... was given to the investors, all of the information that was out there, and the credit rating
Id. Mr. Sparks went on to say: “I 2177 mentioned we made some bad business decisions. These deals performed
horribly. That is bad. ... [T]hat said, just because one person in my business unit or a few people might have had one
view, I can tell you there were a lot of people in my business unit that had a very different view, and there were a lot
of investors that had a very different view.” Id.
2178 See 5/19/2007 draft Goldman presentation, “Mortgages CDO Origination – Retained Positions & Warehouse
Collateral, May 2007,” GS MBS-E-010951926.
2179 2/8/2007 email from Mr. Sparks, “Post,” Hearing Exhibit 4/27-7.
2180 3/3/2007 email from Mr. Sparks, “Call,” Hearing Exhibit 4/27-14.
2181 3/8/2007 email from Mr. Sparks, Mortgage risk,” Hearing Exhibit 4/27-75.
2182 3/12/2007 Goldman memorandum to Firmwide Risk Committee, “March 7th FWR Minutes,” GS MBS-E-
00221171, Hearing Exhibit 4/27-19.
2183 2/21/2007 email from Daniel Sparks to Jon Winkelried, “Mortgages today,” GS MBS-E-010381094, Hearing
2184 April 27, 2010 Subcommittee Hearing at 66. The Anderson, Timberwolf, and Abacus 2007-AC1 CDOs were
issued in March and April 2007.
2185 4/27/2007 email from Deeb Salem to Michael Swenson, GS MBS-E-012432706 (6 of 20 deals in ABX Index
put on watch or downgraded).
2186 7/10/2007 email to George Maltezos, “GS Cashflow/ABACUS CDOs Mentioned in S&P Report on CDO
Exposure to Subprime RMBS,” GS MBS-E-001837256; 7/10/2007 email from Goldman Sachs analyst to Goldman
Sachs Japan salesman, “GS Cashflow/ABACUS CDOs Mentioned in S&P Report on CDO Exposure to Subprime
RMBS,” GS MBS-E-001990255 (updating percentage exposure from 25% to 35%).
Mr. Sparks: Well, I generally feel that the disclosure for the new issues that Goldman Sachs
brought was good.”2177
While Mr. Sparks testified that, in 2007, the Mortgage Department expected its CDOs “to
perform,” a contemporaneous draft presentation that he helped prepare in May 2007 stated that the
“desk expects [the CDOs] to underperform.”2178 Many other emails provide his negative views of
the CDO market at the time, including emails in which Mr. Sparks described the subprime market
as “bad and getting worse,”2179 and directed Goldman’s mortgage traders to “get out of
everything,”2180 and “stay on the short side.”2181 He wrote, among other things: “Game over,”2182
“bad news everywhere,” and “the business is totally dead.”2183 As Senator Tester noted, many of
Mr. Sparks’s dire predictions were made before three of the four CDOs discussed at the hearing
were even offered to customers.2184
Mr. Sparks also testified that the Mortgage Department did not expect the Goldman-issued
CDOs to be downgraded, but all were within a year of issuance. In April 2007, for example, six of
the 20 RMBS deals that comprised the ABX Index were downgraded,2185 and the Hudson CDO that
referenced them followed soon after. Many of the RMBS securities referenced in the other three
CDOs were downgraded within three months of the issuance of the last CDO in April 2007, making
the downgrade of the CDOs themselves all but inevitable.2186 For example, when Moody’s and
S&P announced their first mass downgrades of RMBS securities on July 10, 2007, the S&P
downgrade affected 35% of the assets in Timberwolf.2187 Ultimately, all of the CDOs discussed at
the Subcommittee’s hearing were downgraded to junk status. On October 26, 2007, a Goldman
employee sent an email about Abacus 2007-AC1, even noting this dubious distinction:
10/26/2007 email from Goldman salesman to Michael Swenson, “ABACUS 2188 2007-AC1 – Marketing Points
(INTERNAL ONLY) [T-Mail],” GS MBS-E-016034495.
2189 See “List of WaMu-Goldman Loans Sales and Securitizations,” Hearing Exhibit 4/13-47b.
2190 2/13/2006 Goldman chart, “Current Warehouse Facilities and Funded Balances,” GS MBS-E-001157934.
2191 See Chapters III and IV, above.
“This deal was number 1 in the universe of CDO’s that were downgraded by Moody’s and
S&P. 99.89% of the underlying assets were downgraded.”2188
(b) Goldman’s Conflicts of Interest
In late 2006 and 2007, Goldman’s securitization business was marked, not just by its hard
sell tactics, but also by multiple conflicts of interest in which Goldman’s financial interests were
opposed to those of its clients. The following examples illustrate the problem.
(i) Conflicts of Interest Involving RMBS Securities
In 2006 and 2007, Goldman originated 27 CDO and 93 RMBS securitizations. Beginning in
December 2006, Goldman originated and aggressively marketed some of these securities at the
same time that subprime and other high risk loans were defaulting at alarming rates, the subprime
and CDO markets were deteriorating, and Goldman was shorting subprime mortgage assets. At
times, Goldman originated and sold RMBS securities that it knew had poor quality loans that were
likely to incur abnormally high rates of default. At times, Goldman went further and sold RMBS
securities to customers at the same time it was shorting the securities and essentially betting that
they would lose value. Two examples illustrate how Goldman constructed and sold poor quality
RMBS securities and profited from the decline of the very securities it had sold to its clients.
Long Beach RMBS. The first example involves Washington Mutual Bank (WaMu) and its
subprime lender, Long Beach Mortgage Corporation. WaMu, Long Beach, and Goldman had
collaborated on at least $14 billion in loan sales and securitizations.2189 In February 2006, Long
Beach had a $2 billion warehouse account with Goldman, which was the largest of Goldman’s
warehouse accounts at that time.2190
Long Beach was known within the industry for originating some of the worst performing
subprime mortgages in the country. As explained in Chapter III, in 2005, a surge of early payment
defaults in its subprime loans required Long Beach to repurchase over $837 million of
nonperforming loans from investors, as well as book a $107 million loss.2191 Similar EPD problems
affected its loans in 2006 and 2007. WaMu reviews and audits of Long Beach, as well as
examinations by the Office of Thrift Supervision, repeatedly identified serious deficiencies in its
lending practices, including lax underwriting standards, unacceptable loan error and exception rates,
weak risk management, appraisal problems, inadequate oversight of third party brokers selling loans
to the firm, and loan fraud. While these reviews were not available to the public, the performance
of Long Beach paper was. Long Beach securitizations had among the worst credit losses in the
See, e.g., 4/14/2005 OTS email, “Fitch,” OTSWME05-2192 012 0000806, Hearing Exhibit 4/13-8a.
