Case studies · 1998
Long-Term Capital Management
A hedge fund used huge leverage to bet on small pricing gaps closing, and a shock in an unrelated market widened every spread at once.
What happened
A Wall Street pedigree
Long-Term Capital Management was founded in February 1994 by John Meriwether, formerly the head of bond trading at Salomon Brothers. Its partners included Myron Scholes and Robert C. Merton, who would share the 1997 Nobel Prize in economics, and David Mullins, a former vice-chairman of the Federal Reserve Board. The fund started with equity of $1.3 billion and quickly became the most admired hedge fund on Wall Street. After fees, it returned about 21% in 1995, 43% in 1996 and 41% in 1997 (an earlier year, 1994, had returned just under 20%). By the end of 1997, equity had grown past $7 billion. LTCM itself judged that its trading opportunities had shrunk relative to its size, so in December 1997 it returned $2.7 billion to outside investors, leaving equity of about $4.7 billion at the start of 1998.
A small edge, held in enormous size
By the start of 1998, the fund had borrowed more than $124.5 billion against that equity, taking total assets to about $129 billion, a debt-to-equity ratio of over 25 to 1. On top of that balance sheet, LTCM held off-balance-sheet derivatives, mostly interest-rate swaps and exchange-traded futures, with a combined notional value of about $1.25 trillion. None of this was hidden or unusual for the strategy: LTCM ran what is called convergence, or relative-value, arbitrage. It looked for pairs of very similar securities trading at slightly different prices, then bought the cheap one and sold the dear one, expecting the small gap to close. Each individual mispricing was only a few basis points, so turning it into a meaningful profit required very large positions, funded almost entirely with borrowed money rather than capital.
A default in Moscow, a rush to safety everywhere
The trigger came from a market LTCM had barely touched. On 17 August 1998, Russia devalued the rouble and declared a moratorium on 281 billion roubles, about $13.5 billion, of its domestic Treasury debt. Investors around the world reacted by fleeing to the most liquid, safest assets they could find and dumping anything less liquid, a pattern known as a flight to quality. That was the exact opposite of what LTCM's positions needed. The fund lost about 44% of its capital in August alone. Nearly every convergence trade it held moved the wrong way at the same time. Underneath their different labels, bond spreads, swap spreads, equity-index volatility, all shared the same underlying bet, that liquidity premia would shrink over time. When investors everywhere wanted liquidity at once, that bet failed everywhere at once.
The rescue
By mid-September 1998, LTCM's equity had fallen to about $600 million, and the fund had lost about $4.6 billion since the start of the year, in under four months. Early in September, LTCM told the Federal Reserve Bank of New York it was in trouble. On the evening of 22 September, the president of the New York Fed summoned more than a dozen senior executives from LTCM's lending banks and warned them that the systemic risk of a disorderly default was real. On 23 September, a consortium of 14 banks and securities firms, including Goldman Sachs, Merrill Lynch, J.P. Morgan, Morgan Stanley and UBS, agreed to inject about $3.6 billion of new capital in exchange for 90% of the fund's remaining equity. For the people who built the fund, the consequence was stark: the stakes of LTCM's 16 general partners, worth about $1.6 billion earlier in the year, were worth about $30 million by the end of September.
On 1 October 1998, Federal Reserve Chairman Alan Greenspan testified to Congress that the rescue "was not a government bailout." Federal Reserve funds, he said, "were neither provided nor ever even suggested." The distinction mattered: the New York Fed organised the meetings among the banks, but the banks supplied every dollar of new capital themselves. Greenspan's stated reason for getting involved at all was that a forced, disorderly liquidation of LTCM's book could have "impaired the economies of many nations, including our own," given how large its positions were relative to the markets it traded in. The fund itself was wound down under the new ownership consortium over the following months, and was fully liquidated by around early 2000.
The trades behind the collapse
Two of LTCM's core positions map directly onto this course. One bought the older, less liquid "off-the-run" long-dated Treasury bond and sold short the newest, most liquid "on-the-run" issue of the same maturity, the same on-the-run and off-the-run structure that sits behind cheapest-to-deliver bond pricing. The two bonds pay almost identical cash flows, so in normal markets the yield gap between them is small and tends not to persist. A second, related position bet that the spread between swap rates and government bond yields would narrow: a swap-spread trade like this pays or receives on an interest-rate swap while taking the opposite exposure in government bonds, so it lives or dies on exactly the same convergence logic as the bond-basis trade, just built with a different instrument.
What broke LTCM was never really a wrong view on where these spreads would end up. Most of them did eventually narrow again, in the years after 1998, just as LTCM had expected. What broke the fund was funding risk. A position that is marked to market every day turns any adverse move into an immediate demand for cash, through margin calls and tighter collateral haircuts. LTCM could not raise cash that fast, and it could not simply sell its way out either, because its positions were so large relative to the markets it traded in that selling would have pushed prices further against it. Leverage of over 25 to 1 meant that a single-digit percentage fall in the value of the whole portfolio was enough to destroy almost all of the fund's equity, long before the underlying spreads had any chance to converge back to normal.
| Date | Event |
|---|---|
| Feb 1994 | John Meriwether founds LTCM with equity of $1.3 billion. The board includes Myron Scholes, Robert C. Merton and former Fed vice-chairman David Mullins. |
| 1994-1997 | LTCM returns about 21% (1995), 43% (1996) and 41% (1997) after fees. Equity grows past $7 billion. |
| Dec 1997 | LTCM returns $2.7 billion to outside investors, leaving equity of about $4.7 billion for 1998. |
| Start 1998 | Borrowed funds exceed $124.5 billion; total assets are about $129 billion; derivatives notional is about $1.25 trillion. Leverage exceeds 25 to 1. |
| 17 Aug 1998 | Russia devalues the rouble and declares a moratorium on 281 billion roubles (about $13.5 billion) of domestic Treasury debt. |
| Aug 1998 | Investors flee to the most liquid government bonds and dump illiquid ones, the reverse of what LTCM's trades needed. The fund loses about 44% of capital in the month. |
| Mid-Sep 1998 | Equity has fallen to about $600 million; the fund has lost about $4.6 billion since the start of the year. |
| 22-23 Sep 1998 | The New York Fed convenes LTCM's creditor banks. On 23 September, 14 institutions agree to inject about $3.6 billion for 90% of the fund's remaining equity. |
| 1 Oct 1998 | Greenspan testifies to Congress that the rescue used no Federal Reserve money and was organised, not funded, by the New York Fed. |
| 2000 | LTCM's portfolio is fully wound down under the creditor consortium's ownership. |
The mechanics, in course language
Strip away the reputations and the size, and LTCM's trades were textbook convergence arbitrage: two nearly identical assets priced slightly apart, with opposite positions taken to profit when the gap closed. On the Treasury basis, LTCM was long the illiquid off-the-run bond and short the liquid on-the-run bond of the same maturity. In normal conditions, the two should trade close together because their cash flows are nearly the same; the small yield gap between them reflects a liquidity premium on the more heavily traded, on-the-run issue. LTCM's swap-spread positions used the same logic with a different instrument, taking opposing exposures in swaps and government bonds to profit if that spread narrowed too.
Leverage turned a tiny, reliable-looking mispricing into a fortune, and then into a catastrophe. A yield gap of a few basis points is worth almost nothing unless it is held in enormous size, which is exactly what LTCM's borrowing allowed. Equity of $4.7 billion supported roughly $129 billion of assets and about $1.25 trillion of derivatives notional. When the Russian default hit on 17 August 1998, it did not change anyone's view of US Treasury or swap fundamentals. What it changed was what investors wanted to hold: liquid, safe assets, immediately, everywhere. LTCM was short the liquid bonds, which rallied on that demand, and long the illiquid ones, which fell further as everyone tried to sell them at once. Many of LTCM's seemingly diverse positions, Treasury basis, swap spreads, equity volatility, shared this same underlying liquidity bet, which is why they all lost money together instead of offsetting each other the way genuine diversification would.
Daily mark-to-market accounting is what turned this into a survival problem rather than a paper loss. Every adverse move generated a real cash demand through margin calls and collateral top-ups. LTCM had almost no spare capital, because its business model was to hold the minimum equity that its lenders would tolerate against maximum borrowed size. It could not raise cash quickly enough to meet those demands, and selling its huge positions into a market that already knew it was in trouble would have pushed prices even further against it. Leverage does not only amplify how much a position makes or loses on a given move; it amplifies how fast a paper loss becomes a demand for real cash the fund may not have.
The mathematics
One figure fixes the scale of the fragility: how small a move in LTCM's overall portfolio was enough to destroy almost all of its equity, given the leverage it carried at the start of 1998.
$$\text{Leverage} = \frac{\text{Assets}}{\text{Equity}} = \frac{129.0}{4.7} \approx 27.4 \text{ to } 1$$
$$\text{Implied fall in portfolio value} = \frac{\text{Loss}}{\text{Assets}} = \frac{4.6}{129.0} \approx 3.57\%$$
$$\text{Share of starting equity wiped out} = \frac{\text{Loss}}{\text{Equity}} = \frac{4.6}{4.7} \approx 97.9\%$$
equity_start = 4.7 # USD bn, equity at start of 1998
assets = 129.0 # USD bn, total assets at start of 1998
actual_loss = 4.6 # USD bn, total loss over 1998 (to mid-Sep)
leverage = assets / equity_start
implied_move_pct = actual_loss / assets * 100
equity_remaining = equity_start - actual_loss
pct_equity_wiped = actual_loss / equity_start * 100
print(f"Leverage: {leverage:.1f} to 1")
print(f"Implied fall in portfolio value: {implied_move_pct:.2f}%")
print(f"Equity remaining: ${equity_remaining:.2f}bn")
print(f"Share of starting equity wiped out: {pct_equity_wiped:.1f}%")Output
Leverage: 27.4 to 1
Implied fall in portfolio value: 3.57%
Equity remaining: $0.10bn
Share of starting equity wiped out: 97.9%These numbers set the scale of the fragility. At the start of 1998, LTCM held about $129 billion of assets against only $4.7 billion of equity, leverage of about 27 to 1. This is often cited more loosely as "over 25 to 1" for the fund; that is the same fact, just rounded down. Its actual 1998 loss of about $4.6 billion corresponds to a fall in the value of its whole book of only about 3.6%, an ordinary-looking move for a bond and swap portfolio. At 27-to-1 leverage, that single-digit percentage move destroyed roughly 98% of the fund's starting equity. The arithmetic carries the lesson on its own: a trade can be right about where prices will eventually settle and still prove fatal, because leverage means a small asset-price move is a huge move in equity.
Data and facts
| Quantity | Value | Source |
|---|---|---|
| Founding equity, Feb 1994 | $1.3bn | Edwards (1999), p.196 |
| Returns after fees, 1995 / 1996 / 1997 | ≈21% / 43% / 41% | Edwards (1999), p.196 |
| Equity at start of 1998 | ≈$4.7bn | Edwards (1999); Federal Reserve History |
| Borrowed funds, start of 1998 | over $124.5bn | Edwards (1999), p.198 |
| Total assets, start of 1998 | ≈$129bn | Edwards (1999); course figures |
| Debt-to-equity / leverage | over 25 to 1 | President's Working Group, Apr 1999 |
| Derivatives notional, start of 1998 | ≈$1.25tn (mostly interest-rate) | Edwards (1999), p.198 |
| Russian Treasury moratorium, 17 Aug 1998 | 281bn roubles (≈$13.5bn) | Edwards (1999), p.199 |
| Loss in August 1998 alone | ≈44% of capital | Greenspan testimony, 1 Oct 1998 |
| Total loss, start of 1998 to mid-Sep | ≈$4.6bn | Course figures; corroborated by Edwards (1999) |
| Recapitalisation, 23 Sep 1998 | ≈$3.6bn, 14 institutions | GAO Report GGD-00-67R; President's Working Group |
| General partners' stakes, early vs late 1998 | ≈$1.6bn → ≈$30m | Edwards (1999), p.200 |
The lesson
- A leveraged bet can be right about the destination and still fail on the journey. LTCM's convergence trades mostly narrowed again in the years after 1998, but the fund was gone before that happened. Being correct about fair value does not protect you if you cannot fund the position long enough to get there.
- Margin and mark-to-market convert a market-risk view into a liquidity problem. A position marked to market daily turns any adverse move into an immediate cash demand. Leverage does not just multiply profit and loss, it multiplies how fast a paper loss becomes a real one.
- Diversification across trades is not diversification across risk factors. LTCM held many different-looking positions, the Treasury basis, swap spreads, equity volatility, that all shared one underlying bet: that liquidity premia would shrink. When a single shock made investors want liquidity everywhere at once, every position lost money simultaneously.
- Concentrated leverage turns a survivable move into a fatal one. At leverage of over 25 to 1, a single-digit percentage move in the whole portfolio was enough to destroy almost all of the fund's equity. The same move at a much lower leverage ratio would have been a manageable loss.
- Models price historical relationships, they do not guarantee against genuinely new shocks. LTCM's models, built by some of the most respected minds in finance, priced the past behaviour of these spreads well, but a sovereign default and a global flight to quality fell outside the range they had been calibrated on.
Where it appears in the course
Think about it
- LTCM's models judged its trades to be very low-risk because the underlying spreads had always converged before. What made the events of August 1998 different from the historical range those models were built on?
- If LTCM's core view (that these spreads would eventually narrow) turned out to be broadly correct, in what sense did the fund still fail? What would need to be true about its funding for a "correct but early" view to survive?
- The rescue consortium argued that letting LTCM default in a disorderly way risked wider damage to the financial system. What is the difference between rescuing a firm because it is important, and rescuing it because unwinding it badly would hurt other firms and markets too?
Sources
- Franklin R. Edwards, "Hedge Funds and the Collapse of Long-Term Capital Management," Journal of Economic Perspectives, Vol. 13, No. 2 (Spring 1999), pp. 189-210. jstor.org
- Alan Greenspan, Testimony before the House Committee on Banking and Financial Services, "Private-sector Refinancing of the Large Hedge Fund, Long-Term Capital Management," 1 October 1998. federalreserve.gov
- President's Working Group on Financial Markets, Hedge Funds, Leverage, and the Lessons of Long-Term Capital Management, April 1999. home.treasury.gov
- US General Accounting Office, Letter Report GGD-00-67R, on the LTCM recapitalisation, 2000. gao.gov
- Federal Reserve History, "Near Failure of Long-Term Capital Management." federalreservehistory.org
- Contemporary reporting on the 22-23 September 1998 rescue meeting, as cited in Edwards (1999), p.200, drawing on Michael Siconolfi's coverage in the Wall Street Journal.
