Fisher organized a breakfast in New York the next morning, September 21, with the heads of JPMorgan, Goldman, Citibank, and Merrill Lynch. The group knew that when LTCM failed, the same disappearing hedge that Bear Stearns faced would emerge at every big bank in every market in the world. This is how net risk in calm markets morphs into gross risk when markets are in distress. The conclusion was inescapable: LTCM was too big to fail. The bailout began, but Wall Street was not really bailing out LTCM; it was bailing out itself.
By Wednesday, September 23, the group expanded to include other major Wall Street firms. That evening we received a term sheet from what was now called the Consortium. Their deal provided a $4 billion cash injection, shared among sixteen banks, divided $250 million per bank. There was also a ray of hope. The Consortium would keep the LTCM team intact to unwind the trades, as if we had built a nuclear reactor and were the only ones who knew how to work the control rods without a meltdown. This was largely true.
The Consortium was willing to value the fund at $400 million. This meant we could keep ten cents on the dollar of the $4 billion we held just six weeks before. Ten cents on the dollar was generous because we were days away from zero. Still, for some partners who saw their net worth drop from $300 million to $30 million, the trauma was intolerable. Over the next few days, we had occasion to call ambulances for partners suffering from stress. Some signed the deal agreements through tears. The situation had the elements of Shakespearean tragedy without the blood.
The deal was well on its way to a soft landing when Goldman pulled the pin on another grenade. Even as it participated in the Consortium from the New York Fed’s tenth-floor conference room, Goldman secretly engineered a rival bid in league with Hank Greenberg of AIG and Warren Buffett. Goldman and AIG persuaded Buffett they had the derivatives expertise to unwind LTCM’s trades. Buffett would reap the unwind profits without getting his hands dirty. The gang of three—AIG, Goldman, and Buffett—had one condition: everyone at LTCM was to be fired immediately. They wanted sole control of the trades and the embedded future profits. With this secret bid Goldman was no longer front-running LTCM, it was front-running the Fed.
A term sheet signed by Jon Corzine, Hank Greenberg, and Warren Buffett came off the fax machine in the Greenwich office. Meriwether handed the fax to me still warm from the machine. “What should we do with this?” he asked. I knew we were fiduciaries for the fund and had to consider all offers; we could not pick and choose based on personal preferences. Getting fired was irrelevant from a fiduciary perspective. I told JM I would handle it.
I called a senior partner at Sullivan & Cromwell, the white-shoe law firm that represented the bidders. I said, “Look, you want to buy the fund, but the fund is owned by feeder funds.” A feeder fund is a legal entity that takes money from investors and funnels it to the main fund. They are organized in tax havens so foreign investors are not taxed on profits in their home countries. “I’d have to arrange votes of the feeder fund investors. We don’t have time for that. But there’s another way. You could invest in a new feeder and it could buy control of the fund. You could amend the partnership agreement, and cash out the feeders, then you’d own the whole fund.” The Sullivan & Cromwell partner said he would call me back.
In the next hour he frantically tried to reach Buffett, who was on a fishing trip with Bill Gates in a remote part of Alaska. They had no cell phone service, and satellite phones were not working either. The Sullivan & Cromwell partner called me back and said, “I can’t reach Buffett, and I have no authority to alter the terms of the deal.” I said, “Let me be clear. I’m not saying ‘no’ to your deal; I’m saying it’s not feasible—it can’t be done the way you want.” The lawyer said, “I can’t change the proposal.” I said, “Okay, then there’s nothing done,” and hung up. Goldman’s gambit was foiled for want of a phone. It was back to the Consortium.
On September 24, we were making steady progress with the Consortium. Still, I should have known better than to underestimate greed on Wall Street. My phone rang—it was Warren Spector, one of the top officials at Bear Stearns. He wasted no time. “We’re going to put you into default. I’m on my way to the Fed to tell them. We’re pulling out of the Consortium. I’m just calling to tell you first.” Bear was pulling its ace from the house of cards, letting the whole house hit the ground.
