JP Morgan's London Whale and Tolerable Imperfection

The two companies on almost everyone's mind this past week are Facebook and JP Morgan. Certainly both companies have provided ample reason for that unwelcome focus. Facebook's botched IPO brings back echoes from the dot-com bust when Wall Street insiders were able to manipulate and control the mechanics of getting rich off these hyped new issues. But that was the second blow to Wall Street's image, and perhaps not the worst one. JP Morgan's "London Whale" run aground in IG9 has captivated insiders and related observers in a game of trying to reconstruct just how "economic hedging" could lead to $3 billion in marked losses (and likely more). There is an almost whimsical collective effort underway to see who can unravel the mystery most completely, knowing full well that it may be years before the full story is revealed from the inside. Captivating stuff; almost like a murder mystery.

Apart from the rubber necking, and even schadenfreude, that comes from these kinds of financial disasters, the far more serious inferences have already started to be voiced and expounded. First and foremost, the Volcker Rule is again to be re-evaluated some many months after it was thought settled. The industry itself circles the wagons by suggesting that JP Morgan's whale was nothing more than one rotten apple in the large barrel of hedged good apples. That is hard to argue against on its face since hedging is a "normal" and integral part of the modern financial system, and, in principle, appears mostly uncontroversial.

We do not yet know the full extent of the losses JP Morgan will ultimately take on its activities, but we know from experience that the losses themselves rarely matter. The system of financial hedging is not really vulnerable in the sense that losses cause systemic breakdowns, it is the dynamic application of mathematical predictions about the probability of future losses that causes liquidity problems. Large current losses, even those that would otherwise be fully and readily absorbed by a bank's capital structure and liquidity profile, are not the problem, per se. It is the re-evaluation of counterparty risk that comes with these kinds of "surprises". In other words, mathematical parameters and predictions have to be adjusted or recalculated, causing problems where the math was originally imprecise.

In mid-September 2008, AIG was not pushed into the government's arms, on the brink of bankruptcy, by the direct expectation of losses. AIG was destroyed by collateral calls as perceptions of potential eroded the confidence of counterparties to accept AIG's financial commitments at face value, or at least on terms close to the recent past. Just a month before, in August 2008, AIG disclosed in its second quarter earnings report that it had charged $6.8 billion of other-than-temporary impairments on the values of its super senior portfolios, securities that continued to perform (that is, these were non-cash impairments in valuations on securities that were still "money good"). That led to $6 billion in collateral calls on two AIG subsidiaries, calls that ended up being funded by the parent.

By September 15, 2008, those two subsidiaries were no longer able to access the commercial paper market, forcing the parent to advance another $2.2 billion. The same day, AIG's securities lending business saw an unwind of $5.2 billion in repos, meaning counterparties to that business no longer accepted AIG collateral at consistent haircuts. Given that desperate liquidity profile, and the failure of a Wall Street syndicate to "rescue" the firm, AIG was downgraded at least two notches (three by S&P). The downgrade, as it always does, led to the death knell for AIG - $20 billion in additional collateral needs were created by the downgrades, increasing to $32 billion a few weeks later. At that moment in time, its major assets were still money-good (and, as much as we have been able to determine, have remained money-good at the Federal Reserve's Maiden Lane sequestration).

It is not the losses that kill the firm; it is the collateral calls that do it. Operating on such tight cash flow tolerances as these firms do, there is little to no room for error. There are no bank vaults full of cash that can be sourced in an emergency. The interbank markets are the 21st century equivalent of the bank vault, yet that is exactly where the crisis continues to be located. Every cash asset in a financial firm is leveraged out to generate a return, often encumbered as a pledge in a liquidity arrangement in just these interbank markets. The only saving grace and emergency source of liquidity for many firms is an unencumbered parent, but after 2008 that really no longer exists (see Bank of America's change in its derivative portfolio housing).

In a fully functioning financial system, where interbank wholesale money flows freely on reasonable terms, downgrades and the subsequent liquidity calls may not always be fatal. In a tight market, however, liquidity is really binary - do or die. As it was for AIG, that was the story for Dexia in 2011 (some lessons are never learned, or maybe moral hazard is all too real). Conventional wisdom still holds that it was exposure to PIIGS debt that brought Dexia to near-insolvency, but in reality it was, again, hedging and liquidity.

