What Happened to Long-Term Capital Management (LTCM)?

Long-Term Capital Management was a Greenwich, Connecticut hedge fund that lost about $4.6 billion in 1998 and was taken over by a consortium of 14 banks and brokers on 23 September 1998. It made small convergence bets on bond spreads at more than 25 times its capital. When Russia defaulted in August 1998, nearly every spread it was betting on widened at once, and it ran out of capital before they converged.

The fund's investors had every reason for confidence. John Meriwether had run the most profitable trading desk on Wall Street, Salomon Brothers' bond-arbitrage group. His partners included Myron Scholes and Robert Merton, who would share the Nobel prize in economics in October 1997, and David Mullins, a former vice-chairman of the Federal Reserve. The fund returned about 43% in 1995 and 41% in 1996. On 2 September 1998, Meriwether wrote to investors that the fund had lost 44% in August alone.

What was Long-Term Capital Management?

LTCM was founded in February 1994 with the core of Meriwether's former Salomon Brothers arbitrage group. Meriwether had resigned from Salomon in August 1991 after one of his traders was found to have submitted false bids at Treasury auctions. He rebuilt the team outside a bank, raising about $1.25 billion at launch, as reported.

The method was the Salomon desk's. Buy a cheap security, sell an expensive, nearly identical one, and wait for the gap to close. The fund's best-known positions show the idea.

The loss breakdown comes from Roger Lowenstein's When Genius Failed (2000), a secondary account rather than an audited one. Its total is about $4.6 billion.

The common thread is easy to miss. Almost every line sold liquidity or insurance. LTCM held the less liquid, less popular or riskier leg and was paid a small premium for it. That premium is real, which is why the trades worked from 1994 to 1997. It is also the premium that disappears all at once when investors flee to safety.

How did LTCM make 40% a year?

With small spreads and very large leverage. A convergence trade might earn tens of basis points a year on the gross position. Turning that into about 40% a year on equity required leverage of 25 to 30 times plus derivatives.

The President's Working Group on Financial Markets, which reported on the fund in April 1999, gave the numbers. At the end of 1997 LTCM had $129 billion of balance-sheet assets on capital of $4.8 billion, leverage of about 28 to 1, and about $1.3 trillion of off-balance-sheet derivatives notional. It dealt with about 75 repo counterparties and about 50 derivatives counterparties. At 25 to 1, a 1% fall in asset values costs a quarter of capital.

The leverage was cheap for a reason that had nothing to do with the trades. Many counterparties, impressed by the partners, took little or no initial margin and settled for daily variation margin. In the language of The Edge Equation, that was a funding subsidy. It raised returns in calm years and hid the risk.

The signature trade shows how thin the underlying edge was. The companion's catalogue grades on-the-run versus off-the-run Treasuries C/F. Krishnamurthy's test (Journal of Financial Economics, 2002) found that the spread does converge, but the cost of shorting the new bond in the repo market eats the profit: the trade earned 11.1 basis points a year with a standard error of 34 basis points, before bid–ask costs. LTCM's whole business model is in that paper. A small gross edge was made worthwhile only by leverage.

The decision that raised the risk

At the end of 1997, after a year of about 17%, LTCM returned about $2.7 billion to outside investors. It did not cut its positions. The same book now sat on about a third less capital. The investors who left with that distribution kept large gains. Those who stayed lost most of what they had.

Why did LTCM fail?

The losses began in May 1998, and capital was $4.1 billion at 31 July. Then Russia devalued and defaulted on its domestic debt in mid-August. Investors everywhere sold whatever was illiquid, unpopular or risky and bought whatever was not. LTCM was long the first kind and short the second, almost everywhere.

Capital from the President's Working Group, except the September figure; leverage is the companion's arithmetic on those figures.

Leverage rose as the fund lost money, because capital fell faster than positions could be sold. Nobody added risk. The losses did it automatically. That is the first step of the liquidity spiral the study of blow-ups and liquidity spirals describes.

Four weaknesses decided the outcome.

Correlation under stress. Swap spreads, Italian bonds, Royal Dutch and Shell, and equity volatility had little to do with one another in normal times. They all lost at once because the same investors fled the same kind of risk. The diversification in LTCM's risk model was a calm-weather property.