Bear was uniquely positioned. As LTCM’s prime broker, it held $500 million of cash collateral from the fund at all times. Other banks received mark-to-market collateral, but this just kept you even, it did not put you ahead. Bear’s collateral was free and clear. It was willing to seize it to protect itself. This would hang the rest of Wall Street out to dry. Bear’s brokerage contract with LTCM contained subjective language. Despite ambiguity, Spector now jumped on the contract to call a default.
I had seconds to save our deal with the Consortium and to save global markets. I said, “Warren, maybe you’re right, maybe you can do this. But maybe not. If you put us in default, I’ll wake up tomorrow with one asset: a $4 billion lawsuit against Bear for breach of contract. That’s the amount of embedded arbitrage profit that will be lost if we fail. The rest of Wall Street will join the lawsuit. I can’t stop you, but you’d better hope you’re right because you’re betting your firm.” I knew Spector was one of Bear’s largest stockholders. My tactic was to target his wealth; Bear’s stock would suffer if our lawsuit succeeded. Personal wealth is the only language Wall Street understands. Spector blinked.
Bear Stearns did not call a default, but it refused to join the Consortium. This was not forgotten on Wall Street. Ten years later when Bear Stearns failed, no tears were shed. As far as Wall Street was concerned, Bear’s 2008 collapse was payback for its 1998 stab in the back.
From September 25 to September 27, we worked nonstop to document a deal. Early on September 28, markets held their breath. Either LTCM would be saved, or a global panic would ensue. In the midtown deal rooms at rescue law firm Skadden Arps, there were last-minute dramatics. Lehman pleaded with the Consortium for relief because it was near bankruptcy itself. Lehman reduced its pledge from $250 million to $100 million. Goldman and JPMorgan ponied up the difference. The money moved, the deal was done.
The next morning, September 29, was my birthday. I had barely spoken to my family or friends for six weeks. At LTCM, we had worked around the clock, first trying to save the fund, then trying to save the world. My wife secretly organized an email effort among everyone I knew to send me birthday wishes. I went to my office, still numb from the trauma that ended the night before. I opened my computer. I had forgotten it was my birthday. My in-box exploded with greetings. I looked at the screen and cried.
Lessons (Not) Learned
The lessons of LTCM’s rescue were clear. Derivatives density and opacity meant neither regulators nor banks knew where risk lay. Derivatives allowed massive leverage because the collateral required was minute relative to their gross value. For LTCM, leverage was infinite because the fund refused to post initial margin; it only offered variation margin on profit or loss after the trade was entered.
Yet there was a deeper peril than obvious leverage and transparency issues. The great menace, one Wall Street still does not comprehend, is that risk resides in gross positions, not net. A simple example suffices to illustrate.
Goldman Sachs might enter a $1 billion swap contract with Citibank where it agrees to pay an overnight interest rate based on a U.S. dollar deposit in London, and receive from Citibank a fixed rate of interest at a spread to a five-year Treasury note. This fixed/floating swap means the bank initially profits based on the difference between the overnight rate it pays and the fixed rate it receives. For Goldman, the swap is roughly equivalent to buying $1 billion of five-year notes and financing the position overnight in the repo market. But there’s no note involved, just a contract calling for two-way payments on a $1 billion notional amount.
Now Goldman enters into another $1 bill
ion swap contract, this time with Bank of America, in which Goldman receives the overnight floating rate and pays a fixed rate based on a two-year Treasury note.
Putting the two trades together, Goldman is receiving (from Bank of America) and paying (to Citibank) the overnight rate. Those cash flows net out close to zero. Goldman is also synthetically long $1 billion of five-year Treasury notes, and short $1 billion of two-year Treasury notes. Those notional positions net out close to zero (depending on agreed spreads). Both swap trades are off-balance-sheet, invisible to outsiders.
