By Thomas Adams, an attorney and former monoline executive, and Yves Smith
Our accompanying post at Naked Capitalism describes, at a high level of abstraction, a data compilation and analysis that shows that a substantial majority of the transactions in Maiden Lane III are in the public domain. These transactions have long been a focus of controversy. When the Fed arranged for AIG to pay out 100% on credit default swaps with bank counterparties, part of that payment came through the sale of the underlying CDOs to Maiden Lane III (yes, that’s a simplification, see here for details). Moreover, the redactions from AIG’s SEC filings tried to keep these same transactions secret.
As that article describes, this analysis is significant because the Fed has claimed the Maiden Lane III transaction-level information is confidential, when as we demonstrate, substantial transaction level detail about AIG’s CDOs is already public, and the gaps can be identified and may be able to be filled in via other sources.
We discuss in more detail here the information we have developed on these deals.
We have put a model summary and detail from a subset of transactions on Scrib ID. This information provides insight into the Maiden Lane CDO portfolio, on issues such as vintages, collateral types, deal managers, bankers and counterparties, that previously could only be surmised from partial disclosures. The detail also shows how much can be gleaned from public sources and and illustrates vividly how bogus the Fed’s call for secrecy is.
Note this model is “confirmed data only” where all items come from public sources. We have a second model, where we make reasonable extrapolations (for instance, there is no public source data for par values on CP, or commercial paper “classes”, meaning tranches, but they can be computed readily. Par values for CP classes (8 transactions in total) and Goldman Sachs Abacus deals (its synthetic CDO program) are the main gaps in transaction level detail in the “confirmed data only” spreadsheet.
The AIG memo identifies $66 billion in par value in “multi sector” CDOs held by 9 counterparties, which consisted mainly of so-called “high grade” ABS CDOs, which had AA and A tranches of subprime bonds as a major constituent (the rest were “mezz” CDOs). Despite the name, high grade CDOs have on the whole not fared much better than so-called “mezz” CDOs, due to the fact that high grades were more CDO-squared like than “mezz” (they frequently contained 15% to, in the last gasps of the bubble era, as much as 40% of “inner”, meaning AA or lower CDO tranches). The other constituent elements of a high-grade CDO were generally worse than the non-subprime parts of a “mezz” CDO, since they were mainly seen as providing more diversification, while for a “mezz” deal (which consisted heavily of BBB subprime tranches), some better-quality instruments were included to improve credit strength.
The $66 billion is more than the amount of principal value of high-grade and mezz CDOs that went into Maiden Lane III. The total principal amount was $62.1 billion, and of that, roughly $10.8 billion was commercial real estate CDOs (see p. 3, we believe the missing par value can be extrapolated). That would leave $51.3 billion in high grade and “mezz” CDOs. While some of the oldest CDOs could have terminated between November 2007 and the time of the rescue, the bulk, probably all, of the difference is due to selection: the worst CDOs went to Maiden Lane III, the rest presumably are at AIG. This idea is confirmed by the ratings. As we showed earlier this week, virtually all of the high-grade CDOs had been downgraded to BB+ or lower (and although we have yet to add current ratings to the spreadsheet, so far most of the ones we have found are CCC or DDD, total washouts), which confirms the notion that the strongest ones are at AIG. The notion that some CDOs were kept at AIG is also confirmed by the fact that the Fed disclosed that the Maiden Lane III ABS CDO payouts were made on a portfolio of ABS CDO exposures whose size was larger than the par value of the Maiden Lane facility.
We have a second model, with information that builds on the initial model by adding extrapolated or derived information. In that model, we have also estimated the current value of each transaction, using the ABX index and constructing implied portfolios by year of issue and type of CDO, and using a subprime index, the ABX, as a guide. While the ABX is not a perfect proxy for the value of the underlying collateral, it is the most transparent indicator available today and it is widely used as a tool to estimate mortgage and CDO deal values. We tested this approach by applying it to the transactions at a date we thought would represent when marks were made for the AIG memo, and it delivered prices pretty close to those AIG and its counterparties used. A professional who is currently valuing CDOs also reviewed the methodology and results. Again, this confirms how much can be gleaned from information in the public domain.
This analysis still has some anomalies and gaps:
Our information is based on par at origination. Some of the exposures, especially for the older transactions, may have amortized since closing, although this amortization would be slowed by the mechanics of principal lock out for subordinated bonds in MBS transactions.
We have not included the commercial real estate CDO transactions yet. This is a comparatively small portion of the entire portfolio. The bulk of the value is in two large transactions (identified) that total $7.5 billion of par value. It appears tht the par value of the “other” CRE CDOs can be extrapolated and is roughly $3 billion of par value of commercial real estate CDOs.
Deutsche Bank is listed as a counterparty on only $600 million of ABS CDOs under discussion in November 2007, yet received payment of over $6 billion from the Maiden Lane transaction. Perhaps this is because Deutsche Bank was a counterparty on many of the CRE CDOs. If this is correct, this strongly implies that the $3 billion of “other” CRE CDOs mentioned immediately above are the only transactions in this portfolio that have not been identified.
The AIG memo indicates $5.2 billion of exposure to individually identified Abacus deals but publicly available information on these deals appears to indicate a smaller amount of total issued par, potentially indicating incomplete information in the AIG memo. The Abacus deals were sponsored by Goldman, though some used other third party deal managers.
