

Page 50: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Used for rank-ordering purposes despite the fact that log-odds map to distinct delinquency and default rates.

• Input for many modeling frameworks (joint-odds), and estimating provisions.

• Requires specialized staff and resources (e.g. software, hardware, and data access). Smaller institutions rely more on third party vendors.

• Built for several purposes (delinquency, default, bankruptcy, attrition, profitability, solicitation response, etc.), but the power of scorecards is the ability to build them at a segment level.

• Can capture most factors if properly calibrated (i.e. regional economic data can be appended to account files, indicator variables can proxy account strategies and other exogenous factors) and segmented.

• Macroeconomic information is rarely considered in scorecard modeling

Appendix – Notes on Scorecards

50

Page 51: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Static, capturing the risk profile of the portfolio across a minimum of two dimensions, constructed at the segment or portfolio level.

• Combining both an external (FICO) and internal (behavioral score) scorecards, it can identify exogenous factors (systemic-idiosyncratic) impacting behavior across a particular segment such as delinquency, debt management, and over-limit.

• In particular, each scorecard takes account of n risk factors and summarizes credit risk into two dimensions.

• Each cell represents a loss rate for a group of accounts with a particular FICO and behavioral score (e.g. the estimated net loss rate for accounts with a 660 FICO + 900 behavioral score is 5.23%).

• A 12m forecast is determined by applying the distribution of 1 yr historical loss rates to the current distribution of outstanding loans.

• These models doe not consider attrition, management strategies, and economic factors in the upcoming 12-month forecast horizon.

Appendix – Notes on Matrix models

51

Page 52: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Matrix model exampleHenderson – FRB Philadelphia (2009)

No_score 1-849 850-899 900-924 925-949 950-959 960-969 970-979 980-984 985-989 990-992 993+No_score 15.98% 24.38% 16.89% 8.45% 7.60% 6.84% 6.16% 5.85% 5.56%5 28% 5 01% 4 76%000-590 16.74% 36.57% 21.38% 10.69% 9.62% 8.66% 7.79% 7.40% 7.03% 6.68% 6.35% 6.03%591-610 11.58% 20.41% 19.12% 9.56% 8.60% 7.74% 6.97% 6.62% 6.29% 5.98% 5.68% 5.39%611-630 11.01% 16.29% 15.67% 7.84% 7.05% 6.35% 5.71% 5.43% 5.15% 4.90% 4.65% 4.42%631-650 10.57% 14.20% 14.38% 7.19% 6.47% 5.82% 5.24% 4.98% 4.73% 4.49% 4.27% 4.06%651-660 8.68% 10.22% 10.45% 5.23% 4.70% 4.23% 3.81% 3.62% 3.44% 3.27% 3.10% 2.95%661-770 8.66% 8.45% 9.83% 4.92% 4.42% 3.98% 3.58% 3.40% 3.23% 3.07% 2.92% 2.77%671-690 6.59% 7.68% 7.56% 3.78% 3.40% 3.06% 2.76% 2.62% 2.49% 2.36% 2.24% 2.13%691-710 5.11% 5.22% 5.33% 2.67% 2.40% 1.26% 1.94% 1.85% 1.75% 1.67% 1.58% 1.50%711-730 4.32% 3.25% 1.60% 0.80% 0.72% 0.65% 0.58% 0.55% 0.53% 0.50% 0.47% 0.45%731-750 2.13% 0.10% 0.33% 0.16% 0.15% 0.13% 0.12% 0.11% 0.11% 0.10% 0.10% 0.09%751+ 0.59% 0.00% 0.31% 0.16% 0.14% 0.13% 0.11% 0.11% 0.10% 0.10% 0.09% 0.09%

Behavioral Score

FICO S

core

Joint-odds Matrix

High Risk Low Risk

52

Page 53: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Measures the percentage of accounts or dollars that “roll” from one stage of delinquency to the next.

• Individual accounts are not tracked, only the volume for a particular bucket.

• The roll rates are averages across risk segments or for the total portfolio.

• The critical roll rate is the ‘net charge-off’ rate since it gives you the amount of charge-off at the end of the next month (t+1).

• Roll rate models fit the retail credit business model well as call centers and collection departments are commonly aligned by stage of delinquency.

