Credit Scoring & Its Applications (Monographs on Mathematical Modeling and Computation) [Paperback]

Lyn C. Thomas (Author), David B. Edelman (Author), Jonathan N. Crook (Author)

Book Description

ISBN-10: 0898714834 | ISBN-13: 978-0898714838 | Publication Date: June 15, 2002 | Edition: 1st

Tremendous growth in the credit industry has spurred the need for Credit Scoring and Its Applications, the only book that details the mathematical models that help creditors make intelligent credit risk decisions. Creditors of all types make risk decisions every day, often haphazardly. This book addresses the two basic types of decisions and offers sound mathematical models to assist with the decision-making process. The first decision creditors face is whether to grant credit to a new applicant (credit scoring), and the second is how to adjust the credit restrictions or the marketing effort directed at a current customer (behavioral scoring). The authors have filled an important niche with this groundbreaking book. Currently, only the most sophisticated creditors use the models contained in this book to make these decisions, but all creditors can know these aids to successful lending.

Editorial Reviews

Book Description

Tremendous growth in the credit industry has spurred the need for Credit Scoring and Its Applications, the only book that details the mathematical models that help creditors make intelligent credit risk decisions. Using the models contained in this book, all creditors can learn to make better decisions

About the Author

Lyn C. Thomas is a Professor of Management Science at the University of Southampton.

Jonathan N. Crook is Reader in Business Economics at the University of Edinburgh.

David B. Edelman is Credit Director of Royal Bank of Scotland, Edinburgh.

Product Details

• Paperback: 246 pages
• Publisher: Society for Industrial Mathematics; 1st edition (June 15, 2002)
• Language: English
• ISBN-10: 0898714834
• ISBN-13: 978-0898714838
• Product Dimensions: 9.9 x 6.8 x 0.6 inches
• Shipping Weight: 1.1 pounds (View shipping rates and policies)
• Average Customer Review: 4.0 out of 5 stars  See all reviews (2 customer reviews)
• Amazon Best Sellers Rank: #870,029 in Books (See Top 100 in Books)

More About the Author

Discover books, learn about writers, read author blogs, and more.

Customer Reviews

Most Helpful Customer Reviews

14 of 14 people found the following review helpful:

4.0 out of 5 stars Good overview, but needs case studies, June 27, 2006

By 

Dr. Lee D. Carlson (Baltimore, Maryland USA) - See all my reviews
(VINE VOICE)    (HALL OF FAME REVIEWER)    (REAL NAME)   

Amazon Verified Purchase

This review is from: Credit Scoring & Its Applications (Monographs on Mathematical Modeling and Computation) (Paperback)

It would seem that credit scoring has become a separate branch of computational finance, and given its importance and use throughout the credit industry this is not surprising. There have been many articles written on credit scoring since its inception in the 1950's, but until four years ago there has not been a book available on the subject that summarized the different approaches to credit scoring. Although somewhat short, this book does that, and readers who need to understand credit scoring will find the book very helpful in laying out what tools have been used over the years. Credit analysts and institutions typically use proprietary packages to do their credit scoring, and so the approaches used in these packages remain unknown. However, this book discusses approaches to credit scoring that have been developed by academic researchers, some of which are no doubt used by some of the vendors. Readers will need a fairly strong mathematical background to read the book. It will definitely be of interest to those readers, such as this reviewer, who have this background but are new to credit scoring. The authors include a chapter on the macroeconomics of credit so as to allow the reader to gain an appreciation of the different factors that govern the demand (and supply) of credit. Most of this material can be found in any textbook on economic theory of macroeconomics. The goal of credit scoring is of course to distinguish between "bad" and "good" credit customers. A "good" credit customer is one whose performance over time is deemed acceptable by the lending institution while a "bad" credit customer has performance deemed unacceptable. The authors discuss many of the credit scoring techniques available from the academic literature and the advantages and disadvantages of each. The most popular of course are statistical methods, and the authors discuss discriminant analysis, logistic regression, nonlinear regression, and classification trees. The authors elaborate in some detail on these techniques, but they never include any discussion on real-world case studies that make use of them. In fact nowhere in the book do the authors include these case studies, and this lack is the only source of disappointment in the book.

More "exotic" techniques that are classified as `nonstatistical', such as linear and integer programming, neural networks, genetic algorithms, and expert systems, are also discussed in the book and compared with each other. The latter three of course are thought of as being from the field of artificial intelligence, and so it would have been very interesting (and useful) to discuss in more detail whether these techniques have major advantages over ones that are not deemed intelligent. The artificial intelligence academic community, as well as commercial vendors of intelligent applications, are claiming that these techniques are more advantageous, but to date this reviewer does not know of any objective and scientific study that gives weight to these claims.

Since credit worthiness can change with time, it is natural to want to develop methods by which credit scoring can monitor this change. In the book the authors refer to this as `behavioral scoring', and devote a chapter to its elucidation. As expected, Markov chains are the primary mathematical tools used to describe the repayment and usage behavior of credit customers. This naturally involves the identification of the different states that a customer's account can be in, and the estimation of the probabilities of moving from one state to another. Assuming that these probabilities are independent of previous histories gives rise to Markov chain models for behavioral credit scoring. When reading this part of the book, one might be reminded of similar techniques used in bond ratings. Credit customers could be classified according to a finite set of labels in a way that is similar to the rating of bonds.

The Markov chain approach can be generalized to situations where the lender's decisions can influence the transitions between states. The authors discuss this approach (an example of `stochastic dynamic programming') in fair detail in the book. Both of these approaches have the Markovian assumption built in, and so the authors discuss various techniques to check whether it holds in realistic scenarios. In addition, they examine cases where the probabilities of transition are viewed from a Bayesian perspective (called `Bayesian Markov models' by the authors). Thus the probabilities will not be fixed as they are in the ordinary case, but instead will be updated as more information becomes available. The model parameters will therefore be dependent on the total credit history of the customer. The use of Bayesian Markov credit scoring models will be more attractive to those financial institutions who have massive amounts of credit data and who desire to put more stringent bounds on risk. The authors also give examples on how the Bayesian approach can lead to problems, these arising from issues of computational complexity.

If a credit institution is to use credit scoring profitably, they must make use of some methodology for measuring the performance of scorecards. The authors devote a chapter to performance measures, and since credit scorecards are essentially a classification scheme into two groups, it is to be expected that the performance measures would draw on techniques from the literature on pattern recognition. This is the case in the chapter as the authors use measures of separation, such as Mahalanobis distance and Kolmogorov-Smirnov statistics, and their generalizations: the ubiquitous receiver operating characteristics (ROC) curve. But they also discuss other methods, such as bootstrapping and jackknifing, which are commonly used for scenarios where data sample sizes are small.

2 of 2 people found the following review helpful:

4.0 out of 5 stars Good math, still have rooms for improvement, September 7, 2007

By 

Ng Hon Ming "Book Worm Jeff" (Hong Kong) - See all my reviews
(REAL NAME)   

This review is from: Credit Scoring & Its Applications (Monographs on Mathematical Modeling and Computation) (Paperback)

This book illustrated the complete mathematical background and other implementation concerns for a credit risk scorecards(the way a bank accesses the approval of any lending). Pretty impressive and only of its kind in the market for credit scorecard building, validation, calibration...with complete math backing

My comment on different aspects of the book:

On more technical aspect: I wish the book could have models on Loss-given-default model since it's also part of credit risk model...

On the style: The book is too concise on its maths. Wish to have more elaboration or references so that I can drill down its math.