“Potential large subprime trade and impact on Firmwide VAR,” GS MBS-E-012606879 (“We have the 13.5Bln IO
which makes us less short by $1-1.5 Bln . . . .”) (IO is an asset arising from differences in the duration and timing of
CDS premium payment streams); 8/21/2007 email from Michael Dinias, “Trading VaR Analysis,” GS MBS-E-
009993267 (“the mortgage desk has questioned the size of the implied short exposure (i.e. desk believes they are less
short than implied by the VaR model). We are reviewing the current VaR methodology with the mortgage
traders/strategists and assessing the impact of various potential enhancements.”); Subcommittee interview of Joshua
Birnbaum (10/1/2010).
1828 For more information on the Mortgage Department’s VAR levels, see below.
1829 On August 7, 2007, a senior risk executive wrote to Messrs. Cohn, Viniar, and McMahon: “We are doing a
detailed analysis of mortgage and credit trading var contributors to see where we might get the most efficient var
reduction.” 8/7/2007 email from Bruce Petersen, “VaR,” GS MBS-E-009775575.
1830 See, e.g., 8/21/2007 email from Joshua Birnbaum, “Potential large subprime trade and impact on firmwide
VAR,” GS MBS-E-016359332, Hearing Exhibit 4/27-34 (favoring preservation of the $3.5 billion BBB/BBBshort).
1831 Risk controller Robert Berry explained to Messrs. Viniar, McMahon, and Broderick that VAR was becoming
acutely sensitive to estimates of correlation implicit in the data used in the VAR computation: “[T]here will be days
where ‘nothing happened’ – positions didn’t change, markets were quiet, but small changes in correlation will mean
the difference between 140 and 160. It will really be difficult to explain to you and to the desks why we were under
[VAR limits] yesterday but over today.” 8/15/2007 email from Robert Berry, “MarketRisk: End of Day Summary –
measure of risk,1827 the gyrations in daily profit and loss figures demonstrated that the $9 billion net
short posed real risk to Goldman.
VAR depends primarily on the aggregate size of net asset positions and the volatility of the
relevant market. Goldman was holding several large net short positions, the subprime market had
become extremely volatile, and the Department’s efforts to cover its shorts further complicated
matters. Although Goldman’s covering of its $9 billion net short AAA position would ordinarily
be expected to reduce risk, the massive size of the position, combined with unprecedented levels of
volatility in the market, the speed of Goldman’s covering, and the size of the trades involved,
resulted in steep increases in the Department’s VAR.
The Mortgage Department’s VAR had increased from around $13 million in mid-2006,
well under the Department’s permanent limit of $35 million, to $85 million in February 2007, to a
high of around $113 million in August 2007.1828 The $113 million figure exceeded the
Department’s $35 million permanent VAR limit by more than 350%. Moreover, the Mortgage
Department, which had never contributed more than about 2% of firmwide net revenue prior to
2007, was generating some 54% of all the risk incurred by the firm in August 2007. The Mortgage
Department’s risk level, however profitable, was of increasing concern from a firmwide
perspective.1829
In an effort to maximize the profit potential from its net short positions, SPG personnel in
the Mortgage Department argued against indiscriminately covering all of the Department’s
shorts.1830 But the Mortgage Department’s VAR level proved to be both intractable and highly
unpredictable.1831 It also contributed to record high levels of firmwide VAR, a figure carefully
442
cob 8/10/07,” GS MBS-E-009779885. Mr. McMahon responded: “var can swing between 140 and 160 without any
changes in the complexion of our trading books – essentially it is just noise.” 8/15/2007 email from Bill McMahon
to Gary Cohn and David Viniar, “Trading VaR $165mm,” GS MBS-E-009778573.
See, e.g., 8/15/2007 email from Michael Dinias, “1832 Hedge Analysis cob 8/13/07,” GS MBS-E-010678553
(analyzing 6 potential VAR-reducing trading scenarios featuring different proposed transactions); 8/21/2007 email
from Michael Dinias, “Trading VaR Analysis,” GS MBS-E-009993267 (analyzing the VAR-reducing impact of
going long in various classes of assets); 8/22/2007 email from Daniel Sparks to Tom Montag and Donald Mullen,
“VAR reduction possibilities,” GS MBS-E-010619824 (proposed transactions in single name CDS protection to
reduce VAR).
1833 8/15/2007 email from Gary Cohn, “Trading VaR $165mm,” GS MBS-E-009778573. Mr. Viniar agreed. Two
days earlier, after seeing a daily risk report showing Firmwide VAR at $159 million, Mr. Viniar emailed senior risk
executives: “No comment necessary. Get it down.” 8/13/2007 email from David Viniar to Bill McMahon, Craig
Broderick, and Robert Berry, “MarketRisk: End of Day Summary – cob 8/10/2007,” GS MBS-E-009779885. Even
after Mr. Cohn’s order to “get down now,” Firmwide VAR continued to increase. On August 16, Firmwide VAR
rose from $165 million to $171 million. 8/16/2007 email from Michael Dinias, “Trading VAR $171mm,” GS MBSE-
099775568. On August 17, it rose to $174 million. 8/17/2007 email to David Viniar, Gary Cohn, Jon Winkelried,
Tom Montag, Bill McMahon, Craig Broderick et al., “Weekly Market Risk Summary as of 8/17/07,” GS MBS-E-
009756424.
1834 On August 22, 2007, Mr. Montag emailed Mr. Blankfein: “[W]e are covering a number of shorts in mortgages
today and tomorrow–probably $1.5 billion worth–will reduce mortgages [VAR] hopefully to below 80.” 8/22/2007
email from Tom Montag to Lloyd Blankfein, “Trading VaR $144mm,” GS MBS-E-009605812, Hearing Exhibit
4/27-36. The next morning Mr. Montag emailed Mr. Blankfein that the Department “covered about 700 million in
shorts in mtgs last night–500 million in single names . . . . lots more to go–but they fortunately had bought back 9
billion of AAA abx index over last two weeks.” 8/23/2007 email from Tom Montag to Lloyd Blankfein, GS MBSE-
009585951, Hearing Exhibit 4/27-37.
1835 8/23/2007 email from Michael Swenson, “Harbinger Post,” GS MBS-E-009739009.
1836 8/23/2007 email from Daniel Sparks, “Current Outstanding Notional SN ames,” GS MBS-E-010621231.
monitored by the firm’s most senior management. The Mortgage Department and its risk
controllers tried unsuccessfully to reduce the Department’s VAR, and were discussing new
alternatives,1832 when Goldman’s Co-President Gary Cohn intervened. On August 15, 2007, the
same day that Goldman’s firmwide VAR hit a then-record high of $165 million, Mr. Cohn emailed
the Mortgage Department’s senior managers, risk analysts and controllers: “There is no room for
debate – we must get down now.”1833
The Mortgage Department took immediate action. To reduce its VAR, the Department had
to sell at least some of its BBB/BBB- net short position.1834 On August 23, 2007, Mr. Swenson
reported that Harbinger had bought a large block of single name CDS contracts shorting
BBB/BBB- rated RMBS securities which allowed Goldman to take the long side: “We just printed
$400mm of Singles with Harbinger. 130mm of BBB and 270mm BBB-.” Mr. Montag passed the
message on to Messrs. Cohn and Viniar: “Finally actuall[y] covering singles.” Mr. Cohn
responded: “Great.”1835
But even after that $400 million sale, Mr. Sparks reported to Mr. Montag: “[W]e are short
... about $3BB single names.”1836 Mr. Montag responded that the $3 billion net short position was
“huge and outsized” and $800 million had to be sold: “[I]f I make you sell a whopping 800
[million] out of 3 billion which is less than 30% how can anyone complain–the position is huge
443
1837 Id.
1838 8/23/2007 email exchange between Tom Montag and Daniel Sparks, “Current Outstanding Notional SN ames,”
GS MBS-E-010621324.
