alan flores - tdc sao paulo 20111017

Upload: jtnylson

Post on 03-Apr-2018

219 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    1/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    2/65

    Systemic estimation of PD, LGDand EAD for credit card as a

    reserve requirement andvalidation method.

    Mexican Experience

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    3/65

    Mexican Banking SystemThe Mexican banking system is concentrated in 7 banks that accumulate 87% of

    the total loan portfolio and are mainly foreign owned (US, Spain, UK, Canada).

    Jul-11 Aug-11 Jul-11 Aug-11 Jul-11 Aug-11 Jul-11 Aug-11

    Banking System 2,271 2,298 2,510 2,542 12.53 12.32 16.39 16.21

    BBVA Bancomer 581 588 620 634 20.65 19.95 15.58 15.51

    Banamex 359 359 447 446 8.97 8.53 17.04 16.90

    Banorte 271 275 288 282 14.59 14.23 15.94 15.67

    Santander 297 301 316 338 15.77 15.61 14.81 14.24

    Inbursa 153 155 123 120 8.48 8.21 24.82 23.21

    HSBC 182 184 279 281 1.41 1.64 14.27 14.78

    Scotiabank 114 115 121 121 10.20 10.04 16.60 16.43

    Loan Portfolio Deposits

    Balance sheet (Billions MXN)ROE (12m) (%) Capital Ratio (%)

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    4/65

    Mexican Banking SystemHowever the rate of deterioration of the credit card portfolio showed a

    significant increase in 2009.

    DelinquencyIndex (Past DueLoans / TotalBalance) of

    consumer loansportfolio

    Credit Cards

    3.2%

    9.8%

    4.5%

    2.7%

    6.6%

    1.4%0%1%2%3%4%5%

    6%

    7%8%9%

    10%

    Jan-05 Jan-06 Jan-07 Jan-08 Jan-09

    ConsumerCredit

    Other**

    11%

    Mortgage ConsumerCredit

    CorporateCredit

    Loan loss provisionAnnual Cash Flow

    0

    25,000

    50,000

    75,000

    100,000

    125,000

    mar-06 sep-06 mar-07 sep-07 mar-08 sep-08 mar-090.0

    1.0

    2.0

    3.0

    4.0

    5.0

    ** Includes Personal credit, leases, and other consumer credits

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    5/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    6/65

    Mexican Banking SystemThis deterioration was mostly explained by a systemic increase in household

    indebtedness

    Monthly Average Debt-Capacity per Debtor(Sample)

    0

    100,000

    200,000

    300,000

    400,000

    500,000

    600,000

    700,000

    800,000

    900,000

    1 2 3 4 5 6 7 8

    Number of Credit Cards

    Available CreditLimit (left axis)

    Dec. 2005 Nov 2007

    MXN pesos

    Number of debtors(right axis)

    Credits per Person by Credit Type

    3

    3.2

    3.4

    3.6

    3.8

    4

    4.2

    4.4

    D J F M A M J J A S O N D J F M A M J J A S O N

    2005 2006 2007

    1.00

    1.05

    1.10

    1.15

    1.20

    1.25

    1.30

    1.35

    1.40

    Bank credit cards (left axis) Mortgage (right axis) Car (right axis)

    Source: Credit bureau

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    7/65

    G20 and Basel recommendations on

    loan loss provisionsIn April 2009 the G20 issued recommendations on financial supervision and regulation that led

    the Basel committee to propose the following recommendations related to loan loss

    provisions:

    1. Loan loss provisioning should be robust and based on sound methodologies that

    reflect expected credit losses in the banks existing loan portfolio over the life of

    the portfolio.

    2. The accounting model for provisioning should allow early identification andrecognition of losses by incorporating a broader range of available credit

    information than presently included in the incurred loss model.

    3. The new standard should utilize approaches that draw from relevant

    information in banks internal risk management and capital adequacy systems

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    8/65

    IASB and Basel recommendations on

    loan loss provisionsRecommendations were sent to IASB in order to promote the necessary coordination between

    standard setters, supervisors and regulators on their respective efforts to implement the G20

    recommendations.

