Unfortunately, experience suggests that once the financial press gets hold of a story, we should not expect calm and reasoned debate. "Small earthquake, nobody hurt" is not a headline that sells newspapers. And so it is proving with the hitherto arcane topic of Collateralized Debt Obligations, which has recently emerged from the Credit Markets section of the Wall Street Journal to the part that most people actually read.
Add to the picture the typically opaque language of EITF 99-20 (a refinement to GAAP accounting for riskier ABS), and the desire for oversimplification becomes even more understandable. So now we have a motive (the greed of gullible investors), a weapon (obtuse financial engineering), and a suspect (the new favorite scapegoat the rating agencies).
Case closed? This article will seek to suggest otherwise, without either restating the old standby caveat emptor or deeming the rating agencies completely blameless.
The specific focus herein is cashflow arbitrage CDOs backed by high yield collateral, because this has been the primary area of ratings activity, though some of the conclusions are also applicable to CDOs backed by investment grade collateral. (We had better hope the financial press doesn't figure out that synthetic CDOs involve the use of credit derivatives, otherwise it will all be front page news).
There are at least three primary reasons why the extent of ratings downgrades should be less of a surprise than it seems to be:
*CDO ratings are intended to be consistent with other ratings: This means that we should not consider investment grade ratings to be immune from downgrade, or even default. If we consider a 5-year timeframe (CDO issuance really took off in 1997), S&P statistics tell us that over the last 20 years, on average 18% of all BBB-rated obligors still rated at the end of 5 years would have been downgraded (inclusive of default).
Over a one year period, the average downgrade rate is 5.6%, but the highest downgrade rate for a one-year period was 11.2%, so we also know that averages tell just part of the story. Default statistics reveal the same pattern: according to Moody's, the average cumulative 5-year default rate for Baa-rated obligors (1970-2000) was 1.82%, yet the highest was almost three times this level, 5.27%.
Even Triple-A ratings are not immune: S&P tells us that on average only 61% of AAA-rated obligors still have this rating after five years. So, if we consider the point we have reached in the credit cycle, and we add to this the well-known default-rate lifecycle peak ("seasoning effect") for high yield collateral, we should not be surprised at the volume of CDO ratings actions.
*CDO structures can do little to mitigate ratings migration: Having become accustomed to excellent ratings stability in most traditional ABS, some market participants have been surprised by its apparent absence in this new segment of the market.
However, this ignores the existence of dynamic credit protection mechanisms in revolving ABS transactions (e.g. credit cards, trade receivables securitizations...), which are explicitly designed to promote ratings stability, or the rapid build-up of credit protection in short average life ABS such as auto loans. Any failure to look through the structure to the underlying collateral is short-sighted.
There is a material difference in risk between a credit card receivable, with an average life of less than 12 months and flexible pricing, and a pool of high yield obligations with 5-7 year average lives, basically fixed pricing, and an exponential increase in default risk when subject to their inevitable ratings migration.
That ABS backed by the latter should be subject to much greater ratings volatility should be obvious, and the consistently tighter spreads applicable to traditional ABS have long been a signal (or should have been) that CDO investors are being paid a premium in order to accept this volatility. Something for nothing rarely works out elsewhere, so why should it for CDOs?
*Rating agency methodology does not fully capture all risks: One of the basic premises of the ratings process is, naturally enough, that debt ratings are the most accurate measure of default risk.
However, the development and refinement of Merton-based default models such as KMV CreditMonitor strongly suggest that the incorporation of equity market price data can improve upon the predictive power of a purely ratings-based approach.
Similarly, as noted above, ratings migration and default statistics are subject to significant variation over time, which is typically explained by macroeconomic factors, but in fact for high yield borrowers is just as often due to idiosyncratic factors, which are by definition harder to predict. However, rating approaches (for example Moody's binomial expansion methodology) understandably seek to simplify the equation by stressing average default rates and measuring default correlation through a measure such as the diversity score.
Even for broad market aggregate data, the variability (as measured by the standard deviation) of both default and recovery data is very significant. When we consider that an individual CDO may contain fewer than 100 obligors, and further, that the assumption of a normal distribution implicit in the ratings models is demonstrably not applicable to credit risk, it becomes clear that a more refined approach to understanding CDO risks is appropriate.
Some of the tools increasingly being used to manage credit portfolios, which permit us to change our assumptions about the shape of loss distributions and run Monte Carlo simulations accordingly, refocus attention on risk as the volatility of expected loss. This would clearly demonstrate that the incidence and severity of "unexpected losses" for these relatively small portfolios is greater than is suggested by the ratings process, and the small size of most CDO tranches below the senior level merely serves to compound the severity of impact when performance diverges from average expectations.
Credit portfolio management methodologies, and more recently CDO experience, clearly reveals that the concept of stressed expected loss that is central to the ratings process should therefore be supplemented by a more accurate analysis of risk-adjusted returns in a manner very comparable to the Sharpe ratio traditionally used in the asset management world.
Thus we can perhaps conclude that the rating agencies provide a standardized framework for the analysis of CDOs, but innovation outside the CDO market indicates that there are more finely calibrated tools available.
Furthermore, to give credit where it is due, we should note that the agencies themselves are not oblivious to this. Moody's, for example, has developed its own version of a Merton model for default prediction, and in developing a portfolio risk model for improved analysis of the financial guarantors, they have recognized the limitations of oversimplified default correlation models.
However, because of inherent conflicts between these concepts and those that underpin their core ratings methodology, the latter remains essentially unchanged.
In summary, we have sought to demonstrate that the market's newfound surprise at downward ratings actions in the cashflow arbitrage CDO market is not justified if one has a clear understanding of the underlying collateral, and the methodology underlying CDO ratings.
The rating agencies, investment banks, and most CDO investors continue to operate in a highly compartmentalized manner which fails to capture the true nature of risk, and therefore of value, in this important market. For investors that bring a cross-disciplinary approach to the topic, not only does this permit a better vision of risks, but also an ability to enhance risk-adjusted returns. For well-armed and astute players, this improves the quality of due diligence on primary market deals, as well as enhancing the CDO portfolio management process. It also suggests that the secondary market is potentially an interesting place right now.
Disclaimer: The views expressed here are those of the writer, and do not purport to represent those of NBP.