1839 8/23/2007 email exchange between Gary Cohn and Lloyd Blankfein, “Trading VaR $133mm,” GS MBS-E-
009643469.
1840 Id. Goldman senior executives continued to monitor the Mortgage Department’s actions. For example, on
August 31, 2007, Mr. Montag emailed Mr. Lehman: “abx hurting us today? Mkt over reacting?” 8/31/2007 email
from Tom Montag, “Structured Products Weekly Update – 08/30/07 (Internal Use only),” GS MBS-E-009589083.
Mr. Birnbaum responded: “Based on the markets we are making, we would be ok in this move. We are +50mm in
AAAs alone right now.” Mr. Montag forwarded the emails to Mr. Blankfein who wrote: “Thanks. Appreciate the
posts, I’m watching the financial news.”
1841 12/2007 Quarterly Market Risk Review, GS MBS-E-009586222, Hearing Exhibit 4/27-54f (third quarter
mortgages average VAR was $68 million; fourth quarter mortgages average VAR was $75 million).
1842 Subcommittee interview of David Viniar (4/13/2010); Subcommittee interview of Daniel Sparks (4/15/2010);
Subcommittee interview of Joshua Birnbaum (4/22/2010).
1843 8/21/2007 email from Michael Swenson, “Mortgage Commentary for Firmwide Risk Committee,” GS MBS-E-
010619375 (“Current SPG Trading Desk Position Summary: – RMBS AAA – long $2.2bb”); Michael J. Swenson,
responses to Subcommittee QFRs at PSI_QFR_GS0474.
1844 In the week of December 6, 2007, the Mortgage Department’s risk managers noted “an increase in net short
position in the residential sector.” 12/7/2007 email to David Viniar, “Weekly Market Risk Summary as of
12/06/07,” GS MBS-E-009708911, Hearing Exhibit 4/27-53. See also 9/5/2007 email from Michael Swenson to
Tom Montag, “ABS Update,” GS MBS-E-012386239, Hearing Exhibit 4/27-39; 9/23/2007 email to David Viniar,
“Mortgage P&L for the Week Ended 9/21,” GS MBS-E-009732431, Hearing Exhibit 4/27-43 (Finance division
reported that, in week of September 21, the Mortgage Department reported $63 million from “short synthetic
positions including SPG Trading (+$40M), CDO (+22M) and Residential Credit (+12M).”).
and outsized.”1837 At the end of the trading day on August 23, Mr. Montag asked Mr. Sparks:
“How much did we cover–shooting for $1 billion.” Mr. Sparks responded: “Not much more today
- trying.”1838
Although the Mortgage Department did not cover its shorts as quickly as Mr. Montag
wanted, within days of Mr. Cohn’s August 15 order to “get down,” Goldman’s firmwide trading
VAR had dropped by $40 million and the Mortgage Department’s VAR had dropped below $100
million.1839 On August 23, 2007, Mr. Cohn forwarded a VAR report to CEO Lloyd Blankfein:
“The message got through.” Mr. Blankfein responded: “Good job.” Mr. Cohn wrote: “Down 40
in 2 days.”1840
Although the Mortgage Department’s VAR measure fell substantially, it remained well
above its permanent risk limit of $35 million throughout the rest of 2007.1841 In addition, although
Goldman senior executives rejected the Mortgage Department’s plan to buy $10 billion in AAA
rated RMBS securities,1842 the Department ultimately purchased over $2.2 billion in AAA rated
RMBS securities1843 and continued to hold and strategically sell off its BBB and other short
positions through the end of 2007.1844
444
1845 Salem 2007 Self-Review.
1846 Id. [emphasis in original].
1847 Id. Mr. Salem described the trading strategies as follows:
“a. The Dispersion Trade
More than a year ago, Edwin [Chin] and I realized that the dispersion amongst single-name subprime CDS was
grossly mispriced. Bad names (tier 4), that the market almost universally disliked traded only 50 bps wider than
securities that we all agreed were superior. ... So for the past year, we bought protection on tier 4 names at every
chance possible . . . When we wanted to flatten out the book, we just wrote protection on the tier 1 names.
Amazingly we did this for most of the year without a usable model to value different single names. ... We were very
aggressive with pricing and only shared risk with smart guys if they gave us insight on names to go short or go long
in return. Even when the dispersion widened out . . . by the end of February, Edwin and I refused to monetize the
trade .... With 3-4bb of notional in this dispersion trade, we have recognized $750mm of P&L so far on this trade.
“b. Our single name market share:
Having traded over $200bb of notional in SN [single name] CDS, GS is the dominant market-maker in single name
CDS. We approximate our market share to be greater than 33%. ... This market share and willingness to trade size
proved invaluable in November, December and January when we decided to make the HUGE directional bet of
being long SN CDS protection [or “short risk,” i.e., taking the short side of a CDS transaction]. ... If we did not have
such presence in the SN CDS market, it is unlikely that we would have achieved the size short that we desired and
eventually put on. We’re up $1.7bb in RMBS SN CDS!
(f) Profiting from the Big Short: Making “Serious Money”
Goldman’s short positions continued to produce profits for the firm, as mortgage related
assets continued to lose value. In September 2007, one of the key traders in the Mortgage
Department, Deeb Salem, summarized four key short-side trading strategies that enabled the
Mortgage Department to profit from the collapse of the subprime mortgage market:1845
(a) a “dispersion trade,” in which Goldman took advantage of “mispriced” single name
CDS contracts that referenced very poorly performing RMBS securities, but were not
priced much lower than better performing RMBS securities, resulting in profits of
$750 million “so far”;
(b) a massive purchase of single name CDS contracts referencing a variety of RMBS
securities, taking advantage of Goldman’s estimated 33% “dominant” market share in
single name CDS to make a “HUGE directional bet” against the subprime mortgage
market, resulting in profits of $1.7 billion;1846
(c) the purchase “at a discount” of single name CDS contracts referencing Alt A and A
rated RMBS securities, which “others don’t want/know where to price,” resulting in
profits of $400 million “so far”; and
(d) the purchase of single name CDS contracts on A, AA and AAA rated CDO securities
“in size,” amassing an “enormous” market share whose profits had “exceeded all of
our high expectations” at $900 million “so far.”1847
445
“c. Willingness to put on trades that others don’t want/know where to price
The 2 most successful trades in this regard have been our $70mm long Alt-A protection and our $1-2bb long subprime
Single-A protection [i.e., being “short risk” or taking short side in a CDS transaction]. When CDOs wanted to
sell Alt-A and single-A protection ... the rest of the street was tentative. ... We viewed this as a tremendous
opportunity to buy cheap out-of-the-money options at a significant discount to fair-value. Our exit strategy was
always that if sub-prime fundamentals got bad enough, and they did, that the contagion would have to spread up the
capital structure because cum loss vol/uncertainty has to go up thus. . . .$400mm so far for the desk ....
“d. CDO CDS trade:
This trade has made $900mm so far, which exceeded all of our high expectations for the trade. I had the confidence
and desire for the desk to buy A, AA, and AAA CDO CDS protection in size during the fall of 2006 for several
reasons: ... [T]he rating agencies’ correlation assumptions were out of whack with the growing concern about the
housing market and the remarkable similarity between the bonds in a CDO, the difficulty that GS was having in
placing such mezzanine liabilities and the complexity of such securities scared hedge funds from buying the
protection themselves. Edwin and I used the same aggressive strategies in purchasing protection that we used to
dominate the SN CDS market. Our market share was enormous ... and we did every ... trade possible.” Id.
See, e.g., 10/11/2007 email from Michael Swenson to Donald Mullen, “1848 Early post on P and L,” GS MBS-E-
016031234, Hearing Exhibit 4/27-69. For more information about this downgrade, see Chapter V.
1849 Id.
1850 Id.
1851 10/12/2007 email exchange between Michael Swenson and Donald Mullen, “P and L,” GS MBS-E-018936577,
Hearing Exhibit 4/27-68.
1852 Id.
1853 See 10/3/2007 Goldman presentation, “SPG Trading – 2007,” GS MBS-E-015654036. See also 11/16/2007
Goldman chart, “Mortgages Weekly Metrics 16-November-2007,” GS MBS-E-015863620 (“P&L YTD Annl
3,742,461”).
1854 See, e.g., 4/2010 “Goldman Sachs Mortgage Department Total Net Short Position, February - December 2007 in
$ Billions,” chart prepared by the Subcommittee, Hearing Exhibit 4/27-162 (compiled from Top Sheets); examples
of Top Sheets include GS MBS-E-010369585 and GS MBS-E-010850895, Hearing Exhibit 4/27-162.
In October 2007, after the credit rating agencies downgraded hundreds of CDO
securities,1848 Mr. Swenson wrote to Mr. Mullen that the downgrades would eventually cause
defaults in many CDO securities that Goldman had shorted, meaning that Goldman’s “CDS
protection premiums paid out will go to zero,”1849 and Goldman would receive substantial CDS
payments on those CDOs from the long parties. Mr. Mullen responded: “Looks like we’re going
to make some serious money.” Mr. Swenson replied: “Yes, we are well-positioned.”1850
Indeed, the Mortgage Department made $110 million the very next day.1851 Mr. Swenson
explained: “65mm was from yesterday’s downgrades which lead to the selloff.” Mr. Mullen
replied: “Great day!”1852 By the end of 2007, the Mortgage Department had brought in
approximately $3.7 billion in net revenues from its SPG Trading Desk.1853
(g) Goldman’s Records Confirm Large Short Position
In addition to contemporaneous emails, presentations, and reports, Goldman’s large net
short position is reflected in its own financial records, including its Mortgage Department Top
Sheets.1854 Goldman’s Mortgage Department kept records of its overall net short positions across
446
1855 Id.
1856 See 2/26/2007 Mortgage Department Top Sheet, GS MBS-E-010388545, and 6/25/2007 Mortgage Department
Top Sheet, GS MBS-E-010850895, Hearing Exhibit 4/27-162.
1857 Subcommittee interview of Daniel Sparks (4/15/2010).
1858 Subcommittee interview of Daniel Sparks (4/15/2010); 2/26/2007 email to Daniel Sparks, Michael Swenson,
and others, “New cross desk Top Sheet risk report,” GS MBS-E-010386051.
1859 Subcommittee interview of Daniel Sparks (4/15/2010).
1860 Subcommittee discussions with Goldman Controller staff on Responses to Questions for the Record and related
documents. Despite the fact that the Mortgage Department Top Sheet was the primary comprehensive position
report available to Mortgage Department managers, some more senior Goldman executives appeared to be unaware
of its existence. Chief Risk Officer Craig Broderick, for example, testified that he was unaware of a Mortgage
Department Top Sheet and he thought the “top sheet” concept might not be suitable for the Mortgage Department.
Subcommittee interview of Craig Broderick (4/9/2010). His risk office produced its own daily risk reports that
translated the Mortgage Department’s electronic trading data directly into risk measures, such as “VAR.” Id.; see
also, e.g., 8/9/2007 email from risk manager, “MarketRisk: Mortgage Risk Report (cob 08/08/2007),” GS MBS-E-
010674894 (attached Goldman report, “Mortgage Risk Report – 8/9/07,” GS MBS-E-010674895). The Controller’s
office also produced its own reports based on the Mortgage Department’s electronic trading data, such as the daily
profit and loss report for the Mortgage Department. Subcommittee discussions with Goldman Controller staff on
responses to Subcommittee QFRs and related documents. See also 7/20/2007 email from controllers, “Mortgages
Estimate,” GS MBS-E-009640287.
various subprime asset classes on a daily basis throughout most of 2007.1855 Those records indicate
that the Mortgage Department had large net short positions in the subprime mortgage market
throughout most of 2007. Goldman’s documents reveal that it had a $10 billion net short position
at the end of February 2007, and a $13.9 billion net short position in mid-June 2007.1856
(i) Top Sheets
Daniel Sparks told the Subcommittee that, in 2006 and earlier, he was frustrated because he
felt Goldman had inadequate systems to track and aggregate the daily positions of the various
desks in the Mortgage Department.1857 At the end of each trading day, he had to review up to a
dozen separate reports from the desks to develop a composite mental picture of the Mortgage
Department’s overall positions. To remedy this shortcoming, in February 2007, Mr. Sparks and
the Department’s strategic analysts, sometimes called “strats,” developed a single-page report
called the Mortgage Department Top Sheet.1858
The Mortgage Department Top Sheet became the primary report through which the
Mortgage Department tracked its daily positions in various classes of mortgage related assets. It
provided a comprehensive listing of the Mortgage Department’s long and short positions in
different asset classes across all desks at the end of each trading day. The Top Sheet drew data
from up to a dozen separate reports generated by Goldman’s electronic systems. Mr. Sparks told
the Subcommittee that the Top Sheet evolved over time to become a good tool that provided a
comprehensive record of the Department’s positions.1859 The Goldman Controller’s office told the
Subcommittee that Goldman did not maintain any other type of comprehensive daily report
regarding the Mortgage Department’s net positions in various asset classes.1860
447
2/26/2007 email to Daniel Sparks, Michael Swenson, and others, “New cross desk 1861 Top Sheet risk report,” GS
MBS-E-010386051. The Mortgage Department shared its Top Sheet and several sub-reports for September 25,
2007, with Mr. Viniar around October 1, 2007, in a presentation that summarized the changes in the Mortgage
Department’s long/short positions across all desks and all assets since the initiation of the Gameplan in May 2007.
10/1/2007 Goldman internal document, “September Risk Pack Viniar,” GS MBS-E-013693128 (attached to
10/1/2007 email, “Viniar Risk Pack - Sept,” GS MBS-E-013693127).
1862 The balance of the page provided detail regarding the amount of each asset class held by each of the Mortgage
Department’s desks or business units. Behind the top sheet, the report provided detailed data regarding the various
desks and provided different data analyses, such as changes in position from week to week. The last page of each
report listed all of the electronic data sources and separate reports from which the Top Sheet was compiled. See,
e.g., 2/5/2007 Mortgage Department Top Sheet, GS MBS-E-010369585, Hearing Exhibit 4/27-162.
1863 See chart entitled, “Goldman Sachs Mortgage Department Total Net Short Position, February-December 2007 in
$ Billions,” prepared by the Permanent Subcommittee on Investigations, Hearing Exhibit 4/27-162 (April 2010
version), updated January 2011 (updated version), derived from Goldman Sachs Mortgage Strategies, Mortgage
Dept Top Sheets provided by Goldman Sachs (hereinafter “PSI Net Short Chart”).
The Top Sheet was literally the top sheet, or cover page, to a comprehensive report
distributed daily to Mr. Sparks and other Mortgage Department managers.1861 The top line of the
Top Sheet listed the aggregate total amounts held by the Mortgage Department in various asset
classes, including AAA, AA, BBB, and BBB- rated assets.1862 During most of 2007, the Top Sheet
also provided an overall net short or long figure that summed across the top line totals for each of
the respective asset classes, providing an overall figure by which the Mortgage Department was net
long or net short across all asset classes each day. By plotting this overall net figure for each daily
Top Sheet, the Subcommittee developed a graph of the Mortgage Department’s net position during
2007, as indicated on the chart on the following page.1863
[SEE CHART NEXT PAGE: Goldman Sachs Mortgage Department
Total Net Short Position, prepared by the Permanent Subcommittee
on Investigations.]
Goldman Sachs Mortgage Department Total Net Short Position, February - December 2007 in S Billions
(Market Value, Including All Synthetic and Cash Positions in Mortgage Related Products)
5.0
0.0
2/ /07 3/5/07 4/5/07 5/5/07 6/5/07 7/5/07 8/5/07 9/5/07
-5.0
-10.0
-15.0
10/5/0
Prepared by the U.S. Senate Permanent Subcommittee on Investigations, April 2010. Updated January, 2011.
Derived from Goldman Sachs Mortgage Strategies, Mortgage Dept Top Sheets provided by Goldman Sachs.
11/5/07 12/5/07
449
8/17/2007 Goldman internal chart, “RMBS Subprime Notional H 1864 istory (Mtg Dept - Mtg NYC SPG Portfolio),”
GS MBS-E-012928391, Hearing Exhibit 4/27-56a. The title of the Goldman chart, “NYC SPG Trading” is
confusing, as Mr. Birnbaum asked the analyst, Kevin Kao, for a chart of all synthetic positions across the entire
Mortgage Department (including areas other than SPG Trading). In an email to Mr. Birnbaum, Mr. Kao confirmed
that despite the title, the chart actually included all synthetic positions across the entire Mortgage Department, and
not just the SPG Trading Desk positions. 8/17/2007 email from Mr. Kao to Mr. Birnbaum, GS MBS-E-012929469.
Mr. Kao explained that his chart did not include cash positions, meaning long positions in mortgage loans or RMBS
from any remaining warehouse inventory. Id. That omission was not significant, however, since Goldman had
rapidly sold off the vast bulk of its cash inventory starting in November 2006. 4/2010 “Goldman Sachs Long Cash
Subprime Mortgage Exposure, Investments in Subprime Mortgage Loans, and Investments in Subprime Mortgage
Backed Securities November 24, 2006 vs. August 31, 2007 - in $ Billions,” chart prepared by the Subcommittee,
Hearing Exhibit 4/27-163. In any event, the Subcommittee’s net short chart includes the Department’s cash
positions and demonstrates that any long cash positions were insufficient to offset the shorts, as the Mortgage
Department was massively net short throughout most of 2007. See PSI Net Short Chart.
In his Supplemental Responses to Questions for the Record from the Subcommittee, Mr. Birnbaum
conceded that the Subcommittee’s net short chart includes cash positions and therefore fixed the problem of being
limited to synthetics as was the case with Goldman’s own net short chart. See 8/17/2007 Goldman internal chart,
“RMBS Subprime Notional History (Mtg Dept - Mtg NYC SPG Portfolio),” GS MBS-E-012928391, Hearing
Exhibit 4/27-56a; Birnbaum responses to Subcommittee QFRs at PSI_QFR_GS0509. Mr. Birnbaum maintained his
objection that the Subcommittee’s net short chart improperly summed the notional amounts of different asset classes.
Id. However, as noted in the text, the Mortgage Department’s Top Sheet actually converted the notional amounts of
each asset class into their market values, making it fair to sum across all asset classes.
The Subcommittee’s net short chart is generally consistent with a Goldman chart that Mr.
Birnbaum asked one of the Mortgage Department’s analysts to prepare for him in August 2007,
which can be seen on the following page.
[SEE CHART NEXT PAGE: RMBS Subprime Notional History,
prepared by Goldman Sachs.]1864
The “zero” line in the middle of the chart represents a neutral trading position that is neither
net long nor net short. To use Mr. Viniar’s description, the zero line represents “home.” The area
below the zero line represents net short positions; the area above the line represents net long
positions. The chart shows that the Goldman Mortgage Department was net short throughout
2007, with a total net short position that reached $13.9 billion in July.
451
See Birnbaum responses to Subcommittee QFRs at PSI_QFR_1865 GS0509. Subcommittee interview of Joshua
Birnbaum (10/1/2010).
1866 Subcommittee interview of Joshua Birnbaum (10/1/2010); Subcommittee interview of Michael Swenson
(10/8/2010).
1867 Subcommittee interview of Joshua Birnbaum (10/1/2010).
1868 Subcommittee interview of Joshua Birnbaum (10/1/2010).
1869 Id. Mr. Swenson made similar statements in his second Subcommittee interview. “Were we ever net short? It’s
hard to say what ‘net short’ means really.” Subcommittee interview of Michael Swenson (10/8/2010). Mr. Swenson
had no such difficulty, however, in his first Subcommittee interview. When asked the greatest amount by which he
remembered his desk being net short, Mr. Swenson replied: “$9 billion.” Subcommittee interview of Michael
Swenson (4/16/2010).
1870 As noted above, the risk management and controller areas at Goldman kept their own daily reports based on the
Mortgage Department’s electronic trading data, but they did not track Goldman’s aggregate positions in each asset
class on either a notional or market value basis. Rather, the risk area converted the Mortgage Department data into
risk measures, while the controllers’ area converted the data into profit and loss reports. Though based on the
Mortgage Department’s reported daily trading, these reports did not directly reflect specific net positions in subprime
mortgage assets.
Some Goldman representatives told the Subcommittee that they objected to both the
Subcommittee’s net short chart, and Goldman’s own similar chart, on the grounds that the charts
are too simplistic and fail to account for the relative “weight” or value of short positions in
differing asset classes.1865 For example, in a 2010 interview with the Subcommittee, Mr. Birnbaum
said that because of their differing “weights” or values relative to one another, the notional
amounts in each asset class could not simply be summed up together. For example, he said the
notional amount of AAA assets cannot be added to the notional amount of BBB assets, because the
market assigns each of those positions different weights or values relative to each other.1866 He
said that these relative weights or values are often expressed as specific “hedge ratios” which
Goldman traders can call up on their computer screens in real time to tell them, for example, the
amount of AAA assets they should buy long to hedge a given net short position in BBB assets.1867
Each class also poses a different level of risk than the others. For example, short positions in AAA
assets were relatively inexpensive, as the risk of loss in that tier was considered relatively low,
while short positions in BBB assets would have higher prices, because the risk of loss was
considered high. Mr. Birnbaum said the only chart of Goldman’s net short position that he would
accept as fairly representative would be one that was “Beta-adjusted” to account for differences in
the relative weights, but he admitted that he was unaware of any such charts or reports created by
Goldman during 2007.1868 Without considering these relative weights (which continually change
based on market price movements in various asset tiers), Mr. Birnbaum said it would be
impossible to say that Goldman was ever “net short” at all or in what amount.1869
Goldman’s own documents provide a different picture, however. The Top Sheet was the
primary comprehensive record used by the Mortgage Department to track all net positions across
the Mortgage Department. Goldman did not keep any other records that identified the Mortgage
Department’s net positions in various asset classes,1870 and did not keep any records that used the
“Beta-adjustment” method advocated by Mr. Birnbaum. By early March 2007, the Top Sheet
generally expressed the market values of positions in different asset classes, rather than the
452
In general, positions were expressed in market values for all liquid assets and 1871 in “hedge equivalents” of illiquid
assets, meaning the market value of a position that could be used to hedge that asset. See e.g., 3/5/2007 Mortgage
Department Top Sheet, GS MBS-E-010630691 (Legend: “Market Value (MM$)(Hedge Equiv MV for CDS)”);
9/11/2007 email, GS MBS-E-010690522, attaching 9/10/2007 Mortgage Department Top Sheet, GS MBS-E-
010690523 (“Market value of bonds/loans, bond equiv mkt val for synthetics.”).
1872 See e.g., 3/5/2007 email from Daniel Sparks to Tom Montag and others, GS MBS-E-010646842 (“We think the
overall business is net short”); 3/8/2007 email from Daniel Sparks to Jon Winkelried and others, “Mortgage Risk,”
GS MBS-E-002211242 (“We are still net short”); 7/21/2007 email from Daniel Sparks to Donald Mullen and others,
“Mortgages Estimate,” GS MBS-E-009640287 (“There is also a large net short that we are chipping away to cover”).
1873 1/10/2011 Concordance search of all documents produced to the Subcommittee by Goldman for the phrase “net
short.”
1874 Subcommittee interview of Joshua Birnbaum (10/1/2010).
notional amount of the positions.1871 By using the market values of positions in each asset class on
a given day, the Mortgage Department did, in fact, create a reasonable method for summing across
all asset classes in a single common denominator – the amount of cash the assets would bring in
the market that day. Since the Subcommittee’s net short chart relies on the Department’s daily Top
Sheet figures as expressed in market values rather than notional amounts, it does not suffer from
the incompatibility defect pointed out by Mr. Birnbaum.
Mr. Birnbaum’s position that the notional amounts of each asset class cannot be summed
up together is also inconsistent with what Goldman’s Mortgage Department did on a day-to-day
basis in 2007. Mr. Sparks, the Mortgage Department head who helped design the Top Sheet, used
it to measure the Department’s daily position. Mortgage Department personnel and other Goldman
executives also discussed the Department’s being “net short”– either overall or in specific asset
classes – on a daily basis.1872
Indeed, the phrase “net short” appears more than 3,400 times in the documents produced to
the Subcommittee by Goldman.1873 Since Mr. Birnbaum claimed that one could never really say
whether Goldman was truly “net short,” the Subcommittee asked him why the phrase “net short”
appeared so many times in Goldman’s documents. Mr. Birnbaum said that the use of the phrase
“net short” was a “shorthand” that was used internally in the Mortgage Department.1874 In fact,
neither Mr. Birnbaum, Goldman, nor the Subcommittee could identify any specific document or
instance from 2007 in which the weighting exercise advocated by Mr. Birnbaum in his September
2010 interview was ever used or suggested.
Aside from the Top Sheet that summed up totals across different asset classes to obtain an
aggregate total, there are also many documents in which Goldman personnel described the total
amounts by which Goldman was net short in respective asset classes. For example, Mr. Swenson,
and later Mr. Lehman, compiled an email summary each week that described each Mortgage
Department desk’s net position in different asset classes. A typical format was as follows:
“Current [SPG Trading] Desk Position Summary:
• RMBS Single-As - net short 900mm 100% in single-name CDS
453
8/5/2007 email from David Lehman to Tom Montag, “Mtg Department 1875 Weekly Update,” GS MBS-E-
010671564.
1876 Id.
1877 Id.
1878 See, e.g., 8/23/2007 email from Tom Montag to Daniel Sparks, “Current Outstanding Notional in SN ames,” GS
MBS-E-010621231 (Mr. Montag said, “so if I make you sell a whopping 800 out of 3 billion which is less than 30%
how can anyone complain- -the position is huge and outsized.”); 3/14/2007 email from Tom Montag to Lloyd
Blankfein, “Cactus Delivers,” GS MBS-E-009632839 (“Covered another 1.2 billion in shorts in mortgages”);
8/23/2007 email from Tom Montag to Gary Cohn and David Viniar, “Harbinger post,” GS MBS-E-009739009 (“We
had bought back almost 9 billion of aaa abx over last two weeks though”); 2/27/2007 email from Richard Ruzika to
Tom Montag, GS MBS-E-002204942 (“I want to see us getting the short down to 4.5 bil[lion] net”).
1879 Subcommittee interview of David Viniar (4/13/2010).
1880 See, e.g., Testimony of David Viniar, April 27, 2010, Subcommittee Hearing at 98, 99-100; 4/7/2010 Blankfein
& Cohn Letter to Shareholders (quoted in 4/2010 Goldman report, “Risk Management and the Residential Mortgage
Market,” at 12, Hearing Exhibit 4/27-161).
1881 9/26/2007 email from Lloyd Blankfein, “Fortune: How Goldman Sachs Defies Gravity,” GS MBS-E-
009592726. Mr. Blankfein went on to explain what he meant by “hedge”: “Ie, the avoidance of a bet. Which is why
for a part it subtracted from var, not added to var.” Mr. Blankfein did not explain the “part” in which the big short
subtracted from VAR. In February and August 2007, most of Goldman’s senior management expressed concern that
the large net short positions added to VAR, rather than subtracted from it, and indeed, pushed the firmwide VAR to
record levels.
• RMBS BBB/BBB- - net short 3,000mm (80% in single-name CDS - 50% 2005
vintage
• Correlation Desk - net short $400mm of ABX 06-1 BBB and BBB-
• Mortgage Department short approx $4bb AAA ABX”1875
Mr. Swenson’s email summaries were forwarded to Mr. Montag and other senior executives to
keep them apprised of the Department’s positions.1876 While these documents did not provide an
aggregate overall net short position for the Mortgage Department, they did establish net short
positions in each of the various specific asset classes described, and they were apparently
acceptable to and relied upon by various Goldman executives. Throughout most of 2007, these
weekly reports clearly show very large net short positions in mortgage related assets.1877
Email correspondence among Goldman’s senior executives also routinely referred to the
Mortgage Department’s net short positions in “billions.”1878 CFO David Viniar told the
Subcommittee that a loss (or gain) in the amount of even $25 million was considered “large” and
would immediately be brought to his attention.1879 By that standard, net revenues of $2 billion
from the Mortgage Department’s net short positions in the third quarter of 2007 would have been
considered significant and monitored by senior management, as indeed they were.
In testimony before the Subcommittee and other public statements, Goldman attempted to
minimize the size of its net short position by suggesting to the Subcommittee that the short
positions held by the Mortgage Department were hedges for other, unidentified long assets.1880 Mr.
Blankfein even made that argument to his colleagues in September 2007: “The short position
wasn’t a bet. It was a hedge.”1881 But Mr. Birnbaum contradicted that position just a few days
later. On October 4, 2007, Mr. Birnbaum prepared on behalf of the SPG Trading Desk a
454
10/3/2007 Goldman presentation, “SPG Trading – 2007,” GS M 1882 BS-E-015654036. Mr. Swenson and Mr.
Lehman reviewed and commented upon the “SPG Trading – 2007 ” presentation. Mr. Birnbaum revised the
presentation to make the changes Mr. Swenson and Mr. Lehman suggested. See, e.g., 10/2/2007 email from David
Lehman to Joshua Birnbaum, “SPG Trading - 2007.ppt,” GS MBS-E-015653681 (providing comments and
suggesting text for presentation). Mr. Birnbaum replied and copied Mr. Swenson on the email chain as well: “I
added your bullet and one more.” Id. See also 10/4/2007 email from Joshua Birnbaum to Michael Swenson and
David Lehman, “How’s this?,” GS MBS-E-015712249 (forwarding revised bullet point for presentation). This
collaboration indicates the presentation was made on behalf of the SPG Trading Desk as a whole. Mr. Birnbaum,
however, told the Subcommittee that the SPG Trading Desk ultimately did not use the presentation in connection
with its proposal for 2007 compensation. See Birnbaum responses to Subcommittee QFRs at PSI_QFR_GS0509.
1883 10/3/2007 Goldman presentation, “SPG Trading – 2007,” GS MBS-E-015654036 [emphasis in original].
1884 See Birnbaum responses to Subcommittee QFRs at PSI_QFR_GS0509.
1885 2/28/2007 email from David Rosenblum to Peter Ostrem, “Pete– pls send me cob 2/28 MTM for SP CDO WH’s
first thing in the morning,” GS MBS-E-001800707; 2/28/2007 email from David Rosenblum,“Hedges Status
Report,” GS MBS-E-002640538 (new ABX hedges added and other hedges allocated to ensure that CDO warehouse
risk was fully hedged).
1886 Id.
presentation entitled, “SPG Trading – 2007,” in which he advocated that SPG Trading Desk
personnel should be compensated like hedge fund managers and not as ordinary participants in
Goldman’s traditional bonus pool system.1882 The presentation, which Mr. Birnbaum prepared to
anticipate and refute counter-arguments that might be made by Goldman senior executives,
explicitly addressed and rejected the contention that the profitable net short positions managed by
the SPG Trading Desk were hedges for long assets held by other desks. Mr. Birnbaum, one of the
chief architects of Goldman’s big short, stated:
“By June [2007], all retained CDO and RMBS positions were identified as already hedged.
... SPG trading re-initiated shorts post BSAM [Bear Stearns Asset Management] unwind
on an outright basis with no accompanying CDO or RMBS retained position longs. In
other words, the shorts were not a hedge.”1883 (Emphasis in original).
In response to Questions for the Record from the Subcommittee, Mr. Birnbaum once again stated
that SPG Trading initiated the “shorts on an outright basis with no accompanying CDO or RMBS
retained position longs following the Bear Stearns Asset Management unwind in 2007, and that
these positions were not a hedge.”1884
Mr. Birnbaum’s statement that “all retained CDO and RMBS positions were identified as
already hedged” by June 2007 is also supported by other documents provided by Goldman’s
Mortgage Department. On February 28, 2007, for example, David Rosenblum, Co-Head of the
CDO/CLO Origination Desk, initiated a project to ensure that all of the CDO Origination Desk’s
warehouse accounts were “fully hedged.”1885 As a result, the CDO Origination Desk initiated new
hedges and specifically allocated others to cover all the desk’s warehouse risk. Learning of these
actions, Mr. Rosenblum responded: “Great. Getting pretty nailed down.”1886 In addition, in a
February 12, 2007 email, Mr. Sparks reported to senior management: “Loan and residual books
flat,” and indicated that the Department’s long positions were fully hedged with the short ABX
455
2/12/2007 email from Daniel 1887 Sparks, “Post today,” GS MBS-E-002202310.
1888 Id.; PSI Net Short Chart.
1889 Subcommittee interview of Craig Broderick (4/9/2010). See also Philippe Jorion, “Value at Risk: The New
Benchmark for Managing Financial Risk,” at 20, (3d ed. 2007).
1890 See Mortgage Department Quarterly Average VAR. (Third Quarter 2006 - $13 million; Fourth Quarter 2006 -
$14 million; Second Quarter 2007 - $63 million; Third Quarter 2007 - $68 million; Fourth Quarter 2007 - $75
million). See 9/2007 Quarterly Market Risk Review, “Market Risk Management and Analysis,” GS MBS-E-
009590673, Hearing Exhibit 4/27-54e; see also 12/2007 Quarterly Market Risk Review, “Market Risk Management
and Analysis,” GS MBS-E-009586222, Hearing Exhibit 4/27-54f.
1891 8/9/2007 email from Joshua Birnbaum to Deeb Salem, “MarketRisk: Mortgage Risk Report (cob 08/08/2007),”
GS MBS-E-012927202.
1892 2/27/2007 email from Richard Ruzika to Tom Montag and others, GS MBS-E-002204942; 8/15/2007 email
from Gary Cohn, “Trading VaR $165mm,” GS MBS-E-016344758; Subcommittee interview of Joshua Birnbaum
(10/1/2010).
positions.1887 In a third example, after conclusion of the CDO valuation project, all of the
remaining CDO assets were transferred from the CDO Origination Desk to the SPG Trading Desk
in May 2007. Even after the transfer of those long assets, the SPG Trading Desk remained short,
suggesting again that the CDO assets may have already been hedged.1888 These Goldman
documents all support Mr. Birnbaum’s statement that all retained CDO and RMBS positions were
identified as already hedged when the SPG desk started rebuilding its net short position in June
2007.
(ii) Risk Reports
The Mortgage Department’s large net short positions were also demonstrated by the
periodic risk reports that recorded Goldman’s key risk measure, “Value at Risk” or “VAR,”
throughout 2007. At a 95% confidence level, VAR represents the dollar amount a business unit,
here the Mortgage Department, could expect to make or lose once every 20 trading days – or about
once a month.1889
The Mortgage Department’s VAR skyrocketed during 2007, climbing far beyond the
Mortgage Department’s permanent VAR limit of $35 million and hitting a record high of $113
million in mid-August 2007.1890 In 2007, with notable exceptions discussed below, whenever the
Mortgage Department’s trading exceeded its risk limits, Goldman’s risk managers simply assigned
the department a new, higher “temporary” risk limit to accommodate the Department’s trading.1891
At the same time, the higher VAR did signal the presence of a large net short position and
functioned to limit the size of that position. Twice during 2007, in late February and again in late
August, Goldman’s senior executives ordered the Mortgage Department to reduce its large short
positions in order to bring the firm’s overall VAR measure down.1892 The senior executives were
aware of the Mortgage Department’s large net short positions, in part, because those positions had
contributed to increases in Goldman’s firmwide or trading VAR.
456
Subcommittee i 1893 nterview of Craig Broderick (4/9/2010).
1894 8/8/2007 email from Tom Montag to David Lehman and Michael Swenson, GS MBS-E-011311633.
1895 Subcommittee interview of David Viniar (4/13/2010).
1896 Mr. Birnbaum frequently argued that VAR, or at least Goldman’s then-current model of VAR, was an
inappropriate risk measure for the Mortgage Department’s shorting activities. 8/9/2007 email from Joshua Birnbaum
to Deeb Salem, “MarketRisk: Mortgage Risk Report (cob 08/08/2007),” GS MBS-E-012927200 (“These VAR
numbers are ludicrous, btw. Completely overestimated for SPG trading, underestimated for other mortgage desks.”).
He contended that for various reasons, the Mortgage Department was not actually as short as the VAR measure
reflected. See, e.g., 9/26/2007 2007 MD Reviews, Joshua Birnbaum Self-Review, GS-PSI-01956, Hearing Exhibit
4/27-55c. In arguing for a special compensation model for the SPG Trading Desk, Mr. Birnbaum pointed out that
the desk never lost as much money as the firm’s VAR measure predicted it would. 10/3/2007 Goldman presentation,
“SPG Trading – 2007,” GS MBS-E-015654036. Accordingly, Mr. Birnbaum may have been correct that the firm’s
VAR measure did not accurately measure risk to the firm. Id. On the other hand, since VAR rests on probability
theory which is based on a “normal” distribution of profits and losses (the typical “bell curve”), the Mortgage
Department may have benefitted from the extraordinary and continual one-way movement downward in the subprime
mortgage markets.
One risk management expert testified before Congress: “It is well known that VaR cannot measure crisis
risk. During periods of crisis the relationship between securities changes in strange and seemingly unpredictable
ways. VaR, which depends critically on a set structure for volatility and correlation, cannot provide useful
information in this situation. It contains no mechanism for predicting the type of crisis that might occur, and does
not consider the dynamics of market crises. This is not to say that VaR has no value or is hopelessly flawed. Most
of the time it will provide a reasonable measure of risk – indeed the vast majority of the time this will be the case. If
one were forced to pick a single number for the risk of a portfolio in the near future, VaR would be a good choice for
the job.” Prepared statement of Richard Bookstaber, “The Risks of Financial Modeling: VaR and the Economic
Meltdown,” before the U.S. House of Representatives Committee on Science and Technology, Subcommittee on
Investigations and Oversight, H.R. Hrg. 111-48 (9/10/2009), at 4.
1897 Subcommittee interview of Craig Broderick (4/9/2010).
1898 The Mortgage Department had other risk limits aside from VAR, including Credit Spread Widening or “CSW,”
which seeks to measure what would happen if credit spreads suddenly widened by large amounts (called “shocks”);
“dv01,” which measures the dollar amount by which a security’s value would change based on a 1 basis point change
in the relevant index or interest rate; and balance sheet limits, meaning the amount of Goldman’s balance sheet the
relevant business unit would be permitted to consume. Subcommittee interview of Craig Broderick (4/9/2010).
The firm tolerated the exceptionally high levels of VAR generated by the Mortgage
Department, despite the risks reflected in those high VAR levels.1893 On the few days when the
market rallied, the Mortgage Department incurred huge losses from its net short position that were
immediately reported up the chain to senior management. For example, upon hearing of a $100
million loss in a single day, Mr. Montag asked: “Okay, who lost the hundy?”1894 Mr. Viniar told
the Subcommittee that he was notified of even a $25 million loss in a single day.1895 Allowing the
Mortgage Department to maintain high VAR levels meant that Goldman’s large net short positions
left the firm exposed to large losses, which in some instances did occur, though not as often as its
VAR predicted they might.1896
Risk Management at Goldman. Every business unit and trading desk at Goldman had a
counterpart in the firm’s risk management area.1897 Risk managers were assigned to “shadow” the
relevant business unit and trading desk operations to ensure that their respective trading activities
did not exceed pre-determined risk limits.1898 Separate VAR limits were set for the firm as a whole
and for each division. The division then allocated its VAR limit among each department or
457
9/2006-12/2007 Goldman Sachs Quarterly Market Risk Review, “M 1899 arket Risk Management & Analysis,”
Hearing Exhibits 4/27-54a-f and GS MBS-E-010674895.
1900 11/13/2007 Goldman email, GS MBS-E-010023525 (attachment, 11/14/2007 “Tri-Lateral Combined
Comments,” GS MBS-E-010135693-715). In November 2007, Goldman’s Chief Risk Officer, Craig Broderick,
Controller Sarah Smith, and senior members of their staffs met with the Tri-Lateral Review Group, which included
representatives of the Federal Reserve Bank, the SEC, and the United Kingdom’s Financial Services Authority, to
discuss risk management during the financial crisis. The “Trilateral Combined Comments” are the talking points
prepared by Goldman senior executives for that meeting, in order to respond to specific written questions Goldman
and other firms had received from the Trilateral Review Group.
1901 Id.
1902 Id. The reference to the 30th floor was to the floor on which Goldman’s senior executives then had their offices
in Goldman’s New York headquarters. Goldman also noted that “sr mgmt participated actively in all of the
significant exposure management including when / how to reduce positions.”11/13/2007 Goldman email, GS
MBS-E-010023525 (attachment, 11/14/2007 “Tri-Lateral Combined Comments,” GS MBS-E-010135693-715 at
695).
1903 Subcommittee interview of Joshua Birnbaum (10/1/2010); Subcommittee interview of Craig Broderick
(4/9/2010).
1904 Subcommittee interview of Joshua Birnbaum (10/1/2010).
1905 See, e.g., 2/8/2007 email from Michael Dinias to Robert Berry, and Craig Broderick, GS MBS-E-009980807;
2/24/2007 email from Robert Berry to David Viniar, Craig Broderick and Bill McMahon, “Mortgage VaR,” GS
MBS-E-009778897 (recommending model parameters for VAR calculations); 4/18/2007 email from Jeremy Primer
to Joshua Birnbaum, “Resolution from MRMA meeting?,” GS MBS-E-012868698 (discussing further adjustment to
business area within a division. Some trading desks within a department also had assigned risk
limits.
At the end of a typical trading day, Goldman’s Risk Department prepared risk reports
containing some or all of the risk measures used for the Mortgage Department. The Risk
Department prepared daily, weekly, and quarterly risk reports, as well as reports for the Board of
Directors and various management committees.1899 Goldman’s Firmwide Risk Committee
(“FWRC”) was co-chaired by David Viniar and its weekly meetings were often attended by Co-
President Gary Cohn and CEO Lloyd Blankfein.1900 In preparation for a meeting with regulators,
Goldman senior executives noted that the Mortgage Department was discussed at every meeting of
the FWRC throughout 2007.1901 Goldman also noted that risk “[l]imits are set by the FWRC.
Decisions regarding, e.g., short positions in mortgages taken by business units but with full
knowledge of the 30th floor.”1902
Goldman executives told the Subcommittee that, in general, risk limits were firm and
compliance was mandatory.1903 At the end of each day, department and divisional personnel, their
respective risk managers, and other executives were provided with risk reports from which they
could readily see whether a trading desk had breached its limits. In the event of a violation, the
desk would be directed to curtail its activities or to take whatever steps were necessary to bring its
trading within the applicable limits. Once a desk was notified of its breach of a limit, it was
generally required to act immediately to comply with the existing limits.1904
At times, the Mortgage Department proposed various modifications or alternatives to its
existing risk measures. Some of these proposals were adopted and some were not.1905 But from
458
VaR calculation); 4/23/2007 email from Robert Berry to Daniel Sparks and SPG Trading Desk, “Mortgage VaR,”
GS MBS-E-009708690 (adjustments to VAR calculations).
See, e.g., 8/21/2007 email from Tom Montag 1906 to Gary Cohn, GS MBS-E-016344758; 8/21/2007 email from
Michael Dinias, “Trading VaR Analysis,” GS MBS-E-009742070; Subcommittee interview of David Viniar
(4/13/2010) and Joshua Birnbaum (10/1/2010); 2/24/2007 email from Robert Berry to David Viniar and others,
“FW: Mortgage VaR,” GS MBS-E-009778897.
1907 See generally “The Risks of Financial Modeling: VaR and the Economic Meltdown,” before the U.S. House of
Representatives Committee on Science and Technology, Subcommittee on Investigations and Oversight, H.R. Hrg.
111-48 (9/10/2009), at 4 (Written Testimony of Richard Bookstaber); Philippe Jorion, “Value at Risk: The New
Benchmark for Managing Financial Risk (3d ed. 2007) at 64.
1908 See id. Based on these factors, VAR is generally lower if a desk has many small and non correlated (well
diversified) positions that are trading at steady and predictable price levels. By contrast, VAR is higher if a desk is
holding a small number of very large and highly correlated positions that are trading at volatile and unpredictable
price levels. The position size/correlation and volatility factors may also operate independently. Thus, even if the
desk has a large number of relatively small and non correlated positions, VAR tends to rise as the volatility of
trading rises. Similarly, even if there is little volatility in trading, VAR tends to rise as the desk’s position sizes
become larger and increasingly correlated. The two factors also reinforce one another – large position sizes
combined with high volatility tend to increase VAR dramatically.
1909 Market Risk Management & Analysis, Quarterly Market Risk Review, December 2006, GS MBS-E-009583144,
Hearing Exhibit 4/27-54; see also 2/6/2007 email from MarketRisk,“MarketRisk: Mortgage Risk Report (cob
02/06/07),” GS MBS-E-009980807 (Mortgage SPG VaR Limit 20).
1910 2/8/2007 email from Michael Dinias, “VaR limit for Mtg SPG,” GS MBS-E-009980807.
1911 Id.
1912 Id.
the perspective of the firm’s most senior executives, VAR appeared to have been the predominant
risk measure by which the Department’s activities were judged.1906
VAR Levels Show Net Short. The primary factors that influence VAR are: (1) the
relative size and correlation of positions, and (2) the volatility of trading.1907 While VAR is
computed by applying a complex algorithm to trading data, the VAR measure directly reflects
position size and correlation, and volatility – or a combination of both factors.1908 Changes in
VAR levels over time can also provide information about the general magnitude and direction of
trading positions.
In the fourth quarter of 2006, the Mortgage Department’s permanent VAR limit was $20
million, of which it consumed only $13 million.1909 Early the next year, on February 5, 2007, the
Mortgage Department exceeded its limit with a VAR of $20.5 million.1910 On February 8, 2007, a
senior risk manager recommended that the Mortgage Department’s permanent VAR limit be
increased to $30 million to accommodate anticipated increased price volatility in the mortgage
markets that year.1911 A senior manager concurred and increased the Mortgage Department’s
permanent VAR limit to $35 million, which remained the Mortgage Department’s “permanent”
limit throughout 2007.1912
Almost immediately, however, the Mortgage Department breached its new limit, and its
VAR continued to climb. Over the course a single quarter, the Mortgage Department’s VAR
jumped from $13 million at the end of 2006, to $85 million in the first quarter of 2007 – a 550%
459
Subcommittee i 1913 nterview of Craig Broderick (4/9/2010).
1914 2/27/2007 email from Richard Ruzika to Tom Montag, others, GS MBS-E-002204942; Subcommittee interview
of Joshua Birnbaum (4/22/2010); Subcommittee interview of Daniel Sparks (4/15/2010); 8/9/2007 email from
Joshua Birnbaum to Deeb Salem, “MarketRisk: Mortgage Risk Report (cob 08/08/2007),” GS MBS-E-012927202;
8/16/2007 email from Michael Dinias, “mortgage var contribution,” GS MBS-E-011247689 (“Mortgage Trading has
53.8% marginal contribution to Firmwide VaR and removing the entire mortgage business reduces Firmwide VaR by
$53mm (from $165mm to $112mm). ... As expected SPG Trading desk dominates this risk with 56% contribution
and adds $49mm to Firmwide VaR. This is primarily driven by ABS Synthetics and Correlation book which have
the bulk of the mortgage shorts.”); 8/21/2007 email from Michael Dinias, “Trading VaR Analysis,” GS MBS-E-
009742070 (Mortgage VAR “primarily driven by mortgage shorts on the ABS Synthetics and Correlation desks.”);
8/22/2007 email from Tom Montag to Lloyd Blankfein, “Trading VaR $144mm,” GS MBS-E-009605812, Hearing
Exhibit 4/27-36 (“we are covering a number of shorts in mortgages today and tomorrow–probably 1.5 billion
worth–will reduce mortgages [VAR] hopefully to below [$]80 [million]”).
1915 Subcommittee interview of Joshua Birnbaum (10/1/2010).
1916 8/9/2007 email from Joshua Birnbaum to Deeb Salem, “MarketRisk: Mortgage Risk Report (cob 08/08/2007),”
GS MBS-E-012927198. It is unclear the extent to which Goldman’s regulators were aware of the Mortgage
Department’s VAR levels. In November 2007, Goldman met with the Tri-Lateral Review Group, which included the
Federal Reserve Bank, the SEC, and the United Kingdom’s Financial Services Authority, regarding its risk
management during the financial crisis. Talking points prepared by Goldman personnel for that meeting stated that
with respect to VAR: “Exposures were managed responsibly by the business units within agreed limits. Highs were
set in Mortgages .... Benign P&L allowed for a disciplined but measured response.”11/13/2007 Goldman email, GS
MBS-E-010023525 (attachment, 11/14/2007 “Tri-Lateral Combined Comments,” GS MBS-E-010135693-715 at
707). This description did not disclose that the Mortgage Department, for nearly the entire year, had routinely
exceeded its $35 million VAR limit by significant amounts for months on end.
1917 2/8/2007 email from Michael Dinias, “FW: VaR limit for Mtg SPG,” GS MBS-E-009980807.
1918 2/13/2007 email, “MarketRisk: End of Day Summary - cob 02/12/2007,” GS MBS-E-009716432; see also
2/14/2007 Goldman internal email, “Increase in Mortgage VaR,” GS MBS-E-010374687 (“MTG SPG Desk VaR
increase from $21 mm to $48mm from cob Feb 6 to cob Feb 13, driven primarily by SPG Trading desk”).
increase. Goldman’s Chief Risk Officer, Craig Broderick, told the Subcommittee that he would be
concerned about any breach of a VAR limit, and would certainly investigate the doubling of a
business unit’s VAR, but he admittedly took no action when the Mortgage Department’s VAR
more than quintupled over the course of a single quarter. Mr. Broderick attributed the steep rise in
VAR almost exclusively to unprecedented market volatility,1913 although other Goldman officials
No comments:
Post a Comment