2193 2/8/2007 email from Goldman analyst to Mr. Sparks, Mr. Gasvoda, and others, “2006 Subprime 2nds Deals
Continue to Underperform **INTERNAL ONLY**,” GS MBS-E-003775340, Hearing Exhibit 4/27-167d.
2194 5/17/2007 email from Deeb Salem to Michael Swenson, “FW: LBML 06A,” GS MBS-E-012550973, Hearing
industry from 1999-2003; in 2005 and 2006, Long Beach securities were among the worst
performing in the market.2192
Nevertheless, in May 2006, Goldman acted as co-lead underwriter with WaMu to securitize
about $532 million in subprime second lien mortgages originated by Long Beach. Long Beach
Mortgage Loan Trust 2006-A (“LBMLT 2006-A”) issued approximately $495 million in RMBS
securities backed by those Long Beach mortgages. The top three tranches, representing about 66%
of the principal loan balance, received AAA ratings from S&P, even though the pool contained
subprime second lien mortgages – loans which could recover funds in the event of a default only
after the primary loan was repaid — and even though the loans were issued by one of the nation’s
worst performing mortgage lenders. Yet Goldman was able to use two-thirds of that extremely
risky debt to issue AAA rated securities which Goldman then sold to its customers.
In less than a year, the Long Beach loans began incurring delinquencies. In February 2007, a
Goldman analyst reported internally that all of Goldman’s 2006 subprime second lien RMBS
securities were deteriorating in performance, but “deals backed by Fremont and Long Beach
collateral have generally underperformed the most.”2193 The analyst predicted “lifetime losses in the
teens, and over 20% in some deals.” By May 2007, the cumulative net loss on the LBMLT 2006-A
mortgage pool had climbed to over 12%, eliminating most of the financial cushion protecting the
investment grade securities from loss. That month, S&P downgraded six out of the seven credit
ratings for the mezzanine tranches of the securitization. The Long Beach securities plummeted in
Goldman held some of the unsold Long Beach mezzanine securities on its books, meaning
LBMLT 2006-A securities that carried credit ratings of BBB or BBB-. Goldman had also
purchased the short side of a CDS contract that would pay off if those same securities lost value.
On May 17, 2007, Deeb Salem, a trader on the Mortgage Department’s ABS Desk, learned of
additional losses in the Long Beach securitization and wrote to his supervisor Michael Swenson
with the news:
“[B]ad news … [The loss] wipes out the m6s [mezzanine tranches] and makes a wipeout of
the m5 imminent. … [C]osts us about 2.5 [million dollars]. … [G]ood news ... [W]e own
10 [million dollars] protection at the m6 … [W]e make $5 [million].”2194
In other words, Goldman lost $2.5 million from the unsold Long Beach securities still on its books,
but gained $5 million from the CDS contract shorting those same securities. Overall, Goldman
profited from the decline of the same type of securities it had earlier sold to its customers.
2195 See wamusecurities.com.
2196 For more information about Fremont, see Chapter IV, Section D(2)(d).
2197 3/7/2007 Fremont General Corporation 8-K filing with the SEC.
2198 11/16/2006 Goldman internal email, “ACA and Freemont [sic] deal,” Hearing Exhibit 4/27- 173.
2199 3/14/2007 Goldman email, “NC Visit,” GS-MBS-E-002048050.
2200 Id.; see also 8/10/2007 email from Michelle Gill, “Fremont - Incremental Information,” GS MBS-E-009860358
(Goldman’s repurchase claims against Fremont would have amounted to a 9% ownership stake in Fremont after a
proposed buyout by investor group; Goldman was not the largest purchaser of Fremont loans but its repurchase
claims were 3-4 times larger than the claims of the nearest counterparty).
2201 See Goldman Sachs response to Subcommittee QFR at PSI_QFR_GS0040.
2202 2/20/2007 Goldman memorandum to Mortgage Capital Committee, “Request for renewal of the existing $1
billion ... 1-year revolving warehouse facility,” GS MBS-E-001157942. Goldman wrote the Fremont produced
“revenues totaling $13.38 million in 2006 of which $620,000 came in the form of warehouse usage and commitment
By May 2008, even the AAA securities in LBMLT 2006-A had been downgraded to default
status. By March 2010, the securities recorded a cumulative net loss of over 66%.2195
Fremont RMBS. The second example involves Fremont Loan & Investment, another
subprime lender notorious for issuing poor quality loans.2196 In March 2007, Fremont reported in an
8-K filing with the SEC that the FDIC filed a cease and desist order to which the company
consented. Among other matters, it order Fremont to stop “[m]arketing and extending
adjustable–rate mortgage products to subprime borrowers in an unsafe and unsound manner that
greatly increases the risk that borrowers will default on the loans or otherwise causes losses.”2197
Even before the actions taken by regulators in March 2007, Goldman was aware of the poor
quality of at least some of Fremont’s loans. In a November 2006 exchange of emails, for example,
two Goldman sales representatives were discussing trying to sell Fremont RMBS securities to a
client. One salesperson forwarded to the other the client’s explanation of why it did not want to buy
the securities and its low opinion of Fremont’s loan pools: “Fremont refused to make any forward
looking statements so we really got nothing from them on the crap pools that are out there now.”2198
In March 2007, Goldman initiated a detailed review of its Fremont loan inventory to identify
deficient loans that it could return to the lender for a refund. Mr. Gasvoda placed a priority on
reviewing Fremont loans “since they still have cash but may not for long.”2199 One loan pool review
conducted on March 14, 2007, found that “on average, about 50% of about 200 files look to be
repurchase obligations,” meaning that fully half of the reviewed loans should be returned to the
lender.2200 Goldman eventually made about $46 million in repurchase requests to Fremont, which
was one of the top five mortgage originators from whom Goldman made repurchase requests in
2006 and 2007.2201
Despite these and other indications of Fremont’s poor quality loans, Goldman continued to
underwrite and market securities backed by Fremont loans. In an internal February 2007
memorandum to its Mortgage Capital Committee, Goldman wrote that it had a “significant
relationship with Fremont,” based upon past securitizations, whole loan purchases, and warehouse
fees.2202 In March 2007, at the same time it was sending millions of dollars in loan repurchase
requests to Fremont, Goldman securitized over $1 billion in Fremont subprime loans in one of its
See 2/2/2007 Goldman memorandum to the Mortgage Capital Committee, 2203 “Agenda for Monday, February 5,
2007,” GS MBS-E-002201064; 2/5/2007 Mortgage Capital Committee Memorandum regarding GSAMP Trust
2007-FM2, GS MBS-E-002201055 - 58.
