Exclusive: A Forensic Reconstruction Of Goldman's 2008 Prop Trading

Tyler Durden's picture

Lately, Goldman has been extolling the virtues of its theological affiliations and humanistic aspirations to, well, high heaven. Curious to dig deeper through the firm's purported philanthropic efforts, we decided to take a detailed look at the 2007 and 2008 tax records of the charitable Goldman Sachs Foundation. We will not comment on the performance of the actual Foundation: to the chagrin of many needy children who look up to the St. Goldman cathedral in anticipation of a generous holiday season, the Goldman Sachs Foundation has lost gobs of money in the past two years: the fund started off with $275 million in 2007, $269 million in 2008 and ended the year with $161 million. Of course, it is Goldman's prerogative to trade with its money as it desires: while this loss is deplorable, its only outcome will be that fewer Cap 'N' Trade propaganda initiatives will get the due "charitable funding" courtesy of Goldman. Yet what the foundation's tax record do provide, is a very unique and open glimpse in the myriad trading patterns of Goldman's proprietary trading operations... And boy does the firm trade.

A quick tutorial into trade allocation.

Whenever a huge hedge fund such as Goldman's prop trading desk prints, or executes, a transaction, the physical or electronic ticket will request a split of the capital allocation to the various entities, funds, or LPs that make up the firm's "AUM" umbrella. Sometimes no capital is allocated to excluded strategies, but usually, and especially for product agnostic funds such as Goldman, each entity will be allowed its pro rata share based on the "fungible" capital that makes up the firm's entire Assets Under Management. Therefore, the GS Foundation ("GSF"), with its $270 million of capital at the beginning of 2008, would likely get its pro rata allocation as a percentage of the total capital backing the Goldman hedge fund (which can come from such places as Goldman Sachs Asset Management, and Goldman Sachs & Co., which in turn gets it funding via such taxpayer conduits as the Fed's repo operations and the Discount Window). So if Goldman for example had access to total capital of $50 billion last year (roughly), each trade, when allocated to GSF, would account for about half a percent (0.5%), absent special treatment, of the total capital invested or disposed. As an example, if Goldman were to trade $100 million notional in 10 year Index Swaps, GSF would thus be allocated about $500,000 of the trade.

Why is all this relevant?

Were one to comb through GSF's tax filings, one would uncover in 2007 over 500 pages worth of single-spaced trades, and over 200 in 2008, across absolutely every single asset class: equities, indices, futures, fixed income, currencies, credit swaps, IR swaps, FX, private equity, hedge fund investment, you name it (oddly absent are CDS trades). And this is in 2007 alone. These are a one-for-one proxy of absolutely every single trade that Goldman executed in its capacity as a prop trader in the last two years. The only question is what is the proration multiple to determine what the appropriate P&L for the entire firm would have been based on any one single trade allocated to GSF, and subsequently, disclosed in the foundation's tax forms.

We have compiled an extensive amount of the data and have attained some startling conclusions. In essence Goldman's prop group is woefully bad when it comes to trading, or at least it was in 2007 and 2008. Not only that, but Goldman is likely the biggest trader of futures in the world. The several product groups that account for the highest variation in P&L and are attributed the greatest amount of notional are exactly the high frequency traded ETF's that Zero Hedge has been focusing on since inception, and, to a lesser degree, trading in UST futures and 3 Month LIBOR futures. Goldman also executes numerous Interest Rate swaps daily, thousands each year, whose detail we will however ignore for the time being, although whose nominal gross value exceeds hundreds of billions of dollars (curiously in 2007 Goldman would disclose the counterparties on each and every IR swap trade; it has ceased to do so with the 2008 data).

Let's dissect these one at a time.

First, we take a look at the key ETFs/futures which Goldman trades on a daily basis, whose notional of tens if not hundreds of millions of dollars is traditionally broken down by "child" algos, both on open exchanges and on ATS's (think Sigma X), into millions of smaller blocks, thereby influencing markets either up or down, as momentum traders sense a given upcoming trade, and where, in the absence of incremental liquidity, the market has the potential to get caught up in a momentum vortex, which kills all non-Goldman participants, who lose not only capital if the market turns against them subsequently, but who get killed on their execution price, which is traditionally VWAP-based, and, as we will demonstrate tomorrow, VWAP algos end up being among the worst and most expensive market strategies available. In the meantime, we focus on Goldman's daily trading in the following products:

  • S&P 500 E-Mini
  • Dow Jones Mini
  • Russell 1000 Futures
  • Russell 2000 Mini

Zero Hedge has compiled the daily trading data as representative of GSF. We leave it up to our readers to apply to proper multiple to estimate what the full impact for GS is per each and every trade.

As the chart shows, the daily P&L variance is roughly +/-$800,000 as demonstrated by the P&L allocated to GSF. As Goldman's overall 2008 VaR was in the $200 million ballpark, our 0.5% pro rata estimate to get a reading of what the daily variance as represented at the mother ship is intuitively plausible.

How does this data look when extended on a cumulative basis? Not good, and especially not good if you are the trader who is in charge of the Russell 1000 Futures, which seems to have had the greatest capital allocation and (negative) P&L impact, at least in 2008.