    Consequently IASB issued in November 2009 a proposal to modify loan loss provisionsaccounting.

    Incurred loss model (current IASB standard)

    Assumes that loans will be paid until evidence on the contrary is identified (loss event).

    The financial crisis has evidenced that these models are characterized by evaluating

    optimistically the loan portfolio and are suddenly followed by large credit losses.

    Expected loss models (proposal adopted by Basel and IASB)

    Losses are estimated on a forward looking basis according to the quality of theportfolios. It implies an approximation of the PD, LGD and EAD.

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    9/65

    Content

    1. Introduction

    2. Systemic estimation of PD, LGD and EAD

    3. Risk Analysis of the Credit Card Loans Portfolio

    4. Conclusions

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    10/65

    Incurred LossesUntil August 2009, Mexico had an incurred loss type regulatory provisioning

    model. Provisions were created by applying fixed percentages to loans in

    different levels of delinquency.

    Number of delinquent

    periods

    % of Provisions

    0 2.5%

    1 19%

    2 48%

    3 58%

    4 62%

    5 85%

    6 95%

    7 100%

    8 100%

    9 or more 100%

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    11/65

    Incurred Losses12 month write offs were significantly higher than the balance of loan loss

    provisions revealed on banks balance sheets.

    Q y C / C V

    1 2 M E S E S

    0 %

    2 0 0 %

    4 0 0 %

    6 0 0 %

    8 0 0 %

    E ne -0 6 A g o -0 6 A b r-0 7 D ic -0 7 A g o -0 8

    Write offs (next 12 months) / loan loss provision balance

    Jan-06

    Credit card overall portfolio

    1 Write Offs in the following 12 months,.

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    12/65

    Incurred LossesSome disadvantages of the Incurred Loss Model are:

    Loans are provisioned when factual evidence is available that

    a loan or portfolio of loans will not be repaid in full. (Laterecognition of losses)

    Generate provisions for no concrete time horizon.

    Show pro-cyclicality since they generate largest amount ofprovisions when there is evidence that loans will not be

    repaid.

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    13/65

    Incurred LossesSome disadvantages of the Incurred Loss Model are (continued):

    Do not generate provisions for loans with no delinquency

    even though these loans have a positive expected loss.

    They do not consider the potential growth of the exposure atthe time of default (relevant feature in revolving credit).

    Similar financial assets in different entities may generate losscoverage for different periods of time.

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    14/65

    Expected LossesInternal Models are characterized for estimating the components of the

    expected loss.

    Expected losses standardize the time horizon and the default definition

    for all institutions.

    Expected Loss = PD * LGD * EAD

    Where:PD = Probability of default

    LGD= Loss given default

    EAD = Exposure at default

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    15/65

    PD estimationA window of 25 months was established to carry out the analysis of the

    behavior "profile" of each credit card.

    Historical Period (T-12, T0).

    Performance Period (T0, T12)

    Reference Point = T0

    t= -12 t= 0 t= +12

    Historical Period Performance

    Period

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    16/65

    PD estimation

    Default is declared when a borrower attains a past due status on his

    payment obligations of 90 days.

    t= -12 t= 0 t= +12

    Performance Period

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    17/65

    Sample for analysisA sample of loan-level data representing 97.3% of total credit card loans in thesystem was extracted from banks.

    # Contratos % SistemaInformacin

    Recabada

    % Informacin

    Recabada

    % Sistema

    Ajustado

    Inbursa 712,034 2.03% 712,034 2.03% 2.09%

    AMEX 1,025,869 2.93% 1,025,869 2.93% 3.01%Banorte 1,256,316 3.59% 1,256,316 3.59% 3.69%

    Santander 3,448,425 9.85% 3,448,425 9.85% 10.13%

    Banamex 8,971,960 25.63% 8,971,960 25.63% 26.34%

    BBVA Bancomer\Finanzia 13,999,809 40.00% 13,999,809 40.00% 41.11%

    GE Money 796,019 2.27% 796,019 2.27% 2.34%

    Invex 853,343 2.44% 853,343 2.44% 2.51%

    Scotiabank 790,995 2.26% 790,995 2.26% 2.32%

    HSBC 2,201,229 6.29% 2,201,229 6.29% 6.46%

    Otros 946,902 2.71% - - -

    TOTAL 35,002,901 100.00% 34,055,999 97.29% 100.00%

    The sample allowed a maximum error of 40 basis points on a PD estimation with (1- ) =

    99% confidence.

    ( ) > 1 dppP

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    18/65

    DatabaseData were organized in 12 windows of information spanning 25 months ofpayment history.

    WINDOW FROM TO

    Apr05

    May05

    Jun05

    Jul05

    Aug05

    Sep05

    Oct05

    Nov05

    Dec05

    Jan05

    Feb06

    Mar06

    Apr06

    May06

    Jun06

    Jul06

    Aug06

    Sep06

    Oct06

    Nov06

    Dec06

    Jan07

    Feb07

    Mar07

    Apr07

    May07

    Jun07

    Jul07

    Aug07

    Sep07

    Oct07

    Nov07

    Dec07

    Jan08

    Feb08

    Mar08

    1 Apr 05 Apr 07 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

    2 May 05 May 07 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

    3 Jun 05 Jun 07 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

    4 Jul 05 Jul 07 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

    5 Aug 05 Aug 07 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

    6 Sep 05 Sep 07 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

    7 Oct 05 Oct 07 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

    8 Nov 05 Nov 07 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

    9 Dec 05 Dec 07 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

    10 Jan 06 Jan 08 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 1211 Feb 06 Feb 08 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

    12 Mar 06 Mar 08 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    19/65

    DatabaseThree sources of information were used for the analysis.

    DATABASE 2Credit BureauInformation

    DATABASE 1Bank Information

    DATABASE 3Social securityInformation

    t= -12 t= 0 t= +12

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    20/65

    PD

    BorrowerBehavior

    Product Policy CreditBureau

    t= -12 t= 0 t= +12

    Social securityinformation

    10 Variables 46 Variables 9 Variables 10 Variables

    Examples:

    - Minimum payment amount of

    the last 12 months.

    -Maximum credit limit in the last

    12 months.

    -Theoretical time that will take

    to repay the balance according

    to the minimum payment and

    the interest rate of the product.

    Examles:

    - Average use of the credit limit in

    the last 12 months.

    -Percentage of payment over the

    balance.

    - Number of times that the

    borrower paid the total balance.

    - Number of non-payments in the

    credit card.

    Examples:

    - Number of credit cards

    opened in the period.- Record of payment in other

    open accounts

    - Number of credits that the

    borrower has at the reference

    point.

    Examples:

    -The borrower has formal or

    informal employment

    - The borrower has a mortgage

    with the Social Housing

    institute (INFONAVIT)

    - Borrowers Income at the

    reference point (measured in

    minimum wage).

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    21/65

    PDFor the estimation of the systemic PD standard logistic regression was used tocorrelate the historical period constructed variables with the observation

    period binary event of default.

    )...(1 1101

    1),...,(

    nnxxn exx

    ++++

    =

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    22/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    23/65

    Percentage of paymentAmount of payment made by the cardholder as a proportion of the outstandingbalance at the reference point.

    Credit Card Portfolio

    Interval PD

    0% 26.00%

    10% 20.94%

    20% 16.65%

    30% 13.10%

    40% 10.21%

    50% 7.90%

    60% 6.07%

    70% 4.65%

    80% 3.55%

    90% 2.70%

    100% 2.05%

    110% 1.55%

    % PMT (T0)

    % default

    Ave 0.3695

    Std Dev 0.4051

    Max 1.1000Q75 0.7109

    Median 0.1464Q25 0.0557

    Min 0.0000

    Frecuency %

    Default

    %PMT (T0)

    0%

    20%

    40%

    60%

    80%

    100%

    0%

    10%

    20%

    30%

    40%

    50%

    60%

    70%

    80%

    90%

    100%

    110%

    0.0%

    10.0%

    20.0%

    30.0%

    40.0%

    50.0%

    60.0%

    Pmt = Payment

    Bal = Balance

    0

    0)(%0

    T

    T

    Bal

    PmtTPMT =

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    24/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    25/65

    Credit Limit ConsumptionOutstanding balance of the credit card as a percentage of the credit limitoffered by the bank to the client.

