I'm working through a deep dive of AMBAC's insured positions, which they have helpfully provided in great detail. Before I get to that, I'd like to go over how bond insurance works in the ABS world as well as what the prospects are for recovery post-default from the insurer's perspective.
First of all, insurance written against ABS, MBS and CDOs (which I'm just going to refer to as ABS unless otherwise specified) comes in the form of a pay-as-you-go CDS. Put simply, this means that the insurer pays out to bond holders as interest or principal shortfalls occur. A "principal shortfall" would include any write-down, not just when the principal is legally due. This differs from a classic CDS contract, where the seller of protection (in essence, the insurer) buys the defaulted security at par. This means that the insurer will payout claims over time, not all at once. In addition, like any CDS, the insurance premium is collected over time as well. This is in contrast to municipal insurance policies, where the premium is generally collected up front.
It should also be noted that the insurers account for these policies as derivatives under GAAP guidelines. Pertinent here is that means they are marked to market. You can judge for yourself whether they are likely to be marked properly or not, but fair to say that the insurers are likely to mark down the CDS contract in advance of paying out any actual cash.
Most of the bond insurers' exposure to ABS is at the top of the structure. Here I will present a simple model for how bonds they might have insured would perform under stressed default scenarios. While this will not be at all comprehensive, it should frame the discussion of insurer losses going forward.
OK, so let's start with a simple model of $100 million RMBS off subprime collateral. We'll use the same subordination levels as in my previous post on the dangers of structure squared...
- Senior: 5.75% coupon, $80 million
- Mezz: 6.50% coupon, $15 million
- Subordinate: 8.00% coupon, $5 million
Let's say there was originally a net 6.75% coupon on the collateral portfolio. Now, let's assume that 20% of the collateral portfolio defaults immediately with no recovery, but the rest pays normally. Let's assume that the senior bond was insured, and look at what the insurer's loss position would look like.
Remember that just because 20% has defaulted, technically the obligation to pay the Mezz and Sub pieces doesn't go away. They might never get any cash flow, but the obligation would remain.
So the deal would collect 6.75% * $80 million ($20 million defaulted) or $5.4 million per year (ignoring paydowns for the moment). The Senior tranche is owed 5.75%*$80 million or $4.6 million. So even though the Mezz and Sub bonds are seeing very little of the interest they were originally owed, and are unlikely to ever see any principal, the Senior is doing fine. The insurer would not wind up paying out any cash on this deal.
Of course, the CDS on the Senior would have risen in value (bad for the insurer, who is short the CDS) substantially. The Senior bond has gone from having 20% subordination to zero, obviously the risk profile has increased markedly.
Now let's consider a deal with less subordination at origination. This would probably be because the collateral was considered stronger at the outset. Maybe it was a deal made up of "prime" loans, some of which were stated income. How the market ever looked at stated income as "prime" I'll always wonder. Anyway, let's say the structure looked like this:
- Senior: 5.75% coupon, $90 million
- Mezz: 6.50% coupon, $7 million
- Subordinate: 8.00% coupon, $3 million
Since this deal had "stronger" collateral, the coupon would be lower, say 6.25%. Again, let's say that 20% defaults immediately with no recovery.
So the deal collects $5 million in interest per year (6.25% * $80 million). The Senior is owed $5.175 million. If this Senior was wrapped the insurer would have to pay the $175,000 each year. If we assume the $20 million defaulted was written down to zero, the insurer would likely have to pay $10 million to the Senior holders (the other $10 million is Mezz and Sub's problem). In that case then the Senior would only be owed 5.75%*$80 million, or $4.6 million, in interest each year, because the $10 million of Senior notes written down is in effect a paydown of principal. That means that the insurer would have no on going interest expense but would continue to collect premiums on the CDS contract.
What happens if the loss occurs over time, which is obviously more realistic? This will especially be true with RMBS deals with longer-reset ARMs as collateral. Most deals with a 5 or 7 year fixed period won't reset for several years yet, and it may be that defaults will remain manageable until we get closer to reset. Normally, every dollar of principal repaid goes to pay off the Senior note holders until those notes are retired, which is called sequential pay. Some deals pay pro rata, but even those usually switch to paying sequentially once the deal suffers a certain number of defaults. Given the environment, the odds are good most deals will hit this trigger.
