Active Credit Portfolio Management in Practice

JEFFREY R. BOHN, PHD, leads the Financial Strategies group at Shinsei Bank in Tokyo. Previously, he led Moody's KMV's (MKMV's) Global Research group and MKMV's Credit Strategies group. After Moody's acquired KMV, he and Roger Stein coheaded MKMV's research and product development. ROGER M. STEIN, PHD, is Group Managing Director of the newly formed Quantitative Research and Analytics group at Moody's Investors Service in New York. Previously, he was head of research for Moody's Risk Management Services. After Moody's acquired KMV, he and Jeffrey Bohn co-headed MKMV's research and product development.

• Foreword xiPreface xiiiAcknowledgments xxviiChapter 1 The Framework: Definitions and Concepts 1What Is Credit? 2Evolution of Credit Markets 7Defining Risk 11A Word about Regulation 13What Are Credit Models Good For? 14Active Credit Portfolio Management (ACPM) 16Framework at 30,000 Feet 19Building Blocks of Portfolio Risk 23Using PDs in Practice 32Value, Price, and Spread 34Defining Default 38Portfolio Performance Metrics 38Data and Data Systems 42Review Questions 43Chapter 2 ACPM in Practice 45Bank Valuation 50Organizing Financial Institutions: Dividing into Two Business Lines 52Emphasis on Credit Risk 57Market Trends Supporting ACPM 59Financial Instruments Used for Hedging and Managing Risk in a Credit Portfolio 60Mark-to-Market and Transfer Pricing 63Metrics for Managing a Credit Portfolio 68Data and Models 72Evaluating an ACPM Unit 75Managing a Research Team 77Conclusion 86Review Questions 87Exercises 87Chapter 3 Structural Models 89Structural Models in Context 91A Basic Structural Model 95Black-Scholes-Merton 100Valuation 107Modifying BSM 117First Passage Time: Black-Cox 118Practical Implementation: Vasicek-Kealhofer 124Stochastic Interest Rates: Longstaff-Schwartz 145Jump-Diffusion Models: Zhou 150Endogenous Default Barrier (Taxes and Bankruptcy Costs): Leland-Toft 151Corporate Transaction Analysis 156Liquidity 159Other Structural Approaches 161Conclusion 171Appendix 3A: Derivation of Black-Scholes-Merton Framework for Calculating Distance to Default (DD) 171Appendix 3B: Derivation of Conversion of Physical Probability of Default (PD) to a Risk-Neutral Probability of Default (PD Q) 177Review Questions 179Exercises 179Chapter 4 Econometric Models 183Discrete-Choice Models 186Early Discrete-Choice Models: Beaver (1966) and Altman (1968) 191Hazard Rate (Duration) Models 196Example of a Hazard-Rate Framework for Predicting Default: Shumway (2001) 204Hazard Rates versus Discrete Choice 206Practical Applications: Falkenstein et al. (2000) and Dwyer and Stein (2004) 207Calibrating Econometric Models 215Calibrating to PDs 216Calibrating to Ratings 227Interpreting the Relative Influence of Factors in Econometric Models 234Data Issues 238Taxonomy of Data Woes 241Biased Samples Cannot Easily Be Fixed 244Conclusion 249Appendix 4A: Some Alternative Default Model Specifications 249Review Questions 252Exercises 252Chapter 5 Loss Given Default 255Road to Recovery: The Timeline of Default Resolution 258Measures of LGD (Recovery) 260The Relationship between Market Prices and Ultimate Recovery 265Approaches to Modeling LGD: The LossCalc (2002, 2005) Approaches and Extensions 273Conclusion 285Review Questions 286Exercises 286Chapter 6 Reduced-Form Models 289Reduced-Form Models in Context 291Basic Intensity Models 296A Brief Interlude to Discuss Valuation 310Duffie, Singleton, Lando (DSL) Intensity Model 312Credit Rating Transition Models 329Default Probability Density Version of Intensity Models (Hull-White) 340Generic Credit Curves 348Conclusion 353Appendix 6A: Kalman Filter 354Appendix 6B: Sample Transition Matrices 357Review Questions 358Exercises 358Chapter 7 PD Model Validation 361The Basics: Parameter Robustness 367Measures of Model Power 371Measures of PD Levels and Calibration 379Sample Size and Confidence Bounds 396Assessing the Economic Value of More Powerful PD Models 418Avoiding Overfitting: A Walk-Forward Approach to Model Testing 431Conclusion 437Appendix 7A: Type I and Type II Error: Converting CAP Plots into Contingency Tables 438Appendix 7B: The Likelihood for the General Case of a Default Model 440Appendix 7C: Tables of ROC ε and n max 441Appendix 7D: Proof of the Relationship between NPV Terms and ROC Terms 441Appendix 7E: Derivation of Minimum Sample Size Required to Test for Default Rate Accuracy in Uncorrelated Case 446Appendix 7F: Tables for Lower Bounds of ε and N on Probabilities of Default 447Review Questions 452Exercises 452Chapter 8 Portfolio Models 455A Structural Model of Default Risk 460Measurement of Portfolio Diversification 460Portfolio Risk Assuming No Credit Migration 461Structural Models of Default Correlation 465Credit Migration 470A Model of Value Correlation 475Probability of Large Losses 481Valuation 484Return Calculations 488Risk Calculations 491Portfolio Loss Distribution 498Capital 514Economic Capital and Portfolio Management 519Improving Portfolio Performance 521Performance Metrics 526Reduced-Form Models and Portfolio Modeling 530Correlation in Intensity Models 531Copulas 534Frailty 536Integrating Market and Credit Risk 541Counterparty Risk in Credit Default Swaps (CDS) and Credit Portfolios 544Conclusion 546Review Questions 547Exercises 548Chapter 9 Building a Better Bank: A Case Study 551Description 552Current Organization 554Transforming the Capital Allocation Process 556Portfolio Analysis 558Active Credit Portfolio Management (ACPM) 562Data, Systems, and Metrics 563ACPM and Transforming the Bank 566Appendix: Figures 569Exercises 574References 575About the Authors 589Index 591