2204 1/24/2007 S&P internal email, “Quick Question: Fremont,” Hearing Exhibit 4/23-93b. See also 2/1/2007 S&P
internal email, “Defaults cause Fremont to end ties to 8,000 brokers,” Hearing Exhibit 4/23-93d (S&P analysts
circulated an article about how Fremont had stopped using 8,000 brokers due to loans with some of the highest
delinquency rates in the country).
2205 1/24/2007 S&P internal email, “RE: Quick Question: Fremont,” Hearing Exhibit 4/23-93c.
2206 4/17/2010 S&P downgrade of GSAMP Trust 2007-FM2 containing Fremont mortgages, Hearing Exhibit 4/23-
2207 See 3/21/2007 Goldman spreadsheet, “RMBS CDS Trade History 19Jan06 - 19Mar07 v3,” GS MBS-E-
2208 See Standard & Poor’s www.globalcreditportal.com.
2209 In some cases, Goldman used assets from its own inventory or warehouse accounts, so that it could transfer
assets with falling value to the CDO and the investors who purchased the CDO securities. The Subcommittee
examined seven of those CDOs, and of those, 57% of the CDOs’ assets were sourced from Goldman, including over
$3 billion in synthetic assets in which Goldman was the short party, and therefore stood to profit from a decline in
the value of the underlying assets.
warehouse accounts, originating GSAMP Trust 2007-FM2.2203 At Goldman’s request, Moody’s and
S&P rated the securities, even though analysts at both rating firms expressed concern about the
quality of Fremont loans. At S&P, for example, in a January 2007 email to his supervisor, a credit
ratings analyst wrote: “I have a Goldman deal with subprime Fremont collateral. Since Fremont
collateral has been performing not so good, is there anything special I should be aware of?”2204 One
supervisor told him: “No, we don’t treat their collateral any differently,” while another wrote that,
as long as the analyst had current FICO scores for the borrowers, he was “good to go.”2205 Both
agencies gave AAA ratings to the top five tranches of the securitization.2206
Goldman marketed and sold the Fremont securities to its customers, while at the same time
purchasing $15 million in CDS contracts referencing some of the Fremont securities it
underwrote.2207 Seven months later, by October 2007, the ratings downgrades had begun; by August
2009, every tranche in the GSAMP securitization had been downgraded to junk status.2208
In both examples involving Long Beach and Fremont RMBS securities, Goldman obtained
CDS protection and essentially bet against the very securities it was selling to clients. In each case,
Goldman profited from the fall in value of the same securities it sold to its clients and which caused
those clients to suffer substantial losses.
(ii) Conflicts of Interest Involving Sales of CDO Securities
As with some of its RMBS securities, Goldman at times originated CDO securities using
assets that it believed were of poor quality and would lose value, and sold the securities at higher
prices than it believed they were worth.2209 In addition, Goldman took steps that created multiple
conflicts of interest with the clients to whom it sold the CDO securities, and placed its financial
interests ahead of those of its clients. Four examples involving the Hudson 1, Anderson,
Timberwolf, and Abacus 2007-AC1 CDOs illustrate the problems. Those problems include
troubling and sometimes abusive practices related to how Goldman designed the CDOs and selected
Mezzanine subprime RMBS assets are RMBS securities that carry a credit 2210 rating of BBB or BBB- or CDS
contracts that reference those types of RMBS securities. Mezzanine RMBS assets are riskier than AAA, AA, and A
rated RMBS securities, but less risky than those that carry, for example, BB, B, or CCC ratings.
2211 The ABX Index tracks the performance of a designated basket of 20 subprime RMBS securitizations. It consists
of five separate indices, each of which tracks a different subset of the RMBS basket, divided according to credit
ratings. The indices that track the mezzanine RMBS securities, for example, track the 20 RMBS securities that carry
BBB and BBB- credit ratings. In ABX assets, investors enter into CDS contracts in which one party takes the long
side, essentially betting that the ABX indices tracking the mezzanine RMBS securities will increase in value, while
the other party takes the short side, essentially betting that the indices will fall in value. Prior to establishing the
Hudson CDO, Goldman had taken the long side in a number of CDS contracts linked to the ABX indices tracking
mezzanine RMBS securities.
2212 10/2006 Goldman document, “Hudson Mezzanine Funding 2006-1, Ltd.,” Hearing Exhibit 4/27-87. The first
Hudson CDO, Hudson High Grade, was issued in September 2006. The second, discussed here, was Hudson
Mezzanine 2006-1, issued in October 2006. The third was Hudson Mezzanine 2006-2, issued in February 2007.
2213 Id. For more information about Goldman’s role as the liquidation agent in Hudson, see Section C(5)(b)(iii)AA,
2214 In synthetic CDOs, the cash proceeds from the sales of the CDO securities were used to purchase “collateral
debt securities.” Later, when cash was needed to make payments to a long or short party, those collateral securities
were sold, and the cash was used to make the payments. In the event the collateral securities could not be sold for
face (par) value, the collateral put provider paid the difference to the CDO. For more information about the role of
collateral put providers, see Section C(5)(b)(iii)BB, below.
2215 See 2/18/2008 Goldman document, “CDO Transactions (July 1, 2006 - December 31, 2007) in which Goldman
Sachs acted as underwriter,” GS MBS 0000004337 at 4338. Acting as the “senior swap counterparty” meant that
Goldman served as an intermediary between the Hudson CDO and the super senior investor. Acting as “credit
protection buyer” meant that Goldman initially took the short side of the CDO and, in the case of Hudson 1, kept
100% of the short side during the life of the CDO.
their assets; marketed and sold the CDO securities; designated the value of the CDO securities preand
post-sale; and executed its duties as liquidation agent and collateral put provider.
AA. Hudson Mezzanine Funding 2006-1
Hudson Mezzanine Funding 2006-1 (Hudson 1) was a $2 billion synthetic CDO that
referenced mezzanine subprime RMBS assets2210 and assets linked to the ABX Index.2211 This CDO
was the second in a series of three “Hudson” branded CDOs, which according to Goldman
marketing materials were intended “to create a consistent, programmatic approach to invest in
attractive relative value opportunities in the RMBS and structured product market.”2212 One key
feature of the three Hudson CDOs was that Goldman itself, without any third party participation,
selected the CDO’s assets, which were supposed to remain with the CDO until they reached
maturity or were deemed “credit risk assets,” at which point Goldman, acting as the liquidation
agent for the CDO, was responsible for selling them.2213 In each of the Hudson CDOs, Goldman
played multiple roles in its formation and administration, including selecting assets and serving as
the underwriter, initial purchaser of the CDO securities, collateral put provider,2214 senior swap
counterparty, and credit protection buyer.2215 In Hudson 1, Goldman took 100% of the short side of
the CDO, and when the Hudson 1 securities declined in value, Goldman made a $1.35 billion profit
at the expense of the clients to whom it had sold the Hudson 1 securities.
