Download - Credit Score 02
-
8/8/2019 Credit Score 02
1/52
12/01/10 1Retail Decision Models
Group Risk - Retail Risk
Credit ScoringDevelopment and Methods
James Marinopoulos
Head of Retail Decision Model
-
8/8/2019 Credit Score 02
2/52
12/01/10Retail Decision Models
Group Risk - Retail Risk
Alan Greenspan:
President, Federal Reserve Board
May 1996
We should not forget that the basic economic function of these
regulated entities (banks) is to take risk. If we minimise risktaking in order to reduce failure rates to zero, we will, bydefinition, have eliminated the purpose of the bankingsystem.
-
8/8/2019 Credit Score 02
3/52
12/01/10Retail Decision Models
Group Risk - Retail Risk
Risk Families
We are managing different groups of Risk
Customer fails
to pay
Losing moneyWrong Strategy
Change in
market
prices
Processing failures and
frauds
Regulatory compliance
Customer fails
to pay
Losing moneyWrong Strategy
Change in
market
prices
Processing failures and
frauds
Regulatory compliance
-
8/8/2019 Credit Score 02
4/52
12/01/10 4Retail Decision Models
Group Risk - Retail Risk
Retail Decision Models Responsibilitiess Policy
Set Group policy on Decision Models Approve Decision Model policy changes
s Monitor, Validate and Approve New Scorecard Developments
Existing Scorecard Functionality
Proposed changes to Decision Models Processes
New Decision Models Systems functionality
Decision Models Systems functionality changes
s Governance Monitoring
Undertake bank validations, reports and presentations for APRA
s
Risk Measurement Set risk benchmarks for scorecards Risk grading models
s Advise Worlds best practice in Decision Models
Risk related issues surrounding Decision Models
-
8/8/2019 Credit Score 02
5/52
12/01/10 5Retail Decision Models
Group Risk - Retail Risk
RDM Structure and Responsibilities
G r a d u a tJ a n e t L o
S e n i o r D e c i s i o n M o l
( D e v e l o p m
Q u y e n P h a
M a n a g eD e c i s i o n M o l l i i
K a t h y Z o
M a n a g eD e c i s i o n M o l i i
V a l e n t i n a
G r a d u a tM a r i a D e m i
S e n i o r D e c i s i o n M o l
( V a l i d a t i o
N i c h o l a s Y i
S e n i o r S y s t e m s A s s
G r a e m e J
H e a d o f
R e t a i l D e c i s i l
J a m e s M a r i l
Relationship
Developments
Change Requests
SystemsOngoing Validations
Monitoring
Data Analysis
-
8/8/2019 Credit Score 02
6/52
12/01/10 6Retail Decision Models
Group Risk - Retail Risk
Presentation Topics
Scorecard Modelling
Scorecard Modelling
Business Objectives
Business Objectives
World Banks
World Banks
Monitoring
Monitoring
Future Direction
Future Direction
Overview of scoring
Overview of scoring
-
8/8/2019 Credit Score 02
7/5212/01/10 7
Retail Decision Models
Group Risk - Retail Risk
What is credit scoring?
s A statistical means of providing a quantifiable risk factor for a givencustomer or applicant.
