Meriwether and his partners performed value-at-risk calculations for every position in their fund, then combined each potential loss into a total for the whole portfolio. The trick was to estimate the correlations among the various trades. For example, two positions that were perfectly correlated would require a straightforward addition of exposures—if you risked $100 million in California state bonds and $100 million in New York state bonds and the two moved in lockstep, your total risk came to $200 million. But if two positions were uncorrelated, a fall in one would sometimes be cushioned by a rise in the other, and the combined risk would be smaller. For instance, if you risked $100 million on a convergence bet involving mortgage prepayment rates and another $100 million on something completely unrelated, such as Italian retail bonds, your total exposure would come to $100 million multiplied by the square root of the number of positions—in this case, $141 million. The more you introduced new uncorrelated trades to the portfolio, the more risk could be dampened. The third uncorrelated position would add only $32 million of risk to the portfolio, even if, taken by itself, it threatened a loss of $100 million.19 The fifth uncorrelated position would add $24 million of risk; the tenth would add only $16 million; and so on. Through the magic of diversification, risk could almost disappear. Trades that seemed crazy to others on a risk/return basis could appear highly profitable to Meriwether and his partners.
Ten years later, when the credit bubble imploded in 2007–2009, value-at-risk calculations fell out of favor. Warren Buffett admonished fellow financiers to “beware of geeks bearing formulas.” Nevertheless, Meriwether’s metric represented an advance on the traditional leverage ratio as a way of gauging risk. The traditional ratio failed to account for swaps and options, even though these could be a huge source of risk to a portfolio. It failed to distinguish between hedged bets and unhedged ones, so that an outright bet on Italian bonds would be treated as no more risky than a long/short bet on two similar bonds’ convergence. Most fundamentally, the traditional leverage ratio compared capital to assets, whereas value at-risk compared capital to potential losses. There was no doubt that LTCM had chosen the more relevant yardstick: The whole point of capital was to serve as a cushion against losses, so it was the size of potential losses that determined whether a fund’s capital was adequate.20 To exaggerate only slightly, the traditional approach of comparing capital to assets was like measuring the size of the U.S. armed forces relative to the number of foreigners on the planet. It was inconceivable that every foreigner would attack the United States at once, just as there seemed to be virtually no prospect of all LTCM’s assets losing value simultaneously.
There were obvious objections to value-at-risk calculations, but LTCM’s brain trust was quite aware of most of them. For example, it knew full well that the models forecast the biggest loss that could happen on ninety-nine out of a hundred days; by definition, some days would be worse than that. But a basic insight in financial economics is that markets tend to self-correct, so LTCM was confident that a period of unusual losses would be followed by compensating gains in the portfolio. If a normal price relationship broke down, hitting LTCM with a big loss, other arbitrageurs would see a profitable opportunity to invest; their buying would drive prices back into line, and LTCM’s portfolio would bounce back again. In years to come, critics ridiculed value-at-risk calculations for ignoring the worst day in a hundred, likening them to car air bags that are designed to work all the time except during a collision. But because of the corrective power of arbitrage, ignoring the worst day in a hundred was less reckless than it appeared to be. A bad day for Long-Term would be followed by a better one. The market always tended to spring back. This was the Slinky effect to which Larry Hilibrand alluded.
Of course, the tendency to self-correct was only a tendency. An extreme event could force prices out of their habitual patterns, for a long time or even permanently. LTCM’s models were unlikely to predict such shocks: Like all models, they were based on historical data so could not be expected to anticipate extraordinary events that would by definition be unprecedented. But LTCM’s partners understood this Achilles’ heel too. They sought to compensate for the imagination deficit in their models, brainstorming about potential surprises that could throw off their calculations. For example, they could see that their bond trades would go haywire if the planned European Monetary Union was derailed, and while derailment seemed highly improbable, LTCM was not going to rule out the possibility. So the partners repeatedly stress-tested their portfolio to see how it would perform if monetary union blew up. If they found that the resulting losses could threaten the survival of the fund, they cut back their positions.
