Two Nobel Laureates Ran a Hedge Fund at 25-to-1 Leverage. It Lost 90% in Four Months, and 14 Banks Had to Save It.

John Meriwether's bond-arbitrage group at Salomon Brothers, two Black-Scholes Nobel winners, and a former Fed vice chairman found tiny mispricings between near-identical bonds, hedged out the directional risk, and leveraged the slivers 25 times. The trades were market-neutral and, held to maturity, could not lose. The problem was affording to hold them that long: Russia defaulted, every spread widened at once, and the leverage that minted the returns turned margin calls into a death spiral before the fund could be proven right.

In 1994 John Meriwether founded Long-Term Capital Management, bringing along the bond-arbitrage desk he had run at Salomon Brothers until the 1991 Treasury-auction scandal pushed him out. He staffed it with those traders, plus Myron Scholes and Robert C. Merton, who would share the 1997 Nobel Prize in Economics for the Black-Scholes option-pricing framework, and David Mullins, a former vice chairman of the Federal Reserve. The credibility was the product. Counterparties let LTCM enter interest-rate swaps with zero initial margin and lend it the full value of any top-grade collateral, on the strength of the names.

The strategy was convergence arbitrage. Find two bonds that should trade at nearly the same price, buy the cheap one, short the expensive one, wait for the gap to close. The textbook trade was the on-the-run vs. off-the-run Treasury spread: a freshly issued 30-year Treasury yielding 5.50% against an older 29-year issue at 5.62%. The 0.12% gap was pure liquidity premium, and it narrowed within six months when the Treasury issued a new bond and the old benchmark's premium faded. Pair both legs and you remove the directional risk; you are betting not on whether rates rise but on whether two near-identical instruments re-price toward each other. The edge per trade was a few basis points.

So LTCM borrowed. A few basis points becomes a 20-to-30% return on equity when you lever it 25 times. At the start of 1998 the fund held about $4.7 billion of equity, had borrowed roughly $125 billion, sat on $129 billion of assets, and carried off-balance-sheet derivatives with a notional value of $1.25 trillion. The leverage was not accidental. At the end of 1997 LTCM returned about $2.7 billion to investors without shrinking its positions, pushing the ratio from roughly 17-to-1 to 25-to-1. Alan Greenspan said it plainly: LTCM "reached further for return over time by employing more leverage and increasing its exposure to risk, a strategy that was destined to fail." The fund had earned 43% and 41% in its second and third full years. Now it had less cushion and the same size book.

On August 17, 1998, Russia defaulted on its debt. Investors fled to the safest, most liquid assets they could find, the one thing convergence arbitrage cannot survive. Every spread LTCM had bet would narrow widened at once: the on-the-run premium blew out, European sovereign yields diverged from German bunds, swap spreads gapped. The correlations the models assumed stable went to one, and the diversification that looked real in calm markets vanished. William McDonough, president of the New York Fed, told Congress the diversification "failed them utterly." LTCM lost 44% of its value in August, $1.8 billion that month.pdf). Equity collapsed from $2.3 billion to $400 million by late September while liabilities stayed above $100 billion, an effective leverage above 250-to-1.

Leverage converts a small adverse move into a margin call, and a margin call into a forced sale at the worst price, which deepens the loss and triggers the next call. The fund could not hold the trades long enough for them to be right. On September 23 the New York Fed convened its largest creditors. A last-minute bid led by Warren Buffett, Berkshire putting up $3 billion, AIG $700 million, and Goldman Sachs $300 million, with $250 million to buy out the partners, expired at a 12:30 p.m. deadline over legal hurdles. By 6 p.m., fourteen banks and brokerages had put up $3.625 billion for 90% of the fund, eleven of them at $300 million each. The Fed supplied no public money, though Greenspan conceded the episode stretched the lender-of-last-resort tradition to a private partnership for the wealthy.

The sting is the epilogue. In the year after the recapitalization the fund earned about 10%, and by early 2000 the consortium was repaid in full. The trades had been right. The spreads did converge. LTCM could not afford to be right on a calendar it did not control. As Meriwether wrote to investors on September 2, "the Fund added to its positions in anticipation of convergence, yet... the trades diverged dramatically." The models priced the probability of convergence. They did not price the cost of surviving the divergence.

The takeaway for a modern operator: convergence arbitrage is sound and leverage is how you scale it, but leverage changes the binding constraint from "am I right?" to "can I stay solvent long enough to be right?" At 25-to-1, a 4% adverse move erases your equity, and a crisis makes every position move against you at once because the correlations you diversified across all go to one. The edge is real, the model is correct, and you still go broke, because the people you borrowed from can seize your collateral before your thesis matures. In a leveraged carry trade the binding constraint is never the quality of the analysis. It is liquidity and time: how long you can hold before someone forces you to sell at the price that makes you look wrong.