Credit Intelligence & Modelling: Many Paths through the Forest of Credit Rating and Scoring

Cover
Credit Intelligence & Modelling: Many Paths through the Forest of Credit Rating and Scoring
Copyright
Dedication
Foreword
	Shifting Seas
	The Toolkit
	Forest Paths
Acknowledgements
Language & Syntax
	Presentation—Warnings
	Kindle e-Book—Warnings
Table of Contents
Module A: Introduction
1: Credit Intelligence
	1.1 Debt versus Credit
	1.2 Intelligence!
		1.2.1 Individual Intelligence
		1.2.2 Collective Intelligence
		1.2.3 Intelligence Agencies
		1.2.4 Intelligence Cycle
	1.3 The Risk Lexicon
		1.3.1 What is . . .?
			1.3.1.1 Credit Intelligence
			1.3.1.2 Credit risk
		1.3.2 The Risk Universe
			1.3.2.1 Business Risk
			1.3.2.2 Risk by Nature
			1.3.2.3 Rumsfeld Matrix
			1.3.2.4 Black Swans and Other Strange Creatures
		1.3.3 Measure
			1.3.3.1 Assess, Measure, Communicate
			1.3.3.2 Time Horizons
			1.3.3.3 Desired Rating Properties
		1.3.4 Beware of Fallacies
			1.3.4.1 Argument
			1.3.4.2 Evidence
			1.3.4.3 Appeals
	1.4 The Moneylender
		1.4.1 Credit’s 5 Cs
		1.4.2 Borrowings and Structure
		1.4.3 Engagement
			1.4.3.1 Relationship Lending
			1.4.3.2 Transactional
			1.4.3.3 Providers—Benefits or Not
			1.4.3.4 Customers—Benefits or Not
		1.4.4 Retail versus Wholesale
			1.4.4.1 Wholesale
			1.4.4.2 Retail
			1.4.4.3 Grade Presentation
			1.4.4.4 From Rules and Judgment to Models
			1.4.4.5 Empirical Ratings
		1.4.5 Risk-Based Pricing (RBP)
	1.5 Summary
2: Predictive Modelling Overview
	2.1 Models
		2.1.1 Model Types and Uses
		2.1.2 Choices and Elements
			2.1.2.1 Functional Form
			2.1.2.2 Methodology
			2.1.2.3 Parameter Estimation
		2.1.3 Model Lifecycle
	2.2 Model Risk (MR)
		2.2.1 Categories
		2.2.2 Management
		2.2.3 Models on Models
	2.3 Shock Events
		2.3.1 Past Events
		2.3.2 COVID-19 Views
	2.4 Data
		2.4.1 Desired Qualities
		2.4.2 Sources
			2.4.2.1 Homophily
		2.4.3 Types
			2.4.3.1 Honouring Obligations
			2.4.3.2 Transactions
			2.4.3.3 Non-Financial Behaviour
			2.4.3.4 Disclosed
			2.4.3.5 Investigation
			2.4.3.6 Providing Comfort
	2.5 Summary
3: Retail Credit
	3.1 Scorecard Terminology
		3.1.1 Targeting Rare Events
		3.1.2 Functional Forms
	3.2 Retail Model Types
		3.2.1 When?—Credit Risk Management Cycle
		3.2.2 What?—The Four Rs of Customer Measurement
		3.2.3 Who?—Experience, To Borrow or Not To Borrow
		3.2.4 How?—Empirical versus Judgment
		3.2.5 The Commonest Types
	3.3 Data Sources
		3.3.1 Credit Bureaux
		3.3.2 Ownership Types
		3.3.3 Credit Registries
	3.4 Risk ‘Indicators’
		3.4.1 Types of Risk Indicators
		3.4.2 Banding Presentation
	3.5 FICO Scores
		3.5.1 Scaling Parameters
	3.6 Summary
4: Business Credit
	4.1 Risk 101
		4.1.1 Credit Risk Analysis
		4.1.2 Data Sources
		4.1.3 Risk Assessment Tools
		4.1.4 Rating Grades
			4.1.4.1 External Grades
			4.1.4.2 Standard & Poor’s (S&P)
			4.1.4.3 Internal Grades
			4.1.4.4 Master Rating Scales
		4.1.5 SME (Small and Medium Enterprises) Lending
	4.2 Financial Ratio Scoring
		4.2.1 Pioneers
		4.2.2 Predictive Ratios
		4.2.3 Agency Usage
		4.2.4 Moody’s RiskCalcTM
		4.2.5 Non-Financial Factors
	4.3 Use of Forward-Looking Data
		4.3.1 Historical Analysis
		4.3.2 Structural Models
		4.3.3 Reduced-Form Models
			4.3.3.1 Credit Spreads
	4.4 Summary
	Module B: The Histories
5: Side Histories
	5.1 The Industrial Revolutions
		5.1.1 Authors and Players
		5.1.2 Further Details
		5.1.3 Implications
	5.2 Booms and Busts; Bubbles and Bursts
		5.2.1 17th Century
		5.2.2 18th Century
		5.2.3 19th Century
			5.2.3.1
			5.2.3.2 (1857–59) Panic of 1857
		5.2.4 (1873–96) The Long Depression
