Customer Reviews

Credit Risk Scorecards: Developing and Implementing Intelligent Credit Scoring (Wiley and SAS Business Series)

5 star:		(4)
4 star:		(3)
3 star:		(0)
2 star:		(1)
1 star:		(1)

 

 

I am a model expert in credit risk and it happened that I was looking for a text book that offers different approaches in calculating the score band. Offcourse, based on the great review and on the SAS icon on the back cover it seemed that this book is a good base line.(I was not able to read inside the book and now I know why the publisher didn't add this feature with amazon).It was really a disappionting buy for me, the author was talking about basics that every one in the industry should know??if there is any value of this book is the value of the sas code. AS,a buyer who is looking for a sas code to execute the very generic knowledge of the book,let me tell you that will see literary three small tables of four lines of code in each in the whole book...for practionares don't ever buy this book,for managers I don't know, they are managers and they should kow these basics, for people who knows nothing about the consumer industry then it is a good start.
Usually,I don't list any comments for any book becasue i research the book before the buy, but this book tricked me and it deserves these couple minutes of comment to pay back the publisher...Please don't list this book as a SAS book and remove the SAS icon from the back cover it is embaressing!!

Well, I worked for a major consumer bank in the world. This book does not target risk manager who want to know more about scorecard strategies. This book is good for a new entrant staff or students who want to be familiar with scorecard developement. I think this book is valuable because it shows that you do not have to use comprehansive tools to develope a scorecard. All you need is scorecard knowledge, basic statistics software(SAS, STATA, SPSS), and excel for reports.

This allowed me to work with the developers and be on top of the action items that were critical to implemetation.

I got this book abroad while on a business trip and started reading it right away. I believe it gives a very nice overview of the process involved into building scorecards (from planning to completion). It really covers all bases. My only problem with it is that some topics I consider important were not developped further (such as ROC curves). Instead there are quite a few pages (counted 14) on Reject Inference which although important, did not need so much emphasis. In any case, it is nice to see a book devoted solely to scorecards. If you want to know what it takes to build one or if you want to build your first scorecard, this is definitely a good book to start.

This review is written by the author of a competing book, so any comments made must be taken with a grain of salt.

Mr. Siddiqi's book focuses on the development of in-house scoring capabilities, covering a broad range of development and implementation issues. Unfortunately, very few references are provided. The author obviously has a lot of experience (he works for SAS Canada), but at times the limits show. There are times when I feel that without the backing of SAS, and the current high level of interest in this topic, this book would not have the same level of acceptability.

I would recommend the book, but with qualifications--basically, it should only be read in conjunction with other works on the topic, especially if you want a view on related topics (economics, history, law, credit strategy, etc.).

A fantastic book for managers. The development of scorecards involves seriously sophisticated statistics, such that without a Phd in statistics, there is no point in trying. Due to this, most books on this topic fall into one of two categories: 1. Far too mathematically advanced for ordinary readers. 2. So basic as to be of no practical use. This book is somewhere in the middle. Thus, it caters for managers, providing easy to understand, yet thorough plain English explanations of all key topics, while not losing the reader in advanced maths. At the same time, it contains enough maths to provide managers with basic maths training an insight into the statistical techniques involved in scorecards. Although far too basic for statistical experts who have the job of developing scorecards, this is a fantastic book for managers who want to gain a solid grasp of the issues.

This is an excellent resource for people and organizations creating scorecards. It covers a wide assortment of topics in just enough detail to provide the "nuts and bolts." The author provided much of the material for The SAS Institute's training program on scorecard development as well. NOTE: the volume is somewhat slim and should not be considered an "encyclopedia," so if that's what you need, you'll need to augment this with other resources.

After reading this book, I just want to point out couple of points that I think is beneficial for me as an experience analytical expert:
1 - How to choose a "Bad" definition and how to choose the proper performance window. The author explained and illustrated them in detailes, which I think makes a lot of sense and we actually should apply his way to do it. sometimes when we build the model, the bad definition or the performance window seems too arbitrary or given by "others", but it is our responsibility to probe into it and see if it makes sense, that's what I learned from this book.

2 - Reject reference. The author provided a comprehensive explanation and application on how to do the reject reference, which is quite valuable for me, and I am sure it is true for other readers as well.

Enjoy reading and Good luck!

