Wednesday, July 6, 2011

Built to Fail CDOs 101: How Well Do You Know Your CDS Counterparty?

The Abacus CDO story of 2010 brought to the fore a worrisome scenario in which it could be argued that the arranging bank (Goldman Sachs) played two different roles at once, potentially serving one particular client (Paulson) at the expense of other clients (investors in the Abacus CDO). Goldman settled the case with the SEC for $550mm. What could be worse than participating in such a conflicted scenario? We are concerned that in a number of deals the arranging bank may have positioned itself directly against the CDO investors. In other words, the bank, like Paulson, may have been betting against its clients.

But first, let’s take a step back to explain how this all works…

A CDO is called a “synthetic CDO” when the underlying assets are “synthetically” referenced, rather than being held like physical corporate bonds. The underlying assets are often referenced by way of credit default swaps, or CDSs, and are called “reference obligations.” These CDSs may reference several types of asset classes, but in the synthetic CDO setting they typically reference either corporate debt or structured finance securities, such as commercial mortgage-backed securities (CMBS), residential mortgage-backed securities (RMBS), or even other CDO tranches. In the Abacus deal, the reference obligations were credit default swaps struck on RMBS.

Instead of buying physical assets that pay coupons (when current), the synthetic CDO sells protection on a portfolio of reference obligations. Much like insurance contracts, the buyers of protection on each underlying CDS make periodic “premium” payments to the CDO in exchange for compensation if and when a default, or credit event, occurs with respect to the obligation being referenced.

The CDO’s immediate counterparty on each CDS – typically the arranging bank – often plays an intermediary role between the CDO and each of its CDS transactions. It buys protection from the CDO and sells protection to the end buyer. This layout allows for the CDO to focus solely on the counterparty risk (i.e., the risk that a party will fail to fulfill its obligations under the CDS agreement) of a single party – in this case the arranging bank – as opposed to that of each end buyer (of protection).

Ideally, this dynamic ought to create an environment in which the immediate CDS counterparty (the arranging bank) is neutral to the performance of the CDO as the bank is fully hedged (as long as end buyers do not default).[1]

The imposition of an intermediary CDS counterparty often masks the identity of the end buyers from those who invest in CDO notes, potentially rendering CDO investors unable to discern which parties are ultimately short their portfolio.

Goldman Sachs’s now famous Abacus CDO illustrates a serious danger that can arise from the above confusion. The argument could be made that had investors known that Paulson was the end buyer of protection on a significant portion of Abacus’ CDS portfolio, they may have reconsidered the prudence of their investment, and potentially shunned it.

Built to Fail, Profitably

But what happens if the arranging bank chooses not to off-load all positions to an end buyer? In other words, what happens if the bank retains some or all of the short exposures to the underlying reference obligations? Here, the end buyer of protection, and the immediate CDS counterparty are one and the same: the arranging bank. The bank is now effectively short the CDO.

For example, the plaintiff in re: Space Coast Credit Union vs. Barclays Capital et al argues that:
“[the] facts here leave no doubt there was clear intent to create a very large short bet through Markov against Mezzanine CDO risk”
and that:
the “Defendants were extraordinarily determined to stuff Markov [CDO] with Mezzanine CDO risk.”
Plaintiff argues that:
“most stunning of all, [the Defendant] was so intent on Mezzanine CDO failure that it custom-built $300 million of built-to-fail Mezzanine CDOs … that [the Defendant], through Markov, could then bet against.”

While we do not seek to verify the accuracy of their contention, we are keenly aware of the material conflict such a scenario would present: the arranging bank is short the securities, meaning it would be financially rewarded if those securities were to plunder. The bank would benefit from selecting poor-quality assets. At the same time, the arranging bank is selling CDO notes, supported by these assets, to its clients. If the assets fail, the bank profits at the expense of the CDO noteholders – its clients. If the assets perform well, the bank would suffer financially.

From a higher level fiduciary perspective, the bank’s financial motive would not be aligned with the well-being of its client. Nor would the bank be even indifferent to the performance of its client. Rather, the bank’s profitability would be in direct opposition to that of its client.

While their clients were losing money on the trade, how much were bank profiting?

Removing the time value of money and the default timing as inputs to the model, we can create a simple model to estimate the bank’s profits from this trade. The model assumes that 100% of the assets are synthetically referenced.

Suppose the total premiums being paid were P, and that a bank held the super senior swap, with attachment point AP. The higher the attachment point, the greater the potential for the bank to make money: if losses exceed AP, the bank's profits are capped, as the profits from its short positions mimic identically the losses from its super-senior position.

In dollar terms, suppose the deal is of size $1bn, with an average 1% credit premium (P) on the reference obligations and a super-senior attachment point (AP) of 50%.

Suppose for simplicity that all losses occur within the first year.

If losses (AL) are lower than 1%, say they’re 0%, the bank loses 1% x $1bn = $10mm. Thus, if the portfolio is well selected, the bank stands to loses up to $10mm.

But if the portfolio is poorly selected, and suffers losses over 1%, the bank cashes in handsomely. At 5% losses, the bank makes 4% of $1bn, or $40mm. At 50% losses, the super senior attachment point, the bank caps out at 49% of $1bn, or $490mm. (Profits are maxed out at the 50% AL level as, in this example, the bank holds the super senior swap.)

A bank can either lose up to $10mm for doing a really good job of diligently selecting good assets, or the bank can make as much as $490mm for selecting really bad assets. Would you expect any bank to do the former?


[1] If anything, the CDS counterparty ought to have a slight preference for the continued performance of each CDS contract, as a default would cause settlement and thereby cut short any intermediation fees it may be earning as a middle-man.


Anonymous said...

"A bank can either lose up to $10mm for doing a really good job of diligently selecting good assets, or the bank can make as much as $490mm for selecting really bad assets. Would you expect any bank to do the former?"

And yet they did. In droves. They let super senior tranches of CDOs and synthetic CDOs stay on balance sheet or parked them in SPVs - Citi was notoriously bad.

This risk retention was massive compared to the problems you describe. I'm afraid you have it entirely backwards.

PF2 said...

@ Anonymous: I think we would disagree on this.

Our posting doesn't take a position on the frequency with which banks held super senior swaps. We're also not seeking to evaluate their profits or losses on deals in which they're not short (any or all of) the underlying.

We're commenting only on a subset of the deals originated, and the potential from profiteering within that band; I think if you look through this again with our perspective in mind, you may come to agree with us on this one.

Thanks for your comments