s Credit scoring is a process whereby information provided is converted intonumbers that are added together to arrive at a score. (Scorecard)
s The objective is to forecast future performance from past behaviour.s Credit scoring developed by Fair & Isaac in early 60s
Widespread acceptance in the US in early 80s and UK early 90s FICO scores make 75% of US Mortgage loan decisions
Behavioural scoring accepted as more predictive than applicationscoring
s Decision Models are used in many areas of industries:
Banking and Finance
Insurance Retail
Telecommunicationss
-
8/8/2019 Credit Score 02
8/52
12/01/10 8Retail Decision Models
Group Risk - Retail Risk
Application Scoring
s Application scoring is a statistical means of assessing risk at the point ofapplication for credit
The application is scored once
s Application scoring is used for:
Credit risk determination
Loan amount approval
Limit setting
Credit
Decision
-
8/8/2019 Credit Score 02
9/52
12/01/10 9Retail Decision Models
Group Risk - Retail Risk
Behavioural Scoring
s Behavioural scoring is a statistical means of assessing risk for existing customersthrough internal behavioural data
Customers/accounts scored repeatedly
s Behaviour scoring is used for:
Authorisations
Limit increase/overdraft applications
Renewals/reviews Collection strategies
Risk
Grading
Debit $1344. 12
Debi t $234. 01
Debit $987.56
Debit $6543.22
Debit $32423.11
Total $2556.00
Debit $1344. 12
Deb it $234. 01
Debit $987.56
Debit $6543.22
Debit $32423.11
Total $2556.00
Debit $1344. 12
Debit $234. 01
Debit $987.56
Debit $6543.22
Debit $32423.11
Total $2556.00
-
8/8/2019 Credit Score 02
10/52
12/01/10 10Retail Decision Models
Group Risk - Retail Risk
Sample scorecard characteristics
Financial Assets
Liabilities
Monthly repayment
Total Monthly income
Bureau
No. of bureau defaults
Adverse ANZ behaviour
Application
Purpose of loan
Deposit
Security
s Characteristics used in scorecards are similar to those used intraditional judgemental lending, e.g.:
s The difference being that attributes within these characteristics are givenformal weights (scores) and added to produce a resulting score
Character
Time at current employment
Residential status
Time at current address
-
8/8/2019 Credit Score 02
11/52
12/01/10 11Retail Decision Models
Group Risk - Retail Risk
Scorecard points (example)
Residential statusOwner Renter LWP/Other
+25 -30 +10
Time in employment (years)
-
8/8/2019 Credit Score 02
12/52
12/01/10 12Retail Decision Models
Group Risk - Retail Risk
Other Types of Scoring
s Attritions Authorisations
s Recovery
s Response
s
Profitabilitys Customer
-
8/8/2019 Credit Score 02
13/52
-
8/8/2019 Credit Score 02
14/52
-
8/8/2019 Credit Score 02
15/52
12/01/10 15Retail Decision Models
Group Risk - Retail Risk
'Good/Bad' Discrimination
s The objective of a scorecard is to have characteristics whichdiscriminate between Good and Bad accounts with a sufficientlyhigh probability. Some characteristics are legally or ethically not used
s The score will be a measure of the probability of being a Good orBad performer.
s If the scorecard is performing well then the average scores of Badsare lower than the average scores of the Goods.
040
80
120
160
200
240
280
320
360
400
440
480
520
560
600
640
680
720
760
800
Score
Number
Of Clients
Goods
Bads
-
8/8/2019 Credit Score 02
16/52
12/01/10 16Retail Decision Models
Group Risk - Retail Risk
Performance Charts
s The Good/Bad Odds ateach score can bedetermined and plottedonto a Performance chart
0 4 0
8 0
1 2 0
1 6 0
2 0 0
2 4 0
2 8 0
3 2 0
3 6 0
4 0 0
4 4 0
4 8 0
5 2 0
5 6 0
6 0 0
6 4 0
6 8 0
7 2 0
7 6 0
8 0 0
Score
NumberOf Clients
Goods
Bads
8
1
Graph 2 - Log Odds Performance Chart
0
5
25
128
645
3250
16400
0 4 0
8 0
1 2 0
1 6 0
2 0 0
2 4 0
2 8 0
3 2 0
3 6 0
4 0 0
4 4 0
4 8 0
5 2 0
5 6 0
6 0 0
6 4 0
6 8 0
7 2 0
7 6 0
8 0 0
Good/BadO
dds
0
2
4
6
8
10
12
14
LogGBOs(Base
2)
8 to 1 2 to 13
-
8/8/2019 Credit Score 02
17/52
12/01/10 17Retail Decision Models
Group Risk - Retail Risk
Application Scorecard Construction
Flow Chart
Characteristic AnalysisMultivariate model build
Reject Inference
Statistical Analysis
Customised Scorecard
Product IdentificationFile Data AvailabilitySamplingData Extraction/Cost
Data Integrity
Set cut-off Score
Implementation
Validation
Generic Scorecard
External Data SourceScorecard Vendor
Outsourcing
Scorecard Monitoring
-
8/8/2019 Credit Score 02
18/52
-
8/8/2019 Credit Score 02
19/52
12/01/10 19Retail Decision Models
Group Risk - Retail Risk
Models
s Expert Systemss Decision Trees
s Linear Regression
s Logistic Regression has the following form:
s Neural Networks ==
k
j jjx
p
p01
ln ( )( )
=
=
+=
k
j jj
k
j jj
x
xp
0
0
exp1
exp
-
8/8/2019 Credit Score 02
20/52
12/01/10 20Retail Decision Models
Group Risk - Retail Risk
Model Build
0
0.2
0.4
0.6
0.8
1
0 200 400 600 800 1000
s The model is built on dichotomous data. In this case a 1 for Goodcustomers and a 0 for Bad customers.