The company, demonstrating fully just how interconnected global finance has become, is (was) ostensibly European, but, owing to a 2000 takeover of US bond insurer FSA, had extensive ties to the US muni bond market as a guarantor in various forms. This business line also extended all over the globe and put Dexia front and center in the global public debt markets. Because of this global insurance-like business, Dexia was exposed to all manner of interest rate and currency risks. The risks were, of course, hedged, and the company made all sorts of pronouncements as such.

In either an extreme fit of utter incompetence or due to some undisclosed malfeasance, Dexia was counterparty to huge notional amounts of interest rate swaps (noted to be around EUR63 billion in net notional) that were mostly originated in the 2004-2008 period when the company's bond insurance business was robust. Typically, Dexia would offer municipalities that had borrowed money from a bank "protection" against interest rates rising. Dexia would enter into a swap with that municipality, paying Euribor + a spread (for European municipalities, a different reference rate would have been used for US munis) and in return the municipality would pay Dexia a fixed Structured Coupon. The municipality was then locked into a fixed rate, allowing Dexia to take on interest rate risk (which it saw as a profitable opportunity because it modeled expectations about the movements of interest rates and currencies), meaning Dexia was exposed to rising interest rates (since it was paying the muni the variable Euribor rate).

To offset this rising rate risk, the company would hedge into interest spreads, often financing these hedges by selling options on rates coming down. It was the lowest cost method of hedging what Dexia believed was its biggest risk: rising interest rates. But the financing leg of that hedge strategy opened up risks in the opposite direction of the hedges. By writing options on rates coming down, Dexia was wholly exposed to exactly that, and by huge amounts as it turned out. I don't believe it was stupidity nor fraud - I believe that Dexia took on such massive risk because its models saw the extreme low rate environment as a less than trivial probability. This would put Dexia's models on par with the Federal Reserve's models in 2008.

This fix-for-variable interest rate swap business was also dangerous in that it was asymmetrical in terms of collateral (another risk to be managed). When rates moved, the muni counterparties were free and clear from having to post collateral to Dexia, regardless of where rates would move. Dexia had no such luxury. Its hedges and options required unilateral postings of incremental collateral, meaning that even though it believed it was hedged, its hedges were not hedged enough. Further, the financing of the hedges offered more exposure that needed to be hedged and then modeled into the firm's liquidity profile.

Of course, as 2011 progressed into a Euro-directed crisis, shortage of quality collateral led to higher than expected demand for German bunds - squeezing German interest rates far lower than anyone's model believed likely. Dexia, its hedges "faced" in the wrong direction, was increasingly exposed to collateral calls without any offset protection. Not only did Dexia put itself into a short-term funding shortfall of about EUR80 billion, a 1% decline in the 10-year German interest rate led to a EUR13 billion collateral shortage. Between June 30, 2011, and October 5, 2011, when Dexia was downgraded, the 10-year German interest rate swap had fallen by about 1%. By the end, Dexia's collateral needs had risen to EUR46 billion (up from EUR30 billion a few months before). The bank did not have the reserves or any plausible means to obtain them. Once the "less than trivial" probability actually came to fruition, the written options that were intended as nothing more than financing the main part of the firm's hedging activities ended up being the weak link that destroyed the bank.

Dexia was eventually bailed out and taken over by its true parents, the countries of Belgium and France (and Luxembourg). Nowhere in the case of Dexia were losses prominent, at least not losses as they are traditionally understood. As much as the media tried to blame Greek debt, that was not the fatal trigger. Instead, the company was exposed (exponentially exposed, as it turned out) to having to fund and collateralize interest rate options that were, at best, tangential to its main business and where it thought its main risks lay. In other words, the company destroyed itself by trying to save money in its hedging strategies because the firm's math did not adequately describe risks and probable risks.

