Two Nobel Laureates, Five Years of Data, and a $4.6 Billion Hole

John Meriwether built Long-Term Capital Management in 1994 with a roster that read like a finance dream team: bond traders from Salomon Brothers, plus board members Myron Scholes and Robert Merton, who would share the 1997 Nobel in economics for the Black-Scholes options model. The first three years returned 21%, 43%, and 41% after fees.

The strategy was to find pairs of bonds whose prices had drifted apart, bet they would converge, and lever the bet enough to make the small spreads matter. By the end of 1997 LTCM was running roughly $30 of debt for every $1 of capital. As long as the pairs converged, the math compounded into very large numbers.

In August 1998 Russia devalued the ruble and stopped paying its debts. Investors everywhere fled to the safest assets they could find. The convergence trades LTCM owned didn't converge; they diverged in unison, exactly the scenario the models had treated as a multi-sigma improbability. The fund lost 44% of its value that month and around $4.6 billion across the four months of the crisis.

The historian Niall Ferguson noted afterward that LTCM had calibrated on five years of market data. Eighty years would have included the post-1917 Soviet repudiation of Tsarist debt — the previous time a major power had walked away from its bonds and dragged global spreads with it. The Federal Reserve Bank of New York convened 14 banks to put up $3.6 billion in September 1998, less to save LTCM than to unwind its book without breaking the rest of Wall Street.