LTCM, 1998

Two Nobel laureates, a decade of theory, thirty times leverage — and four months in which the trades that could not lose, lost.

What happened

  • 1994 — John Meriwether founds LTCM with a roster including Myron Scholes and Robert Merton, who win the Nobel prize in 1997.
  • 1994–97 — spectacular returns from convergence trades: buy the cheap-but-nearly-identical security, short the expensive one, wait for the gap to close.
  • August 1998 — Russia defaults on domestic debt and devalues. Investors sell everything hard to sell and buy the most liquid asset on earth: on-the-run US Treasuries.
  • August–September 1998 — every LTCM position moves the wrong way at once. Losses reach roughly $4.6bn, wiping out most of the fund's capital.
  • 23 September 1998 — the New York Fed convenes fourteen banks, which recapitalise the fund with around $3.6bn of their own money. The Fed organised the room; it did not put in public money.

The mechanism

The trade's premise was that the spread converges. Under forced liquidation it did the opposite first — and leverage meant the fund could not wait for the model to be right.
Russia defaultsActual spreadWhere the model said it goesTime (1998)Spread that 'must' converge
  • The edge was real but tiny — a few basis points per trade. Making it a business required leverage of roughly 25–30 times on the balance sheet, plus a derivatives book with far larger notional.
  • Leverage converts "wait" into "cannot wait": with that gearing, a small adverse move triggers margin calls, and meeting them means selling exactly the positions everyone else is selling.
  • Diversification failed by construction: dozens of trades across countries and asset classes looked independent, but every one of them was long illiquidity and short liquidity. When investors fled to safety, they were one position wearing many names.
  • Being famous made it worse: rivals knew roughly what LTCM held and traded ahead of the forced unwind.

What it teaches

  • Solvency and liquidity are different problems. LTCM's trades largely converged eventually — the fund simply no longer existed to collect (see margin & collateral).
  • Correlation is a peacetime measurement. Historical correlations described the world that existed before the stress, not during it (see diversification).
  • Leverage is a time limit. It does not change whether you are right; it changes how long you are allowed to be wrong.
  • Models describe; they do not oblige. The mathematics was excellent. The assumption that markets would keep behaving as the sample period suggested was the position.