LTCM also took pains to consider a related risk that was much discussed in later years—the risk to the fund’s liquidity rather than its solvency. The fund’s trades often involved buying an illiquid instrument and hedging it with a more liquid one. For example, off-the-run Treasury bonds were cheap precisely because they were less liquid than on-the-run ones. By buying illiquid securities in numerous markets, Long-Term was exposing itself to a particular risk: In a panic, the liquidity premium would rise everywhere, and apparently uncorrelated bets would simultaneously lose money. At precisely such a moment, brokers might withdraw credit from LTCM, forcing the fund to liquidate its positions in the worst possible circumstances. Investors might withdraw capital too, compelling further fire sales. The corrective power of arbitrage might turn out to be the equivalent of an ambulance that is located too far away. Even if markets swung back to equilibrium in the end, LTCM might not survive to reap the benefits.
Meriwether and his partners took this liquidity risk seriously. Their fund was called Long-Term Capital Management for a good reason: It was only over the long term that markets could be relied upon to spring back to their efficient level. It followed that LTCM needed a capital structure that would protect it from the calamity of fire sales. So rather than following the usual practice among hedge funds and borrowing short-term money from brokers, Meriwether made at least some effort to secure longer-term loans and special lines of credit that he could call on in a crisis.21 Driven by the same logic, Long-Term broke with the usual hedge-fund practice of allowing investors to withdraw capital on a monthly or quarterly basis. Instead, it insisted on a three-year commitment, later shifting to a policy of allowing investors to withdraw up to one third of their stake annually.22
In the wake of LTCM’s failure, it was easy to forget these multiple precautions. Meriwether and his partners came to be seen as the victims of a reckless faith in their models—which fueled the belief that some modest addition of caution could prevent similar disasters. But the truth was both more subtle and harder to live with. LTCM’s risk management was more nuanced and sophisticated than critics imagined, and the lessons drawn from its failure included several prescriptions that LTCM itself had implemented. In the course of the ensuing decade, there were calls for value-at-risk calculations to be supplemented with stress tests; LTCM had done that. There were calls for financial institutions to pay attention to liquidity risks; LTCM had done that. And yet LTCM failed anyway, not because its approach to calculating risk was simplistic but because getting the calculations right is extraordinarily difficult. To stress-test your portfolio, you have to conceive of all the inconceivable shocks that could occur; to compute your fund’s value-at-risk and liquidity risk, you have to estimate the correlations among your various positions and guess how these could change in extreme circumstances. The real lesson of LTCM’s failure was not that its approach to risk was too simple. It was that all attempts to be precise about risk are unavoidably brittle.
SOON AFTER THE BANK OF CHINA PARTY, IN OCTOBER 1997, Merton and Scholes received the news that they had won the Nobel Prize for economics. The award was greeted as a vindication of the new finance: The inventors of the option-pricing model were being thanked for laying down a cornerstone of modern markets. By creating a formula to price risk, the winners had allowed it to be sliced, bundl
ed, and traded in a thousand ways. The fear of financial losses, which for centuries had acted as a brake on human endeavor, had been tamed by an equation. The Merton-Scholes hedge fund, Long-Term Capital Management, was celebrated as a stunning fulfillment of the professors’ vision. By calculating the risk of losses, Long-Term could hold the capital it needed and no more, turning minuscule price anomalies into fabulous profits.
Yet as the scholars savored their glory, Long-Term reached a fateful crossroads. Back in the 1980s, Meriwether and his professors had been the upstarts on Wall Street; one decade on, nearly all investment banks had fixed-income arbitrage desks that competed with them directly.23 In the first half of 1997, LTCM’s profits had started to slow down, and the partners began to do some soul-searching. One response to shrunken opportunity was to shrink the fund, returning money to investors. But LTCM was not ready to shrink to the point of giving up. To keep the money machine going, Meriwether and his team began to venture into equities.