Anyway, the insurer's position is improved by prepayments. Not only is the par amount insured decline, but the percent subordination also improves. Consider a deal with $80 million in the Senior note and $20 million in other notes, for 20% subordination. If $5 million in principal is repaid, that leaves $75 million in Senior notes and still $20 million in other, or 21% subordination. The older the deal, the more this element is benefiting Senior holders.
Finally, let's consider CDOs. I think a CLO (which has bank loans as collateral) would perform similarly to what's presented above from an insurance perspective. CLO deals own the bank loans directly, just like a RMBS deal owns the mortgage loans directly. Same goes for TRUP deals and some commercial real estate deals. However, an ABS CDO is structure built on top of structure, which creates new problems, as discussed here.
So here is a sample ABS CDO structure. Again, we'll assume the Senior is wrapped and look at the loss situation.
- Senior: 5.45% coupon, $80 million
- Mezz: 6.00% coupon, $12 million
- Sub: 8.00% coupon, $4 million
- Equity: $4 million
If we assume the deal had 50/50 Mezz and Sub pieces, then the coupon would probably be about 7.25%. This time, we assume that 20% of the actual loans default. Assuming the underlying ABS were structured like the first RMBS deal discussed above (the 80/15/5) then 20% loan defaults would cause an interest short fall looking like this:
UNDERLYING DEAL CASH FLOW:
- DEALWIDE INTEREST: 6.75% * $80 million = $5.4 million
- Senior: 5.75% * $80 million = $4.6 million (satisfied in full, leaves $800,000)
- Mezz: 6.50% * $15 million = $975,000 (suffers $175,000 shortfall, or 18% of what's owed)
- Sub: Nada
CDO CASH FLOW
- DEALWIDE INTEREST: $50 million in Mezz pieces gets $2.667 million in interest (~82% *$50 million * 6.50%) and el zilcho on the $50 million in Subs.
- Senior: 5.45% * $80 million = $4.36 million (paid only $2.667 or about 61% of what's owed)
- Mezz and Sub = Confederate money.
The insurer has to make up the ~$1.7 million in interest short fall. More likely is that the insurer winds up paying out some amount of principal to the Senior note holders up front, as described above. Unfortunately, it isn't as clear when (or how much in) write downs need to occur in CDOs, because complications like overcollateralization and triggers make the principal repayment position of the Senior note holders more complicated. In fact, technically a pay-as-you-go CDS can result in the insurer paying principal to note holders, but later note holders paying some back due to better-than-expected recovery.
Anyway, what have we learned here? Well, starting with a very bearish scenario (20% immediate defaults with zero recovery), the performance of Senior notes in most ABS sectors is pretty good. The insurers would likely suffer significant write downs, because the CDS contracts will rise in value, forcing the insurer to mark-to-market their short CDS position. But their actual cash flow won't be too bad. The sectors that should perform OK are:
- Senior notes with RMBS, HELOC, and other ABS collateral.
- Direct pools of RMBS where the insurer has some subordination and/or overcollateralization as protection.
- Older ABS deals, where the insured position has already enjoyed some paydowns.
- CLO and other CDOs where the collateral is direct credit exposure, and not a repackaging of structure.
The exception is in ABS CDOs, where the structure squared problem really could hit hard. I showed a stylized example of what a 20% default rate in underlying collateral would look like, but that doesn't tell the whole story. Defaults could be higher, but occur over time. Defaults could be lower, but concentrated in higher coupon debt and therefore still cause large interest shortfalls. Defaults could be lower causing cash to flow to the junior tranches, only to later see defaults ramp up. Or defaults could be better in some deals and worse in others.
You can also see that if the ABS CDO had been made up of more Mezz and less Sub, the cash flow shortfall would be greatly reduced. A 100% Mezz deal would actually manage to pay the Senior holders in full.
The problem in trying to model losses in ABS CDOs is that even the most minute change in structure, default rate, recovery rate, and default timing makes a giant difference in CDO performance.
So when I finally finish my look at AMBAC, I'm going to assume the worst for all their ABS CDO, but something more modest for other CDO exposures. I'm also going to assume defaults come in at extreme levels for RMBS deals, but given the analysis above, I think those losses will be manageable.
Disclosure: No holdings in any bond insurer directly. I own various municipal credits which have been wrapped.