Review of The Credit Scoring Toolkit: Theory and Practice for Retail Credit Risk Management and Decision Automation (Hardcover)

As a professional statistician working in the banking industry I highly recommend this book as an invaluable reference for getting an overview of the credit risk management scenario as of today, as well as having an insight into its history and recent developments.

This book fills a long awaited gap for a comprehensive, thorough and complete outlook of the booming area of quantitative credit risk management. Since I graduated and started my professional career in the banking industry I was looking for such a text, both for personal use - as I tried to make a bridge between the academic knowledge acquired in the university and the business world - , as well as a pointer to indicate to colleagues from graduate school or other professionals wanting to start working on the area.

After buying lots of different books and spending hundreds of dollars I can finally say my search is over. This is a sound reference on credit scoring, serving both for the savvy risk professional wanting to brush up his skills as well as for the starter coming from the academy or other professional areas.

The book is outlined in eight thematic sections, which helps delivering the huge amount of information covered in a manageable way. Some of the highlights/weaknesses of the book are the following:

Section A: Setting the scene. This section covers an overview of credit, scoring and credit scoring setting the ground for what is coming next in the book. It glances through the Credit Risk management cycle (CRMC) and the reasons for and against use of scoring in credit retail operations. It also has a chapter on the history of credit as well as an introductory chapter on the mechanics of credit scoring, which summarizes the main technical aspects of scoring in the day-to-day business operation.

Section B: Risky Business. In this section the author contextualizes credit risk into the larger framework of risk management, relating it to the other three primary risks in the banking industry environment, namely business, market and operational risk. A nicely put overview of the philosophies of science on the Chapter on Decision Science is one of the highlights. It ends with a chapter on assessment of enterprise risks, from SME lending to middle and large corporations.

Section C: Stats and maths. This is the densest part of the book, covering in less than 100 pages and 4 chapters a content that spans over a dozen of statistics text-books. Kudos to the author for the extensive research carried but the section also has its shortcomings. Chapter 7 - ZPredictive statistics 101' abuses of mathematical notation and mixes some statistical concepts in its explanation of modeling techniques (specially regarding LPM - Linear Probability Modeling). Nonetheless, the author got the core concepts right and the overall coverage of the statistical methods (such as Logistic Regression and Regression Threes) in the other chapters are very good, with highlights to the definition of information value in terms of the Kullback-Leiber distance. The last Chapter on the section, on software and people resources is also very useful, going into some of the technical aspects for model implementation.

Section D: Data!: Data is the single most important aspect of any statistical analysis, and it is not different for credit scoring and decision automation. So it's not surprising that this is the largest section of the book and it deserves your special attention no matter your background. The chapters in this section cover the most relevant issues with data treatment, quality assessment and preparation that you are faced with in the industry. The chapter on Data preparation is particularly enlightening and discusses important decisions in credit scoring modeling, such as staggered versus static outcome windows, good/bad/indeterminate definitions, observation excludes, sampling considerations and use of external data.

Section E: Scorecard development. This section covers the practical aspects of day-to-day scorecard development, discussing transformation of variables (and statistical methods for it), characteristic selection, segmentation, reject inference, calibration, validation and development management issues. Highlights are its discussions of characteristic selection and reject inference, both comprehensive and filled with examples and references for further reading. It's the section that the scoring analyst will refer to the most while at the development of a new scoring model.

Section F: Implementation and use. A scorecard that's not implemented is useless by itself. This section covers the implementation of one or more developed scorecards, along with all the impacts to portfolio and business, and the monitoring of the implemented model. The Chapter on Monitoring is a highlight of this section, covering all the major issues with scorecard monitoring, such as misalignment, population and score drifts, stability reports, book rates, selection process, policy rules and proper treatment of overrides. It also has a nice chapter on finance (26), which gives insights into how to analyze financial impacts of current and tentative decision processes. There is also a linkage to Basel II model parameters, such as the Loss Given Default (LGD).

Section G: Credit Risk management cycle. This is a brief overview of other components of the CRMC, such as marketing, application processing, account management, collections and recoveries and fraud. The chapters in this section are somewhat concise but it seems to be intentional, as the focus of the book is specifically on credit scoring. Nonetheless, enough of the topics is covered to give the reader a good notion of how scoring fits into the credit cycle and other potential applications of the statistical techniques and process improvement. The chapter on Fraud is fairly comprehensive if you take in account that this is a book about credit risk (and fraud, strictly speaking is a concern of operational risk).

Section H: Regulatory environment. As the author poses, since the 1960s there has been an increasing regulation of financial institutions, and retail consumer credit did not escape it. This section of the book covers issues such as the Equal Credit opportunity Act (ECOA) in the US and other Fair Lending related legislation, as well as data privacy, capital adequacy and anti-discrimination. Of particular interest for practitioners worldwide are the comparisons of such legislation against a set of English-speaking countries, such as the US, Canada, Australia, UK and the Republic of South Africa. Although it does not cover the country I'm working on currently (Brazil), its broad coverage and analysis gave insights into understanding Brazil's current legislation on the matter and implications on credit risk model development.

I believe the author has done an excellent job on assembling all this information together in this format and highly recommend it for beginners and practitioners alike. The list price is somewhat above average but it's definitely worth it, as the only other option I have found so far to acquire the information covered by the book is to work in a retail credit area of a big player in the financial industry.



Click Here to see more reviews about: The Credit Scoring Toolkit: Theory and Practice for Retail Credit Risk Management and Decision Automation (Hardcover)