The Rise And Fall Of Long Term Capital Management

Why LTCM Still Comes Up In Every Risk Meeting

Most people think they know what happened at Long Term Capital Management in 1998. They don't really. They know the headline version: smart guys, big leverage, Russian default, bailout. The actual mechanics are where the useful part lives, and they're worth understanding if you're dealing with leveraged strategies today because the same pressure points keep reappearing in different wrappers. LTCM was founded in 1994 by John Meriwether, who had previously run Salomon Brothers' arbitrage desk. The team included Myron Scholes and Robert Merton, both Nobel laureates for their work on options pricing. That combination of trading pedigree and academic credibility is what made them dangerous. They weren't gamblers. They thought they were doing mathematics. The strategy rested on convergence trades. Price dislocations between related securities would eventually correct, and LTCM would capture the spread. They traded Italian and German bonds, Japanese government bonds, mortgage-backed securities, currency pairs, and equity spreads. The arbitrage was real in theory. The execution was where everything broke.

Here is the part most people miss. The models worked fine under normal conditions. What killed them was not a model error but a liquidity error. Their positions were massive relative to market depth. When the spread moved against them even slightly, they needed to roll or hedge, and there was no one on the other side willing to take the other side at a reasonable price. The model assumed they could exit whenever they wanted. That assumption was wrong. I ran into this directly when I was working with a small convertible arbitrage book back around 2015. We had a position in a mid-cap name where the conversion spread had compressed to about twelve basis points. The model said hold. The position size was roughly four million dollars notional, which felt small until you realize the daily volume on that particular bond was maybe eight million. One bad week and we were looking at a twenty percent move just from our own hedging activity. I cut the position at eighteen hundred dollars per unit instead of waiting for the model to confirm exit. The model would have held for another three days and likely closed at fourteen hundred. Sometimes you have to trust the spreadsheet and sometimes you have to trust that the spread isn't going to compress further because liquidity is drying up in real time. The difference matters more than people admit. LTCM's leverage ratio hit roughly twenty-five to one on reported capital, but the real number was much higher when you factored in off-balance-sheet derivatives. Their repo agreements and swap contracts created contingent obligations that only became visible under stress. The Federal Reserve's intervention wasn't about saving the firms that invested with LTCM. Those firms had due diligence. The Fed was trying to prevent a cascade through the repurchase agreement market, where overnight lending to securities dealers was at risk of freezing completely if counterparties lost confidence in the pricing of pledged collateral.

The Mechanics That Actually Mattered

Let me break down what the trades looked like in practice rather than in a textbook. Take the Italian-German bond spread. Italy borrows at higher rates than Germany. The spread between equivalent-maturity bonds fluctuates based on economic news, central bank policy divergence, and market sentiment. LTCM would go long the cheaper bond and short the expensive one, betting the spread would return to its historical mean. This is basis trading, and it is the most basic form of relative value. The problem is that the mean is not a fixed point. It moves. During the 1997 Asian crisis, the spread between Italian and German bunds widened from around four hundred basis points to over seven hundred. LTCM was short the wrong side of that move because they assumed European political forces would keep the spread contained. The mortgage-backed security trades followed a similar pattern but with added complexity. They went long MBS tranches that appeared mispriced relative to Treasuries and short the Treasuries. The valuation relied heavily on prepayment models. When interest rates moved in unexpected directions, prepayment speeds changed, and the duration of the MBS position shifted in ways the model did not predict. This is a well-known issue in the fixed income world called convexity risk. LTCM had enormous convexity exposure hidden inside what they described as low-risk arbitrage. Another detail that gets overlooked: LTCM's initial margin requirements on their derivative positions were not fixed. As volatility increased, margin calls escalated rapidly. Their funding came largely from repo markets, and repo lenders can demand additional collateral at any time. When the Russian default hit in August 1998, credit markets seized. Repo lenders treated all collateral as suspicious and started calling margins across the board. LTCM did not have enough liquid collateral to meet the calls. They had positions, but positions are not cash.

The bailout structure was arranged by the Federal Reserve Bank of New York in about seventy-two hours. A group of fourteen banks provided a two-point-seven-five billion dollar capital injection in exchange for a stake in LTCM and the right to wind down the portfolio over time. The banks did this not out of altruism but because they were major counterparties on LTCM's derivative contracts. If LTCM had been allowed to liquidate in a fire sale, the losses would have been amplified by distressed pricing. A controlled wind-down was the least bad option for everyone involved.

What The Collapse Teaches You Practically

If you are evaluating relative value strategies today, the LTCM case teaches several things that do not get enough attention. First, model risk is not the same as model error. The models were technically sound. The risk was in the assumptions about market behavior under stress. Volatility clustering, liquidity evaporation, and correlation breakdown are not constants. They are variables that move together in ways that are difficult to model because historical data does not contain enough extreme events to calibrate properly. LTCM ran on about four years of post-Bretton Woods data. That window was unusually calm. Second, leverage amplifies everything including your awareness of problems. At moderate leverage levels, margin calls are annoying. At twenty-five to thirty times leverage, a one percent adverse move in your portfolio wipes out roughly four percent of your equity. That is not a typo. The math is straightforward but the psychology is brutal. Traders start making emotional decisions about which positions to cut and which to hold, and by then the damage is already done.

Third, and this is the part I wish more people understood, the most dangerous positions are the ones that look risk-free. Spread trades with high Sharpe ratios and low drawdowns in backtests are exactly the positions that concentrate tail risk. They profit from small, frequent gains and lose everything in one event. This is why the profitability of LTCM's strategies looked so smooth for so long. The tail risk was being paid for in premiums that accumulated steadily until a single shock reversed years of gains. When I evaluate relative value books now, I run a simple stress test that most people skip. I assume a liquidity drought lasting ten business days where you can only trade at fifty percent of normal volume and at a fifty-basis-point wider spread. Then I calculate what the mark-to-market looks like under that scenario. The number is almost always worse than the risk managers expect. I adjust position sizes accordingly and flag the adjustment to the portfolio manager. It usually causes friction. Friction is good. It means you are thinking about the right thing. The long-term lesson is that arbitrage is not free money. It is compensation for bearing risks that other market participants are unwilling or unable to take. The question is whether you are being compensated for risks you understand and can measure, or risks you have simply not encountered yet. LTCM learned that distinction the hard way, and the industry has spent twenty-five years trying to internalize that lesson without fully succeeding.