		5.2.5 20th Century
		5.2.6 21st Century
	5.3 Registration
		5.3.1 Social Relationships
			5.3.1.1 Hierarchies
			5.3.1.2 Obligations
		5.3.2 In History
			5.3.2.1 Ancient China
			5.3.2.2 Ancient Rome
			5.3.2.3. Early-Modern Europe
		5.3.3 Evidence
			5.3.3.1 Passports
			5.3.3.2 References
			5.3.3.3 Certificates
			5.3.3.4 Tokens
	5.4 Identification
		5.4.1 Visual
			5.4.1.1 Traces
			5.4.1.2 Worn and Borne
			5.4.1.3 Mannerisms
		5.4.2 Oral
			5.4.2.1 Voice
			5.4.2.2 Language and Accent
			5.4.2.3 Unique Knowledge
		5.4.3 Disclosed
			5.4.3.1 Names
			5.4.3.2 Numbers
		5.4.4 Authenticators
		5.4.5 Invasive
	5.5 Summary
6: Credit—A Microhistory
	6.1 The Ancient World
		6.1.1 Mesopotamia
		6.1.2 Greece
		6.1.3 Roman Empire
	6.2 The Mediaeval World
		6.2.1 Early Middle Ages
		6.2.2 Churches and Holy Men
		6.2.3 Pawnbroking
		6.2.4 Vifgage and Morgage
		6.2.5 Merchant Banking
		6.2.6 Bankruptcy Legislation—16th through 18th Centuries
	6.3 Credit Evolution
		6.3.1 Trade Finance and Investment
		6.3.2 Personal Credit—Pre-1880
		6.3.3 Personal Credit—1880s Onwards
		6.3.4 Instalment Credit
	6.4 Credit Vendors
		6.4.1 Tallymen, Credit Drapers and Travelling Salesmen
		6.4.2 Department Stores
		6.4.3 Mail Order
		6.4.4 Mobile Network Operators (MNO)s
		6.4.5 Internet Service Providers
	6.5 Credit Media and Assets Financed
		6.5.1 Promissory Note and Bill of Exchange
		6.5.2 Cheques and Overdrafts
			6.5.2.1 Cheques
			6.5.2.2 Overdrafts
		6.5.3 Charge and Credit Cards
		6.5.4 Car Loans and Consumer Durables
		6.5.5 Home Loans
		6.5.6 Student Loans
	6.6 Summary and Reflections
	Questions—History of Credit
7: The Birth of Modern Credit Intelligence
	7.1 Pre-Revolution
	7.2 United Kingdom
	7.3 United States
		7.3.1 Early America
			7.3.1.1 Mercantile Reporting Agencies
			7.3.1.2 The Tappans
			7.3.1.3 Mercantile to 1859
			7.3.1.4 Dun versus Bradstreet to 1933
			7.3.1.5 19th-Century Operations
			7.3.1.6 Publishing and Rating
			7.3.1.7 Dun & Bradstreet
		7.3.2 Credit Men and Information Exchanges
		7.3.3 Credit Bureaux
	7.4 The ‘Big Three’ Credit Bureaux, Plus Some
		7.4.1 Equifax
		7.4.2 Experian
			7.4.2.1 Commercial Credit Nottingham (UK)
			7.4.2.2 Thompson Ramo Wooldridge (USA)
			7.4.2.3 Experian
		7.4.3 TransUnion
		7.4.4 Centrale Rischi Finanziari (CRIF)
		7.4.5 CreditInfo
		7.4.6 Others
		7.4.7 Current Spread
		7.4.8 Economics and Statistics
	7.5 Rating Agencies
	7.6 High-Level Observations
	7.7 Summary and Reflections
	Questions—History of Credit Intelligence
8: The Dawn of Credit Scoring
	8.1 Before Statistics
	8.2 Statistical Experiments: 1941–1958
	8.3 Rise of the Scorecard Vendor
		8.3.1 Fair, Isaac & Co. (FICO)
		8.3.2 VantageScore Solutions
		8.3.3 Management Decision Systems (MDS)
		8.3.4 Scorelink and Scorex
	8.4 Rise of the Corporate Modeller
		8.4.1 JP Morgan
		8.4.2 Kealhofer McQuown Vašícek (KMV)
		8.4.3 Moody’s Analytics
	8.5 Regulation
		8.5.1 Privacy—Fair Credit Reporting Act (FCRA) (1970)
		8.5.2 Privacy—OECD and European Legislation
		8.5.3 Anti-Discrimination Equal Credit Opportunity Act
		8.5.4 Capital Requirements—Basel II, III, IV
		8.5.5 Accounting—International Financial Reporting Standards (IFRS)
	8.6 Borrowed Concepts
	8.7 Statistical Methods
		8.7.1 Linear Programming (FICO)
		8.7.2 Discriminant Analysis
		8.7.3 Linear Probability Modelling (LPM)
		8.7.4 Logistic Regression (Independents and Others)
		8.7.5 Neural Networks
		8.7.6 Other Non-Parametric Techniques
	8.8 Summary and Reflections
	Questions—History of Credit Scoring
	Module C: Credit Lifecycle
	Shared Service Centres (SSCs)
