Long-Term Capital Management did not fail because anyone made a single bad bet.
It failed because the bets were a single bet, the model behind them was treated as a fact instead of a model, and the people running the fund were the ones least likely to be challenged. By the time the position unwound in 1998, three Nobel prizes, two former Federal Reserve officials, and a long Wall Street résumé could not make the loss small.
The useful question was never whether the convergence trade was a good idea in 1994. It was whether a fund built on it could survive a world where convergence stopped working for one quarter.
The fund that was built like a small investment bank
John Meriwether ran Salomon Brothers' bond arbitrage desk from the late 1980s until 1991. The desk was responsible, on its own estimate, for nearly all of Salomon's global earnings for several years. After Meriwether left Salomon during the Treasury bond scandal, he spent two years putting together a hedge fund that looked less like a trading shop and more like a small Wall Street department.
Long-Term Capital Management began trading on February 24, 1994. Its first capital was just over $1 billion. Meriwether had recruited not just traders but also academics and former regulators. Myron Scholes and Robert C. Merton were board members; both would share the 1997 Nobel Prize in Economics for the Black-Scholes work that defined a generation of derivatives thinking. David W. Mullins Jr., a former Federal Reserve Vice Chairman, sat on the board. Larry Hilibrand, Victor Haghani, Eric Rosenfeld, and a small group of former Salomon traders rounded out the partnership.
The infrastructure around the fund was unusually stripped-down. Bear Stearns handled execution. Merrill Lynch handled client relations. There were about 150 employees. Costs were kept deliberately low so that nearly every basis point of return could flow back to partners and investors.
The first three years were, by the headline numbers, extraordinary. Annualized returns after fees were roughly 21% in 1994, 43% in 1995, and 41% in 1996. In 1997 the number dropped to 17%. Most of the trading public treats that year as the warning. Inside the fund it was treated as a reason to look harder for trades.
The trade
The core strategy was convergence trading. The idea was simple in theory and very hard in practice. Two bonds whose prices should track each other will sometimes trade apart because of supply, demand, liquidity, or technical noise. The fund would buy the cheaper one, short the more expensive one, and wait for the relationship to return to its historical pattern.
For convergence trades to produce meaningful returns, the fund had to take large positions. The mispricings were small, often a handful of basis points. To turn small mispricings into fund-level returns, the desk applied leverage. The standard textbook version of LTCM's portfolio by early 1998: equity of about $4.7 billion, debt of about $124.5 billion, total assets around $129 billion. Off-balance-sheet derivative positions had a notional value of roughly $1.25 trillion, most of it in interest rate swaps.
The deal with the universe was plain. If the spreads converged, the fund earned a steady, almost boring stream of small gains. If the spreads diverged unexpectedly, the losses on a leveraged book of that size were not small.
The market, by 1998, was no longer offering many cheap convergence trades. Other desks had seen the same model. The easy gains had already been taken. LTCM responded the way ambitious funds often respond when their best lane narrows. It widened what it was willing to do. It added emerging-market debt, merger arbitrage, equity volatility, and a large net short position on S&P 500 volatility. By the second quarter of 1998, the portfolio was more concentrated than the founding thesis had ever intended.
The slow bleed
In May 1998, LTCM lost 6.42%. In June, it lost 10.14%. Together that was about $461 million, against a starting equity of around $4.7 billion at the year. July was worse.
In July 1998, Salomon Brothers — then part of Travelers — announced it was exiting the bond arbitrage business. The liquidation of Salomon's own book hit the same spreads LTCM was in, and forced LTCM to compete with a former partner's wind-down. By the time Russia defaulted on its domestic GKO bonds on August 17, 1998, the fund had already lost more than a third of its capital.
Russia was the trigger, but the shape of the move mattered more than the headline. In a classic flight to quality, the most liquid assets get bid up and the less liquid ones get sold. LTCM's convergence book had been structured to be roughly neutral to ordinary moves. It was not structured to be neutral to a sudden, large, global move toward liquidity. The fund was long illiquid and short liquid in dozens of positions at once. When liquidity became the only thing the market wanted, every leg of the book moved against the fund at the same time.
Roger Lowenstein's reconstruction of the unwind describes a Royal Dutch / Shell pair that captures the problem in miniature. LTCM had built a $2.3 billion dual-listed company position in 1997 on the bet that the Royal Dutch premium over Shell would return to a small single-digit number. After the August shock, the premium widened to about 22%. LTCM was forced to unwind at exactly the moment it could least afford to. The single trade accounted for a large share of a $286 million loss on the pairs book. Other positions were worse.
The week the model broke
By the end of August, LTCM had lost $1.85 billion. In the first three weeks of September, equity fell from $2.3 billion to about $400 million. On the same liabilities, that translated into an effective leverage ratio of more than 250 to 1.
Counterparties started pulling financing. Repo lines were not extended. Margin calls multiplied. The fund tried to raise fresh capital. On September 23, 1998, Warren Buffett, Goldman Sachs, and AIG offered to buy out the partners for $250 million, inject $3.75 billion, and run the portfolio inside Goldman. Meriwether was given less than an hour to accept. The window closed.
The fund was not rescued by its investors. It was rescued by the institutions that had been its counterparties, organized by the Federal Reserve Bank of New York.
The bailout that changed the rules
On the evening of September 23, 1998, the New York Fed under William J. McDonough assembled representatives of fourteen financial institutions. The deal that emerged provided about $3.625 billion in new capital. The banks received 90% of the fund; the partners kept 10%, worth about $400 million on paper but consumed entirely by their own debts. Bear Stearns and Crédit Agricole declined to participate.
No public money was used. The Fed did not lend a dollar and did not guarantee a dollar. What the New York Fed did was more unusual. It organized the private rescue and used moral suasion to keep the counterparties in the room. The general counsel for LTCM during those negotiations was James G. Rickards.
The justification Greenspan gave to the House Banking Committee on October 1, 1998 was that an LTCM default would force a disorderly unwind of a portfolio so entangled with the rest of the financial system that the damage would not stay inside one hedge fund. Critics called it the Greenspan put in embryo: a quiet assurance that the central bank would not allow a large, sophisticated counterparty to fail on its own.
What the lesson actually is
LTCM is often filed under "smart guys got unlucky." That is the comfortable reading. It is not the useful one.
The fund had been warned. Seth Klarman declined to invest on the grounds that the leverage plan did not account for rare events. Warren Buffett and Charlie Munger both looked at the structure in 1993 and walked away. Paul Samuelson, Eugene Fama, and Mitch Kapor all questioned the model. Buffett later put the dynamic in his own words: the strategy was "picking up nickels in front of a bulldozer."
Niall Ferguson offered the most uncomfortable summary in The Ascent of Money. The fund's value-at-risk models had implied that the loss LTCM took in August 1998 was so improbable it should not have happened in the lifetime of the universe. That was because the models were calibrated on five years of data. Eleven years would have caught the 1987 crash. Eighty years would have caught the last great Russian default after the 1917 revolution. Meriwether, born in 1947, was later quoted as saying that if he had lived through the Depression, he would have understood events better.
Ferguson's line is the cleanest version: the Nobel laureates had known plenty of mathematics, but not enough history.
The lesson is not that convergence trading is wrong. It is that the size of a convergence book has to be set against the size of the move that breaks convergence. The size that makes the strategy profitable in normal times is the same size that makes it fatal when the regime changes. Leverage turns a small, well-known risk into a large, occasionally ruinous one, and the people closest to the model are usually the last ones willing to ask whether the model still describes the market.
Disclosure: Margin of Pain publishes research and commentary about traders, markets, and risk. This article is not investment advice or a recommendation to buy, sell, short, or hold any security, derivative, futures contract, currency, commodity, or asset.
Source trail
- Roger Lowenstein, When Genius Failed: The Rise and Fall of Long-Term Capital Management (Random House, 2000). Royal Dutch / Shell pair, August–September unwind, Salomon exit, internal partner commentary.
- Niall Ferguson, The Ascent of Money: A Financial History of the World (Penguin, 2008). Value-at-Risk calibration window, Meriwether on the Depression, "mathematics, but not enough history."
- Alan Greenspan, Private-Sector Refinancing of the Large Hedge Fund, Long-Term Capital Management, Testimony before the House Committee on Banking and Financial Services, October 1, 1998. Federal Reserve Board.
- Federal Reserve History, Near Failure of Long-Term Capital Management. federalreservehistory.org.
- "Hedge Funds and the Collapse of Long-Term Capital Management," Journal of Economic Perspectives 13(2), 1999. aeaweb.org.
- Long-Term Capital Management L.P., Wikipedia entry, used for cross-checking founding date, capital base, leverage ratio, loss attribution table, and counterparty list. en.wikipedia.org.
- Buffett / Goldman / AIG buyout offer and timing, secondary citations via Lowenstein and New York Times coverage of the September 1998 weekend.
- Credit Suisse's later Archegos post-mortem is included here only as a reminder that the lesson traveled; it is not a primary LTCM source.