In his latest Bloomberg column on Goldman Sachs, William Cohan makes an important point:
"Goldman hasn't made its internal analysis public, nor does it intend to, but it has showed the documents around town (including to me)."
In other words, the answer shall remain secret. Only those deemed worthy by Goldman may see its data, which purportedly refutes the Levin report . The rest of us are kept in the dark. We cannot challenge Goldman's claims, because we cannot see what they see. They know what they are talking about; we do not. Instead, we must rely on Andrew Ross Sorkin , Holman Jenkins , Dick Bove , and others to reveal the truth.
Except you don't need to read these secret documents to figure out what's going on. You need simply read the publicly disclosed documents to see how Goldman's defense is built on sand. Consider Goldman's accusation, dutifully reported by The Wall Street Journal, that Levin's subcommittee used "sloppy math and incomplete analysis" too determine Goldman's net short positions. That's not really true. Senate staffers used no math and performed no analysis. They simply copied Goldman's numbers.
Copying Goldman's Numbers
The Senate subcommittee referenced Goldman's own "top sheets," which were daily reports prepared for the benefit of Goldman's head of the mortgage unit, Dan Sparks, who wanted a single-page summary of the firm's most critical mortgage risk concentrations. As the Levin report makes clear, these top sheets represented a comprehensive measure of "the Mortgage Department's net positions in variou s asset classes." Contrary to the false insinuations made in the media, the subcommittee carefully avoided any suggestion that these top sheets captured all of Goldman's positions in all types of mortgages. Nor did the subcommittee state that the top sheets used the exact same categories of mortgage exposure over time.
During the investigation, Goldman made a critical point, which the subcommittee acknowledged in its report. Not all mortgages are created equal. A long position in prime mortgages does not offset a short position in subprime. A triple-A tranche does not offset a triple-B tranche. Therefore, no single dollar amount, long or short, adequately captures all that's going on in a mortgage portfolio. Goldman trader Josh Birnbaum suggested that the number should be "beta adjusted." While noting Goldman's important caveat, the subcommittee also noted that the top sheet totals were regularly referenced in contemporaneous correspondence, and that, at the time, Goldman had no other comparable calculation for companywide exposures. So the subcommittee determined that the internally generated top sheet totals represented the best available proxy for how Dan Sparks was evaluating his firm's overall long or short positions. Investigators conspicuously declined to apply any math or analysis to put everything on an apples-to-apples basis, which might have invited criticism about manipulating the evidence. Unlike Goldman, the subcommittee was upfront about what it did. It never held anything back.
Anyone reviewing the top sheets published by the subcommittee in April 2010 can see that, over time, Sparks wanted additional categories added to his one-page summaries. Some commercial real estate securitizations were added in June 2007, and by September 2007 prime mortgages were added as well. That made sense, because investor distaste for residential securitizations began to spill over into commercial deals. Sparks was very aware that CMBS CDOs are stuffed with subprime bonds and vice versa. In addition, home prices decline throughout 2007 meant that problems in the subprime sector could migrate into prime markets.
So the Journal writes, " For June 25, 2007, Goldman officials believe Senate investigators didn't take into consideration more than $5 billion of prime, or high-quality, mortgage-backed bonds held by the firm at the time, another document shows." Andrew Ross Sorkin references the same document, and goes further to impugn the subcommittee:
"But in studying the document, the subcommittee may have mixed apples and oranges. It added in $4.1 billion worth of short positions for commercial real estate to residential real estate. And the subcommittee ignored the footnote on the bottom of the document that the 'top sheet' had not included long positions in other parts of the business that people close to the firm said were in excess of $5 billion."
As it happened, some staffer made one of the dumb typographical mistakes with regard to this particular top sheet. He accidentally doubled the top sheet total, which showed a net short of just over $6.9 billion, and presented it as $13.9 billion. Embarrassing, especially since the mistake had been hiding in plain sight since April 27, 2010.
Net Totals Are For Dummies
Birnbaum was right. No singular net number adequately captures all that what is going on. The suggestion that the inclusion or exclusion of prime loans somehow debunks the Levin report or vindicates Goldman is, to put it charitably, a red herring. Suppose gas trader Brian Hunter said: "I have been unfairly maligned by the Levin report, which accuses me of destroying my employer, Amaranth Advisors. The report says I took enormously risky bets on the price of gas, I can point to evidence showing that the opposite was true. Whenever I bought gas futures for January 2007, I simultaneously sold futures for November 2006. In other words, I was fully hedged!"
No gas trader would be stupid enough to believe that, and no mortgage trader would be stupid enough to believe, as Goldman now suggests, that a long position in prime mortgages represented any kind of hedge against a short position in subprime mortgages. As noted here earlier, the difference between prime mortgages and subprime mortgages is like night and day.
It's All About The Triple-Bs, Stupid
But there's a far more salient point, which Birnbaum noted and which Goldman's defenders now choose to overlook. Even within the same subprime mortgage securitization, a long position on the triple-A tranche can never effectively hedge a short position on the triple-B tranche. A lot of these "correlation trades" are oxymoronic. The triple-B tranches are heavily reliant on excess spread, which can be wiped out in a heartbeat. (Excess spread is the cash left over after: (a) the interest coupon for all the tranches is fully paid, and (b) all the principal for the 94% of all debt senior to the triple-Bs is fully paid.) Subprime deals were set up so that the size of the initial equity cushion below the triple-Bs was always smaller than the expected losses on the collateral. If the mortgage pools prepaid faster than expected, or if delinquencies spiked faster than expected, the excess spread got wiped out, so that the triple-Bs got wiped out. Conversely, the triple-As, with six times the equity cushion of the triple-Bs, had almost no prepayment risk; early prepayments meant these investors got their cash back sooner rather than later.
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).