Skip to main content
Risk-Management Case Studies

LTCM: Leverage, Correlation, and the Bailout

Pomegra Learn

How Did LTCM's Leverage and Correlation Risks Collide in 1998?

Long-Term Capital Management's spectacular failure in 1998 was not the result of a single bad bet. It was the convergence of two independent risks that the fund's models had not adequately measured: unsustainable leverage and the assumption that correlations remain stable during market stress. When the Russian government defaulted on its debt in August 1998, both risks activated simultaneously, creating a perfect storm that nearly fractured the global financial system.

The Lede: Two Risks, One Catastrophe​

LTCM leverage risk was hidden in plain sight. The fund's fundamental equation was simple: borrow $96 for every $4 in capital, then deploy that $100 across hundreds of positions. This 24:1 leverage ratio works beautifully in calm markets where positions move as the models predict. It becomes lethal the moment something unexpected happens. Simultaneously, LTCM's mathematical models rested on the assumption that certain asset pairs—ones the fund believed were economically identical—would maintain a stable price relationship. This assumption broke catastrophically during the August 1998 crisis. The collapse of leverage and correlation simultaneously produced losses far exceeding what any single risk could have caused alone.

Quick definition: Leverage is amplified risk. Correlation is the degree to which two assets move together. LTCM combined high leverage with the assumption of stable correlations. When correlations broke, the leverage turned every small loss into a massive one.

Key Takeaways​

  • LTCM's 24:1 leverage meant that a 4-percent portfolio loss would erase 100 percent of its equity—a razor's margin for error.
  • The fund's strategy assumed "statistical arbitrage" pairs would maintain historical price relationships, but correlation breakdown invalidated this assumption entirely.
  • Prime brokers had lent LTCM billions without requiring leverage limits or stress testing, exposing 14 major banks to simultaneous losses.
  • The Federal Reserve's bailout was not charity; it was a recognition that LTCM's failure would cascade across the over-the-counter derivatives market and threaten banking system stability.
  • The crisis demonstrated that leverage multiplies correlation risk. At lower leverage, a correlation breakdown is manageable; at 24:1 leverage, it is fatal.

Understanding LTCM's Leverage Structure​

LTCM did not borrow transparently. The fund raised capital from investors ($4.7 billion) and then used that capital as collateral to borrow from prime brokers. The borrowing happened through two main mechanisms: repurchase agreements (repos) and derivatives counterparty relationships.

Repo borrowing: LTCM would sell a bond to Goldman Sachs for $100 and agree to repurchase it tomorrow for $100.01. The difference was interest. This structure allowed LTCM to finance positions at low cost. By stacking repos on top of each other, the fund could borrow far more than its capital base. A prime broker was happy to lend because they held the bond as collateral. If LTCM failed to repay, the prime broker kept the bond.

Derivatives borrowing: LTCM also entered into swap contracts and other derivatives with multiple counterparties. These positions created contingent exposures—if rates moved a certain way, LTCM owed money; if they moved another way, the counterparty owed LTCM. The fund did not report these as liabilities in the traditional sense, but they were real obligations.

By September 1998, LTCM's total leverage was not 24:1 but closer to 25:1 when derivatives were included. This meant the fund controlled $125 billion in notional positions with just $5 billion in capital remaining.

The Correlation Problem: Theory vs. Reality​

LTCM's strategy relied on identifying pairs of assets that were economically identical but trading at different prices. For example:

  • On-the-run vs. off-the-run Treasuries: The most recently issued 30-year Treasury bond (on-the-run) sometimes trades at a premium to an older 30-year bond (off-the-run) even though they have nearly identical cash flows. LTCM would buy the cheap off-the-run bond and short the expensive on-the-run bond, betting the spread would narrow.

  • Corporate bond vs. Treasury plus swap: A corporate bond can be replicated by buying a Treasury and entering a swap. If the two strategies produce different prices, LTCM would buy the cheaper and short the more expensive one.

  • Emerging-market convergence plays: LTCM also bet that spreads between Russian government bonds and other credit instruments would narrow as the economy improved.

The mathematical logic was sound: if two assets have identical cash flows and risks, they should trade at identical prices. When they do not, an arbitrage opportunity exists. A trader who buys the cheap one and shorts the expensive one locks in a profit once prices converge.

LTCM's error was assuming correlations would remain stable. Specifically, the fund assumed:

  1. The price relationship between paired assets would persist.
  2. In any given month, no more than a small percentage of pairs would fail to converge.
  3. Liquidity would always be available to exit positions at prices close to market.

In reality:

  1. During market stress, price relationships break. Investors flee risk and move to safety. On-the-run Treasuries become safe; everything else becomes risky.
  2. The distribution of outcomes shifted. Instead of 1 percent of pairs failing, 60 percent failed simultaneously.
  3. Liquidity evaporated. When LTCM tried to sell positions to raise cash, bid-ask spreads widened from basis points to hundreds of basis points. The fund was forced to accept fire-sale prices.

The August 1998 Correlation Spike​

On August 17, 1998, Russia defaulted on its ruble-denominated debt. This was not a surprise to economists—Russia's fiscal crisis had been building—but the market did not fully price the risk until it happened.

The immediate effect was a run on emerging-market assets. Investors who owned Russian bonds, Brazilian bonds, and other emerging-market debt rushed for exits. Capital that had been slowly trickling out became a flood. Emerging-market spreads (the yield premium over U.S. Treasuries) widened from 250 basis points to over 1,000 basis points in a matter of days.

For LTCM, this was the first shock. The fund owned emerging-market positions and emerging-market convergence trades. Spreads were widening, not narrowing as the models predicted.

But the contagion did not stop at emerging markets. Global investors, having taken losses on Russian debt, became more risk-averse. They sold anything that seemed risky and bought anything that seemed safe. This included:

  • Corporate bonds: Flight to safety caused spreads to widen on even investment-grade corporate debt.
  • High-yield bonds: Spreads exploded. Some high-yield issues traded only at distressed levels.
  • Mortgage-backed securities: The widening in spreads spread to U.S. mortgages, a sector LTCM had not anticipated would be affected.
  • Currencies: The Japanese yen strengthened as investors unwound yen-carry trades and moved to safety. The Brazilian real weakened as the contagion spread.

For LTCM, the critical moment came when correlations between supposedly independent positions broke down. The fund had assumed its strategy pairs would move independently of broader market risk appetite. In reality, all risk assets became correlated. Everything the fund had shorted (the "risky" side of each pair) fell in value. Everything it had bought (the "safe" side) rose. But the magnitudes were reversed from what the models predicted.

The Leverage Multiplier Effect​

Imagine LTCM held $100 million in positions financed by $4 million in capital and $96 million in borrowing. The models predicted maximum monthly loss of $2 million (2 percent). In August 1998, the fund lost $4.6 million (4.6 percent) in a single month—more than double the predicted maximum.

With 24:1 leverage, a 4-percent loss wipes out all capital:

Loss: $4 million on $100 million portfolio = 4%
Impact on capital: $4 million / $4 million capital = 100% loss

But LTCM's actual losses were worse because the fund had to mark positions to market as prices moved. When a bond that LTCM had borrowed fell 10 percent, the prime broker would demand additional collateral. LTCM would have to sell other positions to raise cash. Those sales would be at fire-sale prices, crystallizing losses and cascading damage across the portfolio.

This is leverage's dark side: it does not just amplify gains; it accelerates losses in markets under stress. It also forces the leveraged investor to become a forced seller at the worst times.

Prime Brokerage Exposure and Systemic Risk​

LTCM did not work in isolation. It was tightly embedded in the global financial system through prime brokerage relationships. Fourteen major banks had collectively lent LTCM tens of billions of dollars:

  • JPMorgan: $2.3 billion exposure.
  • Goldman Sachs: $2 billion exposure.
  • Merrill Lynch: $1.5 billion exposure.
  • Morgan Stanley: $1.5 billion exposure.
  • Citigroup, Chase, Bank of America, others: Additional exposures in the hundreds of millions each.

These were not simple loans. They were exposure to LTCM through repos, derivatives contracts, and securities lending relationships. Each bank viewed LTCM as a diversified, sophisticated counterparty managed by geniuses. No single bank expected LTCM to fail. No single bank thought carefully about what would happen if all fourteen banks needed to unwind their LTCM exposures simultaneously.

By late September 1998, that scenario was not hypothetical. LTCM was approaching insolvency. Prime brokers realized that if the fund failed:

  1. They would suffer direct losses on unsecured exposure.
  2. They would be forced to liquidate $137 billion in LTCM positions simultaneously.
  3. That forced selling would crash bond prices, credit spreads, and equity markets.
  4. The cascading losses would potentially exceed the capital of any single bank.
  5. If a prime broker failed, the contagion would spread to depositors, counterparties, and clearing systems.

This was systemic risk in real time.

The Correlation Cascade Diagram​

The Hedge That Failed​

LTCM's core assumption was that its positions were hedged. The fund claimed its strategies were "market-neutral"—they did not profit from broad market direction, only from the convergence of specific mispricings. This sounded safe.

In reality, the hedge was only effective if correlations remained stable and liquidity remained available. When both broke, LTCM discovered it had no hedge at all. Selling the short positions (which should have made money) was impossible because there were no buyers at reasonable prices. And buying additional long positions (which might have offset losses) would have required cash that the fund no longer had.

This is a critical lesson: a hedge is only as good as the assumptions underlying it. When assumptions break, the hedge vanishes.