-
8/8/2019 Credit Score 02
21/52
// Retail Decision Models
Group Risk - Retail Risk
Logistic Regression
0
0.2
0.4
0.6
0.8
1
0 200 400 600 800 1000
Good/Bad Probability
Logistic
Linear (Good/Bad Probability)
s The logistic regression fits the probability better than Linearregression.
-
8/8/2019 Credit Score 02
22/52
12/01/10 22Retail Decision Models
Group Risk - Retail Risk
Reject Inference and Validation
s Reject Inference Reject Inference is only necessary for scorecards were there is no
performance information for rejected applications
Applications that are rejected must be included in the final model.
Behavioural scorecards deal only in existing customers, therefore
do not require reject inference.s Validation
A randomly selected control group (hold out sample) or proxyportfolio to test the model.
-
8/8/2019 Credit Score 02
23/52
12/01/10 23Retail Decision Models
Group Risk - Retail Risk
Measures of discrimination
s Receiver Operating Curve (ROC) The Receiver Operating Curve is the area under the curve generated when
the cumulative Bads are plotted against the cumulative goods (Lorenz Curve).
s Gini coefficient (G) This discrimination measure is geometrically defined as the ratio of the area A
of the shaded semi-circular area to the area B of the triangle in the Lorenzdiagram.
s PH (percentage Good for 50% Bad) This is defined as the cumulative proportion of Goods up to the median value
of the Bads.
)1(2
1+= GROC
Gini.xlsGini.xls
http://gini.xls/http://gini.xls/http://gini.xls/http://gini.xls/http://gini.xls/http://gini.xls/ -
8/8/2019 Credit Score 02
24/52
-
8/8/2019 Credit Score 02
25/52
12/01/10 25Retail Decision Models
Group Risk - Retail Risk
Measures of discrimination (II)
s Discrimination measures should be determined for discreteattributes
Chi-Squared
Fico (Kullback Divergence)
i
iii
B
GBG ln)(100
Exp
ExpObs 2)(
Based on a book by SolomonKullback
Information Theory and Statistics
-
8/8/2019 Credit Score 02
26/52
12/01/10 26Retail Decision Models
Group Risk - Retail Risk
Issues for Successful Implementation
s Cultural Changes Requires top management support
s Operational process
Redesign to minimise manual intervention and maximise costsavings.
s Data Integrity
Quality of the overall decisions, and subsequently the Portfolio, isdependant upon the accuracy of the data input. The first time!