If indeed our 200x GSF to GS step-up pro-rata factor is correct, then Russell 1000 Futures trading alone cost Goldman approximately $1.2 billion in 2008, and maybe much, much more (see below). The other ETF's had an average positive daily variance and in fact had a positive cumulative net contribution for 2008. We should note that one thing that the chart above omits is one trade that Goldman executed on November 17, 2008, precisely in Russell 1000 Futures, which cost the Goldman Sachs Foundation -$6,938,775 dollars! Again, this was merely one trade which was likely caused by the firm's need to exit all underwater Russell futures positions ahead of the November 30 fiscal year end (shortly to be rolled to December 31). By applying the same logic as above, did Goldman Sachs lose $1.4 billion trading Russell 1000 Futures in one single day almost a year ago? All signs point to that having in fact been the case. The chart below includes the outlier errant trade from November 17, 2008:

Continuing to other trading products, we next analyze Goldman's performance in the US Treasury futures arena. The chart below captures the trading performance among the core UST Futures categories:

  • 2 Year UST Futures
  • 5 Year UST Futures
  • 10 Year UST Futures
  • Treasury BD Futures
  • 10 Year Swap Notes Futures

It is notable that like in the example above where the biggest P&L delta was attributable to Russell 1000 Futures, in the Treasury arena, Goldman's 5 Year UST Futures did an admirable job in 2008. With all other products being effectively flat if not negative for the year, the 5 Year trader did an admirable job of keep his head above water. Yet after September 11 it seems virtually all products experience a substantial capital loss, most pronounced at the 5 Year Futures and 10 Year Swap Notes desks, even as 10 Year Futures managed to stage an impressive comeback. Nonetheless, Goldman's UST Futures trading desk ended the year down by almost $400,000 at GSF, which based on our back of the envelope calculation would mean at least a $800 million principal loss for the entire firm.

Daily UST Futures P&L below:

Cumulative UST Futures P&L below:

Next we look at what is apparently another favorite of Goldman's Futures trading desk: the 3 Month LIBOR futures. While the variance here is smaller than at both ETF futures and UST futures, the performance was substantially worse, and much more adversely impacted by Lehman's collapse.

And with hundreds of other data points, much more granularity can be derived. We challenge our more inquisitive readers to perform comparable analyses on some of the other Futures and non-Futures based products to obtain their own conclusions. Yet what is obvious no matter how the data set is sliced and diced, is that the firm was bleeding money across virtually all prop-traded groups in 2008. Is it any wonder that the firm's only source of revenue is courtesy of i) the near-vertical treasury curve (thank you taxpayers) and ii) the ability to demand usurious margins on Fixed Income and other products from clients trading in bulk who have no other middleman choices. We would not be surprised if a comparable analysis of the Goldman Sachs Foundation's 2009 results when these become available in one year demonstrates just how "lost" Goldman's prop trading desk has been once again, with the only true backstop to record, record revenues being the elimination of competitors Lehman and Bear, and the ability to demand monopolistic terms from anyone who wishes to transact in size in whatever passes for modern capital markets.

One last notable observation, is the data provided by Goldman's Futures cash Collateral margin: this is the money Goldman collects from Futures counterparties when various variation margin thresholds are impacted (or determined to have been so, as the case may be, if say a major investment bank were to file for Chapter 11). It would not be surprising to see a huge spike in the Cash Collateral Margin series just after Lehman implodes, as Goldman panics into survival mode and demands any and all cash owed immediately become payable. Indeed, looking at the chart below, one sees a spike which generated nearly $5 million in collateral collections for GSF in the two days after Lehman's bankruptcy. Pro rated, this implies Goldman proper pocketed over $1 billion in due variation margins once it got on the calls with every single counterparty it deemed unfit. And this is only in the futures arena! One wonders just how much of this amount was AIG. Curiously, the series ends on September 17th, a day when GSF foundation sucked up $2.2 million, and when Goldman Sachs pocketed who knows how much more via collateral calls (hint: orders of magnitude more), and, when AIG was nationalized. One can imagine what the collateralization drama must have been like in CDS, where the real feces were hitting the fan. Alas, that data set is not available courtesy of GSF.

2008 Futures Cash Collateral Margin

September 2008 futures collateral detail:

The observations above are troubling: Goldman's trading is by no stretch of the imagination better than average. In fact, in 2008, the firm's prop trading was on par with some of the worst performers on Wall Street. Which begs the question: just how has Goldman managed to transform itself into a behemoth that over the past 6 months has had only three trading days of losses? The answer is simple: with no Lehman and no Bear to curb its tentacular dominance of all aspects of the Fixed Income market, Goldman can now rely almost exclusively on its monopolist agency position vis-a-vis mutual, pension, and hedge funds who are desperate to maintain a good relationship and an open dialog with the firm which rewards its best clients with market moving information ahead of all others peasants. In exchange, Goldman can collect an arm and a leg in the form of wide spreads, child algos that get executed efficiently and, always, profitably, and a trading platform (REDI) which has become ubiquitous, and in which Goldman preaches the mantra of VWAP trading. Tomorrow we will analyze why VWAP has cost many asset managers hundreds of millions of dollars, and once and for all provide a quantification of the ever increasing cost of High Frequency Trading.