Failed Them Utterly

Anyone who has sat through a finance course has met the Black-Scholes formula, the equation that puts a price on an option. On 14 October 1997 two of the men behind it, Robert Merton and Myron Scholes, won the Nobel prize in economics. Eleven months later the hedge fund in which both were partners was being handed to fourteen banks in a conference room at the Federal Reserve Bank of New York, because the people in that room believed that if it failed on its own it could jam a good part of the world's bond markets.

The fund was Long-Term Capital Management, LTCM to everyone who dealt with it, founded in February 1994 by John Meriwether, who had run the famous bond arbitrage group at Salomon Brothers and brought most of it with him.

What the fund did

The trade was simple. Find two securities that are nearly the same thing but priced slightly differently, buy the cheap one, sell the dear one, and wait for the gap to close.

The example the General Accounting Office, as Congress's auditors were then called, used in its 1999 post-mortem was American government debt. A two-year Treasury note issued this month and one issued a few months ago are the same promise from the same borrower, but the new one is easier to trade, so investors pay a little extra for it. LTCM would buy the old note and short the new one, borrowing it and selling it with a promise to hand it back later. As the new note aged the prices had to meet, and the difference was profit. Looser versions paired things alike in most ways but not all, such as the shares of Royal Dutch and Shell, two listed halves of one oil group that traded at different prices.

Such gaps are tiny, so they pay only with enormous borrowing. LTCM borrowed. It returned 43 per cent in 1995 and 41 per cent in 1996. At the end of 1997 it handed about $2.7 billion back to investors without shrinking its positions, leaving capital of $4.8 billion. By 31 August 1998 it carried more than $125 billion of assets, leverage of more than 25 to 1 even against that January capital, according to the President's Working Group on Financial Markets. Off the balance sheet sat more than $500 billion of futures, $750 billion of swaps and $150 billion of options and other contracts, measured by notional value.

What the models got right

The easy story is that professors trusted their equations and the equations lied. The record doesn't support it.

LTCM measured risk with value-at-risk, the standard method at big banks, which uses past prices to estimate how much a portfolio could lose. For a fund with hundreds of positions the number that matters most is correlation, the chance that one trade goes wrong at the same moment as another. Almost nothing connected American mortgage bonds to Royal Dutch and Shell. Five years of data put the correlations at about 0.1 or lower.

The partners didn't take that at face value. They told the sociologist Donald MacKenzie that they ran the model as if correlations were about 0.3, and padded the volatilities too. The model predicted the fund's value would swing by about 14.5 per cent a year. In practice it swung by 11. Investors had been told to expect 20.

Measured against the past, the model was cautious and accurate.

Who else held the trades

What five years of prices could not show was who else had learned to do this.

Alan Greenspan put it in a sentence to the House Banking Committee on 1 October 1998: "it is the nature of the competitive process driving financial innovation that such techniques would be emulated." LTCM's profits were common knowledge, its strategy had to be explained to investors, and any bank on one side of a trade could guess the other side.

MacKenzie calls the result a superportfolio, a mass of overlapping positions across banks and hedge funds, none of which knew how much of the same thing the others held. While everyone was buying, that helped, because the gaps closed faster. No price history could describe what happens when the holders of a consensus trade all sell at once, because it hadn't happened yet.

The first big seller had nothing to do with LTCM. Salomon, by then owned by Travelers, closed its US arbitrage desk in July 1998 after losses and sold its positions fast. "Not only did we lose money as the positions went against us as we were selling them," Costas Kaplanis, Salomon's head of global arbitrage, told MacKenzie, "but all the other funds that also had these consensus trades also started losing money."

August

On 17 August Russia devalued the rouble and stopped paying on its debt. LTCM had little money in Russia, and it didn't need to. Firms that lost money there sold whatever else they could, investors everywhere ran for the safest and most easily traded assets, and almost every gap LTCM had bet would narrow grew wider. William McDonough, president of the New York Fed, later told Congress: "As markets around the world moved in the same direction at the same time, the diversification on which Long-Term had previously relied failed them utterly."

On 21 August the fund lost $550 million in a day. Correlations of 0.1 jumped to around 0.7. In August alone it lost 44 per cent of its capital, which fell to $2.3 billion.

Meriwether wrote to investors on 2 September that the opportunity "is believed to be among the best that LTCM has ever seen". On the trades he was right. But the letter was on the internet within minutes, Merton recalled, and from then on the market traded against whatever it thought LTCM owned. Lenders began valuing its positions at forced-sale prices and calling for collateral to match. By 21 September it could default as soon as the 23rd.

