This is Argyn's blog. I comment on topics of my interests such as software, math, finance, and music. Also, I write about local events in Northern Virginia, USA and all things related to Kazakhstan

Monday, July 06, 2009

Book: Measuring and Managing Credit Risk by Arnaud/ Renault, Olivier and Servigny, Arnaud De De Servigny

I went to a library to pick another book, and saw this one on a shelf, browsed quickly and took it. I think that some material can be outdated, but it seems to cover a lot ground otherwise.

Started reading... The first impression - very nice overview. Got through 2 chapters, not very technical, but technical enough, liked microecon overview with important references... I think I'm buying it after reading...

2nd chapter's on scoring models. If you know econometrics and finance, then there's enough details to get an idea of each model described. Again, many references to original works. Structural models (Merton, KMV etc) as well as some other stuff like vector support models are covered. Authors bring their experience in rating agencies here. The latter's interesting for me, as I'm more familiar with what's used in banking.

The model testing part covers both discriminative and predictive power tests. I found this part a little too light on details. On the other hand, it's easy to follow the references, and find all the details if needed. Overall, I liked the chapter. It's a good survey of what's used or can be used to score and test scoring models.

Chapter 3 is about LGD. Now I see that the book's clearly focuses on corporate debt, some microeconomics background is given on capital structure. There's not much coverage of consumer loans (mortgages) or commercial real estate. It's Ok with me. This chapter touches upon LGD and PD correlations too. There's some empirical data and good references on LGD research. They discuss loss and recovery and how they're different. Some loss distribution models are covered, particularly kernel smoothing method is described. A couple of loss distributions are mentioned. So far a good reading, again, not too technical, easy reading.

Chap 5 is on default correlations. There's very nice and slightly technical intro to copula models. I think it's one of the most accessible text on subj for non-quant folks. There's some description of factor models. Again, there's nothing on consumer loans here, but I liked the content overall. It gives a very good overview of importance of correlations, and current modeling issues.

Chape 6 is on portfolio credit models. They cover 5 commercial products. They also describe how to combine LGD, PD, correlations, rating downgrades and all previous content on individual assets to come up with portfolio level credit risk. Basic RAPM, RAROC, Econ Capital, VaR, Unexpected Loss are explained. I think that UL is explained somewhat strangely, it seems to be based on gaussian distribution, i.e. all about STD. Some basic backgrounder on Monte Carlo simulations is here. This chapter is the most technical so far, especially the Appendices with advanced stuff like generating functions.

I read two next chapters on the beach. Quite an entertaining reading, I mus tsay. Chapter 7 was very interesting. It's about economic capital allocation and credit risk management from strategic point of view, i.e. for a bank in its entirety. They look at two approaches which they call bottom-up and top-down. The example of the latter would RAROC all they way from a loan to a business uint. The former would be to look at what other banks do and allocate capital "proportionally". They describe several methods both static and dynamic, and discuss them. I like the discussion on role of the banking industry and how risk management should be aligned with it. Entire chapter is focused on banking.

Chapter 8 is on credit yield spreads, and their determinants. The appendices are quite technical. They cover basic yield spread models. Structural and reduced form models are covered. There's insightful discussion of empirical results. For instance, they claim that research shows that credit quality doesn't explain much in credit yield spreads, and that liquidity plays is a big factor, which is not modeled very well etc.

Chapter 9 is on credit derivatives. Of course, it's thoroughly outdated. It's not to say that it's useless though. They cover without many details most popular products CDOs, CDS, CLN, basic swaps etc. This chapter is one of the least technical ones. They explain deal structures on very high level, and talk about market growth, distribution of risk etc. This is all before 2007 material, you can see right in first pages of a chapter. They touch upon the risks and disadvantages of credit derivatives. That they don't decrease risk, but return in it back to market in form of increased equity volatility. They say that CDS increase liquidity and other things that don't hold true as we know now. You can almost skip the entire chapter, but there are some bits and pieces of useful information, and the chapter's quite short.

Last chapter is on Basel II accord. Suprisingly, this was good read, despite some obviously outdated content. They cover Basel's history up to 2003. I liked the overview of Basel's criticisms, they sound so valid now. There's brief but thoughtful discussion on liquidity, and how it affects the prices. They started this in previous chapter, but it comes back in a blurb on fair-value accounting and its possible interference with Basel II. All in all, the chapter's worth reading, although there could be more up to date introductions to Basel II.

Finally, my conclusion is that despite some content being outdated, a book could serve as a handbook/introduction to credit risk for quant practitioners in banks. There's not much mortgage related content, which is one of few drawbacks to me.

No comments: