Why Genius Failed: The LTCM Postmortem

One Fund, $125 Billion, No Bids

TL;DR

1. LTCM imploded because its concentrated book of carry trades hit a funding spiral when correlations and volatility jumped, its lenders traded against it and used hostile marks, rising haircuts, margin calls, and forced sales to maximize their own profits which exhausted LTCM’s liquidity before it could get any rescue capital.

2. Sky-high leverage was a necessary but insufficient condition for its failure.
   

“We got through ‘08 because of Long-Term Capital back in 1998…They were down 90[%] before they lost control… They had the most thoughtful structure to how they financed their business and how they managed the various forms of risk in their business, to literally be down 90% and still be in control. And so when Long-Term Capital failed, I actually interviewed a number of senior people at LTCM to understand how they pulled off that miracle..They were in control until almost the bitter end. So that dictated how we created our cap structure, how we funded the business, how we managed our affairs...Big lessons [sic] for us: don’t pretend to be a bank…We are in the moving business, not the storage business.” – Ken Griffin¹

The narrative arc of LTCM as presented in Roger Lowenstein’s When Genius Failed, ‘genius quants, astronomic leverage, and a grey swan shock that made them and their investors look silly’ is directionally right but incomplete.

The full story also includes market microstructure: funding terms, collateral behavior, netting sets, crowding, adversarial marks, and how information reshapes lender and competitor incentives.

LTCM’s partners were deliberate about financing to buy time, but once other market participants smelled blood, those defenses weren’t enough: cash began to leave faster than it arrived and its runway vanished. I’m going to look closer at these dynamics and describe how hedge funds today can and do design around them.

LTCM’s Liquidity and Liability Management

On paper and in ordinary times (famous last words), the firm’s liability design made sense for convergence trades that needed time². But the execution under-modeled path dependence risks.

1. Repo & OTC Structure: By summer 1998 LTCM was financing government securities via repo agreements with 75 counterparties and had over 50 OTC derivatives counterparties, with mark-to-market exposure generally collateralized.

2. Syndicated Credit: LTCM syndicated credit facilities with several dozen banks and much of that capacity wasn’t drawn until the near-failure phase so it didn’t drive the leverage build-up, but it did not help much later with cash squeeze either.

3. Investor Lockups: After the August collapse, existing investors were limited to withdrawing 12% at the calendar year-end, slowing but not stopping capital flight pressure.

   Endogenous Risks and Exogenous Risks

“The hurricane is not more or less likely to hit because more hurricane insurance has been written. In the financial markets this is not true. The more people write financial insurance, the more likely it is that a disaster will happen, because the people who know you have sold the insurance can make it happen. So you have to monitor what other people are doing….We put very little emphasis on what other leveraged players were doing because I think we thought they would behave very similarly to ourselves…It was as if there was someone out there with our exact portfolio, only it was three times as large as ours, and they were liquidating all at once.” – Victor Haghani, former LTCM Partner

I think we over-index on exogenous shocks because they’re available, get a lot of attention, and easy to point to. I think there’s a need for more thinking of a “Funding VaR” and “Funding ES/CVaR” where dispute clocks and settlement lags are factored into capture endogenous risks³.

Exogenous shocks were the catalyst and set the direction, and endogenous loops set the speed of LTCM’s downfall. Russia widened spreads while margin rules, haircut changes, and netting mismatches determined the velocity of cash outflows.

The regime shifts also blew through correlation assumptions. LTCM stress-tested to 0.30 correlation across positions but realized correlations jumped to 0.70 as the crisis unfolded, turning nominally diversified bets into one big trade.

Endogenous factors set the pace: higher measured leverage and volatility led to higher haircuts and collateral calls, which necessitated the sales of liquid hedges first leaving illiquid tails exposed, which forced counterparties’ marks move to defensible liquidation levels which needed more collateral, more forced sales, and repeat. By September 21, 1998, Bear Stearns had tightened intraday/settlement collateral requirements and counterparties were marking toward liquidation values, directly worsening liquidity.

The endogenous risks came from participants’ incentives and market structure: dealers marked to defendable exit levels; LPs paused re-ups; trading desks and rival funds shorted instruments known to be in LTCM’s book.

Five Components of the LTCM Implosion Deserve More Attention

1. Leverage without genuine diversification: LTCM’s leverage (typically 25:1 to 30:1 before its crisis; 130:1 at the peak in the third week of September 1998) mirrored big dealers, but dealers were diversified across businesses. On the other hand, $3B of the $4.4B loss came from just two trades: interest-rate swaps and selling long-term equity options⁴.

2. Many instruments, one trade.”: Thousands of positions mapped to a small set of risk premia. When correlations spiked, “diversified” positions moved together.

3. Crowding Raised Entry Levels: After strong results in its early years (up 43%, 31% and 17% in 1995, 1996, and 1997 respectively), other traders piled into LTCM’s trades so the carry trades got crowded and reduced spreads. LTCM did not reduce the size of their trades when the expected returns/opportunity set was smaller, but instead kept putting on these trades at higher levels even as spreads compressed.

“’Everyone else started catching up to us. We’d go to put on a trade, but when we started to nibble the opportunity would vanish.” – Eric Rosenfeld, LTCM Partner

4. Salomon’s Unwind Reduced Market Depth: In July 1998, Salomon shut its US bond arbitrage desk⁵ and let a separate group liquidate faster than normal, hurting prices and stressing similar books, including LTCM’s.

5. Financing Proved Non-robust In A Hostile Regime: The combination of dozens of bilateral collateral streams, contentious daily MTM, tightened intraday credit, and marks toward liquidation values created a self-reinforcing cash-outflow loop.

The Timeline

1997

• December 1997: LTCM returned $2.7B to investors, citing a thinner opportunity set without proportionally reducing positions, effectively raising leverage.

1998

• July 6: Salomon Smith Barney closed its famous arbitrage unit, and announced it is closing certain trades, and these liquidations reduced depth in trades overlapping LTCM’s.

• August 17: Russia devalued the ruble and declared a debt moratorium; a global flight-to-safety began.

• August 21: LTCM posted a one-day $550M loss (its worst to date).

• Sept 1: The investor letter went out; LP withdrawals capped at 12% and deferred to December.

• September 2: Meriwether’s investor letter went out, disclosing the drawdown and seeking fresh capital; around this time LTCM also alerted Federal Reserve officials to its difficulties.

• Sept 18-20: Fed officials met with LTCM as the scale and scope of LTCM’s positions and interconnections become clearer.

• Sept 21: Another $500M single-day loss; Bear Stearns requires collateralization of potential settlement exposures; counterparties seek maximum collateral; marks drift toward liquidation values; the cash flow strain is acute.

• Sept 22-23: NY Fed convenes firms; Warren Buffett’s bid (with AIG/GS) briefly interrupts talks; 14 firms agree to invest $3.6B for 90% ownership; the Fed facilitates but does not lend.

Q4 1998 to Q2 1999

• Positions are wound down and by June, 1999 the fund was up 14.1% from late-September levels; the banks ultimately earn 10% on their investment; the $3.6B is fully returned by end-1999 and the fund is dissolved.

The Leaked Investor Letter and LP Exits

The firm sent an investor update that described poor performance and argued that spreads had become unusually attractive by September 2, 1998. It was meant to steady nerves and raise capital but it had the opposite effect after it was leaked. Allocators who might have added became cautious and other LPs made redemption requests, while others withdrew support for extensions and avoided fresh commitments.

Lenders Became Counterparties (Predators)

“Financial markets are complex ecosystems, with different species competing, evolving, innovating, and adapting.” – Andrew W. Lo

When LTCM’s book became public, counterparties pounced when they smelled blood in the water because the leaked investor letter had a second effect: it told the Street that LTCM needed capital and is a forced seller and the other players turned their profit maximization dials up to eleven.

Collateral terms tightened, dealers protected themselves by marking to levels they could defend if liquidations occurred, trading desks cut exposure and reduced bids on instruments LTCM wanted to exit, and prop desks and other funds shorted assets the firm was likely to sell.

This act of explaining the book to its brokers and lenders exacerbated the situation because its brokers and lenders were also its trading rivals. The Chinese Walls separating banks’ lending and trading units were a lot more porous, if they existed at all, and markets are ruthless and rapacious (redundancy here).

Cross-Netting & Cash Mismanagement

LTCM routinely split its pair trades by parking the long leg with one broker-dealer and the short leg with another dealer to conceal its trades from copycats and to optimize execution across 75 repo counterparties and 50 OTC counterparties. This tactic backfired when markets thinned because it prevented cross-netting, and the collateral-timing mismatches made its cash needs far more volatile than the model’s P&L paths implied. Counterparties also moved valuations toward liquidation when they feared forced sales.

Models Ignored Path Risk Even If They Were Correct About The Final Price

The funding-liquidity lesson that bites harder than the ex-post valuation story. The expected value case for convergence was mostly right because the post-rescue unwind earned positive returns, but the path dependence was wrong. When correlations spiked, liquidity vanished, and collateral flows dominated the mark-to-model.

LTCM remains a cautionary tale on the limits of models, the risks of concentration & illiquidity, and fairly so, but is also a reminder to be deliberate and rigorous in knowing & deciding your netting, collateral management, cash management, and communications.

1. Ken Griffin also sought out lessons from Enron’s natural gas trading unit after the company collapsed.

2. Hans Hufschmid, co-head of the LTCM London office, who sat on the risk management and management committees, founded GlobeOp Financial Services in 2000, which did outsourced HF admin/investment operations, which sold for $900M to SS&C in 2012.

3. Some have attempted to model crowding behavior by studying the patterns of herds of cows.

4. LTCM had at least 18 major trades across assets and geographies, but reportedly 40% of its profits came from Italian bonds.

5. This is the same desk where the majority of the LTCM partners worked on right before they founded LTCM, and chronicled in Michael Lewis’s Liar’s Poker.