    Sistema

    Interval PD

    0% 5.67%

    10% 6.75%20% 8.02%

    30% 9.51%

    40% 11.24%

    50% 13.24%

    60% 15.53%

    70% 18.13%

    80% 21.06%

    90% 24.33%

    100% 27.92%

    250% 86.38%

    % USE (T0)

    % default

    Ave 0.4851

    Std Dev 0.3976

    Max 2.5000

    Q75 0.8657Median 0.4124

    Q25 0.0984

    Min 0.0000

    Frecuency %

    Default

    % USE (T0)

    0%

    20%

    40%

    60%

    80%

    100%

    0%

    10%

    20%

    30%

    40%

    50%

    60%

    70%

    80%

    90%

    100%

    250%

    0.0%

    5.0%

    10.0%

    15.0%

    20.0%

    25.0%

    30.0%35.0%

    40.0%

    45.0%

    Credit Card Portfolio

    Bal = Balance

    0

    0)0(%T

    T

    LimitCredit

    BalTUSE =

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    26/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    27/65

    Current Non-Payment (ACT)Consecutive periods, in which the cardholder has not covered the paymentobligation.

    Sistema

    Interval PD

    0 9.24%

    1 30.49%

    2 65.39%

    3 89.06%

    ACT% Default

    Ave 0.2386Std Dev 0.6219

    Max 3.0000Q75 0.0000

    Median 0.0000Q25 0.0000

    Min 0.0000

    Frecuency %

    Default

    ACT

    0%

    20%

    40%

    60%

    80%

    100%

    0 1 2 30.0%

    10.0%

    20.0%

    30.0%

    40.0%

    50.0%

    60.0%

    70.0%

    80.0%

    90.0%

    Credit Card Portfolio

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    28/65

    Historical Non-Payment (HIS)Number of periods in which the cardholder has not covered the minimumpayment in the last 6 months.

    Sistema

    Interval PD

    0 6.60%1 14.31%

    2 28.30%

    3 48.26%

    4 68.79%

    5 83.89%

    6 92.49%

    HIS

    % Default

    Ave 0.7226Std Dev 1.1467

    Max 6.0000Q75 1.0000

    Median 0.0000Q25 0.0000Min 0.0000

    Frecuency %

    Default

    HIS

    0%

    20%

    40%

    60%

    80%

    100%

    0 1 2 3 4 5 6 0.0%

    20.0%

    40.0%

    60.0%

    80.0%

    100.0%

    120.0%

    Credit Card Portfolio

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    29/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    30/65

    Type of EmploymentNon formal employment = 0

    Formal employment (social security) = 1

    Sistema

    Interval PD

    0 24.26%

    1 10.77%

    T_Employ

    % Default

    Ave 0.6227Std Dev 0.4847

    Max 1.0000Q75 1.0000

    Median 1.0000Q25 0.0000Min 0.0000

    Frecuency %

    Default

    T_Employ

    0%

    20%

    40%

    60%

    80%

    100%

    0 10.0%

    2.0%

    4.0%

    6.0%

    8.0%

    10.0%

    12.0%

    14.0%

    16.0%

    18.0%

    20.0%

    Credit Card Portfolio

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    31/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    32/65