• The roll rate model do not explicitly incorporate attrition, management strategies, and exogenous factors such as the economy.

Appendix: Notes of Roll-rate models

53

Page 54: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Retail roll-rate methodsTime

Past Due Days

Current

Bucket1

Bucket 2

Bucket2+n

LastBucketProvision(i) = Bucket balance (i) * PDPIT(i) * LGD

Rollrate r(i)

The roll-rate methodology predicts losses based on delinquency. Most roll-rate methodologies assume that delinquency is the only loss event and that significant allowances are not needed until a loan becomes delinquent. Roll-rate methodologies are also known as migration analysis or flow models.

54

Page 55: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Roll rates – exampleFDIC’s Manual Chapter XII

MonthCurrent Balance

30 daysCurrent-to 30d RR

60 days30d-to 60d

RR90 days

60d-to 90d RR

Current to loss factor

Oct. $1,000 $39 $12 $9Nov. $1,000 $40 4.00% $15 38.46% $10 83.33% 1.28%Dec. $1,250 $50 5.00% $20 50.00% $12 80.00% 2.00%Jan. $1,200 $65 5.20% $28 56.00% $17 85.00% 2.48%

Delinquency buckets (FDIC Chapter XII page 6) - Simplified Roll Rate Example

55

Page 56: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• A sequence of random variables is said to form a Markov chain if each time an account is in some initial state I and there is a fixed probability that it will next be in another state J.

• Markov chain models can account for all probabilities within delinquency stages, not just those that move sequentially from one stage of delinquency to the next.

• Rules must be used to limit the dimensions of the Markov chain. For example, an account can only be current (c), delinquent (L), and default (D).

• Transition probabilities are averaged across risk segments or the total portfolio.

• Markov chain models can account for attrition and some account management strategies, but still ignore economic factors.

Appendix – Markov Chain models

56

Page 57: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Vintage models segment the portfolio by either the year (YOB) or month that an account is booked (MOB).

• Once the vintage criteria is determined, the loss performance of the segment is tracked over time.

• Can be further segmented to reflect more granular levels of risk such as delinquent/non-delinquent and bankrupt/non-bankrupt populations.

• Annual loss rates and month-on-book losses usually provide fewer data points so exponential smoothing techniques (weighted averages) are useful.

• Can account for management strategies and exogenous factors by optimally adjusting parameters within an exponential smoothing algorithm.

• Forecasts for the next YOB could simply reflect the previous YOB performance with downward or upward adjustments based on the credit quality of new loans.

Appendix: Notes on Vintage models

57

Page 58: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

Vintages - exampleYear 1 Year 2 Year 3 Year 4 Year 5 Year 6 Year 7

1999 0.68% 1.46% 3.57% 4.58% 5.27% 5.10% 5.00%2000 0.85% 1.82% 4.46% 5.72% 6.59% 6.38%2001 0.89% 1.90% 4.67% 5.99% 6.89%2002 0.78% 1.67% 4.09% 5.25%2003 0.65% 1.39% 3.41%2004 0.59% 1.26%2005 0.54%2006 0.71% 1.58% 4.04% 5.39% 6.25% 5.74% 5.00%

Year Since the Account Was Booked

0.00%

1.00%

2.00%

3.00%

4.00%

5.00%

6.00%

7.00%

0.00%

5.00%

10.00%

15.00%

20.00%

25.00%

Year 1 Year 2 Year 3 Year 4 Year 5 Year 6 Year 7

2005

2004

2003

2002

2001

2000

1999

2006

58

Page 59: Impairment Parameters Look well below the surface Joaquin Gutierrez The World Bank FinSac - Vienne -- October 22, 2014

• Anolli, Mario; Becalli, Elena – Retail credit risk management. McMillan (2013).

• Van Deventer – Credit Risk Models & Basel Accords. Willey (2009).

• Saunders – Credit Risk Measurement – Willey (2010)• Ong – Internal Credit Risk Models – Risk (2005)• Cossin & Pirotte – Advanced Credit Risk Analysis. Willey

(2001).• Henderson – Retail Credit Risk Models. FRB Philadelphia

(2009).• FDIC – Examination manual, Chapter XII, Allowance for

Loan Loss.

References

59