See Goldman 2216 response to Subcommittee QFR at PSI_QFR_GS0239.
2217 9/20/2006 email from Arbind Jha to Josh Birnbaum, GS MBS-E-012685289. See also 9/20/2006 Firmwide
Risk Committee Minutes, GS MBS 0000004472.
2218 See 10/24/2006 email from Jonathan Sobel to Tom Montag, Dan Sparks, and others, GS MBS-E-010919930
(“CDO should price tomorrow and is in good shape. ... We also are starting to see some short covering, which we
will sell into to further reduce our risk toward your 50% goal.”).
2219 Subcommittee interview of Michael Swenson (4/16/2010). Mr. Swenson told the Subcommittee that Goldman
had been long “several billion” in ABX in September 2006.
2220 8/9/2006 Firmwide Risk Committee Minutes, GS MBS-E-009682590; Subcommittee interview of Joshua
Birnbaum (10/1/2010). Mr. Birnbaum recalled a “directive” to reduce ABX exposure in the summer of 2006.
2221 See, e.g., 9/21/2006 email from Jonathan Sobel to Tom Montag, “ABX wider again today,” GS MBS-E-
009739145 (“Down about $10mm.”); 9/12/2006 email from Jonathan Sobel to Michael Swenson and Daniel Sparks,
“ABX,” GS MBS-E-012681410 (“The last post you gave me was this morning when you thought things were ‘firm’.
Now I find out that we’re down $6mm on the day. I understand things move, but you need to post me. Also, I want
to reduce this position.”); 9/9/2006 Firmwide Risk Committee Minutes, GS MBS 0000004468 (“Business continuing
to reduce volatile ABX position.”); 8/23/2006 Firmwide Risk Committee Minutes, GS MBS-E-009615593
(“Mortgages sold down another net 15% of their large ABX position.”).
2222 9/19/2006 email from Jonathan Sobel to Daniel Sparks, Michael Swenson, and Josh Birnbaum, GS MBS-E-
Transferring Risk. Hudson 1 was conceived and designed by Goldman to transfer the risk
associated with a large collection of ABX assets in its inventory, in which Goldman held the long
side of CDS contracts referencing mezzanine subprime RMBS securities that were tracked by the
ABX and that might go down in value. The objective of the CDO was to transfer the risk of
unwanted financial assets off of Goldman’s books. In response to questions from the
Subcommittee, Goldman explained that Hudson 1 was “initiated by the firm as the most efficient
method to reduce long ABX exposures.”2216 Contemporaneous notes from Goldman’s Firmwide
Risk Committee meetings also stated that Hudson 1 was an “exit for our long ABX risk.”2217
Goldman records show that the firm used Hudson 1 to short $1.2 billion worth of the ABX assets in
the firm’s inventory as well as $800 million in single name CDS contracts referencing subprime
RMBS securities carrying mostly BBB or BBB- credit ratings. Hudson 1 was one of several
methods Goldman used to transfer its risk associated with its subprime mortgage holdings during
the fall of 2006.2218
Conceiving Hudson 1. In the months leading up to the creation of Hudson 1, Goldman had
accumulated billions of dollars in ABX assets referencing mezzanine subprime RMBS securities.2219
By August 2006, Goldman management had decided that this ABX trade had “run its course,” and
directed the Mortgage Department’s ABS Desk to sell off its ABX holdings.2220 After several
weeks of effort, however, the ABS Desk was unable to find many buyers, and its ABX mezzanine
assets, which were dropping in value, were losing millions of dollars for the firm.2221
On September 19, 2006, Jonathan Sobel and Daniel Sparks called an 8:00 a.m. meeting to
discuss the ABX problem with Joshua Birnbaum, head of ABX trading, and Michael Swenson, head
of both the Structured Product Group (SPG) Trading Desk and ABS Trading Desk.2222 Mr.
Swenson was unable to attend, and in a later email from Mr. Birnbaum who recounted the meeting
to him, Mr. Sobel and Mr. Sparks wanted to know whether the Mortgage Department should sell all
9/19/2006 email from Josh Birnbaum to Michael Swenson, GS M 2223 BS-E-012683946. Mr. Sobel also informed
Mr. Swenson later that day: “We need to reach a conclusion on the viability of a structured exit.” 9/19/2006 email
from Jonathan Sobel to Michael Swenson, GS MBS-E-012328199.
2224 Subcommittee interview of Peter Ostrem (10/5/2010).
2225 Id.; Subcommittee interview of Darryl Herrick (10/13/2010).
2226 9/19/2006 calendar invite from Michael Swenson, GS MBS-E-012328194.
2227 9/19/2006 email from Peter Ostrem, GS MBS-E-01818608, Hearing Exhibit 4/27-86. Goldman later told the
Subcommittee that Hudson actually offset $1.39 billion in ABX assets on its books. Goldman response to
Subcommittee QFR, PSI_QFR_GS0239.
2228 9/20/2006 email from Michael Swenson to Jonathan Sobel, GS MBS-E-012328203. See discussion of selection
of 60 single names by Mr. Herrick and Mr. Salem, below.
2229 9/20/2006 Firmwide Risk Committee Minutes, GS MBS 0000004472; 9/20/2006 email from Arbind Jha to Josh
Birnbaum, GS MBS-E-012685289.
of its ABX mezzanine holdings or “double down” the ABX holdings, which could happen only if it
found a “structured place to go with the risk.”2223
Later that day, Mr. Sparks approached Peter Ostrem, who headed the desk that originated
CDOs for Goldman, and asked “if there was something [the CDO Desk] could do with ABX.”2224
Mr. Ostrem spoke with Darryl Herrick, who worked for him on the CDO Desk and who eventually
became the Hudson 1 deal captain, about crafting a CDO to reduce the firm’s ABX risk. The two
brainstormed a structure that became the foundation of Hudson 1.2225
That same evening, September 19, 2006, Mr. Swenson scheduled a meeting with several
traders on the ABS Desk he oversaw, Joshua Birnbaum from the ABX Desk, and Mr. Ostrem and
other personnel from the CDO Origination Desk to discuss “ABX and Single-Name
Opportunities.”2226 After the meeting, around 8:00 p.m., Mr. Ostrem sent his CDO team an email
announcing “Hudson Mezz - new”:
“We have been asked to do a CDO of $2bln [billion] for the ABS desk. Approx. $1.2bln
will be CDS off single-names referenced from the AB[X] index 06-1 and 06-2. This is a
trade we need to execute for the desk over the next 4-6 weeks and involves selling half the
equity (at least 30mm to sell) and the seniors and the mezz (at least half of the BBBs to get
true sale). I would like everyone to work together on this one. We expect to charge ongoing
10bp [basis point] liquidation agent fees and 1-1.5pts upfront. ... Obviously important to
overall SP [Structured Product] floor and Sobel and Sparks are focused on this
Also that evening, Mr. Swenson emailed Mr. Sobel to inform him they were “proceeding with the
CDO solution, the CDO team has 60 single-names that they will be able to begin to build a deal
The next day, Mr. Sobel reported to senior executives at Goldman’s Firmwide Risk
Committee that the CDO Desk was working on the first ever ABX CDO, which would function as
an exit for the firm’s long ABX position.2229 Later that same afternoon Mr. Sobel sent an email to
Goldman senior executive Thomas Montag, discussing the Mortgage Department’s ABX losses and
9/21/2006 email from Jonathan Sobel to Tom Montag, G 2230 S MBS-E-009739145.
