Internal Credit Risk Models: Capital Allocation and Performance
Measurement
Michael Ong's Internal Credit Risk Models
was published in 1999 in the midst of ongoing debates as to how to replace the
original Basle accord on bank capital requirements with a new Basle II. In that
context, this book was a piece of propaganda (in a good sense) promoting the use
of internal models for calculating bank's capital for portfolio credit risk. It
provided a soup-to-nuts description of the state of the art in portfolio credit
risk modeling and risk-based capital allocation within banks.
Fast forward a few years, and the book seems dated.
It is hard to believe, but the pace of developments has been that rapid.
Shortcomings aren't glaring. The terminology seems quaint at points, and some
important developments are anticipated more than covered. I think the most
significant shortcoming of Ong's treatment is its general silence about how to
calibrate the proposed models to current market conditions. The book covers in
detail what to do with inputs such as expected default frequencies, but it
offers little guidance on how to construct these inputs. Calibration of credit
risk modeling is one issue that has seen significant improvement since this book
was published.
Ong opens with an insightful history of bank
capital regulations. This motivates the reasons for bank capital as well as the
challenges it poses. Next the book builds, chapter-by-chapter, a general
framework for modeling portfolio credit risk. What I really like are some of the
side comments indicating why this or that issue is important or what some of the
debates were back in 1999. These afford insights that more recent books do not
afford—perhaps because many of the issues, having since been resolved, are less
on people's minds today.
The discussion of portfolio credit risk culminates
with two chapters on constructing a portfolio loss distribution. One is
parametric, fitting a beta distribution to the portfolio. The other uses a Monte
Carlo analysis. There is then a chapter on extreme value theory. Three more
chapters focus on risk-adjusted performance measures and capital allocation
within banks. These cover basic issues, but focus primarily on philosophical
issues of capital allocation based upon risk-adjusted performance measures. The
book closes with an appendix containing relevant papers by several researchers.
Contents
1. On Basle, Regulation and Market Responses Past and Present
2. Overview of Approach
3. Modelling Credit Risk
4. Loan Portfolios and Expected Loss
5. Unexpected Loss
6. Portfolio Effects: Risk Contribution and Unexpected Losses
7. Correlation of Default and Credit Quality
8. Loss Distribution for Credit Default Risk
9. Monte Carlo Simulation of Loss Distribution
10. Extreme Value Theory
11. Risk-Adjusted Performance Measurement
12. Implementing the Internal Model Across the Enterprise
13. Credit Concentration and Required Spread
14. Epilogue: The Next Steps
Appendix
Despite its
limitations, this remains a valuable book. It offers a nice compromise between
mathematical sophistication and accessibility. Compared to books like Saunders and Allen (2002)
or Crouhy et al. (2001) that give
high-level overviews of different portfolio credit risk models, Ong gives an
in-depth look at one unified approach based largely on the asset value (Merton)
model. Default correlations are modeled with a single factor model based on an
obligor's industry. This aspect of the treatment draws on CreditMetrics.
The book's focus on one unified modeling approach
is similar to that of Arvanitis and Gregory (2001).
The latter book is clearly superior—more up-to-date, more sophisticated, and
more practical. However, it is not as accessible.
Ong's treatment of portfolio credit risk is most
similar to that of Bluhm, Overbeck and Wagner
(2003). (Note that later chapters
of the two books are entirely different: Ong goes on to consider capital
allocation. Bluhm, et al. goes on to consider the use of default models in
financial engineering). Both books are reasonably accessible and focus on the
use of asset value models. Bluhm et al. is more up-to-date and discusses
calibration issues. Ong, on the other hand, offers much better context—something
Bluhm et al. does not do a good job with.
In summary, Ong is kind of like the Hull (2005)of credit risk modeling. It was once great, but is now faded. It is, however,
still useful, offering a valuable perspective that can complement other books.
