The Most Famous Quant Disaster
In September 1998, a Greenwich, Connecticut hedge fund called Long-Term Capital Management lost 3.6 billion bailout to prevent the unwinding from triggering a global financial crisis.
LTCM remains the canonical case study in how brilliant quantitative thinking, combined with extreme leverage and underestimation of correlation in stress, can produce catastrophic results.
For broader context on systemic financial events, see our Flash Crash explainer and quantitative easing explained.
The All-Star Lineup
LTCM was founded in 1994 by John Meriwether, the former head of bond arbitrage at Salomon Brothers. The partner roster was extraordinary:
- Myron Scholes - co-creator of Black-Scholes options pricing model, 1997 Nobel laureate
- Robert Merton - foundational figure in continuous-time finance, 1997 Nobel laureate
- David Mullins - former Vice Chairman of the Federal Reserve
- Eric Rosenfeld, Larry Hilibrand, Greg Hawkins - star traders from Salomon's legendary bond arb desk
- Victor Haghani, Hans Hufschmid - from Salomon's London operations
The fund opened with 10 million and 3-year lockups.
The Strategy
LTCM's core strategy was convergence trading: identifying pairs of related securities that had diverged in price and betting they would converge.
Examples:
-
On-the-run vs off-the-run Treasuries: Newly-issued Treasuries trade at slightly lower yields (higher prices) than older but otherwise-equivalent Treasuries due to liquidity preferences. LTCM would short the on-the-run and buy the off-the-run, capturing the convergence as the on-the-run "aged."
-
Sovereign yield curves: Italian and Spanish bonds yielded more than German bonds. LTCM bet the spreads would narrow as European integration proceeded.
-
Mortgage-backed securities: LTCM took complex relative-value positions in different MBS tranches.
-
Equity volatility: LTCM sold long-dated index volatility (S&P 500), betting that long-term realised vol would be lower than implied vol priced in long-dated options.
-
Merger arbitrage: Long-acquirer-short-target spread positions.
The expected return on each individual trade was tiny - often a few basis points. To generate attractive returns, LTCM applied enormous leverage.
The Leverage
LTCM ran with leverage of approximately 25:1 on its balance sheet at the peak. With 125 billion in assets.
But that understates the real exposure. Including off-balance-sheet derivatives positions, LTCM's notional exposure exceeded $1 trillion - roughly 5% of global GDP at the time.
This level of leverage was possible because:
- Counterparties believed the strategies were truly low-risk
- The partners' personal credibility (Nobel laureates!) reassured banks
- Most positions were collateralised, so apparent counterparty risk was managed
- Banks competed to provide the most generous financing terms
The Early Years (1994-1997)
For its first four years, LTCM produced extraordinary returns:
| Year | Net return |
|---|---|
| 1994 (partial) | 20% |
| 1995 | 43% |
| 1996 | 41% |
| 1997 | 17% |
Net of the fund's high fees, investors earned ~20-40% annually. The fund's strategies appeared to be working exactly as the partners had modelled.
By 1997, LTCM had grown to ~2.7 billion of capital to outside investors - they wanted to keep more of the upside for themselves and their own personal investments. This proved disastrous in retrospect: the fund kept the same level of leverage but with a smaller equity base, increasing the leverage ratio.
1998: The Year It Fell Apart
The trigger: Russian default (August 1998)
In August 1998, Russia defaulted on its sovereign debt and devalued the rouble. This was unexpected: even with the IMF rescue package, markets had assumed sovereign defaults were essentially impossible in the late 1990s.
The default created a "flight to quality" panic. Investors dumped riskier bonds (emerging market debt, off-the-run Treasuries, corporate bonds) and bought safe assets (on-the-run US Treasuries, German bunds).
Why this destroyed LTCM
LTCM's positions were structured as: long the "less liquid" / "less safe" asset, short the "more liquid" / "more safe" asset. The flight to quality moved both legs of these trades against the fund simultaneously.
Crucially:
- All of LTCM's positions correlated against the fund
- The historical correlation models had under-estimated tail correlation
- Markets that were "uncorrelated" in normal times became perfectly correlated in stress
This is the canonical lesson of LTCM: correlation in stress is not the same as correlation in normal times.
The cascade
By late August, LTCM was losing $100M+ per day. The fund's positions were so large that any attempt to liquidate would move the markets against itself. Counterparties began demanding more collateral. The fund couldn't reduce leverage because reducing positions would crystallise losses and trigger more margin calls.
By mid-September, LTCM was facing imminent insolvency. The fund had lost ~600 million against $125 billion+ in positions.
The Bailout
If LTCM had been allowed to default and liquidate:
- Counterparties would have been forced to liquidate the collateral they held
- The forced selling would have crashed multiple markets simultaneously
- Other major financial institutions held similar positions and would have suffered cascading losses
- Goldman Sachs estimated total counterparty losses could reach $200 billion+
The Federal Reserve Bank of New York organised a meeting on September 23, 1998. The major investment banks (those that were also LTCM's largest counterparties) were essentially compelled to put up $3.6 billion in fresh equity in exchange for 90% ownership of the fund.
The fund was wound down over the next several years. Outside investors lost most of their capital. The partners themselves lost significant personal wealth.
The Lessons
1. Leverage amplifies everything, including unmodeled risks
A strategy with 1% expected return and 0.5% expected volatility looks attractive. The same strategy at 25:1 leverage has 25% expected return and 12.5% expected volatility - and a tail risk that the modelled distributions don't capture.
2. Correlations change in stress
Historical correlation matrices underestimate tail correlation. When markets panic, "uncorrelated" assets all move together because they're all responding to the same liquidity shock.
3. Liquidity is conditional
LTCM's positions were liquid in normal markets and illiquid in stressed markets. Models that assumed continuous liquidity were wrong.
4. Capacity matters
LTCM's positions were so large they couldn't be unwound without moving markets. Capacity constraints aren't visible in normal times but dominate in crises.
5. Brilliant people can be wrong
The presence of two Nobel laureates and former Fed Vice Chair didn't prevent disaster. Pedigree doesn't protect against systematic risks.
6. Counterparty risk is real
LTCM's bilateral derivatives counterparties had no idea how concentrated their exposure was - because LTCM was running similar trades against multiple banks who didn't share information.
Aftermath and Legacy
Regulatory response
Post-LTCM, regulators tightened oversight of hedge funds, particularly:
- Better disclosure of large positions to regulators
- Improved counterparty risk management at major banks
- More rigorous internal stress testing
- Eventually (post-2008), Dodd-Frank and the Volcker Rule
Sequel: JWM Associates
John Meriwether founded a successor fund, JWM Associates, in 1999. It was wound down in 2009 after losses in the 2008 crisis. The pattern repeated.
Surviving partners' legacy
Several LTCM partners had successful subsequent careers:
- Larry Hilibrand and others went into family office investing
- Myron Scholes consulted and joined boards
- Robert Merton continued academic work and consulted
- Eric Rosenfeld founded JWM Associates with Meriwether
The intellectual contributions of LTCM's partners (Black-Scholes-Merton, ICAPM, continuous-time finance) remain central to modern quant finance.
What Modern Quants Should Take From LTCM
If you're starting a quant career in 2026, the LTCM story is essential context:
-
Stress testing is not optional. Every strategy needs explicit consideration of: what happens if correlations go to 1, liquidity disappears, counterparties withdraw credit?
-
Position sizing matters more than alpha. A great strategy at the wrong size is a disaster.
-
Leverage is a multiplier on hidden risks. It's tempting to use leverage to make a low-edge strategy attractive. The risks compound non-linearly.
-
Models have limits. Every model is a simplification of reality. The simplifications matter most when reality diverges from them.
-
Humility is a survival trait. The smartest people are most at risk of trusting their own models too much.
For more on managing these risks:
- Quant research interview questions (risk section)
- Time series interview questions
- Machine learning finance guide
Further Reading
- When Genius Failed by Roger Lowenstein - the definitive book on LTCM
- Inventing Money by Nicholas Dunbar - earlier and more technical account
- The Quants by Scott Patterson - covers LTCM as part of broader quant history
- Liar's Poker by Michael Lewis - background on the Salomon Brothers culture that produced Meriwether and his team
For other major financial-market events:
Practise the questions Long-Term Capital Management Collapse: The 1998 LTCM Disaster Explained actually asks
Reading about the interview is one thing - sitting one is another. Quantt's interactive coding tests are modelled on the same problem types that show up in firms like Jane Street, Citadel, Hudson River and Optiver. Run real Python in the browser, get instant feedback, and benchmark yourself against the bar.
Free to start - no credit card required