2231 See, e.g., 10/16/2006 email from John Li to Darryl Herrick, “Call Arbind Jha,” GS MBS-E-018209595
(“Regarding Hudson Mezz Risk issue”).
2232 9/20/2006 email from Arbind Jha to Josh Birnbaum, GS MBS-E-012685289. When asked about this email and
Hudson 1 in general, Mr. Birnbaum told the Subcommittee that he had no specific recollection of any involvement
with the Hudson 1 CDO. Subcommittee interview of Joshua Birnbaum (10/1/2010).
2233 10/12/2006 email from Arbind Jha, “Re: Risk Issue,” Hearing Exhibit 4/27-88.
2234 9/20/2006 Firmwide Risk Committee Minutes, GS MBS 0000004472.
2235 See, e.g., 3/23/2007 “CDO Rating Factors: Inclusion of Tranched ABX Indices in ABS CDOs,” Moody’s,
Document No. SF95049.
2236 The ABX 06-1 Index and the ABX 06-2 Index each tracked a completely different set of 20 RMBS
securitizations. Each Index also had its own subset of five indices tracking individual securities issued by those 20
securitizations, divided by credit rating. For example, one of the RMBS securitizations tracked by the ABX 06-2
Index was called CWL 2006-8. One of the sub-indices within the ABX 06-2 Index tracked BBB rated securities that
stating: “I think most hedge funds have been right on this (i.e. they’ve been short). ... The synthetic
CDO seems like a viable takeout here.”2230
A Goldman risk officer, Arbind Jha, began contacting Goldman mortgage traders and CDO
personnel for regular Hudson updates.2231 In fact, the day after the ABS and CDO teams came up
with the Hudson 1 concept, Mr. Jha emailed Mr. Birnbaum asking about the CDO: “Sobel this
morning mentioned in the Firmwide Risk Committee meeting that we are looking at CDO exit for
our long ABX risk. Wanted to get some color on this.”2232 On another occasion, Mr. Jha asked
“Do we really have scenario risk on $2bn [billion] not’l [notional]? $1.2bn of not’l is being
sourced from ABX desk - this risk is already being captured in our risk number for SPG
Trading desk. ... The remaining $0.8bn will be composed of single name CDS. Since we do
not have any pre-existing long (credit), we will be going short after we price this CDO and
therefore will have a risk mitigating impact on our risk. Please correct me if I am getting
Designing Hudson 1. At the time Hudson 1 was conceived, no other investment bank had
issued a CDO in which the majority of assets referenced ABX assets.2234 Prior to that, the major
credit rating agencies refused to rate any CDO with more than a 5% exposure to credit default
swaps (CDS) using an ABX index as the reference obligation.2235 CDS that used an ABX index as
the reference obligation allowed the parties to the CDS to make a pure bet on the composite
performance of a basket of 20 RMBS securities. Credit rating agencies were concerned that the
inclusion of ABX assets in CDOs would increase market-wide correlation and make CDO
performance more volatile.
In order to get around that limitation and create a CDO that the credit rating agencies would
be willing to rate, Goldman took several steps. First, it decided Hudson 1 would reference the two
ABX 06-1 indices and the two ABX 06-2 indices that referenced RMBS securities with BBB and
BBB- ratings. Since each of those four indices tracked 20 subprime mezzanine RMBS securities,
altogether they tracked 80 RMBS securities.2236 Next, Goldman created 80 single name CDS
were issued as part of the 20 securitizations, including the M8 tranche of CWL 2006-8 which carried a BBB rating.
Another of the five indices tracked the securities carrying a BBB- rating, including the M9 tranche of CWL 2006-8
which carried that credit rating. The CWL 2006-8 M8 and CWL 2006-M9 securities were just two of the 40 BBB
and BBB- rated securities issued by the 20 RMBS securitizations in the ABX 06-2 Index. The ABX 06-1 Index
functioned the same way; its sub-indices tracked another 40 mezzanine RMBS securities from the 20 RMBS
securitizations that composed that Index.
A “single name CDS” contract uses a single security as its reference obligation, 2237 such as a specific RMBS
2238 For simplicity, all 20 assets in each of the ABX baskets were given the same weighted price in Hudson 1.
2239 See, e.g., 4/27/2007 email from Fabrice Tourre to Michael Swenson, Deeb Salem, and Edwin Chin, GS MBS-E-
2240 Mr. Swenson described this situation in his testimony before the Subcommittee: “Throughout 2006, numerous
clients wanted to sell the ABX in order to express a negative view on the U.S. residential housing market. As a
result of these trades, we took on long positions. In order to hedge those positions, we began to increase our short
position in single-names. By November 2006, volatility in the ABX increased, pushing prices down. Because our
positions in single names did not match identically the basket of securities that comprised the ABX, the positions
moved at different rates and even different directions, resulting in losses for the ABS desk.” Prepared statement of
Michael Swenson, April 27, 2010 Subcommittee Hearing, at 208-09.
2241 Goldman decided to price the single name assets at one point below the median ABX trading price. See
9/21/2006 email from Darryl Herrick to Deeb Salem, Michael Swenson, Joshua Birnbaum, Peter Ostrem, and Edwin
Chin, GS MBS-E-012685645.
contracts, each of which used as its reference obligation one of the subprime mezzanine RMBS
securities tracked by the ABX indices.2237 Goldman’s ABS Desk took the short position in each of
those 80 contracts, while Goldman’s CDO Origination Desk took the long position.2238 By taking
the short position in the 80 single name CDS contracts, the ABS Desk essentially offset its long
position in the corresponding ABX contracts. Next, the CDO Origination Desk entered into a CDS
contract with Hudson 1, taking the short side while the Hudson CDO assumed the long side of the
80 single name CDS contracts. The end result was that Hudson 1 took the long side and assumed
the risk associated with the long position for $1.2 billion worth of CDS single name contracts
referencing the RMBS securities comprised the ABX 06-1 and 06-2 BBB and BBB- indices.2239 By
transferring the long CDS position to Hudson, Goldman effectively transferred the risk associated
with its ABX long assets to any investors who bought the Hudson securities.
