What Happened
Long-Term Capital Management was founded in 1994 by John Meriwether, the legendary Salomon Brothers bond trader, along with Myron Scholes and Robert Merton who would win the Nobel Prize in Economics in 1997. The fund used quantitative models to identify small pricing inefficiencies across global bond markets and exploited them with enormous leverage.
For four years LTCM generated returns of 21%, 43%, 41%, and 17%. Wall Street banks competed to lend to the fund, allowing it to build $125 billion in assets on just $4 billion in equity — 30-to-1 leverage. The models showed all positions were hedged. The models were wrong.
Russia defaulted in August 1998 and devalued the ruble. The yen carry trade unwound simultaneously. The correlations in LTCM’s models went to 1 — all assets fell together instead of providing diversification. The fund lost $4.6 billion in four months. The Federal Reserve Bank of New York organized a $3.6 billion private rescue from 14 banks to prevent a disorderly unwind that threatened global credit markets.
The Mechanism
The correlation assumption and the limits of historical modelsLTCM’s models were built on historical data that showed asset prices were imperfectly correlated — when some fell, others held steady, providing natural hedges. This assumption holds in normal markets. In a crisis, investors sell everything simultaneously to raise cash. Correlation goes to 1. Every hedge fails at exactly the moment you need it most. LTCM held enormous positions in every major credit market. When they needed to sell, there were no buyers because all potential buyers were in the same trades.
What the Consensus Believed
The prevailing view before the reckoning
The Nobel laureates had solved risk. Quantitative models could identify safe arbitrage opportunities and manage them with precision. The leverage was safe because the positions were perfectly hedged. A fund with this intellectual firepower and this track record was the safest investment in finance.
What the Record Shows
Models are not reality
Historical correlations break down in crises. The data used to build LTCM's models did not include scenarios where all major markets fell simultaneously. They assumed it was essentially impossible. It happened.
Leverage is catastrophic
At 30x leverage, a 3.3% decline in asset values wipes out equity entirely. LTCM's positions fell far more than that. Leverage guarantees catastrophic loss under adverse conditions that always eventually occur.
Liquidity and solvency are different
LTCM was arguably solvent — if given time, many positions would have converged. But it was illiquid — it could not meet margin calls without selling into markets where it was the only seller. Liquidity killed it before the positions could recover.
Systemic risk is invisible until it is acute
No regulator had a complete picture of LTCM’s exposure because it reported to no single regulatory body. The systemic risk was only discovered when the fund was already failing. This lesson was not learned before 2008.
↑ Cognitive pattern: Model Worship — Confusing quantitative precision with accuracy
Key Voices
Called It Right
Warren Buffett
Berkshire Hathaway
“To make money they didn't have and didn't need, they risked what they did have and did need. Foolish.”
Post-crisis Right
Nassim Taleb
Various
“Black swans exist. Models built on historical data cannot capture events that have no historical precedent. LTCM proved this.”
1997 Predicted model failure
Wrong
LTCM Partners
LTCM
“LTCM has the most sophisticated risk management system ever built. Our models have never failed and our positions are fully hedged.”
January 1998 Fund collapsed August 1998
Wall Street Banks
14 major institutions
“LTCM is the safest counterparty in the market. They have Nobel laureates and the best models. Lend them as much as they want.”
1997–1998 Then organized rescue
The Rescue
Alan Greenspan
Federal Reserve
“The Fed organized a private sector rescue. This was not a government bailout — taxpayer money was not used. But the systemic risk was real.”
September 1998 Organized rescue