    Income LevelNumber of times the minimum wage

    Sistema

    Interval PD

    0 40.03%

    1 33.08%

    2 26.79%3 21.32%

    4 16.71%

    5 12.93%

    10 3.20%

    15 0.73%

    20 0.16%

    25 0.04%

    30 0.01%

    35 0.00%

    Income

    % default

    5.4320

    7.3905

    116.91007.7700

    2.6100

    0.0000

    AveStd Dev

    Max

    Q75Median

    Q25

    Min 0.0000

    Frecuenc

    y %Default

    Income

    0%

    20%

    40%

    60%

    80%

    100%

    [0,5

    )

    [5,1

    0]

    >10 0.0%

    2.0%

    4.0%

    6.0%

    8.0%

    10.0%12.0%

    14.0%

    16.0%

    18.0%

    20.0%

    Credit Card Portfolio

    0

    0)(_ 0T

    T

    wageMinimum

    IncomeTMWINCOME =

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    33/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    34/65

    Social Housing Institute BehaviorThe cardholder has (or does not have = 0) a Social Housing Institute mortgage.

    Sistema

    Interval PD

    0 14.81%

    1 17.25%

    CREDIT (T0)

    % Default

    0.1510

    0.3581

    1.0000

    0.0000

    0.0000

    0.0000

    0.0000

    Frecuency %

    Default

    CREDIT (T0)

    0%

    20%

    40%

    60%

    80%

    100%

    0 10.0%

    2.0%

    4.0%

    6.0%

    8.0%

    10.0%

    12.0%

    14.0%

    16.0%

    18.0%

    Credit Card Portfolio

    AveStd Dev

    Max

    Q75Median

    Q25

    Min

    INFONAVIT is the Mexican Social Housing Institute (National Workers Housing Fund

    Institute) that gives mortgage credits and deducts the monthly payment from the worker

    salary.

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    35/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    36/65

    Maturity of the Credit Card in the Bank

    Age of the credit card measured in months

    Sistema

    Interval PD

    12 19.11%

    24 16.87%

    36 14.85%

    48 13.03%

    60 11.40%

    72 9.95%

    84 8.67%

    96 7.54%

    108 6.55%

    120 5.68%

    MAT (T0)

    % Default

    46.8745

    64.1524

    455.5333

    48.7000

    22.0000

    10.80000.2667

    Frecuency %

    Default

    MAT

    0%

    20%

    40%

    60%

    80%

    100%

    12

    24

    36

    48

    60

    72

    84

    96

    108

    120

    >120 0.0%

    5.0%

    10.0%

    15.0%

    20.0%

    25.0%

    Meses

    Credit Card Portfolio

    AveStd Dev

    Max

    Q75Median

    Q25

    Min

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    37/65

    Maturity in the Credit BureauAge on the credit bureau

    Sistema

    Interval PD

    12 24.91%

    24 22.96%

    36 21.11%

    48 19.38%

    60 17.75%

    72 16.24%

    84 14.83%

    96 13.52%

    108 12.31%

    120 11.20%

    MAT_CB (T0)

    % Default

    91.2057

    60.4804

    455.5333

    129.7333

    81.8000

    44.4000

    0.9000

    Frecuency %

    Default

    MAT_CB (T0)

    0%

    20%

    40%

    60%

    80%

    100%

    12

    24

    36

    48

    60

    72

    84

    96

    1

    08

    1

    20

    >1

    20

    0.0%

    5.0%

    10.0%

    15.0%

    20.0%

    25.0%

    30.0%

    Meses

    Credit Card Portfolio

    Ave

    Std DevMax

    Q75

    Median

    Q25

    Min

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    38/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    39/65

    Open AccountsNumber of existing credit cards at the reference point (Open credit cards, notclosed before the reference point).

    Sistema

    Interval PD0 13.61%

    1 13.82%

    2 14.03%

    3 14.24%

    4 14.46%

    5 14.68%

    6 14.91%

    7 15.13%

    8 15.36%

    9 15.59%10 15.83%

    ACCOUNTS TOT% Default

    6.9802

    5.057584.0000

    9.00006.0000

    3.0000

    0.0000

    Frecuency %

    Default

    ACCOUNTS_TOT

    0%

    20%

    40%

    60%

    80%

    100%

    0 1 2 3 4

    5ym

    s 0.0%

    5.0%

    10.0%

    15.0%

    20.0%

    25.0%

    30.0%

    35.0%

    40.0%45.0%

    Credit Card Portfolio

    AveStd Dev

    Max

    Q75Median

    Q25

    Min

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    40/65

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    41/65

    Delinquency in the Credit BureauIndicates if the cardholder presented in the previous 12 months delinquency onany other debt obligation different than the credit card.