In order to attract investors and convince them to buy the Hudson securities, Goldman
decided to make use of a pricing difference between CDS contracts referencing the ABX Index
versus single name RMBS securities.2240 In 2006, at the time Hudson was being constructed, the
price of CDS contracts that used an ABX index as the reference obligation was lower than the price
of CDS contracts that used an individual RMBS security as its reference obligation. Because
Goldman exercised complete control over the CDO and created CDS contracts referencing assets
from its own inventory, Goldman also exercised complete control over the pricing of those
contracts. When Goldman set up the 80 single name CDS contracts, and sold the long side of those
contracts to Hudson 1, Goldman decided internally what it would charge the CDO to acquire the
long side of those contracts. Goldman decided to price the contracts, not according to the cost of a
single name CDS contract on the market, but instead according to the cost of a CDS contract for the
long side of the relevant ABX index that day.2241 By using this pricing method, Goldman enabled
Hudson to purchase the single name CDS contracts at the lower ABX prices, which meant it could
See 10/2/2006 email from Darryl Herrick to Michael Swenson, Joshua Birnbaum, 2242 Deeb Salem, Peter Ostrem,
and Daniel Sparks, GS MBS-E-010913416 (“We plan to announce Hudson Mezzanine Funding tomorrow in the am
for Europe, Asia and the US[.] I’m circulating around to everyone the CDO portfolio and spreads we will be
showing investors and agencies, based on our agreed upon amounts and levels from last week.”). Mr. Swenson
responded: “Darryl we should use the 265 and 245 spread for the ABX2 and ABX1 triple-B minus spreads and 145
and 130 for triple-B ABX2 and ABX1 triple-B spreads.” Id.
2243 The gain generated by the CDS single name contracts was retained on the books of the ABS Desk, and created a
favorable “basis” compared to the cost of the ABX long position held by the desk. See 4/27/2007 email from
Fabrice Tourre to Michael Swenson, Deeb Salem, and Edwin Chin, GS MBS-E-012432742; Goldman response to
Subcommittee QFR at PSI_QFR_GS0239. See also 9/19/2006 email from Michael Swenson to Thomas Cornacchia
and Joshua Birnbaum, GS MBS-E-012684557 (“[ABX] Index to single-name basis is at the wides[t] (ie 40 bp at
BBB- level). Index to cash is even more extreme at 70bp. Bids for cash deals remain strong and have barely
widened.”); Performance Review for Michael Swenson, GS-PSI-02396, Hearing Exhibit 4/27-55b (describing the
mortgage trading desk’s strategy of shorting single-name RMBS to offset long ABX positions: “[D]uring the early
summer of 2006 it was clear that the market fundamentals in subprime and the highly levered nature of CDOs was
going to have a very unhappy ending. The beauty of the CDO short was that it allowed for a very efficient method
for capturing the value in the ABX to single-name basis from the short side.”).
2244 Subcommittee interview of Darryl Herrick (10/13/2010). See also 10/8/2006 email from Darryl Herrick to a
Goldman salesperson, GS MBS-E-017502983 (discussing Hudson 1: “Omar, I realize lack of manager may be
tough hurdle for them [investors]. May be helpful to let Deeb and I get on a call with the investor and discuss our
asset selection criteria and I can go through asset sale criteria.”).
2245 See 9/19/2006 email from Darryl Herrick to Deeb Salem, Peter Osterm, others, GS MBS-E-011402123, with
attachment GS MBS-E-011403442 (Mr. Salem wrote to Mr. Herrick: “Attached are 60 RMBS Ref Obs ... for the
CDO we’re discussing. On the RMBS side, we chose 30 Baa2 and 30 Baa3 CUSIPs evenly split btw 2005 and 2006
vintage. We can add a few alt-a names as well. How many of those would you like?”). When interviewed by the
tell investors that, by purchasing Hudson securities, they would be purchasing the single name CDS
contracts at the discount price at which Hudson acquired them.2242
The pricing differential also benefitted Goldman’s ABS Desk in two ways. First, it created a
modest incentive for investors to buy Hudson securities and take the long position needed to offset
Goldman’s ABX risk. Second, it left Goldman with a short position in the form of CDS single
name contracts, which Goldman expected to become more valuable than a short position in CDS
contracts that referenced ABX indices. The ABS Desk kept both its long ABX positions and its
short single name CDS positions in the same dedicated account. Over time, the short position in
single name CDS did gain in value and boosted the overall value of the portfolio of assets held by
the ABS Desk by more than $1 million, producing additional profits for Goldman.2243
In addition to the $1.2 billion in single name CDS contracts to offset Goldman’s ABX risk,
Mr. Herrick from the CDO Origination Desk and Deeb Salem from the ABS Desk worked together
to select $800 million in additional single name CDS contracts to include in the Hudson CDO.2244
Mr. Herrick told the Subcommittee that he gave Mr. Salem a specific set of criteria for selecting
these CDS contracts, including a list of RMBS names that he wanted to be included, a list of RMBS
names that he did not want to be included, and an acceptable price range for each CDS contract.
Since Goldman planned to take 100% of the short side of Hudson 1, these lists were presumably
used to identify RMBS contracts that Goldman expected to offset Goldman’s long positions. Mr.
Herrick told the Subcommittee that Mr. Salem responded with an initial list of 60 possible RMBS
reference obligations, 52 of which were ultimately included in Hudson 1.2245
Subcommittee, Mr. Salem had no specific recollection of how assets were selected for Hudson 1 and little specific
recollection about Hudson 1 as a whole. Subcommittee interview of Deeb Salem (10/6/10).
See, e.g., 2246 Goldman response to Subcommittee QFR, PSI_QFR_GS0192.
2247 Goldman used the cash paid into Hudson 1 to purchase “collateral securities,” as discussed further below.
2248 See 10/30/2006 email from Peter Ostrem, GS MBS-E-0000057886, Hearing Exhibit 4/27-90 (“Super senior note
($1.2bln in size) was executed in the first week of the transaction and was a key driver of this deal[’]s success.”).
2249 The $1.2 billion investment made by Morgan Stanley did not correspond to the CDO’s $1.2 billion in CDS
contracts referencing RMBS securities in the two ABX indices. Rather, Morgan Stanley was investing in the top tier
of the CDO as a whole, which included CDS contracts referencing both the ABX and other RMBS securities.
2250 Subcommittee interview of Darryl Herrick (10/13/2010). See further discussion of Morgan Stanley investment
in Section C(5)(b)(iii)AA, below.