The market risk in Goldman’s position boils down to the spread between the fixed rate Goldman pays and the fixed rate it receives. The spread between two-year notes and five-year notes is historically low. As a result, Goldman is required to hold very little capital against this risk. Wall Street banks use a formula called value at risk, or VaR, mentioned earlier, which implies Goldman has almost no risk. Under accounting and regulatory rules applied to swaps, the notes disappear, the accounting disappears, and almost all market risk disappears. It’s all good.
Yet it’s not all good. In the real world, when Citibank and Bank of America do these trades with Goldman, they turn around and do trades in the opposite direction to hedge the risk to Goldman. Counterparties to those trades with Citi and Bank of America, which could be JPMorgan or UBS, then do more trades in a mammoth, ever-widening circle of low-risk trading.
What happens if Goldman goes bankrupt? Suddenly, Citibank’s $1 billion hedged position is gross long, because the offsetting short position to Goldman has disappeared. Citi must go into the marketplace and sell $1 billion of five-year notes to rebalance its books. Bank of America is in the opposite situation; it immediately buys $1 billion of two-year notes to offset the net short position that emerged when Goldman disappeared from the synthetic long.
It would be welcome if Citibank and Bank of America had enough information to find each other and replicate the swaps that Goldman defaulted. They can’t do this easily because neither has access to Goldman’s books, and the market is opaque. New clearinghouses mitigate this risk for simple swaps. Still, clearinghouses do not cover more exotic swaps where liquidity is always problematic. Also, clearinghouses merely shift replacement risk from banks to the clearinghouse itself. What keeps a clearinghouse solvent when multiple markets and banks are collapsing?
This example is realistic, albeit simplified. The difficulties of replacing trades of a bankrupt counterparty when notional amounts are in the tens of trillions of dollars, represented by thousands of contracts covering underlying instruments in stocks, bonds, commodities, and currencies, spread across the books of scores of subsidiaries and special purpose entities in multitudinous markets around the world, are extraordinary. This is why select banks are too big to fail. A single point of failure collapses the entire system.
A crack-up has names like “Tipping Point,” “Black Swan,” and “Minsky Moment” given by sociologists, economists, and media. Those concepts, colorful as they may be, are not science. The dynamics of ruin are best understood using complexity theory, a hard science that offers tools to see collapse coming in advance.
The term “complexity” is often used loosely as synonymous with complication or connectedness. In dynamic systems analysis, those terms have quite different meanings. Complication poses challenges, yet does not produce those totally unexpected results associated with complexity called emergent properties. Mere connectedness does not produce complex dynamics without other elements such as diverse agents and adaptive behavior.
The few capital markets experts who grasp complexity are still in the early stages of applying the science to risk management. Emergence, as shown in the LTCM and later Lehman collapses, has a growing following, although it is still terra incognita to regulators constantly taken by surprise. Even advanced practitioners have not yet assimilated the importance of scale.
Scale in complex systems is synonymous with size, and refers specifically to those metrics that generate risk. The cases of LTCM in 1998, and AIG in 2008, as well as the preceding example, show that risk is embedded in derivatives’ gross notional value, not net value as assumed by Wall Street and regulators. Gross notional value is a simple scaling metric (there are others). There is scant recognition that as gross notional value increases, risk goes up in a nonlinear fashion. Put plainly, if you double derivative gross notional value, you do not double the risk, you increase it by a factor which can be ten or one hundred times depending on specific system characteristics. A provisional law of the new science of complexity in capital markets is: Derivatives risk increases exponentially as a function of scale measured by gross notional value.
To illustrate, imagine an office desk with two empty drawers and one file on top of the desk. The drawers are labeled “A” and “B.” An assistant puts the file away in one of the drawers each night. The assistant could put the file back in drawer A on one night, then in drawer B the next night. If he was keeping track, this would produce a time series of A, B. What are the possible sequences of drawers for putting away the file on two consecutive nights? The possible time series are: AA, AB, BB, and BA, a total of four combinations.