    Sistema

    Interval PD

    0 6.59%

    1 24.89%

    DELINQ_CB_HIST

    % Default

    0.4754

    0.4994

    1.0000

    1.0000

    0.00000.0000

    0.0000

    Frecuency %

    Default

    DELINQ_CB_HIST

    0%

    20%

    40%

    60%

    80%

    100%

    0 1 0.0%

    5.0%

    10.0%

    15.0%

    20.0%

    25.0%

    30.0%

    Credit Card Portfolio

    AveStd Dev

    Max

    Q75Median

    Q25

    Min

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    42/65

    PDSelected Variables

    The selected variables in the model are:

    Current Non-Payment (ACT): Number of consecutive periods, up to the reference

    point, in which the cardholder has not covered the minimum payment.

    Historical Non-Payment (HIS): Number of periods in which the cardholder has not

    covered the minimum payment in the last 6 months.

    Percentage of payment (% PAY): Amount of payments made by the cardholder

    over the total balance at the reference point.

    Credit Limit Use (% USE): Percentage that represents the total balance from the

    credit limit at the reference point.

    Maturity (MAT): Number of months elapsed since the opening of the credit card

    to the reference point.

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    43/65

    Average Cardholder

    The average cardholder of the Credit Card Loans Portfolio has the followingvalue for each selected variable in the model.

    Credit CardPortfolio

    Actual Non-Payment (ACT) 0.24

    Historical Non-Payment (HIS) 0.72

    Maturity (MAT) 46.87

    Percentage of payment (%PAY) 36.95%

    Credit Limit Use (%USE) 48.51%

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    44/65

    PD

    Final Estimation of PD

    Receiver operating characteristic ROC = 86%

    Estimator Description Value

    Constant -2.9704

    C1 ACT +0.6730

    C2 HIS +0.4696

    C3 MAT -0.0075

    C4 % PAY -1.0217

    C5 % USE +1.1513

    INCUMP_90D_4 / Coeficientes estandarizados(Int. de conf. 95%)

    MORAMIN_SA

    MORAMIN_HIS

    ANTIG_T0 PJE_PAGO_T0

    PUSO_LINEA_T0

    -0.4

    -0.3

    -0.2

    -0.1

    0

    0.1

    0.2

    0.3

    0.4

    Variable

    Coeficientes

    estandarizados

    )*1513.1*0217.1*0075.0*4696.0*6730.09704.2( 54321

    1

    1CCCCC

    e

    ++++

    =)

    Standardized Coefficients

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    45/65

    Parameters Estimation

    Loss GivenDefault

    (LGD)

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    46/65

    LGD

    3 months

    Frequency of Classes

    The Loss Given Default was established at 81%.

    % Payment % ClassAverage

    Payment% Recovery

    3 months

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    47/65

    Estimation of Parameters

    Exposure atDefault

    (EAD)

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    48/65

    EAD

    Credit limit use at the time of analysis (horizontal axis) and the final exposure atthe time of default (vertical axis)

    0%

    500%

    1000%

    1500%

    2000%

    2500%

    3000%

    0% 50% 100% 150% 200% 250% 300%

    PUSO_LINEA_T0

    F

    actor=EAD/Saldo_

    T0

    Exposureatdefault/

    balancetoday

    %USE: Percentage that represents

    the balance at the reference point

    from the credit limit.

    Factor: Exposure at the date of

    default over the balance at the

    reference point.

    Balance_T0: Balance at the point of

    analysis (date of reference).

    %USE = Balance today / Credit Limit

    %USE is the same variable used to estimate PD

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    49/65

    EAD

    An EAD adjustment factor is deducted from data.