s Setting the Cut-off score correctly
-
8/8/2019 Credit Score 02
27/52
12/01/10 27Retail Decision Models
Group Risk - Retail Risk
Presentation Topics
Overview of scoringOverview of scoring
Scorecard ModellingScorecard Modelling
World BanksWorld Banks
MonitoringMonitoring
Future DirectionFuture Direction
Business ObjectivesBusiness Objectives
-
8/8/2019 Credit Score 02
28/52
12/01/10 28Retail Decision Models
Group Risk - Retail Risk
Business Objectives
s Increase consistency of lending decisions Consistent & unbiased treatment of applicant
Customers with the same details get the same score
Total management control over credit approval systems
Allows for loosening or tightening of lending through credit cycles
Potential increase in approvalss Reduce operating costs
Increase in automated processing
s Improve customer service
Fast and consistent decisions at application point
More appropriate limit and authorisation decisions
Reduction in collection actions on low risk accounts
Risk based allocation of credit limits and issue terms
-
8/8/2019 Credit Score 02
29/52
12/01/10 29Retail Decision Models
Group Risk - Retail Risk
Business Objectives (cont)
s Improved portfolio management Manage credit portfolios more effectively and dynamically
Better prediction of credit losses
Management ability to react to changes fast & accurately
Ability to measure & forecast impact of policy decisions
Quick and uniform policy implementation Improved Management Information Systems (MIS)
Permits MIS to be developed to assist business needs and marketingactivities
MIS can be fed back into future scorecard developments and collectionactivities
-
8/8/2019 Credit Score 02
30/52
12/01/10 30Retail Decision Models
Group Risk - Retail Risk
Presentation Topics
Overview of scoringOverview of scoring
Scorecard ModellingScorecard Modelling
Business ObjectivesBusiness Objectives
MonitoringMonitoring
Future DirectionFuture Direction
World BanksWorld Banks
-
8/8/2019 Credit Score 02
31/52
12/01/10 31Retail Decision Models
Group Risk - Retail Risk
World Banks
s ANZs European Banks
Banking market in Europe is restructuring
Banks are merging across country boundaries
s UK bank visits
Bank A - bank with many recent acquisitions
Bank B - bank dealing with mainly credit cards
Bank C - ex building society now owned by bank
Bank D - large diverse bank
s
National Australia Bank
W ld B k
-
8/8/2019 Credit Score 02
32/52
12/01/10 32Retail Decision Models
Group Risk - Retail Risk
Mortgages - Y Y - Y Y
Personal Loans Y - - Y Y Y
Current Accounts Y - Y - Y Y
Credit cards Y - - Y Y Y
LMI - In House In House - External External
Retail FUM ? 58b 47b 8b $100b+ $60b
Scorecards 20 -? App Scrds
1 Beh Scrds70 ? 50 (12)
App lication scorecards NewUnder
DevelopmentNew New All All
Behavioural scorecards Existing - Best 40%Existing
> 6 months oBooks
P roduct Just D evelope
Data Storage Adequate Good Good Good Good Average
BureauB & W
(Equifax)
B & W
(Equifax)
B & W
(Experian)
B & W
(Experian)
Black (Credit
Advantage)
Black (Credit
Advantage)
Scoring Modelling Staff 20+ 3 30+ ? 40+ 15
World BanksUK Banks AUS Banks
-
8/8/2019 Credit Score 02
33/52
12/01/10 33Retail Decision Models
Group Risk - Retail Risk
Bureauss Fair Isaac is the main bureaus in USA
White and Black data is supplied to and from all financial institution
s Fair Isaac (Equifax) and Experian are the two main bureaus in UK
White data is supplied to a financial institution if the supply to bureau
Currently few banks supply and receive white data
Mergers are leading most banks to look at this option
Fair Isaac is trying to beat Experian in having bureau scores in the UK
This is only possible when all banks supply white data
s Credit Advantage is used in Australia Provides Black data only
Linked with Decision Advantage (previously Equigen)
Bureau scores used for ANZ Small Business
We could use Dunn & Bradstreet for over $250k lending
s Baycorp is used in New Zealand
Provides Black data only Baycorp is also a collections agency
NZ puts the smallest amount lost as a default
s Baycorp and Credit Advantage have just merged
-
8/8/2019 Credit Score 02
34/52
12/01/10 34Retail Decision Models
Group Risk - Retail Risk
Country No ScoringData Collectio
Centralisatio
Generic
Scorecards
ApplicationScorecards
Only
BehaviouralScorecards -
Product Base
e avoura
Scorecards -Customer
Based
CustomerRelationship
Managemen Bureau
UK W & B
USA W & B
Canada W & B
South Africa B
Spain BAustralia B
New Zealand B
Italy B
Germany B
France -
Belgium -
Czech Repub lic -
Hong Kong B
Singapore -
Thailand -
India -
Korea -
Lebanon -
Saudi Arabia -
Credit Scoring & Bureaus Around the WorldWe are not alone!
B
BBBBB
B
-
8/8/2019 Credit Score 02
35/52
12/01/10 35Retail Decision Models
Group Risk - Retail Risk
BASEL - The New Accords The New Accord will give banks with sophisticated risk
management capabilities increased flexibilitys More emphasis on banks internal measures of risk, supervisory
review and market discipline
s Decision support technology has an important role to play
s Incentivise better risk management
s Data warehouses are fundamental to addressing many of therequirements
s SMB sector will be key
s More risk sensitive
s Competitive equality
Paul%20Russell%2013a[1]
The New Basel Capital AccordThe New Basel Capital Accord
Pillar 1 :Pillar 1 :Minimum capitalMinimum capital
requirementrequirement
Pillar 2 :Pillar 2 :SupervisorySupervisory
reviewreview
processprocess
Pillar 3 : MarketPillar 3 : Market
disciplinediscipline
-
8/8/2019 Credit Score 02
36/52
12/01/10 36Retail Decision Models
Group Risk - Retail Risk
Pillar 1 : credit risk
s Internal Rating Based (IRB) approach
Foundation
Bank sets Probability of Default (PD)
Standard Exposure At Default (EAD)
Standard Loss Given Default (LGD)
Advanced
Banks sets PD, EAD & LGD
s Better recognition of credit risk mitigation techniques
s Behavioural scoring
Internal
External
s Data storage
-
8/8/2019 Credit Score 02
37/52
12/01/10 37Retail Decision Models
Group Risk - Retail Risk
Future direction of scoring
s
Adaptive Control first implemented 1985 in USA Champion/Challenger processes for determining actions based on scores
Required 10 years to be widespread in USs Customer Relationship Management
Profitability (NIACC)
Attrition
Propensity to Buy (Cross Sell) Life time revenue
s Recovery scorecardss Operations Research Methods
Simulation modelling
-
8/8/2019 Credit Score 02
38/52
12/01/10 38Retail Decision Models
Group Risk - Retail Risk
Presentation Topics
Overview of scoringOverview of scoring
Scorecard ModellingScorecard Modelling
Business ObjectivesBusiness Objectives
World BanksWorld Banks
Future DirectionFuture Direction
MonitoringMonitoring
-
8/8/2019 Credit Score 02
39/52
12/01/10Retail Decision Models
Group Risk - Retail Risk
Monitoring Examples
s 1. Operation Stability Reports
The four types of front end monitoring reports:
1.1 Approval Statistics Report
1.2 Population Stability Report
1.3 System Rules Referral Report
1.4 Portfolio Statistics Report Operational statistics can be obtained as soon as an automated
decision process is implemented
Early warning indicators of decision functionality error andscorecard validity
Should be produced by Business Units or MIS
-
8/8/2019 Credit Score 02
40/52
12/01/10 40Retail Decision Models
Group Risk - Retail Risk
Loan Approval/Declines by Score
Approva/Declinal Rates by Score
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1000
Score Bands
Percentages
Auto Declined
Manually Declined
Manually Approved
Auto Approved
-
8/8/2019 Credit Score 02
41/52
12/01/10 41Retail Decision Models
Group Risk - Retail Risk
Population Stabilitys Compare each characteristic and attribute
over time against benchmarks
s Plot score distributions over time for potential change
s Indicates potential drift in performance
NO YES
Dec-96 25% 75%
Mar-97 23% 77%
Jun-97 24% 76%
Sep-97 22% 78%
Dec-97 21% 79%
Mar-98 19% 81%
Jun-98 19% 81%
Sep-98 22% 78%Dec-98 20% 80%
Mar-99 20% 80%
Jun-99 18% 82%
Sep-99 18% 82%
Dec-99 17% 83%
Benchmarks 29% 71%
Population Stability
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
NO YES
Dec-96
Mar-97
Jun-97
Sep-97
Dec-97
Mar-98
Jun-98
Sep-98
Dec-98
Mar-99
Jun-99
Sep-99
Dec-99
-
8/8/2019 Credit Score 02
42/52
12/01/10 Retail Decision Models
Group Risk - Retail Risk
Monitoring Requirements
s 2. Performance Analysis
The two types of back end monitoring are:
2.1 Scorecard Performance Report
2.2 Characteristic Analysis Report
2.3 Dynamic Delinquency Report
Performance Analysis is undertaken once a certain level ofcustomer maturity has been established
Should be produced by BU and Group Risk
-
8/8/2019 Credit Score 02
43/52
12/01/10 43Retail Decision Models
Group Risk - Retail Risk
Loans - Approval & Delinquency Rates
Even with manual assessment below the cut-off score of 350 thedelinquency rates are higher
Loans Approval & Delinquency Rates
0%
10%
20%
30%
40%
50%60%
70%
80%
90%
100%
1-300 301-350
351-400
401-450
451-500
501-550
551-600
601-650
651-700
701-750
751-800
>800
Score
ApprovalRates
0%
5%
10%
15%
20%
25%
Delinquen
cyRates
% Approved (LHS)
Delinquency Rates (RHS)
-
8/8/2019 Credit Score 02
44/52
12/01/10 44Retail Decision Models
Group Risk - Retail Risk
Scorecard Performances Scorecard performance based on 30+ delinquency
Good/Bad odds increase as expected by score
Score Distribution & G/B Odds
0
500
1000
1500
2000
2500
3000
3500
4000
1000
Score
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
Non Delinq
Delinq
HL GB Odds
-
8/8/2019 Credit Score 02
45/52
12/01/10 45Retail Decision Models
Group Risk - Retail Risk
Presentation Topics
Overview of scoringOverview of scoring
Scorecard ModellingScorecard Modelling
Business ObjectivesBusiness Objectives
World BanksWorld Banks
MonitoringMonitoring
Future DirectionFuture Direction
-
8/8/2019 Credit Score 02
46/52
12/01/10 46Retail Decision Models
Group Risk - Retail Risk
Future Direction
s
Modellings Experimental Design
Champion/Challenger Strategies
Hypothesis testing (uni & multi- dimensional)
s Quality Control Techniques
Control Chartss Operations Research
Optimisation techniques
Simulation Models
Stress Testing
-
8/8/2019 Credit Score 02
47/52
12/01/10 47Retail Decision Models
Group Risk - Retail Risk
Conferencess Fair Isaac and Experian are the two main credit scoring companies world wide
s Fair Isaac (Every year, alternating in Europe and USA)
Main bureau and FICO Scores in USA
Equifax in UK
Systems included TRIAD
Conference was mainly selling FICO products and systems (but also Technical)
s Experian (Every year, in Europe)
Formerly CCN
Systems include Transact and Hunter
Conference on world wide banking, financial, telecommunications and predictivemodelling usage (Business and/or Management)
s University of Edinburgh (Every 2 year in Edinburgh)
Very technical academic papers
Proposal to run alternate years in a USA university
-
8/8/2019 Credit Score 02
48/52
12/01/10 48Retail Decision Models
Group Risk - Retail Risk
lowlow
highhigh
highhigh
E[Vo lu
me
]
E[Volu
me
]
Three Portfolio Dimensions:
Volume, Loss, and ProfitLowLow
cutoffscutoffs
HighHigh
cutoffscutoffs
-
8/8/2019 Credit Score 02
49/52
12/01/10 49Retail Decision Models
Group Risk - Retail Risk
Efficient Frontiers in two dimensions
OP
HighCutoffs
E[Volume]
E[Loss]
LowCutoffs
0.60.6
0.00.00.20.2
LowCutoffs
HighCutoffs
E[Profit]
E[Loss]
OP
0.90.90.60.6
0.00.00.20.2 0.60.6
HighCutoffs
LowCutoffs
OP
E[Volume]
E[Profit]
0.60.6
0.20.2
0.20.2 0.90.9
Efficient Frontier
-
8/8/2019 Credit Score 02
50/52
12/01/10 50Retail Decision Models
Group Risk - Retail Risk
Improved portfolio performance
OP
HighCutoffs
E[Volume]
E[Loss]
LowCutoffs
0.60.6
0.00.00.20.2
LowCutoffs
HighCutoffs
E[Profit]
E[Loss]
OP
0.90.90.60.6
0.00.00.20.2 0.60.6
HighCutoffs
LowCutoffs
OP
E[Volume]
E[Profit]
0.60.6
0.20.2
0.20.2 0.0.9
Single Score
CombinedScores
Single Score
CombinedScores
Single Score
CombinedScores
Efficient Frontier
-
8/8/2019 Credit Score 02
51/52
-
8/8/2019 Credit Score 02
52/52
Other Techniques
s
Customer Relation Managements Survival Analysis
s Multiple Indicator Multiple Cause
Proportional Hazards.ppt
Measuring Customer Quality.doc