Though primarily a bond fund, LTCM had dabbled in equities before 1997, and in principle the firm could make a case that its skills matched the challenge. The fund was not going to pick stocks or judge companies: That was for Julian Robertson and his Tigers. But it was going to find trades with the mathematical clarity of bonds—trades involving two almost identical securities that were valued differently or trades in which the normal relationship among various prices had broken down, presumably temporarily. The simplest examples involved stocks that traded in two markets at once. Long-Term’s favorite was Royal Dutch/Shell, which was owned by two listed companies, Royal Dutch Petroleum of the Netherlands and Shell Transport of Britain. The Dutch shares and the British shares represented claims on the same flow of profits, but the British ones traded at a discount to the Dutch ones. Like the gap between on-the-run and off-the-run bond prices, this gap seemed irrational. By buying the cheap British shares and shorting the expensive Dutch ones, Long-Term could collect more dividends than it paid out as a result of being short; and because it was hedged, it could be indifferent as to whether Royal Dutch/Shell’s share price rose or plummeted. Meriwether and his partners seized upon this opportunity and played it in vast size. To many traders on Wall Street, the professors in Greenwich were starting to overreach themselves.24
LTCM’s biggest equity venture was a bet on markets’ steadiness—a strategy that caused Morgan Stanley to dub the firm “the central bank of volatility.”25 It was typical of Meriwether’s faculty: Rather than take a view on which way a market might move, the professors bet on how far it would move, never mind the direction. In the fall of 1997, LTCM noticed that jittery investors were paying a fat premium to insure themselves against sharp moves in stocks: Some were buying call options, thinking that the market might spike up; others were buying put options, seeking to offset the risk of an abrupt crash in prices. As a result, five-year options on the S&P 500 were selling at a price that implied the index would fluctuate by 19 percent per year, way higher than the 10 percent to 13 percent that had been typical in the 1990s. True to their usual pattern, Meriwether and his partners saw a chance to act as the balancers of irrational panic. They sold call options and put options, collecting the premiums in the belief that they would not have to pay out to buyers. So long as volatility remained within its usual range, the bet could not go against them.
Even as they stacked up these wagers, the partners remained focused on their prudential guidelines. Their value-at-risk calculations told them that the most they were likely to lose on ninety-nine out of a hundred trading days was $116 million and the most they could lose over the span of twenty-one trading days was $532 million. The odds of losing everything were vanishingly low; the fund had only lost 2.9 percent in its worst-ever month, and the calculations indicated it would take a ten-sigma event, an event that occurs one in every 1024 times, to put them out of business. To be sure, these projections were based on historical prices, and history could be a false friend. But at the twice-weekly risk meetings in Greenwich, the partners imagined all the shocks that could ambush them. They gamed out the consequences of a crash on Wall Street or a Japanese earthquake. “We thought we were being conservative,” Rosenfeld said later.26
The first sign that they were not conservative enough came in May 1998. The IMF’s bailout of Indonesia faltered and the Suharto regime collapsed. The troubles in East Asia spilled into Japan, which suffered a recession—two consecutive quarters of declining output—for the first time since 1974. Russia’s authorities were forced to triple interest rates to halt an exodus of capital, and the volatility of equity indices all over the world shot upward. Investors reacted to these ructions as they generally do. They piled into the safest investment they could find, U.S. government bonds, sending the yield on thirty-year Treasuries down to its lowest point since they were first issued in 1977. In other words, they panicked.