    Factor

    0% 5% 845%

    5% 10% 447%

    10% 15% 333%

    15% 20% 274%

    20% 25% 237%

    25% 30% 211%

    30% 35% 192%

    35% 40% 176%

    40% 45% 164%

    45% 50% 154%

    50% 55% 145%

    55% 60% 138%

    60% 65% 131%

    65% 70% 126%

    70% 75% 120%75% 80% 116%

    80% 85% 112%

    85% 90% 108%

    90% 95% 105%

    95% 100% 101%

    100% 105% 100%

    105% 110% 100%

    110% 115% 100%

    115% 120% 100%

    %USE

    EAD = Balance_T0* Factor

    0%

    100%

    200%

    300%

    400%500%

    600%

    700%

    800%

    900%

    1000%

    0% 20% 40% 60% 80% 100% 120% 140%

    % USE

    Fac

    tor

    { }%100,*0_ 5784.05CMaxTBalance

    C5 = Credit Limit Use (Balance_T0 / Credit Limit) at the reference point.

    %USE is the same variable used to estimate PD

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    50/65

    Final Proposal

    Current regulation for credit card reserve. Entered into effect september 2009

    Credit Card Provisionsi = PDi * LGD * EADi

    Where:

    =

    +< +++

    %1004

    1

    14

    1

    )5*1513.14*0217.13*0075.02*4696.01*6730.09704.2(1

    CSi

    eCSi CCCCC

    PD

    C1 = Number of consecutive periods in which the cardholder didnt cover the minimum payment at the reference point

    C2 = Number of periods in which the cardholder did not cover the minimum payment in the last 6 months

    C3 = Maturity of the credit card in the Institution at the reference point (months)

    C4 = Amount of payment made by the cardholder over the outstanding balance at the reference point

    C5 = Percentage of the outstanding balance at the reference point over the credit limit.

    %75=LGD

    { }%100,*0_ 5784.05 CMaxTBalanceEAD =

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    51/65

    Content

    1. Introduction

    2. Provisions Based on Expected Loss for a Revolving Credit Portfolio

    3. Risk Analysis of the Credit Card Loans Portfolio

    4. Conclusions

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    52/65

    Credit Card Loans Portfolio

    Interest SpreadThe formula allowed for loan level expected loss estimation. Expected loss iscompared to interest rate.

    5%

    10%

    15%

    20%

    25%

    30%

    35%

    40%

    Sep-05 Dec-05 Mar-06 Jun-06 Sep-06 Dec-06 Mar-07 Jun-07 Sep-07 De-07 Mar-08

    Expected Losses Spread

    Expected Losses were estimated for the Mexican credit card loans portfolio using the presented model for provisions.

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    53/65

    Spreads vs Expected LossesBased on the risk-return relationship found for each bank, 4 types of

    institutions were identified.

    0%

    10%

    20%

    30%

    40%

    50%

    60%

    0% 10% 20% 30% 40% 50% 60%

    Spread

    WELL

    ESTABLISHED

    MULTISEGMENT(low & high

    incokme)

    COMPETE IN NEW SECTOR

    PRICECOMPETITION

    InsufficientPrice

    Over-Priced

    Expected Losses

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    54/65

    Maturity and Expected LossesMore recent clients to the financial sector are riskier.

    The size of thecircle indicates theexpected loss

    EL %

    0

    20

    40

    60

    80

    100

    120

    0 20 40 60 80 100 120Maturity in the Financial System (Credit Bureau first record) (months)

    Average

    MaturityintheBank

    (months)

    LOYAL CREDITCARDHOLDERS

    HIGHCOMPETITION

    PRICECOMPETITIONOLD USERS

    CLIENTS THAT HAVENEVER HAD ACREDIT CARDBEFORE

    1/ Expected Loss is the average of the expected losses for the period between April 2006 and March 2007