Credit scoring is of course of enormous importance not only to lending institutions but also to individual borrowers who benefit from high credit scores when obtaining personal, commercial, and mortgage loans. The details of how the credit scores are calculated though are frequently proprietary, so it is imperative to understand the generalities behind credit scoring if one is to make reasonable judgments on its use and its risks. This short book gives an elementary introduction to credit scoring and is useful to those readers who need to understand it without going into complicated mathematics. Readers who require a more in-depth treatment will have to look at other books on the subject.

Risk scorecards are according to the author a powerful solution to the problem of identifying good and bad credit customers. Key to the success of a scorecard is finding the most effective `population performance definitions' which must be flexible enough to be used for different purposes. A good scorecard must also be successful in its ability to distinguish "bad" customers from "good" ones, and this distinction must respect the attitudes of the institution that is deploying it. The author clearly favors an approach to credit scoring that is based on statistical and data mining techniques. However, he does not believe that the mere use of programs or algorithms to process data will result in an effective scorecard. Such a scorecard can only be obtained by applying concepts from business intelligence, where data in his words "is passed through the business-trained mind of the user." In addition, a scorecard must not be so complex so as to stymie real-time decision-making or diagnostics. The scorecard will probably not be the sole arbiter in the decision-making process, but it still needs to be tractable enough to allow this process to reach fruition in reasonable time scales.

The development of scorecards requires, according to the author, that several individuals be involved. One of these is the actual developer, who does the required statistical analysis, thus ensuring that the scorecard development is valid from a statistical standpoint. Another is the risk manager, who is responsible for the company portfolio and the eventual use of the scorecards. Still another is the product manager, who understands the client base that the scorecard will target. In addition to these the operational manager is needed in order to give advice on how the scorecards will be implemented in actual price; a project manager is needed to oversee the management of the scorecard project; and an IT manager is needed to assess and manage the software development that will be needed to create the scorecard. Lastly, the legal staff of the company needs to be involved to ensure that the scoring conforms to existing regulations.

Any scorecard, no matter how sophisticated in its mathematical formalism or complexity, is dealing with incomplete information, and so it will sometimes omit "good" customers and admit "bad" ones. Subjective measures of preference will also come into play, especially with scorecards that implement some sort of a "cutoff" function to rate or classify credit customers. Such preferences will naturally be different depending on the organization that will be deploying the scorecard. In addition, "seasonal" effects and the choice of time window for tracking customer credit behavior are all issues that must be dealt with when designing a scorecard. These issues are discussed in fair detail in the book.

A successful scorecard should at least confirm what customers are unacceptable in the current data pool. This requires of course an independent judgment of what it means to be an unacceptable customer. The author discusses a few methods by which the scorecard's definition of unacceptable can be confirmed.

The credit scorecard is of course relevant to risk management, and the author discusses how they can be used to assess the risk profile of the credit applicant. The most important part of this discussion is the emphasis on the need for the validation of the scorecard using existing data in customer accounts. As a minimum the scorecard should be able to measure the change in this data from the sole standpoint of the scorecard characteristics.

Credit scoring is a predominantly quantitative exercise, and typically involves techniques from mathematical statistics and artificial intelligence. The author does not discuss in great depth these techniques, since he wants to address a wider audience, but he does allow enough of a peek into the mathematical machinery needed to grant the reader an appreciation of what is done and should be done in finding an effective scorecard.

There are many issues, both technical and legal that have arisen in recent years regarding scorecards, both in their development and in their use. Individual customers have become much more knowledgeable about credit scoring, due to the new ways in which credit issuers are making use of credit scores. Customers are finding out that they must monitor their credit scores continuously to make sure they are optimal, since many credit issuers, particularly those in the credit card industry, are changing the terms of credit based sometimes on only minor changes in the credit score. In addition, legal complaints have been lodged against credit reporting agencies, charging that their use of scoring is biased or even discriminatory. In addition, the current drive towards automation in the financial community has had major ramifications to the implementation of scorecards, due in part to the need for granting exceptions to the scoring predictions. The granting of exceptions is an unpredictable activity, and one that relies heavily on subjective judgment. Machine intelligence, advanced as it currently is, is not yet sophisticated enough to handle these kinds of judgment calls. This will no doubt change in the near future, as automation continues to increase. It will be interesting to see whether credit scoring will become fully automated, but if it does it will be not only because the technology is able to support such automation, but also because credit customers trust this technology and are confident in its fairness.