None of this was good for LTCM. If the fund’s trades had one premise in common, it was that cool heads would prevail: Irrational panic would be corrected by the rational calm of arbitrage. Before the May–June turmoil, LTCM had been short Treasuries and long the “swap rate” paid to money-market investors, believing that the abnormally wide gap between risk-free Treasuries and riskier market instruments reflected temporary risk aversion. But the world’s intensifying economic crisis created risk aversion of an extreme form, causing the spread between Treasuries and riskier instruments to widen and wounding LTCM severely. Equally, LTCM had been betting against equity volatility so aggressively that a single-percentage-point rise cost it $40 million, and now the expected volatility of the S&P 500 index shot up from 19 percent on May 1 to 26 per cent by mid-June. As all these bets went wrong, the fund lost 6 percent in May and another 10 percent in June, way more than its value-at-risk calculations had anticipated.27 Long-Term’s partners canceled their vacations and debated whether something fundamental had gone wrong. But then the fund recovered in July. The partners trimmed their positions and relaxed, dismissing the spring shock as a freak misfortune.28
ON MONDAY, AUGUST 17, RUSSIA STUNNED THE WORLD BY defaulting on its debt payments to foreigners. Unlike Soros and Druckenmiller, LTCM had little direct exposure to Russia, but it soon became clear that the indirect effects would be appalling. Russia’s default led to the bankruptcy of an aptly named hedge fund, High-Risk Opportunities, which in turn threatened a handful of financial firms to which High-Risk owed money. By Friday, August 21, there were rumors that Lehman Brothers might go under, and the panic set off another round of flight to quality. Even more viciously than in May and June, safe government securities rose in value and everything remotely risky fell; and Long-Term’s bets on a calm world blew up disastrously. Long-Term had been betting that U.S. Treasury rates and swap rates would converge; but the gap, which typically moved less than a basis point each day, widened by a stunning eight basis points.29 Long-Term had a similar bet in Britain; again, the spread between British government bonds and market rates widened sharply. In emerging markets, LTCM had constructed essentially the same trade: It shorted relatively stable bonds and owned risky ones, and again it lost badly. By the end of that Friday, Long-Term had lost a total of $550 million, 15 percent of its capital.30
It was the middle of August, and most of Long-Term’s senior partners were enjoying the vacation they had deferred earlier in the summer. John Meriwether was in China. Eric Rosenfeld was in Idaho. LTCM’s counsel, Jim Rickards, was with his family in North Carolina. The skeleton crew in Greenwich stared at the trading screens in wonder. It was not just money that was going up in smoke. Long-Term’s confident assumptions were burning too; it was a bonfire of the fund’s own vanities. LTCM had thought its portfolio was safe because relationships in credit markets were generally stable; now they were stormy. LTCM had thought its portfolio was safe because the correlation between its different strategies was low; with panic driving every market the same way, its positions fell in lockstep. LTCM had thought its portfolio was safe because its value-at-risk estim
ates suggested it could lose no more than $116 million in a trading day. But now its estimate was off by more than $400 million. Most fundamentally, LTCM had believed in the corrective power of arbitrage. Markets could be irrational, but a run of bad returns practically ensured that profit-seeking traders would dive in and order would be reestablished. As they watched their portfolio burning, the traders waited in vain for the market to spring back. The Slinky effect had been suspended.
At midday on the East Coast, Eric Rosenfeld called in from a golf course in Idaho. His 9:00 A.M. tee time had been a little delayed, and he just wanted to be sure that everything was fine in Greenwich. When he heard what was happening, he knew his vacation was over, and pretty soon all the partners were hurrying back home.31 Before Meriwether boarded his flight home from China, he took care to call Jon Corzine, the boss of Goldman Sachs, warning him that LTCM had suffered a bad day but promising that there was no cause to worry.
That Sunday evening, the partners discussed how they could stem the losses. They would have to liquidate some positions, but selling on the open market would not be easy in a panic. Rosenfeld phoned Warren Buffett in Omaha to see if he would buy Long-Term’s $5 billion portfolio of bets on stocks of companies involved in mergers. Buffett refused graciously. The next morning Meriwether breakfasted with Soros and Druckenmiller at Soros’s Fifth Avenue apartment: Perhaps Quantum might like to invest in LTCM on special concessionary terms, since Long-Term had hit a rocky patch just lately? Meriwether explained that his positions were a bargain for anyone with the financial muscle to hold on through the turmoil, and Soros and Druckenmiller seemed intrigued.32 But the most that Meriwether could extract from the breakfast was a conditional offer. Soros would put up $500 million at the end of August if Meriwether was able to raise another $500 million from other investors.
More Money Than God_Hedge Funds and the Making of a New Elite Page 27