This is commonplace in modern financial firms. Complex strategies are often employed in this manner, where offsetting "bets" are used to finance other "bets". As long as everything is "netted" out somehow, it all looks proper and near riskless - this is the modern definition of "hedging". In fact, everything about JP Morgan's whale trades looks exactly like this kind of hedging strategy (whether it was really hedging or not is another topic altogether). There have been references in various news outlets of JP Morgan's CIO unit being long and short all sorts of products, across several tenors and maturities. No doubt they were using premiums for derivatives written to finance other pieces of the trade(s) structure, and probably keeping track of these multi-directional trades via some kind of sophisticated delta calculation. In doing so, that meant that there would exist hedges of hedges, and hedges of financing vehicles for hedges. Losing control of this process, through even the slightest imprecision, can easily lead to an overall position where the firm starts out short and ends up long (or vice versa) almost by accident.

In November 2007, Morgan Stanley lost $3.7 billion by being absolutely correct about shorting subprime mortgage products. Yes, you read that correctly, they were right on being short subprime at the right time, and ended up losing $3.7 billion for their trouble. This was one of the first eye-opening events for me during that period, especially in how little handle the big banks had on what was unfolding. Here we had one of the most successful and sophisticated Wall Street banks being absolutely correct in their investment thesis, but getting killed on the mechanics because of their adherence to mathematical interpretations of how to hedge that short thesis.

Without getting too far into the details of what happened, they were so successful in their shorts that they ended up being long - at exactly the wrong time. To put it another way, as the value of the piece(s) they were shorting (probably a smaller trade in a junior tranche of a subprime structure) fell closer to zero, that successful trade was no longer balancing the long side risk they still were exposed to (their hedge), leaving them with greater and greater long exposure with no other way to offset that risk. Just around the time that the short "worked" best, the price of the long piece (likely a super senior tranche with a far greater allocation because that is what the math told them would be a good "hedged short" - meaning the price of the super senior was not expected to be as "volatile" as the target short, leading to a mathematical calculation where a huge imbalance between the size of the target short in favor of the size of the long side hedge indicated well-managed risk) was beset by illiquidity and irregularity, and that downward price pressure was amplified by negative convexity (and correlation smiles common to the extreme tranches of CDO structures) to the point that what looked like a good, winning and fully hedged short position ended up being a costly and "toxic" long subprime bet. Not only did that cost them in terms of charges directly against net income (at exactly the wrong time), the hit to reputation and the incremental fear it created elevated the relatively new bank crisis into something far more sinister and severe. The entire collateral regime of the financial began to shift.

The November 2007 conference call for Morgan Stanley's loss sounded a lot like JP Morgan's from a couple weeks ago. More questions were raised than were answered, and it set the financial community to look for both better answers and other banks that were doing the same thing, or at least were exposed to risks they did not fully understand nor account for. In 2007, into 2008, that was a target-rich environment that "forced" the Fed to stun the marketplace with its first extraordinary liquidity program (announced on December 12, 2007, TAF was deployed a few days later with the first $20 billion 28-day, "fully collateralized" allotment, followed closely with three other allotments of term auctions, a shocking amount for a different and naive age).

The fact that failures are relatively rare is not evidence of infallibility (the black swan). It is a design flaw that leads to hubris. Indeed, the fact that these outsized losses are relatively rare is the primary parameter driving the growth in these types of activities. But does that make sense or is it necessarily a wise course to follow? Would anyone voluntarily participate in a system that is guaranteed to work for 99.9% of the time, but that 0.1% "anomaly" destroys the system? Of course not, that kind of setup is just Russian Roulette on a different scale. It might attract the gamblers, but it is far from appropriate for the global financial system as a whole.

JP Morgan's CEO, Jamie Dimon, calls this kind of analysis a "tempest in a teapot", as if there is nothing abnormal or unnatural to see here. But even if we take Wall Street at its word that gross notional is misleading, that is small comfort when total "market values" still aggregate to the tens of trillions (the BIS estimates that the gross "market value" of just single currency interest rate swaps at the end of 2011 was $18.046 trillion, on a gross notional of $402.610 trillion) and the biggest banks still, to this day, are completely, unconditionally dependent on short-term and overnight liquidity rolls. Even when net exposures are hedged, as the three examples above show, that does not absolve nor completely obliterate all financial risks. Like all things with the banking system, there is no such thing as a reduction of risk, it is only transformed or transferred. No matter how complex the hedging scenario, no matter how experienced and talented the trading team, there are still unknowns and plain old human mistakes. Mathematics is not a perfect substitute for the real world - the variables are far too dynamic to get to 100% precision. JP Morgan's London Whale loss may not lead to panic or systemic deconstruction on its own, but it pushes the system that much further on edge because it demonstrates that precision really is just an illusion.

