Long-Term Capital Management: Two Nobel Laureates, $4.6 Billion, Four Months

Let us establish one thing first: these were not stupid people.

The partner list at Long-Term Capital Management included Myron Scholes and Robert Merton — 1997 Nobel laureates in economics, architects of modern option pricing theory. Alongside them sat John Meriwether, formerly Salomon Brothers’ star bond trading head, and David Mullins, a former Vice Chairman of the Federal Reserve.

It may have been the highest concentration of financial intellect ever assembled in one office. It lost $4.6 billion in four months.

What they actually did

LTCM’s core strategy was convergence arbitrage.

The logic is clean. Markets constantly contain securities that should trade at the same price but briefly diverge for technical reasons. A newly issued 30-year Treasury (liquid, actively traded) versus a 29.5-year Treasury issued six months earlier (nearly identical cash flows, but less liquid). Their yields differ slightly — perhaps a tenth of a percentage point.

Over time that gap almost always closes. LTCM bought the cheap one, shorted the expensive one, and waited for convergence.

The judgement itself was correct. These spreads had converged again and again throughout the historical record.

The problem was the size of the prize

A tenth of a percentage point does not make anyone rich.

To turn tiny, high-probability spreads into attractive returns, there is exactly one option: magnify them.

By early 1998, LTCM held roughly $4.7 billion in equity against about $125 billion in assets — leverage of roughly 25 to 1. Counting the notional value of off-balance-sheet derivatives, the positions involved exceeded $1 trillion.

Twenty-five to one means: a 4% loss on the asset side wipes out the equity entirely.

In their models, this was not a concern. These spreads had historically moved so little that a 4% adverse move was statistically almost impossible.

17 August 1998

Russia devalued the rouble and defaulted on its domestic debt.

The direct exposure was not itself fatal. What was fatal was the reaction: investors worldwide simultaneously dumped risk and rushed into whatever was safest and easiest to sell.

The spread that “almost always converges” began to widen instead.

LTCM held the cheap, illiquid side and was short the expensive, liquid side. In a flight to liquidity, cheap gets cheaper and expensive gets more expensive. Every single position was on the wrong side of the money flow.

Three things the models did not price

One: correlations go to 1 in a crisis

LTCM held hundreds of positions described as uncorrelated, spread across countries and markets. The models therefore treated total risk as far smaller than the sum of its parts.

But those positions were uncorrelated only in calm conditions. They shared one hidden factor: the liquidity premium. When the liquidity crisis arrived, that single factor hit all of them at once. The diversification evaporated exactly when it was needed.

Two: you cannot exit a market with no buyers

As losses mounted and positions had to be cut, LTCM discovered it owned precisely the assets nobody wanted. Worse, the market knew LTCM was in trouble and had to sell — which moved prices further against them before they could act.

Three: when you are large enough to be the market, there is no exit

In several niche markets, LTCM’s positions were most of the market. That means there is no such thing as “selling at the market price” — the moment they started selling, the price collapsed.

Mark-to-market valuation rests on the assumption that you could transact at that price. For a sufficiently large position, that assumption was never true.

How it ended

By late September, LTCM’s capital had fallen from $4.7 billion to roughly $400 million — a 92% loss.

By then the issue was larger than one fund. LTCM had counterparty relationships with nearly every major institution on Wall Street; a disorderly liquidation risked cascading defaults.

On 23 September 1998, the Federal Reserve Bank of New York gathered the heads of 14 financial institutions in a room and brokered a $3.625 billion recapitalisation to take over and wind down the positions in an orderly fashion. No public money was used, but a central bank convening a private rescue was itself an extraordinary step.

One final irony: most of LTCM’s positions did eventually converge. The direction was right. They simply did not survive long enough to collect.

What transfers

This is not a story about avoiding leverage. It is about three more fundamental things:

• Right direction plus an unsurvivable position equals wrong. Markets can stay irrational longer than you can stay solvent. Being correct and forced out along the way produces the same result as being wrong.
• Absence from the historical record is not absence of possibility. Any risk measure built on historical volatility cannot measure events outside the sample — and those are the events that decide survival.
• Diversification assumes correlations are stable. If every position rests on one hidden factor, you do not own N independent positions. You own one position N times.

A closing note: in this case the judgement was not what failed. The sizing was. That is precisely why a site about reading charts has to talk about position sizing and risk too — the value of a read depends entirely on whether you are still around when it pays off.

This is a historical review written for risk education. It is not investment advice and makes no inference about any current market conditions.