According to a recent article in The Financial Times ('Unlocking the Language of Structured Securities') we may finally have arrived at one of the main contributory factors in ushering in the financial meltdown in 2008. Evidently, bond traders and raters took liberties with their computations of mortgage backed securities by taking shortcuts when using Intex.
What is Intex? Actually three things rolled into one, according to Donald MacKenzie, author of the piece:
1) A high level computer language for defining intricate cash flow rules for assorted deals.
2) A graphics interface for defining, illustrating deals
3) A computerized library of the parameters of the underlying asset pools and cash flow rules for more than 20,000 deals.
What does this cost? A cool $1.5 million.
Mortgage backed securities themselves can be frightfully complex, and most of this has to do with the inclusion of derivatives called "credit default swaps". In earlier blogs, I wrote about the illustrious quants who designed these based on a complicated formula known as the Gaussian Copula Formula. Many of these CDS instruments factor in such things as: the rate at which borrowers repay mortgages earlier (lowering the yield on the security), the propensity of borrowers to default, and the proportion of debt that can't be recovered on default. Other CDS also factor in disturbing intangibles that destabilize the whole compact- especially bets on any of the preceding happening.
Intex's output reliability is not at all automatic or a foregone conclusion, even for $1.5 million! It depends entirely on the validity of the user's assumptions and how spot-on or wide of the mark they are. As they say, "garbage in, garbage out" - and even a $1.5 million utility can churn out garbage if not properly prepared beforehand.
Even if you make your proper assumptions, you can still foul up. For example, maybe you do take time and reckon in the securities containing collateralized debt obligations (CDOs) possess not simply packages of assembled mortgage debt - but tranches of other packages of CDOs - each of which may be incorporating further CDOs, like an unending sequence of those Russian dolls (where each large doll has a smaller one within). Well, in the case of CDO tranches, each has millions of smaller ones within!
The effect is that the underlying mortgage pools and ancillary deals are multiplied enormously. While a simple CDO-tranche analysis may take only a few hours, a file with hundreds of nested ones may take days or even weeks. No surprise since time is money, many rating agencies took shortcuts here and - instead of doing a full exhaustive run - they chose to operate one run based on the prepayment, default and severity rates deemed most likely. Since the sidebets weren't also factored in, it's no wonder many of these blew up in their faces.
The error of many other bond rating agencies was to treat the complex, CDS -infested CDOs as if they were corporate bonds. (In which case the properties of the bonds would be simply inferred from the separate parcel ratings - most of which were already AAA). Thus, the ratings agencies operating on this false assumption went on to rate the large CDS-infested, multiple CDO tranche packages as 'AAA' too! Bad mistake! We all paid for it, and are still doing so!
Interestingly, one bank and one alone did analyze the CDO-CDS structures properly, and that was Goldman- Sachs. They were also clever in using a "computer farm" in New Jersey to spread the computations over several machines and thereby save time. Based on what it saw, Goldman ended up liquidating or hedging its positions on mortgage-backed securities, allowing it to survive almost unscathed. (Except for an SEC probe into its selling of securities-mortgages it already knew would fail).
What is the moral of all this? Basically that just because you shell out $1.5 million for a high level program to parse structured securities, you shouldn't expect it to do all your work for you - or save you time. (Well, at least not save you unrealistic time!).
Let's hope the next time bond agencies use Intex they do so honestly, with a view to giving honest bond ratings, not fraudulent ones.