2251 9/27/2006 email from Michael Swenson to Crystal Young, GS MBS-E-012328848. Mr. Swenson told the
Subcommittee that he had no recollection of this conference call, and that it would have been unusual for him to be
involved in the marketing efforts of a CDO. Subcommittee interview of Michael Swenson (10/8/2010).
2252 9/28/2006 email from Darryl Herrick, GS MBS-E-014042217, with attachments GS MBS-E-014042218 and GS
According to Goldman’s contemporaneous records and its responses to Subcommittee
questions, 100% of the CDS contracts included in Hudson 1 were supplied by Goldman’s Mortgage
Department.2246 Because Hudson 1 contained only CDS contracts, it was entirely “synthetic”; it
contained no loan pools or RMBS securities that directed actual cash payments to the CDO.
Instead, the only cash payments made to Hudson 1 consisted of the cash paid by investors making
initial purchases of the Hudson securities and the premiums that Goldman paid into Hudson 1 as the
sole short party.2247
Marketing Hudson. After establishing its basic characteristics and selecting the CDS
assets to be included in Hudson 1, Goldman began to look for investors. A key development took
place early on, when near the end of September 2006, Morgan Stanley’s proprietary trading desk
committed to entering into a CDS agreement with Goldman referencing the “super senior” portion
of Hudson 1, meaning the CDO’s lowest risk tranche that would be the first to receive payments to
the CDO.2248 Morgan Stanley agreed to take the long side of a CDS that represented $1.2 billion of
the $2 billion CDO, while Goldman took the short side.2249 As part of its agreement to invest in
Hudson 1, Morgan Stanley was permitted to review the $800 million in single name CDS contracts
to be included in the CDO and, in fact, vetoed the proposed inclusion of certain CDS contracts
referencing commercial mortgage backed securities.2250
After getting the commitment from Morgan Stanley, Goldman turned its focus to selling the
remaining Hudson securities. On September 27, 2007, Mr. Swenson, the SPG Trading Desk and
ABS Desk head, sent an email to set up a meeting, which later became a conference call, on
“Marketing Strategy for the ABX CDO Trade.”2251 The invitees included Daniel Sparks, Jon Sobel,
Peter Ostrem, Darryl Herrick, and others. Mr. Herrick circulated a draft copy of the Hudson 1
termsheet and transaction overview for review in advance of the call.2252
Goldman’s CDO marketing strategy typically involved its sales personnel sending clients a
marketing booklet outlining different features of a particular CDO. Mr. Herrick drafted the
marketing booklet for Hudson 1, and circulated it for review to Mr. Ostrem and other members of
9/30/2006 email from Darryl Herrick to Peter Ostrem, Benjamin Case, 2253 and Matthew Bieber, GS MBS-E-
014367160, with attachment GS MBS-E-014367161; 9/30/2006 Goldman internal email chain among Darryl
Herrick, Peter Ostrem, Benjamin Case, and Matthew Bieber, GS MBS-E-017504075.
2254 10/2006 Hudson Mezzanine Funding 2006-1, LTD., GS MBS-E-009546963, at 966, Hearing Exhibit 4/27-87.
2255 Id. at 966, 978.
the CDO Origination Desk including Benjamin Case and Matthew Bieber.2253 The executive
summary of the marketing booklet described Goldman’s Hudson CDO program generally and
Hudson 1 in particular:
“Goldman Sachs developed the Hudson CDO program in 2006 to create a consistent,
programmatic approach to invest in attractive relative value opportunities in the RMBS and
structured product market[.]
-We successfully launched Hudson High Grade in September. This is a continuation of the
program using mezzanine Baa2/Baa3 quality RMBS[.]
Hudson CDOs are non-managed and static in nature and provide term non-recourse funding
where Goldman Sachs acts as Liquidation Agent on an ongoing basis. The
Liquidation Agent will be responsible for efficiently selling credit risk assets ...
Goldman Sachs has aligned incentives with the Hudson program by investing in a portion of
equity and playing the ongoing role of Liquidation Agent.”2254
The marketing booklet also described the Hudson 1 assets, and the selection process for
“The portfolio composition of Hudson Mezzanine Funding 2006-1 will consist of 100%
CDS on RMBS.
- 60% of the RMBS will be single name CDS on all 40 obligors in ABX 2006-1 and ABX
- 40% of the RMBS will consist of single name CDS on 2005 and 2006 vintage RMBS ...
Goldman Sachs’ portfolio selection process:
- Assets sourced from the Street. Hudson Mezzanine Funding is not a Balance Sheet CDO
- Goldman Sachs CDO desk pre-screens and evaluates assets for portfolio suitability
- Goldman Sachs CDO desk reviews individual assets in conjunction with respective
mortgage trading desks (Subprime, Midprime, Prime, etc.) and makes decision to add or
See Goldman 2256 response to Subcommittee QFR, at PSI_QFR_GS0223.
2257 Barron’s Dictionary of Finance and Investment Terms defines “the Street” as “referring to the financial
community in New York City and elsewhere. It is common to hear ‘The Street likes XYZ.’ This means there is a
national consensus among securities analysts that XYZ’s prospects are favorable.”
2258 Subcommittee interview of Andrew Davilman (9/30/2010).
2259 Subcommittee interview of Morgan Stanley (6/24/2010).
2260 The Subcommittee asked Mr. Ostrem whether he considered the Hudson 1 assets to be the RMBS securities or
the credit default swaps referencing those securities, and Mr. Ostrem responded that the assets in the CDO were the
RMBS securities which had been originated by a variety of financial institutions on Wall Street. When asked why it
would be important to indicate to investors that not all the underlying RMBS securities were underwritten by
Goldman, given that this information would be clear to a professional reviewing the names of the reference assets in
Hudson 1, Mr. Ostrem replied that he didn’t know. Subcommittee interview of Peter Ostrem (10/5/2010).
2261 Subcommittee interview of Deeb Salem (10/6/2010).
2262 Subcommittee interview of David Lehman (9/27/2010); Subcommittee interview of Matthew Bieber
The marketing booklet statement that “Goldman Sachs had aligned incentives with the
Hudson program by investing in a portion of equity,” was misleading. Goldman did, in fact,
purchase approximately $6 million in Hudson equity.2256 However, that $6 million equity
investment was outweighed many times over by Goldman’s $2 billion short position, which made
Goldman’s interest adverse to, rather than aligned with, the Hudson investors. Neither the
marketing booklet nor other offering materials disclosed to investors the size or nature of
Goldman’s short position in Hudson 1.
The marketing booklet also stated that Hudson’s assets were “sourced from the Street,” and
that it was “not a Balance Sheet CDO,” even though all of the CDS contracts had been produced
and priced internally by Goldman and $1.2 billion of the contracts offset Goldman ABX holdings.