9: Front-Door
	9.1 Marketing
		9.1.1 Advertising
		9.1.2 Two Tribes
		9.1.3 Pre-Screening
		9.1.4 Data
		9.1.5 Summary
	Questions—Marketing
	9.2 Origination
		9.2.1 Gather—Interested Customer Details
			9.2.1.1 Acquire Applicant Details
			9.2.1.2 Paper-based capture
			9.2.1.3 Pre-scoring screening and sanitation
		9.2.2 Sort—Into Strategy Buckets
			9.2.2.1 Enquire—Internal
			9.2.2.2 Enquire—External
			9.2.2.3 Measure and Decide
		9.2.3 Action—Accept or Reject
			9.2.3.1 Declines
			9.2.3.2 Accepts
		9.2.4 Summary
	Questions—Originations
	9.3 Account Management
		9.3.1 Types of Limits
			9.3.1.1 Use of Other Scores
			9.3.1.2 Triage
		9.3.2 Over-Limit Management (Takers)
			9.3.2.1 Cheque Accounts—Pay/No Pay
			9.3.2.2 Credit Cards—Authorizations
			9.2.3.3 Informed Customer Effect
		9.3.3 More Limit and Other Functions
			9.3.3.1 Limit-Increase Requests (Askers)
			9.3.3.2 Limit-Increase Campaigns (Givers)
			9.3.3.3 Limit Reviews (Repeaters)
			9.3.3.4 Cross-Sales (Repayers/Repeaters/Leavers)
			9.3.3.5 Win-Back (Leavers)
		9.3.4 Summary
	Questions—Account Management
10: Back-Door
	10.1 Collections and Recoveries (C&R)
		10.1.1 Overview
			10.1.1.1 Delinquency Reasons
			10.1.1.2 Excuses
		10.1.2 Process
			10.1.2.1 Core Systems Requirements
			10.1.2.2 Agencies
			10.1.2.3 Reporting
		10.1.3 Triggers and Strategies
			10.1.3.1 Strategy Setting
			10.1.3.2 Practical Considerations
		10.1.4 Modelling
			10.1.4.1 Collections and Recoveries (C&R) versus Behavioural
			10.1.4.2 Collections Scorecard Classifications
			10.1.4.3 Champion/Challenger
		10.1.5 Summary
	Questions—Collections
	10.2 Fraud
		10.2.1 Credit Card Fraud Trends
		10.2.2 Definitions
			10.2.2.1 Relationship to Account
			10.2.2.1.1 Kite Flying
			10.2.2.2 Misrepresentation
				10.2.2.2.1 Embellishment/Massaging
				10.2.2.2.2 Social Engineering
				10.2.2.2.3 Identity Theft and Synthetic Identities
				10.2.2.2.4 Property Hijacking
				10.2.2.2.5 Advanced Persistent Threat
			10.2.2.3 Authorization
				10.2.2.3.1 Unauthorized Fraud
				10.2.2.3.2 Authorized Fraud
		10.2.3 Prevention Measures
			10.2.3.1 Manual/Physical Measures
			10.2.3.2 Online/Telephonic Measures
				10.2.3.2.1 Knowledge
				10.2.3.2.2 Visual Biometrics
				10.2.3.2.3 Other Biometrics
				10.2.3.2.4 Tokens
				10.2.3.2.5 Multi-Factor
		10.2.4 Data and Tools
		10.2.5 Summary
	Questions—Fraud
	Module D: Toolbox
11: Stats & Maths & Unicorns
	11.1 Variance and Correlations
		11.1.1 Variance
		11.1.2 Pairwise Correlations
			11.1.2.1 Causation versus Coincidence
			11.1.2.2 Why are Correlations an Issue?
			11.1.2.3 Measures and Thresholds
			11.1.2.4 Variable Types
		11.1.3 Pearson’s Product-Moment
		11.1.4 Spearman’s Rank-Order
		11.1.5 Mahalanobis Distance
		11.1.6 Variance Inflation Factor (VIF)
			11.1.6.1 Greek, Damn Greek and Statistics
			11.2.1 Coefficient of Determination (R-squared
