Page 296 - untitled

LATE  TO EARLY : BILLIONS IN SUBPRIME LOSSES 

Gary Gorton, then a finance professor at the University of Pennsylvania’s Wharton
School, who began working as a consultant to AIG Financial Products in  and
was close to its CEO, Joe Cassano. The Gorton model had determined with .
confidence that the owners of the super-senior tranches of the CDOs insured by AIG
Financial Products would never suffer real economic losses, even in an economy as
troubled as the worst post–World War II recession. The company’s auditors, Pricewa-
terhouseCoopers (PwC), who were apparently also not aware of the collateral re-
quirements, concluded that “the risk of default on [AIG’s] portfolio has been
effectively removed and as a result from a risk management perspective, there are no
substantive economic risks in the portfolio and as a result the fair value of the liability
stream on these positions from a risk management perspective could reasonably be
considered to be zero.” 
In speaking with the FCIC, Cassano was adamant that the “CDS book” was effec-
tively hedged. He said that AIG could never suffer losses on the swaps, because the
CDS contracts were written only on the super-senior tranches of top-rated securities
with high “attachment points”—that is, many securities in the CDOs would have to
default in order for losses to reach the super-senior tranches—and because the bulk
of the exposure came from loans made before , when he thought underwriting
standards had begun to deteriorate. Indeed, according to Gene Park, Cassano put a

halt to a  million hedge, in which AIG had taken a short position in the ABX in-
dex. As Park explained, “Joe stopped that because after we put on the first  . . . the
market moved against us . . . we were losing money on the  million. . . . Joe said,
‘You know, I don’t think the world is going to blow up . . . I don’t want to spend that
money. Stop it.’” 
Despite the limited market transparency in the summer of , Goldman used
what information there was, including information from ABX and other indices, to
estimate what it considered to be realistic prices. Goldman also spoke with other
companies to see what values they assigned to the securities. Finally, Goldman
looked to its own experience: in most cases, when the bank bought credit protection
on an investment, it turned around and sold credit protection on the same invest-
ment to other counterparties. These deals yielded more price information. 
Until the dispute with Goldman, AIG relied on the Gorton model, which did not
estimate the market value of underlying securities. So Goldman’s marks caught AIG
by surprise. When AIG pushed back, Goldman almost immediately reduced its July
 collateral demand from . billion to . billion, a move that underscored the
difficulty of finding reliable market prices. The new demand was still too high, in
AIG’s view, which was corroborated by third-party marks. Goldman valued the
CDOs between  and  cents on the dollar, while Merrill Lynch, for example, val-
ued the same securities between  and  cents. 
On August , Cassano told PwC that there was “little or no price transparency”
and that it was “difficult to determine whether [collateral calls] were indicative of true

market levels moving.” AIG managers did call other dealers holding similar bonds
to check their marks in order to help its case with Goldman, but those marks were
not “actionable”—that is, the dealers would not actually execute transactions at the