The AIG memo also does not mention the total amount of collateral calls for the Société Générale counterparty exposure, whereas this amount is listed for all of the other counterparties. The collateral calls also appear to be a smaller amount than the marks (after taking into account the thresholds) would indicate. Put another way, based on the marks, the collateral calls should have been larger than they were, in aggregate and by counterparty, than they were described in the AIG.
As discussed in the AIG memo, the marks provided by the counterparties ranged widely, even for similar transactions. While the Goldman marks appear to be the most aggressive, other counterparties seemingly lobbed in softball marks of close to 100% even on 2006 mezzanine deals, which were likely to be in distress by late 2007. For example, for 2006 mezzanine deals, the counterparty provided marks ranged from 55% to 100%. Generally, the marks provided for the mezzanine deals were generous, in light of serious deterioration in the underlying collateral by that time. As a result, the discrepancies between our implied marks and the counterparty provided marks are the greatest for the mezzanine deals.
Certain of the insured classes had multiple counterparties, thus obscuring our ability to calculate the exposure by counterparty. Allocation within a class to multiple counterparties is not disclosed publicly.
The data also highlights the distribution of the transactions across deal managers or advisors. Trust Company of the West, TCW, was the deal manager for over $10 billion of AIG’s exposures, over twice the next largest deal manager. The concentration of deals from the top managers highlights the cozy relationships AIG had with these managers and, to a degree, the consequences of such concentrations.
Very little information was available on the Abacus deals in contrast to the other transactions (and even in contrast to information on other Abacus deals).
While our information on the lead bankers for the transactions is not as complete as we would like, it does provide some insight into other interesting relationships in the portfolio. Frequently, the bank which underwrote the transaction would end up as the counterparty with AIG. However, in many cases, other banks would be the counterparty, which provides clues on the interconnections in the CDO market and the lack of distribution for this product. Even when a lead bank found a third party purchaser for the CDO, it wound up being one of a small list of banks already participating in AIG’s transactions. Some banks, such as Merrill and Goldman, were frequent sellers and buyers of the AIG insured bonds.
We also want to challenge the Fed’s oft-touted notion that it is necessary to keep transaction-level detail in the Fed’s various special bailout vehicles (Maiden Lane I, the Bear Stearns vehicle, Maiden Lane II, which holds RMBS related to AIG’s secured lending portfolio, and Maiden Lane III) secret. As we demonstrate here, this information was not secret in the case of Maiden Lane III, but the repeated assertion that it was helped discourage further investigation.
The Fed claims that exposing the original and estimated current value of the Maiden Lane III CDOs would reduce the ability to realize maximum value from the entity. But that is spurious. In real estate, for instance, the ability to determine the owner’s purchase price and its appraised value is distinct from what a buyer might offer and a seller might accept. No one in New York City, for instance, believes that the owner of a condo (where buyer can find what the owner paid) has more negotiating leverage than the owner of an apartment in a co-op (where that information is not available). Any possible buyer of a CDO would do his own due diligence and valuation.
Knowing someone’s position size and cost basis can a give a buyer an advantage in liquid market (where dealers front run, or as they would more politely put it, get out of the way if they see inventory being sold and they think more is coming). But these are bespoke transactions in an illiquid market. The seller will have an idea of what its wares might fetch, and shops for bids. And even if a seller has a large position, if he can afford to wait and signals patience, he becomes much harder to exploit. And who can greater staying power than a central bank?
But more mundane factors make the “traders can take advantage of us” argument even more dubious. There is simply very little in the way of bids for CDOs. They are very difficult and costly to value if one does it correctly, and with so many moving parts, it is easy to be very wrong. And wrong often means you recover nothing. So the implicit idea, that the Fed’s exit is a sale of the CDOs as CDOs, as opposed to via liquidation, is also questionable.
So the reason for secrecy may simply be imperial reflex. But it might be to disguise the fact that the Fed is at risk of taking a loss on the Maiden Lane III loans, by showing that its current marks are unrealistic. While a mere $24 billion loan might seem unimportant in the context of a $2 trillion balance sheet, if Maiden Lane III’s valuation look generous, that calls into doubt the loans made to Maiden Lane I (the entity for Bear Stearns toxic waste) and Maiden Lane II (which holds some problematic AIG mortgage exposures). And the Fed is hoovering up such large amounts of paper on its own balance sheet, some of it outside its traditional, very stringent standards for lending, that small percentage losses there can add up to big dollar amounts. As Willem Buiter, former central banker, now chief economist of Citigroup, has pointed out, if the Fed takes enough losses on its forays into risky transactions, it will need to be recapitalized by the Treasury, meaning in the end by taxpayers. While the Fed in theory can “print” its way out of any credit losses, in practice that is constrained by the Fed’s inflation mandate:
….even if the central bank prices the private securities it purchases appropriately (that is, there is no ex ante implicit quasi-fiscal subsidy involved), it is possible that, should the private securities default, the central bank will suffer a capital loss so large that the central bank is incapable of maintaining its solvency on its own without creating central bank money in such quantities that its price stability mandate is at risk. Without a firm guarantee up front that the Federal government will fully re-capitalise the Fed for losses suffered as a result of the Fed’s exposure to private credit risk, the Fed will have to go cap-in-hand to the US Treasury to beg for resources.
Buiter argued that a recapitalization would cost the Fed its independence. Its high-handedness on secrecy looks likely, and deservedly, to produce the same outcome.