The Collapse of Long-Term Capital Management (1998): When Nobel Laureates Nearly Destroyed Wall Street

LTCM: Finance's Dream Team

In 1994, John Meriwether — the legendary Salomon Brothers trader made famous by the book 'Liar's Poker' — founded Long-Term Capital Management in Greenwich, Connecticut.

His team was simply extraordinary:

• Myron Scholes: 1997 Nobel Laureate, co-inventor of the Black-Scholes model for option pricing
• Robert Merton: 1997 Nobel Laureate, the most prominent mathematician of modern finance
• David Mullins: Former Vice Chairman of the Federal Reserve
• Dozens of physicists, mathematicians, and computer scientists from the world's top universities

Initial results were spectacular:

• 1995: +59%
• 1996: +57%
• 1997: +25%

At its peak, LTCM managed $4.7 billion in equity capital but controlled $125 billion in assets with a 25:1 leverage. Through derivatives, its effective exposure was estimated at $1 trillion — equal to Italy's GDP at the time.

The Strategy: Convergence Arbitrage

The LTCM strategy was based on a simple but powerful idea: financial markets occasionally price similar assets slightly differently. If you buy the 'cheap' asset and sell the 'expensive' one, prices will eventually converge, and you will earn a profit without directional risk.

Examples of positions:

• European sovereign bonds: the spread between Italian BTPs and German Bunds seemed too wide (pre-Euro)
• On-the-run vs off-the-run securities: new US Treasuries vs slightly older ones
• Index arbitrage: differences between futures and spot markets

The problem: the 'risk-free gains' were microscopic. To generate high returns, LTCM needed massive leverage.

1998: The Perfect Storm

Two sudden, correlated events shattered LTCM's models:

August 17, 1998: Russia Defaults Russia — caught between low oil prices and capital flight — declared a debt moratorium and devalued the ruble. It was the first default of an industrialized nation in modern history.

Panic spread across all emerging markets. Investors worldwide sold risky assets and bought 'safe havens' (US Treasuries, gold).

The problem for LTCM: LTCM's quantitative models were built on historical data that did not include such extreme events. Markets stopped behaving 'normally.' The convergence expected by the spreads did not occur — the spreads widened further.

Contrary to what the models predicted, all markets moved in the same direction at the same time — the wrong one for LTCM.

The Death Spiral

August 1998: LTCM loses 44% of its capital in one month Early September: The fund has lost 92% of its capital — from $4.7 billion to $600 million

With 25:1 leverage and $125 billion in assets, LTCM was technically insolvent. The New York Fed summoned the CEOs of the 14 largest American and international banks (Goldman Sachs, Merrill Lynch, JP Morgan, Deutsche Bank, UBS, etc.).

If LTCM had defaulted, it would have been forced to sell $125 billion in assets in a matter of weeks — enough to destabilize every market in which it operated.

The Fed-Orchestrated Bailout

On September 23, 1998, 14 banks agreed to inject $3.6 billion into LTCM in exchange for 90% ownership of the fund. The government did not invest a cent — it was a private rescue organized by the central bank.

LTCM was gradually liquidated over the course of 2000. The banks essentially recovered their investment.

What Went Wrong: Lessons from the Models

The fundamental problem of LTCM revealed three limits of quantitative models:

1. The normal distribution does not describe real markets: LTCM's models assumed that extreme events were extremely rare. In reality, markets have 'fat tails' — events that models define as 'impossible' happen much more often

2. Correlation increases during crises: models assumed that different markets were relatively independent. During crises, all correlations converge toward 1 — all markets move together

3. Liquidity can vanish: models assumed it was always possible to exit positions. In the 1998 panic, LTCM could not find buyers for its assets

The Ultimate Lesson

The story of LTCM is the most powerful demonstration that intelligence and mathematical models do not eliminate risk — sometimes they hide it. Enormous leverage turned a model error into a systemic near-catastrophe.