Now, let’s say we increase the number of drawers from two to three and label the drawers “A,” “B,” and “C.” How many different ways can the assistant put the file away on two consecutive nights? Possible time series are: AA, AB, AC, BA, BB, BC, CA, CB, CC, a total of nine combinations.
In this example, the number of drawers increased 50 percent (from two to three), yet the number of combinations increased 125 percent (from four to nine). The number of possible outcomes increased in a nonlinear manner relative to systemic scale. The relationship between the number of drawers and the number of combinations is exponential.
If one translates these outcomes to market risks (for example, the drawers represent the number of swap agreements, the sequences represent possible paths of contagion including bank failure), it’s clear that increasing derivatives scale increases contagion risk even faster.
Complexity theorists generalize possible paths in the office desk example with the following equation:
P2 = P1 × r × (1 − P1)
In this equation, P1 is the file position at the end of day 1, P2 is the file position at the end of day 2, and r is a variable derived from the dynamics of the system under study. This is a recursive function because the output of one iteration is the input for the next. Each output may be viewed as a part of a path of financial contagion.
For example, assume we were calculating one file position in a tall stack of shelves. An office assistant puts the file on one shelf at the end of each day according to a rule produced by the formula. The top shelf is “1” and the bottom shelf is “0.” Every shelf in the stack has a fractional number between 1 and 0 that corresponds to its place in the stack. A shelf designated 0.5 is halfway up the stack between 0 and 1. A shelf designated 0.25 is one quarter of the way from the bottom of the stack. If there were 100 shelves in the stack, 0.25 corresponds to shelf 25 from the bottom.
If the file was on shelf 0.25 at the end of day 1, and we set r = 3, then P2, the location at the end of day 2, is determined as follows:
P2 = 0.25 × 3 × (1 − 0.25)
P2 ≈ 0.56
This means that at the end of day 2, the assistant puts the file on a shelf 56/100ths of the way up the stack between 0 and 1. If there are 100 shelves in the stack, he puts the file on shelf 56 from the bottom.
To determine where the file goes on day 3, or P3, we take the output from day 2, or 0.56, and repeat the process. The recursive equation looks like this:
P3 = 0.56 × 3 × (1 − 0.56)
P3 ≈ 0.74
At the end of day 3, the assistant puts the file on shelf 74 from the bottom.
We can repeat this process as many times as we like. That is exactly what complexity theorists do using computers. They graph results of long
time series, then observe strange emergent properties in the results. Continuing the example above, the time series produces 0.25, 0.56, 0.74, 0.58. . . . The file bounces around from shelf 25 to shelf 56, and so on without repetition or a discernible pattern. This is called chaos. Now repeat the calculation changing the variable r slightly from 3 to 4. As before, start on shelf 25. Here’s what happens when we run the equations.
The file position at the end of day 2 is given by:
P2 = 0.25 × 4 × (1 − 0.25)
P2 = 0.75
The file position at the end of day 3 is given by:
P3 = 0.75 × 4 × (1 − 0.75)
P3 = 0.75
Doing this repeatedly gives the following time series: 0.25, 0.75, 0.75, 0.75. . . . Using the new inputs, we are stuck on shelf 75. No matter how many times we run the equation, the output equals the input, and the file goes on shelf 75. It’s as if the file were attracted to shelf 75. The name for this is a fixed-point attractor. When r = 3 as in the earlier example, the chaotic result is said to have a strange attractor, because it’s hard to predict where the file ends up.
These examples show two important attributes of complex behavior. The first is that small changes in inputs produce widely divergent outputs. The only difference between initial inputs in the two examples above was a change in the value of r from 3 to 4. Still, r = 4 settled into a stable resting place at shelf 75, while r = 3 produced chaos. The second lesson is that complex systems produce unexpected outcomes. Complexity is full of surprises, what are called emergent properties.
The Road to Ruin Page 16