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    55/65

    IRB model validation

    PD CNBV vs PD Bank X

    Modelo PD CNBV

    IRB

    PD CNBV vs PD BankY

    PD Model IRB

    PD CNBV

    Observed default rate

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    56/65

    Interest rate differentation according to riskcharacteristics of clients

    */ Ex-post (in the next 12 months after the horizontal axis computation)

    Outstanding balance / Credit limit

    Payment / outstanding balance

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    57/65

    PD Through the Cycle vs PD Point in Time

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    58/65

    Content

    1. Introduction

    2. Provisions Based on Expected Loss for a Revolving Credit Portfolio

    3. Risk Analysis of the Credit Card Loans Portfolio

    4. Conclusions

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    59/65

    Credit Card Loans PortfolioThe impact in the system was significant (average 2.14 times the previous

    requirement).

    0%

    10%

    20%

    30%

    40%

    50%

    60%

    70%

    BANK 1 BANK 2 BANK 3 BANK 4 BANK 5 BANK 6 BANK 7 BAN K 8 BANK 9 BANK 10 SISTEMA

    Incurred Loss Expected Loss

    Institution

    Expected Loss

    vs Incurred

    Loss

    BANK 1 2.13x

    BANK 2 1.8x

    BANK 3 1.75x

    BANK 4 2.14x

    BANK 5 2.61x

    BANK 6 2.53x

    BANK 7 2.26x

    BANK 8 3.25x

    BANK 9 1.56x

    BANK 10 2.02x

    PORTFOLIO 2.14x

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    60/65

    Capital Requirement and ProvisionsCredit Card Portfolio

    Basel I & incurred lossprotection

    Expected Loss

    18.4%

    BIS II (1-k=0.001)36.5%

    Provisions (Incurred Loss)9.3%

    BIS I (1-k=0.6487)17.3%

    9.3%

    18.12 %

    % portfolio

    FREC

    UENCY

    18.4%

    36.5%

    17.3%

    8.%

    Basel II & expectedloss protection

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    61/65

    Individual bank estimation

    Best bank

    Expected loss13.70%

    BIS II (1-k=0.001)27.49 %

    Reserves8.10%

    BIS I (1-k=0.6689)16.10%

    8% cartera

    13.79% cartera

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    62/65

    Individual bank estimation

    Worst bank

    Expected loss30.21%

    BIS II (1-k=0.001)54.11%

    Reserves15.84%

    BIS I (1-k=0.6689)23.84%

    8% cartera

    23.90% cartera

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    63/65

    Systemic PD, EAD, LGD as validationmethod

    In Mexico no bank had an internal model before the introduction of this rule

    because the incentive of capital reduction was the opposite.

    This method does not intend to substitute internal models. On the contrary it seeks

    to incentivize its use and development by setting a comparable standard as the

    standard method of reserves.

    One bank has certified its internal model in parallel with the introduction of this

    rule. Loan level comparisons were done and general parameters compared for

    sample portfolio. Differences in parameters were modest and subject to

    explanation.

    This approach allows a rich validation process as specific differences on PD

    estimation can be explained by bank specific policies which are made evident by

    comparing parameter estimates.

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    64/65

    Conclusions

    The National Banking and Securities Commission (CNBV) is gradually changing theregulation from an incurred loss to an expected loss provisioning scheme

    The first portfolio that changed was the revolving consumer loans (credit cards)

    which is now being provisioned with the presented expected loss model. In March

    2011 mortgage and personal loans were introduced. September 2011 saw the stateand municipalities reserve rule change and december 2011 is the objective date for

    corporate and SMEs loans (D&B score).

    The objectives of these reforms are to recognize losses in a timely manner, to

    assure that provisions cover losses for a 12 month horizon, to apply international

    standards and encourage banks to use more information in the process.

    This type of models also represent an incentive for the banks to develop their

    internal rating models for provisioning and capital assessment.

  • 7/28/2019 Alan Flores - TdC Sao Paulo 20111017

    65/65

    Systemic estimation of PD, LGD

    and EAD for credit card as areserve requirement and

    validation method.

    Mexican Experience