The plain meaning of the phrase, “sourced from the Street,” is that the Hudson 1 assets were
purchased from several broker-dealers on Wall Street.2257 Indeed, a former Goldman salesperson
who sold Hudson 1 securities to investors told the Subcommittee that he thought “sourced from the
Street” referred to assets being acquired from a variety of different broker-dealers at the best prices,
and was surprised to learn that all of the Hudson assets had been provided by Goldman’s ABS
Desk.2258 A Hudson 1 investor told the Subcommittee that it had also interpreted the phrase
“sourced from the Street” to mean assets acquired from a variety of different broker-dealers.2259
The Subcommittee asked several Goldman traders involved in Hudson 1 to explain their
understanding of the phrase, and received inconsistent answers. Darryl Herrick, who drafted the
Hudson marketing booklet, stated that “sourced from the Street” meant the assets were “sourced
from a street dealer at street prices.” His supervisor, CDO Managing Director Peter Ostrem, stated
that “sourced from the Street” referred to the fact the underlying RMBS securities were not
originated or underwritten by Goldman.2260 Deeb Salem, a Goldman mortgage trader who selected
40% of the assets in Hudson 1, described “the Street” as simply “short hand for all brokerdealers.”
2261 David Lehman, who became the head of the CDO Origination Desk after Mr. Ostrem,
and Matthew Bieber, who worked for Mr. Ostrem and later Mr. Lehman, claimed that it was
accurate to say the Hudson assets were “sourced from the Street,” even though all the assets were
acquired from the Goldman ABS Desk, because Goldman was part of “the Street.”2262
12/3/2006 Hudson Mezzanine 2006-1, LTD. 2263 Offering Circular, GS MBS-E-021821196.
2264 Id. at 021821241.
2265 See, e.g., 9/19/2006 email from Michael Swenson to Thomas Cornacchia and Joshua Birnbaum, GS MBS-E-
012684557 (“we are going to price an innovative full capital structure $1+bb CDO deal with 60% of the risk in ABX
(no one has done this before).”); 9/21/2006 email from Darryl Herrick to Deeb Salem, Michael Swenson, Joshua
Birnbaum, Peter Ostrem, and Edwin Chin, GS MBS-E-012685645.
2266 9/27/2006 email from Michael Swenson to Joshua Birnbaum, GS MBS-E-012689798.
2267 12/3/2006 Hudson Mezzanine 2006-1, LTD. Offering Circular, GS MBS-E-021821196, at 021821229. This
disclosure related to the master credit default swap, where Goldman Sachs International served as the credit
protection buyer facing the Hudson Mezzanine 2006-1, Ltd., the legal entity that issued the Hudson 1 securities. See
12/1/2006 ISDA Master Agreement, GS MBS-E-021822056. In this role, Goldman was serving as an intermediary,
and was protecting the CDO from credit risk by placing the Goldman Sachs name on the transaction and assuring
investors that a single credit-worthy entity would be making all required payments to the Hudson 1 trust. Having one
By using the phrase, “sourced from the Street,” Goldman may have misled investors into
thinking that the referenced assets had been purchased from several broker-dealers and obtained at
arms-length prices, rather than simply taken directly from Goldman’s inventory and priced by its
own personnel. Moreover, this phrase also appears to hide the fact that Goldman had an adverse
interest to investors and was seeking to transfer unwanted risk from its own inventory to the clients
it was soliciting. By claiming it was “not a Balance Sheet CDO,” Goldman may have misled
investors into believing that Goldman had little interest in the performance of the referenced assets
in Hudson, rather than having selected the assets to offset risks on its own books.
In addition to the Hudson marketing booklet, in December 2006, Goldman issued an
Offering Circular which it distributed to potential investors.2263 The Offering Circular contained the
statement that no independent third party had reviewed the prices at which the CDS contracts were
sold to Hudson 1.2264 In addition to lacking third party verification, no external counterparty had
participated in any aspect of the CDS contracts – all of the CDS contracts had been produced,
signed, and priced internally by two Goldman trading desks which exercised complete control over
the Hudson CDO.
Internally, while Hudson 1 was being constructed, Goldman personnel acknowledged that
they were using a novel pricing approach.2265 At one point, Mr. Swenson sent an email to Mr.
Birnbaum, raising questions about how they could explain some of the pricing decisions.
Mr. Swenson wrote that he was: “concerned that the levels we put on the abx cdo for single-a and
triple-bs do not compare favorably with the single-a off of a abx 1 + abx 2 trade,” telling Mr.
Birnbaum “[w]e need a goo[d] story as to why we think the risk is different.”2266 The prices that
Goldman established for the CDS contracts that Hudson “bought” affected the value of the CDO
and the Hudson 1 securities Goldman sold to investors, but the Offering Circular failed to disclose
the extent to which Goldman had single-handedly controlled the pricing of 100% of the CDO’s
Perhaps the most serious omission from the marketing booklet and other offering materials
was Goldman’s failure to disclose the fact that it would be the sole short party in the entire $2
billion CDO. The Goldman materials told investors that an affiliate, Goldman Sachs International
(GSI), would be the “credit protection buyer” or initial short party for the Hudson 1 CDO.2267 It was
broker-dealer intermediate between the market and a CDO vehicle was desirable to credit rating agencies in order to
minimize risk in the CDO, and Goldman clearly disclosed this role to investors.
Goldman intermediated between other broker-2268 dealers and the CDO vehicle in Anderson Mezzanine Funding
2007-1, Camber 7, Hudson Mezzanine 2006-1, Hudson Mezzanine 2006-2, and Timberwolf I, among several other
CDOs. See Goldman response to Subcommittee QFR, at PSI_QFR_GS0192.
2269 12/3/2006 Hudson Mezzanine 2006-1, LTD. Offering Circular, GS MBS- E-021821196, at 021821251.
2270 10/16/2006 email exchange between Daniel Sparks and Peter Ostrem, GS MBS-E-010916991 (“Cambridge is
upset that we are delaying their deal. They know that Hudson Mezz (GS prop deal) is pushing their deal back.”).
common practice for underwriters to act as the initial short party in a CDO, acting as an
intermediary between the CDO vehicle and broker-dealers offering competitive bids in order to
short the assets referenced in the CDO.2268 The disclosure provided by Goldman contained boiler
plate language suggesting that would be the role played by GSI in the Hudson transaction.
Goldman never disclosed that it had provided all of Hudson’s assets internally, GSI was not acting
as an intermediary, and GSI would not be passing on any portion of the short interest in Hudson to
any other party, but would be keeping 100% of the short position. The Hudson disclosures failed to
state that, rather than serving as an intermediary, Goldman was making a proprietary investment in
the CDO which placed it in a direct, adverse position to the investors to whom it was selling the
Hudson securities.The Offering Circular contained a section entitled, “Certain Conflicts of Interest,” which