	11.2 Goodness-of Fit Tests
		11.2.1 Coefficient of Determination (R-squared)
		11.2.2 Pearson’s Chi-Square
		11.2.3 Hosmer–Lemeshow Statistic
	11.3 Likelihood
		11.3.1 Log-Likelihood
		11.3.2 Deviance
		11.3.3 Akaike Information Criterion (AIC)
		11.3.4 Bayesian Information Criterion (BIC)
	11.4 Holy Trinity
		11.4.1 Likelihood Ratio
			11.4.1.1 Wilks’ Theorem
			11.4.2 Wald Chi-Square
			11.4.3 Rao’s Score Chi-Square
	11.5 Summary
	Questions—Stats & Maths & Unicorns
12: Borrowed Measures
	12.1 Mathematics and Probability Theory
		12.1.1 Logarithms
			12.1.1.1 Archimedes
			12.1.1.2 Napier, Briggs and Bürgi
			12.1.1.3 Bernoulli and Euler
		12.1.2 Laws of Large Numbers
		12.1.3 Bayes’ Theorem
		12.1.4 Laplace—Expected Values
		12.1.5 Kolmogorov–Smirnov—Curve and Statistic
		12.1.6 Gradient Descent
	12.2 Probability Distributions and Hypotheses
		12.2.1 Binomial Distribution
		12.2.2 Normal Distribution and Z-Scores
		12.2.3 Student’s t-Distribution
		12.2.4 Verhulst’s Logistic Curve
		12.2.5 Pearson’s Chi-Square Distribution
	12.3 Economics
		12.3.1 Lorenz Curve
		12.3.2 Gini Coefficient
		12.3.3 Gini Impurity Index
	12.4 Information Theory and Cryptography
		12.4.1 Shannon’s Entropy
		12.4.2 Gudak—Weight of Evidence (WoE)
		12.4.3 Kullback—Divergence Statistic
	12.5 Signal-Detection Theory
		12.5.1 Confusion Matrices
		12.5.2 Receiver Operating Characteristic (ROC)
		12.5.3 Area under the ROC (AUROC or AUC)
	12.6 Forecasting
		12.6.1 Markov Chains
		12.6.2 Survival Analysis
	12.7 Summary
	Questions—Borrowed Measures
13: Practical Application
	13.1 Characteristic Transformations
		13.1.1 Rescale
		13.1.2 Discretize
			13.1.2.1 Dummy Variables
			13.1.2.2 Piecewise
			13.1.2.3 Weight of Evidence (WoE)
	13.2 Characteristic Assessments
		13.2.1 Information Value (IV)
		13.2.2 Population Stability Index (PSI)
		13.2.3 Chi-Square (.2)
	13.3 Model Assessments
		13.3.1 Lorenz and Gini
		13.3.2 Cumulative Accuracy Profile, Accuracy Ratio and Lift
		13.3.3 Divergence Statistic
	13.4 Odds and Sods
		13.4.1 Deviance Odds
		13.4.2 Calinski–Harabasz Statistic
		13.4.3 Gini Variance
	13.5 Summary
	Questions—Power, Separation and Accuracy
14: Predictive Modelling Techniques
	14.1 A View from on High!
		14.1.1 Caveats
		14.1.2 Learning the Language
	14.2 Parametric
		14.2.1 Linear Regression
		14.2.2 Discriminant Analysis
		14.2.3 Linear Probability Modelling (LPM)
		14.2.4 Probability Unit (Probit)
		14.2.5 Logistic Regression (Logit)
		14.2.6 Linear Programming
			14.2.6.1 LP for Classification
	14.3 Non-Parametric
		14.3.1 K-Nearest Neighbours
		14.3.2 Decision Trees
			14.3.2.1 Bootstrap Aggregation and Random Forests (RF)s
		14.3.3 Support Vector Machines (SVM)
		14.3.4 Artificial Neural Networks
		14.3.5 Genetic Algorithms
	14.4 Conglomerations
		14.4.1 Multiple Models
			14.4.1.1 Practical
			14.14.1.2 Parallel