Four days at the Fed

On Sunday 20 September a team led by Peter Fisher of the New York Fed went to LTCM's offices in Greenwich, Connecticut, and found positions "much greater than market participants imagined." On the Tuesday Fisher brought in Goldman Sachs, Merrill Lynch, J.P. Morgan and UBS, and by evening a wider group of firms.

On the Wednesday morning the talks stopped for a rival bid. Goldman, Warren Buffett's Berkshire Hathaway and AIG offered $250 million for the partners' stake plus $3.75 billion for the fund, with a 12:30 deadline. LTCM said it could not legally accept that quickly. The banks went back to arguing for five more hours. Eleven firms put in $300 million each and three put in less, about $3.6 billion in all, for 90 per cent of the fund and control of it.

Two firms refused. One was Bear Stearns, which had cleared LTCM's trades all along. "We never saw the swaps," its chairman Jimmy Cayne told Euromoney. "We didn't know how big they were."

Why not just lend?

McDonough said losses at LTCM's own counterparties, more than 75 of them, were not his main worry. The worry was everyone else: a rush to close out hundreds of billions of dollars of positions at once, hitting investors who had never dealt with the fund, until some credit markets might "cease to function for a period of one or more days and maybe longer." Then the line the Fed most wanted heard: "Not one penny of public money was spent or committed."

The shape was right. The firms that had lent LTCM the money to get that big paid to clean it up, and the partners' stake, about $1.6 billion at the start of 1998, was worth about $30 million by November. It was less clean than the line suggests. Greenspan admitted the Fed had moved faster than it would have "in more normal times", and between 29 September and 17 November it cut interest rates three times. The lesson markets could draw was that when a big enough firm got into trouble, the Fed would at least call the meeting. Ten years later Bear Stearns and Lehman Brothers, both at the table in 1998, were at the centre of the 2008 crisis.

What became of the trades

The consortium had its $3.6 billion back by the end of 1999.

Royal Dutch shares had traded about 8 per cent above Shell's in 1997, and during the crisis the gap went past 17 per cent. LTCM had bet it would close. In the spring of 2001, with LTCM long gone, the premium vanished.

Sources

• Federal Reserve Bank of New York, "Statement by William J. McDonough, President, Federal Reserve Bank of New York, before the Committee on Banking and Financial Services, U.S. House of Representatives", 1 October 1998 (the "failed them utterly" and "not one penny" quotations; the Fisher visit; the meetings of 22 and 23 September and the rival offer; three firms contributing less; two declining; more than 75 counterparties).
• Board of Governors of the Federal Reserve System, "Private-sector refinancing of the large hedge fund, Long-Term Capital Management", Greenspan testimony, 1 October 1998 (techniques being "emulated"; acting faster than "in more normal times").
• President's Working Group on Financial Markets, "Hedge Funds, Leverage, and the Lessons of Long-Term Capital Management", April 1999 (capital returned and remaining; assets, leverage and notional values; August losses; collateral calls; possible default on 23 September).
• General Accounting Office, "Long-Term Capital Management: Regulators Need to Focus Greater Attention on Systemic Risk", October 1999 (convergence trades and the Treasury example; the 550 million dollar day; the Goldman, Berkshire and AIG offer; the partners' stake falling to 30 million dollars).
• Federal Reserve History, "Near Failure of Long-Term Capital Management", Michael Fleming and Weiling Liu (founding date; returns; terms of the rival bid; 90 per cent for 3.6 billion dollars; repayment by end of 1999).
• NobelPrize.org, "The Prize in Economic Sciences 1997: Press release" (the award to Merton and Scholes on 14 October 1997).
• Donald MacKenzie, "Arbitrage", chapter 8 of An Engine, Not a Camera, MIT Press, 2006 (correlations and padded volatilities; predicted, actual and promised volatility; the superportfolio; Salomon's desk and Kaplanis; correlations near 0.7; the 2 September letter; the Royal Dutch and Shell premium to 2001).
• The Wall Street Journal, "A Hedge Fund Falters, So the Fed Persuades Big Banks to Ante Up", 24 September 1998 (eleven firms at 300 million dollars apiece).
• Euromoney, "Bear Stearns: The old firm goes its own way", February 1999 (Bear Stearns refusing to contribute; Cayne's quotation).
• Board of Governors of the Federal Reserve System, press releases of "29 September 1998", "15 October 1998" and "17 November 1998" (the three rate cuts).