Crowding. Banks' proprietary desks ran copies of LTCM's trades. Salomon Smith Barney restructured its US bond-arbitrage desk in July 1998 and sold similar positions. Every seller's sales were every other holder's losses.

Size. Positions were large relative to their markets. Dealers who knew roughly what LTCM held could mark against it and trade ahead of its exit. In long-dated equity volatility the fund was reportedly so large that dealers called it the "central bank of volatility."

Funding. Lock-ups protected the fund from its own investors but not from margin calls and its counterparties' marks. The capital that mattered could be withdrawn by lenders.

How was LTCM rescued?

On 18 September 1998 the fund called the Federal Reserve Bank of New York. LTCM's own estimate, reported by the President's Working Group, was that its top 17 counterparties would lose $3 billion to $5 billion in aggregate in a default. That estimate is why the New York Fed judged a disorderly failure a systemic risk.

On 22 September the New York Fed hosted the banks. A bid led by Warren Buffett, $250 million for the partners' stake followed by $3.75 billion of new capital, lapsed at 12:30 pm on 23 September. At about 6 pm that day, 14 banks and brokers agreed to put in $3.625 billion for 90% of the fund. The original owners kept 10%. Two firms in the talks declined. The Fed convened the rescue and lent nothing, and the consortium was fully repaid by the end of 1999.

Many of the spreads converged after the consortium took control. That is the point of the case. LTCM's models were not obviously wrong about where prices would end up. They were wrong about whether the fund would be allowed to wait.

What does LTCM teach about leverage and survival?

The chapter on arbitrage and relative value separates two kinds of price gap. A convertible gap closes on a known date because a contract forces it to, as an index future converges to the index at expiry. A non-convertible gap has nothing forcing it closed by any date. Most of LTCM's book was the second kind, and that kind can widen first.

Shleifer and Vishny had described the trap the year before, in "The Limits of Arbitrage" (Journal of Finance, 1997). Arbitrage is done by a few specialists, often with other people's money, and it fails exactly when prices are furthest from value, because that is when backers and lenders want their money back.

The comparison with firms that survived the same trades is the clearest lesson.

Berkshire sold the same kind of long-dated equity volatility a decade later and collected premiums of $4.9 billion. It was structured so that no counterparty could force a sale. The premium was the same. The structure decided the outcome.

The President's Working Group drew the policy lesson. It found that the market discipline meant to come from lenders had failed, and it recommended more disclosure and better counterparty risk management. It did not recommend direct leverage limits. The same pattern returned in 2021, when Archegos spread its leverage across about ten banks, none of which could see the whole book.

LTCM in the four factors of the edge equation

The Edge Equation judges every trading business on four factors that multiply: edge, survival, execution, and discipline and adaptation.

Edge was partly there. The spreads were genuine premia for supplying liquidity, and the catalogue grades the modern Treasury cash–futures basis trade, LTCM's descendant, B for institutions. But part of the edge was the financing terms, and those vanished in August 1998.

Survival was the factor that went to zero. Leverage that rose as capital fell, positions too big to sell quickly, correlations that converged to one, and lenders who could leave made the path to being right longer than the fund's capital.

Execution failed with it, because the mark was never the exit price for positions of that size.

Discipline is mixed. The partners had most of their wealth in the fund, which aligned incentives but left no outside cushion. Returning $2.7 billion without cutting positions in late 1997 was the decision that made the book fragile.

The chapter on blow-ups shows the same pattern in other cases, and the chapter on sizing, leverage and survival covers the arithmetic. For more cases, see our list of the biggest trading losses in history and our overview of how trading firms make money.

The takeaway

LTCM did not fail because its models were foolish or its traders careless. It failed because it ran small, real premia at leverage that left no room for the path to convergence, financed by lenders who could change their minds. When a flight to safety hit every position at once, losses raised the leverage, the leverage forced sales, and the sales deepened the losses. A trade that is right at maturity is still a bet on the path if someone else controls the funding. The full account is in Chapter 15 of The Edge Equation, and the company profile of Long-Term Capital Management holds the dated record.