That is what the debate should really focus on - what kind of imprecision can the system tolerate. Given the $18 trillion in aggregate market value of interest rate contracts, how many of these kinds of mistakes (including those where the trader gets the thesis exactly right, but the trades "outgrow" tolerances) can the system tolerate before blowing itself apart? This is certainly a relevant topic to our current predicament, in that interest rates have behaved in ways that were fully unpredicted - and expectations about future interest rates are almost fully binary now. How many swaps will blow apart if rates ever "normalize" (suggesting that perhaps ZIRP is more permanent than anyone cares to admit)? It is a full part of modern financial theory that risk must be managed in this way, but that just may be a false comfort where faith in math has led to unbelievably high risks being imbedded, unseen and unaccounted for, in the very fabric of what is now considered routine.

In the Morgan Stanley trade, as in the total breakdown of the repo system as it existed under the mortgage bond collateral paradigm of the housing bubble, it was the dynamic nature of correlation that "fooled" the models. Correlation was always assumed to be relatively stable and behaved as such throughout the history of structured finance (short as that history really was, especially in liquid and traded forms). Thus the expectation, for Morgan Stanley, that pricing of super senior tranches would likewise be stable - a good "hedge" for a good investment thesis. But, as it turned out, the application of the hedge actually completely transformed the risks in the equation without anyone being aware until it was too late. In terms of the Gaussian copula, which implied correlation from credit spreads, there was no real qualitative analysis as to the veracity of the data interpretations, nor did anyone at the time believe there should be. These losses were rare, so there was no established reason to challenge the data or the pricing. Illiquidity, more than anything, drove credit spreads in the same direction at the same time, but that does not necessarily mean true correlation is rising amongst unrelated loan obligors - and it certainly did not mean correlations were rising to levels that made little real world or common sense. On some level, implied mathematical parameters are themselves unidentified risks since they are subsets of the model, not the real world.

The mathematical system, which worked near perfectly up until 2007, transformed in a direction that was almost completely unpredicted solely because of this recency bias. Statistical models will always be captured by the data series upon which they are based. Anything outside of that data series, is, by definition, thought to be rare. Yet, we were told unrelentingly that everything is hedged and netted, because, more than anything, this hubris leads to false confidence. What is thought to be hedged, even well-hedged, may just be a massive risk in an unforeseen direction, a direction that may not have yet appeared in the data series.

Technological complexity is always a double-edged sword, meaning that care and due consideration should be exercised before it is extended to every facet of society. Maybe the financial system is not the right guinea pig for cutting edge mathematical research and pricing/statistical theory (especially when the entire system is based on random walk statistics). Perhaps, certainly for a depository institution entrusted as a safekeeper for other people's money, pushing the envelope is inconsistent with the mission of the firm and its utility to the larger real economic system. I have said this many times, but true intermediation really does not need this kind of hedge-based hubris. Real intermediation is about segregation - if an investment appears too risky, perhaps it is better left unfunded. False confidence is the path to PIIGS and subprime. Unshakeable faith in hedging has led to a system that believes it can fund anyone, anywhere, and simply "manage" the risk - and at scales that can no longer be adequately processed by the human brain.

The liquidity problems that have plagued the system since 2007 are themselves an outgrowth of faith in risk management. Reliance on overnight repos and wholesale money markets is itself another form of modeled risk, just as susceptible to the unknown as is the modern practice of ultra-hedging a bank book. It is not the losses that kill the firm; it is the downgrades and collateral calls. Ironically, this amounts to model on model violence, as the rating agencies' models lead to the extinguishment of the firm's models. Perhaps that is fitting since the entire system is now fully and solely contained within an utterly massive and often incomprehensible mathematical paradigm. Complexity is now really at uncharted levels, so much so that only math can destroy math.