Do not underestimate the power of structure

Citigroup and Merrill's recent large write downs were largely due to losses on mortgage securities. Now, we know that neither Citi nor Merrill were big originators of subprime loans. Furthermore, neither was taking losses on loans they held on their balance sheet for investment. The losses were primarily in CDOs or ABS securities being warehoused for future CDO issuance.

Its critical for readers to understand how structured investments can experience accelerated losses, and thus why its important to distinguish between firms holding whole subprime loans versus those holding securitized loans. To illustrate this, I've got a little math exercise for your reading pleasure.

Let's start with an ultra simple ABS security with $100 million first lien subprime mortgages with a 6.75% coupon as the collateral. The tranches are as follows:

• Senior: 5.75% coupon, $80 million
• Mezz: 6.50% coupon, $15 million
• Subordinate: 8.00% coupon, $5 million

For the sake of argument, let's assume that all principal is paid sequentially with no other triggers or any such complexities. A little later I'll show what happens if we add that stuff back in. We'll further assume that 3% of the principal pays off each quarter. This means that in the absence of any defaults, the first quarter cash flow for the deal will look like this:

• DEAL WIDE: $3 million principal, $1.69 million interest
• Senior: $3 million principal, $1.15 million interest ($80 million par, 5.75%/4 coupon)
• Mezz: $0 principal, $243,750 interest
• Sub: $0 principal, $100,000 interest

Note that in a sequential pay system, all principal goes to the senior-most tranche until its completely paid off, hence why there is zero principal going to the other tranches. Also note there is extra interest, some of which would normally be paid to a residual holder, and some would be kept as extra cushion for future shortfalls.

Anyway, I've built this hypothetical structure into a spreadsheet and added that the deal suffers 2% defaults each quarter from quarter 3 through quarter 7. So we see 10% total defaults. To make life easy, I assumed no recovery. Anyway, let's not quibble about the exact loss rate, rather focus on the concept I'm presenting here.

So if that happens, by quarter 8, here is the P&I situation by tranche:

• DEAL WIDE: $3 million principal, $1.16 million interest
• Senior: $3 million principal, $850,000 interest (all paid)
• Mezz: $0 principal, $240,000 interest (all paid)
• Sub: $0 principal, $70,000 interest ($100,000 was due)

So in the 8th quarter, the deal isn't producing enough interest to pay all its tranches. The shortage gets larger and larger over time, because as the Senior tranche is paid down, the more expensive junior tranches put more stress on the total interest available. So by the 17th quarter, shortly after the 4th year, there is no interest available for the sub at all. The Mezz tranche starts to see interest shortfalls after that. Furthermore, the Mezz tranche only winds up receiving 2/3 of the principal it was due, while the Sub tranche receives no principal at all.

Compare this result with a bank which had held the same $100 million as old fashioned loans on their balance sheet. Yes, the bank would have suffered a 10% loss on its portfolio, which is bad. But that would be the extent of it. In the case of the structured deal, the Senior holder gets all his principal and interest as expected. But both the Mezz and Sub holders take big losses.

Now consider what happens if a CDO was made of several "Sub" bonds, all had the same loss experience. A CDO of that kind of ABS would probably have similar tranching to our subprime home loan deal:

• Senior: 5.45% coupon, $80 million
• Mezz: 6.00% coupon, $12 million
• Sub: 8.00% coupon, $4 million
• Equity: $4 million

So what happens to the CDO in quarter 8? Remember that was the point at which the Sub bond suffered a 30% interest short fall. In the case of the CDO, the whole deal would suffer a 30% interest short fall. And the short fall accelerates rapidly until the CDO would get no interest at all by quarter 17, and would never get any principal whatsoever.

Now, I admit this example is a gross simplification. In real life, losses might occur over a more drawn out period. We'd expect to get some recovery. In addition, some of the excess interest garnered in the early part of the ABS deal might be used to keep the Sub notes current for a while longer than I'm assuming here.

But the point is that each time you add on structure, the losses get redistributed. When the losses are redistributed, some one has less risk, but some one must have more. Sometimes much more.