			14.4.1.3 Sequential
		14.4.2 Machine Learning
	14.5 Making the Choice
	14.6 Summary
	QUESTIONS—Predictive Modelling Techniques
	Module E: Organizing
15: Project Management
	15.1 Development Process Overview
		15.1.1 Initiation
		15.1.2 Preparation
		15.1.3 Construction
		15.1.4 Finalization
	15.2 Initiation and ‘Project Charter’
		15.2.1 High-Level
			15.2.1.1 Model register
		15.2.2 Making the Case
			15.2.2.1 Business Case
			15.2.2.2 Scope
			15.2.2.3 Assessment Criteria
		15.2.3 Stakeholders and Players
			15.2.3.1 Sponsor and Steering Committee
			15.2.3.2 Project Manager (PM)
			15.2.3.3 Team Lead (TL)
			15.2.3.4 Team Members
		15.2.4 Resources and Timetables
			15.2.4.1 Resources
			15.2.4.2 Project Timetable
		15.2.5 Assumptions, Risks and Constraints
			15.2.5.1 Target Specification
			15.2.5.2 Model Form
			15.2.5.3 Data Sources
			15.2.5.4 Environmental Instability
	15.3 Project Deliverables
		15.3.1 Communication and Documentation
		15.3.2 Model Development Documentation (MDD)
		15.3.3 Implementation Instructions (MIID)
		15.3.4 Project Code
		15.3.5 Data
	15.4 Other Considerations
		15.4.1 Scorecard Development Software
			15.4.1.1 Statistical Analysis System (SAS)
			15.4.1.2 World Programming System (WPS) Analytics
			15.4.1.3 Statistical Package for the Social Sciences (SPSS)
			15.4.1.4 R
			15.4.1.5 Python
			15.4.1.6 Other Proprietary Tools
		15.4.2 Implementation
			15.4.2.1 Decision-Making Stages
			15.4.2.2 Technology Options
			15.4.2.3 Vendors
		15.4.3 Next Steps
	15.5 Summary
16: Data Acquisition—Observation
	16.1 Make a Plan!
	16.2 Gather
		16.2.1 Key Fields
		16.2.2 Matching Keys
		16.2.3 Data Aggregation
		16.2.4 Retention Rules
			16.2.4.1 Retrospective Histories
	16.3 Reduce
		16.3.1 Characteristic Review
		16.3.2 Proscribed Characteristics
		16.3.3 Unand Under-Populated Characteristics
		16.3.4 Correlated Characteristics
	16.4 Cleanse
		16.4.1 Out-of Scope
		16.4.2 Underpopulated
		16.4.3 Duplicates
		16.4.4 Outliers
		16.4.5 Inconsistencies
	16.5 Check
	16.6 Summary
17: Data Acquisition—Performance
	17.1 Planning Extraction
		17.1.1 Minimum Requirements
		17.1.2 Casting the Net
		17.1.3 Basic Checks
	17.2 File Preparation and Review
		17.2.1 Deep Dives of Simple Sorts
		17.2.2 Performance Arrays
		17.2.3 Payment Profile Strings
		17.2.4 Performance Maintenance
	17.3 Window Setting
		17.3.1 Length
		17.3.2 End-of-versus Worst-of-Window
		17.3.3 Fixed vs Variable
	17.4 Summary
18: Target Definition
	18.1 Overview
		18.1.1 Binaries
		18.1.2 Requirements
		18.1.3 Performance Components
		18.1.4 Code Crosschecks
	18.2 Definition Strictness
		18.2.1 Status nodes
		18.2.2 Level of Delinquency
			18.2.2.1 Time 0 vs Time 1
			18.2.2.2 Time 1 vs Time 2
		18.2.3 Trivial Balances
		18.2.4 Closed Accounts
	18.3 Integrity Checks
		18.3.1 Consistency Check
		18.3.2 Characteristic Check
		18.3.3 Swap-Set Check