The Bailout Decision​

On September 23, 1998, the Federal Reserve convened a meeting of the fourteen largest banks. The Fed did not order a rescue; it facilitated one. The Fed's logic was:

  1. LTCM's failure would force simultaneous liquidation of $137 billion in positions.
  2. Forced selling of that magnitude would crash fixed-income markets, particularly credit spreads.
  3. Crashing credit spreads would damage investment-grade corporations, municipal governments, and emerging-market countries.
  4. The damage would spread to equity markets and the economy as a whole.
  5. Preventing this cascade was worth the moral hazard of allowing a consortium of banks to absorb LTCM's losses.

The rescue was not a gift. The consortium invested $3.6 billion and assumed all losses. LTCM's founders and investors lost nearly everything. The fund was liquidated over the next 18 months, allowing positions to be unwound systematically rather than catastrophically.

1. Assuming leverage is safe if returns are consistent: LTCM had exceptional returns for three years. Consistent performance convinced investors and prime brokers that the risk was under control. Consistency in calm markets does not guarantee safety.

2. Believing models more than market reality: The models said LTCM's maximum loss was X. The market delivered 2X. This happens because models are based on historical data, and crises produce unprecedented events.

3. Ignoring counterparty risk from leverage: LTCM's leverage came from fourteen banks. When trouble hit, all fourteen were exposed simultaneously. Concentration of counterparty risk turns an individual crisis into a systemic one.

4. Treating convergence as certain: LTCM assumed pairs would converge. Sometimes pairs that diverged during calm times diverge further during stress. Spreads can stay wide if risk appetite remains depressed.

5. Underestimating correlation instability: Historical correlations are often 0.2 to 0.5. During crises, correlations spike to 0.8 or higher. This is not a rare event; it is a feature of how markets behave under stress.

FAQ​

What is the difference between leverage and margin? Leverage and margin are often used interchangeably, but margin usually refers to the minimum capital requirement for a specific position, while leverage refers to the overall portfolio debt-to-equity ratio. LTCM used leverage of 24:1, meaning 96 cents of debt for every dollar of equity across the entire fund.

Why didn't LTCM's VaR (Value at Risk) model catch this? LTCM likely calculated VaR, which estimates the maximum loss under normal market conditions with high confidence (typically 95 or 99 percent). The August 1998 loss was a tail event—outside the VaR threshold. VaR is useful for routine risk management but fails during crises when the distribution of returns is non-normal.

Could LTCM have survived if leverage was lower? Possibly. If LTCM had operated at 10:1 leverage instead of 24:1, a 4-percent portfolio loss would have wiped out only 40 percent of capital. The fund would have been wounded but potentially survivable. It might have had time to unwind positions without forced selling.

Why did prime brokers lend so much to a single fund? Prime brokers believed LTCM was diversified and that its positions were hedged. No single bank thought it would be wiped out by LTCM's failure alone. But when all fourteen banks tried to liquidate simultaneously, contagion spread.

Did the Fed's rescue create moral hazard? Yes, arguably. By coordinating a rescue, the Fed signaled that large, interconnected financial institutions would not be allowed to fail if their failure posed systemic risk. This may have encouraged future overleveraging. However, the alternative—allowing LTCM to fail in an uncontrolled manner—might have triggered a financial crisis comparable to 2008.

What happened to the hedge fund industry after LTCM's failure? Hedge funds initially faced stricter oversight from prime brokers, who began enforcing leverage limits and stress-testing requirements. Over time, many of these controls loosened as memories of LTCM faded. The industry grew dramatically from 2003 to 2008, only to face renewed scrutiny after the 2008 crisis.

How does LTCM compare to other major financial crises? LTCM was a near-miss crisis. The 1987 stock market crash, the 2000 dot-com bubble, and the 2008 financial crisis were all larger in magnitude. But LTCM was uniquely dangerous because it involved opaque derivative positions and tight interconnectedness with the banking system. A few hours of different decision-making might have allowed LTCM to fail uncontrollably.

Summary​

LTCM's 1998 collapse was the perfect storm of two independent risks converging simultaneously. The fund's 24:1 leverage meant even small portfolio losses became catastrophic. The fund's assumption that correlations would remain stable proved tragically wrong. When the Russian default triggered a risk-aversion shock, correlations spiked, and LTCM's pairs stopped behaving as the models predicted. Leverage turned what might have been a manageable 4-percent loss into a fund-ending event. The Federal Reserve's decision to coordinate a rescue was not based on sympathy for LTCM but on the recognition that systemic risk—the threat of cascading failure across fourteen major banks and the derivatives market—had become intolerable. The lesson is clear: leverage and correlation risk are not independent. At high leverage levels, correlation breakdown is not a survivable event.

Next​

→ Nick Leeson and the Fall of Barings Bank

Sources: