LTCM 1998

Definition

Long-Term Capital Management — a fund staffed with the era’s most credentialed quantitative talent, including Nobel laureates — lost ~90% of its capital in under two months in 1998 and required a Fed-coordinated private recapitalization. It is the KB’s canonical case study in CONVERGENCE-TRADE correlation: dozens of “independent” relative-value positions worldwide were all, in the stress that followed Russia’s default, one position — short liquidity and short volatility at extreme leverage.

How it works / structure

• The strategy set: convergence and relative-value trades (strategy-pairs-trading at institutional scale) — on-the-run/off-the-run Treasuries, swap spreads, merger arb, equity vol — each individually small-edge, high-confidence, and sized on calm-period statistics with balance-sheet leverage estimated above 25:1 (PWG report).
• The failure mechanics: Russia’s August 1998 default triggered a global flight to liquidity; EVERY convergence spread widened simultaneously — historical cross-trade correlations near zero went effectively to one (risk-correlation-exposure’s founding exhibit); losses forced deleveraging into markets that knew LTCM’s positions, worsening the spiral; the fund’s own size made exit impossible.
• The quantitative lessons (Jorion): VaR calibrated on short calm samples understated tail risk structurally; sizing near growth-optimal on estimated edges (risk-kelly-criterion oversized-territory arithmetic — an estimation error at high leverage is fatal); liquidity risk was unmodeled entirely.
• The structural aftermath: counterparty-risk practices, hedge-fund leverage reporting debates, and the PWG report itself.

When it applies

Cited whenever stress-correlation aggregation is discussed (port-correlation-budgets — the calm-matrix flattery warning IS this episode); whenever Kelly-adjacent sizing on estimated edges is evaluated; whenever a strategy’s capacity and its exit liquidity are assumed rather than measured.

Risk profile & failure modes

• The central lesson: diversification counted in trades is fiction when the trades share a hidden driver (short liquidity premium); effective exposures, not position counts (port-diversification-math).
• Credential risk: intellectual pedigree provided no protection and arguably enabled the leverage — bias-overconfidence at its most expensive.
• Size as risk: the fund’s positions were large enough that its stress WAS the market’s stress — capacity limits are risk limits.
• Misuse of the episode: “models failed” is the lazy reading; the PWG/Jorion reading is that the models were asked calm-period questions and the sizing ignored their stated assumptions.

Evidence & limits

The PWG report (1999) is the official record; Jorion (2000) is the standard risk-management autopsy. Exact position and leverage figures are estimates from the workout. The episode predates most current market structure — its mechanics lessons (correlation convergence, liquidity spirals, sizing on estimated edges) are structural, not institutional.

Falsifiable-thesis examples

Illustrations only, not signals:

• “This book’s relative-value positions, stress-correlated at 0.8 (not their calm 0.2), keep the portfolio inside its drawdown budget (LTCM-style aggregation audit)” — falsified by the scenario computation.
• “Every position in this book can be exited within 5 sessions at under 3× calm-period spread cost at current size (liquidity audit)” — falsified by the measured capacity table.

Cross-references

• The aggregation doctrine it founded: risk-correlation-exposure, port-correlation-budgets
• The sizing arithmetic: risk-kelly-criterion (estimation error at leverage), risk-fixed-fractional
• The strategy family: strategy-pairs-trading (its institutional limit case)
• The humility source: lens-quantitative, bias-overconfidence