	18.4 Summary
19: File Assembly
	19.1 Merge Observation and Performance
		19.1.1 Finding Performance
		19.1.2 Outcome Field Merge
		19.1.3 Kill and Other Rules
		19.1.4 Not Taken Up (NTU), Uncashed
	19.2 External Data Acquisition
		19.2.1 Retro History Requests
		19.2.2 Data Security
	19.3 Further Reduction
		19.3.1 Pre-Processing
		19.3.2 Correlated Characteristics
			19.3.2.1 Weak Characteristics
	19.4 Summary
	Module F: Packing
20: Sample Selection
	20.1 Overview
		20.1.1 Terminology
		20.1.2 Optimal and Minimum Sample Sizes
		20.1.3 Law of Diminishing Data Returns
	20.2 Training, Holdout, Out-of Time Recent (THOR) Samples
		20.2.1 Sample Types
		20.2.2 Sampling Guidelines
			20.2.2.1 Training
			20.2.2.2 Hold-out
			20.2.2.3 Out-of Time
			20.2.2.4 Recent
		20.2.3 Observation Windows
		20.2.4 Sampling Plan and Outcome
	20.3 Afterthoughts
		20.3.1 Unand Under-Populated Characteristics
		20.3.2 Exact Random Sample
		20.3.3 Housekeeping
	20.4 Summary
21: Data Transformation
	21.1 Traditional Transformations
		21.1.1 Dummy Variables
		21.1.2 Weight of Evidence
		21.1.3 Piecewise
	21.2 Classing/Binning
		21.2.1 Characteristic Analysis Reports
		21.2.2 Bulk Classing
		21.2.3 Fine Classing
		21.2.4 Coarse Classing
			21.2.4.1 Class Sizes
			21.2.4.2 Monotonicity
			21.2.4.3 Automation
			21.2.4.4 Training versus Hold-Out and Out-of Time
			21.2.4.5 Known versus Inferred
			21.2.4.6 Final Checks
		21.2.5 Piecewise Classing
		21.2.6 Final Transformation
	21.3 Missing Data Treatment
		21.3.1 Traditional
		21.3.2 Missing Singles
		21.3.3 Missing Multiples
	21.4 Summary
	Questions—Data Transformation
22: Segmentation
	22.1 Overview
		22.1.1 Drivers
			22.1.1.1 Common Splits
		22.1.2 Inhibitors
		22.1.3 Mitigators
	22.2 Analysis
		22.2.1 Learning Types
		22.2.2 Finding Interactions
			Open versus closed form
			Open form for binaries
			Closed form for binaries
		22.2.3 Segment Mining
		22.2.4 Boundary Analysis
			22.2.4.1 Boundary Types
			22.2.4.2 An Example
	22.3 Presentation
	22.4 Summary
	Questions—Segmentation
23: Reject-Inference
	23.1 The Basics
		23.1.1 Pointers
		23.1.2 Missing at Random, or Not
		23.1.3 Terminology
		23.1.4 Characteristic Analysis
		23.1.5 Swap-SetAnalysis
			23.1.5.1 Score Level
			23.1.5.2 Characteristic Level
		23.1.6 Population-FlowDiagram
	23.2 Intermediate Models
		23.2.1 Accept/Reject
		23.2.2 Taken Up/Not Taken Up (TU/NTU)
		23.2.3 Known Good/Bad
		23.2.4 Bringing it All Together
	23.3 The Inference Smorgasbord
		23.3.1 Supplementation
		23.3.2 Performance Surrogates
			23.3.2.1 It’s a Bird, it’s a Plane; no, it’s . . . Super Model!
		23.3.3 Reject Equals Bad
		23.3.4 Augmentation
		23.3.5 Weight of Evidence (WoE) Adjustments
		23.3.6 Iterative Reclassification
		23.3.7 Extrapolation
	23.4 Favoured Technique
		23.4.1 Fuzzy-Parcelling
		23.4.2 Extrapolation
			23.4.2.1 Raw Inference
			23.4.2.2 Extra Prejudice
		23.4.3 Attribute-Level Adjustments
	23.5 Let’s Get Practical!
		23.5.1 Variable Names and Codes
		23.5.2 Record-Level Inference Example
	23.6 Summary
	Questions—Reject-Inference
	Module G: Making the Trip
24: Model Training
	24.1 Regression
		24.1.1 Options and Settings
		24.1.2 Regression Outputs
	24.2 Variable Selection
		24.2.1 Criteria
		24.2.2 Automated Variable Selection (AVS)
		24.2.3 Stepwise Output Review
			24.2.3.1 Stepping Summary
			24.2.3.2 Model Coefficients
		24.2.4 Constraining the Beta Beast
			24.2.4.1 Negative Coefficients—Beta < 0
			24.2.4.2 Overprediction—Beta > 1
		24.2.5 Stepping by Gini
	24.3 Correlation and Multi-Collinearity
		24.3.1 Multi-Collinearity
		24.3.2 Pairwise Correlations
	24.4 Blockwise Variable Selection
		24.4.1 Variable Reduction Blocks
		24.4.2 Staged Blocks (Residual Prediction)
		24.4.3 Embedded Blocks
		24.4.4 Ensemble Blocks
	24.5 Multi-ModelComparisons
		24.5.1 Lorenz Curve Comparisons
		24.5.2 Strategy Curve Comparisons
	24.6 Model Calibration
		24.6.1 Simple Calibration
		24.6.2 Piecewise Calibration
		24.6.3 Score and Points CalibrationThe previous section assumes we are adjusting predictions provided
		24.6.4 MAPA Calibration
	24.7 Summary
	Questions—Model Training
25: Scaling and Banding
	25.1 Scorecard Scaling
		25.1.1 Background
		25.1.2 Percentages
		25.1.3 Fixed Ranges
		25.1.4 Scaling Parameters
			25.1.4.1 Basic Formulae
			25.1.4.2 Intercepts and Coefficients → Scorecard Points
			25.1.4.3 Log-Odds↔ Probability ↔ Score
		25.1.5 Other Considerations
			25.1.5.1 Scorecard Presentation
			25.1.5.2 Adverse Reason Codes
	25.2 Risk Banding
		25.2.1 Zero Constraints
		25.2.2 Fitted Distributions
		25.2.3 Benchmarked
		25.2.4 Fixed-BandBoundaries
	25.3 Summary
	Questions—Scaling and Banding
26: Finalization
	26.1 Validation
		26.1.1 High Level
		26.1.2 Independent Oversight
		26.1.3 Quantitative Assessment
		26.1.4 Assessing Misalignment
			26.1.4.1 Score Misalignments
			26.1.4.2 Characteristic Misalignment
	26.2 Documentation
		26.2.1 Possible Outline
		26.2.2 Supplementary Tables and Graphics
			26.2.2.1 Model Development
			26.2.2.2 Period-On-Period—Pre- and Post-implementation
		26.2.3 Selection Strategies
			26.2.3.1 Failure versus Rejection—That is the Question!
			26.2.3.2 Preparing for Post-implementation
		26.2.4 Comparing New Against Old
			26.2.4.1 Rating Transition Matrix
			26.2.4.2 Swap Sets
	26.3 Implementation
		26.3.1 Platform Choice
			26.3.1.1 Criticality
			26.3.1.2 Budget
		26.3.2 Testing
		26.3.3 Further Considerations
	26.4 Monitoring
		26.4.1 Front-End
			26.4.1.1 Overrides
		26.4.2 Back-End
			26.4.2.1 Vintage/Cohort Analysis
			26.4.2.2 Early Monitoring
	26.5 Summary
	Questions—Finalization
Afterword
	Areas for Further Research
	Societal and Histories
	Empirical
Module Z: Appendices
Glossary
Bibliography
Index