Quant Analytics: from scratch to pro · Professional practice and failure case studies · Pro
Case study: Long-Term Capital Management, 1998
A hedge fund with trades that were each modest in size lost most of its capital in 1998 and had to be recapitalised by its creditors. The case shows how leverage, thin liquidity and similar positions held by others turn a small adverse move into a failure.
In this lesson
- Describe how leverage turns a small adverse move in asset values into a large loss of capital
- Explain why convergence trades can lose money before they converge
- Separate liquidity, crowding and model risk as distinct lessons from the case
- Compute the capital loss for a given leverage and price move, and the gain needed to recover
Read first: Leverage and margin, Tail risk: when the bell curve fails, Model risk and live monitoring
Long-Term Capital Management (LTCM) was a hedge fund run by well-known traders and academics. In 1998 it lost most of its capital, and in September its creditors put in new money to avoid a disorderly sale of its positions. The case is taught because the losses came from the interaction of leverage, liquidity and crowding, not from one bad idea.
What happened
LTCM held large, leveraged positions in bond and other markets, many of them convergence trades. A convergence trade buys a cheaper asset, sells a similar expensive one, and earns the gap if the two prices meet. In the summer of 1998 Russia devalued its currency and defaulted on part of its debt. Investors moved to the safest and most liquid assets, and gaps between similar securities widened where LTCM expected them to narrow.
GAO reports that LTCM lost more than 90% of its capital in 1998 (its summary page; the full report speaks of almost 90% by September). On 23 September 1998 fourteen banks and brokerage firms agreed to put $3.625 billion into the fund for 90% of it. The Federal Reserve Bank of New York convened the meetings; the money came from the private firms, not from the government.
The mechanism
Leverage. At the end of 1997 the fund held roughly $30 of debt for every $1 of capital. Leverage multiplies every move in asset values into a move in capital.
Convergence can take longer than funding. Prices that should meet may first move apart. A leveraged holder faces falling capital, rising margin demands and lenders who want collateral. The position may be right in the end and still be closed first.
Liquidity. In calm markets a large position can be sized and exited at small cost. In stress, buyers disappear and exits move prices. The fund's own selling would have widened the gaps further.
Crowding. Other market participants held similar positions, so many holders wanted to sell the same assets at once. The loss then depends on what everyone else does, not only on the position.
Model risk. Jorion (2000) concludes that LTCM underestimated its risk by relying on a short history and by concentrating exposures. Correlations tend to rise in stress, so diversification measured in calm periods is smaller than it looks. See model risk.
Worked example: what the numbers say
Take $1 of capital with $30 of debt, so $31 of assets. A 1% fall in asset values is a loss of 31% of the capital. A fall of 1.5% costs 46.5%. A fall of 1 / 31, which is 3.2%, removes all of it. The chart shows the same arithmetic for three leverage levels.
A loss of 90% of capital needs a gain of 1 / (1 - 0.9) - 1, which is 900%, to recover. Losses at this size are practically permanent.
One inference from the recapitalisation: if $3.625 billion bought 90% of the fund, the fund's capital after the injection was 3.625 / 0.9, which is $4.03 billion, and about $0.40 billion was left before it. This is arithmetic from the two reported figures, not a reported figure itself. It is consistent with the loss of more than 90%.
Lessons for a quant
- Leverage: size positions so that a plausible, not just an expected, adverse move leaves the firm able to hold them. Capital lost grows with leverage times the move. See leverage and margin.
- Liquidity: a price in a model is not a price at which a large position can be sold. Test exits in stress, with the position's own size.
- Crowding: ask who else holds the trade and what happens if they sell. Backtests do not show a crowd. See crowded trade.
- Correlation going to one: measure risk with stress periods and assume diversification shrinks when it is needed most. See correlation and its limits.
- Funding and counterparties: lenders can withdraw credit or raise margin at the worst time. Funding terms are part of the risk.
- Model risk: a short history understates tail risk. Use stress tests, challenge from outside the team and a monitoring band. See tail risk.
What changed afterwards
The President's Working Group on Financial Markets published a report in April 1999 on hedge funds, leverage and the lessons of LTCM. Its focus was on better counterparty risk management, more transparency and ways to constrain leverage across the financial system. GAO found that weak risk management at creditors and counterparties let LTCM's size and leverage grow unchecked, and that regulators lacked a full view of the risk.
Common mistakes
- Reading the case as proof that models fail. The failure was in how risk was sized, funded and tested.
- Assuming a correct long-run trade is a safe one.
- Treating a historic volatility from calm years as the future's.
- Counting diversification across positions that depend on the same stress.
- Citing exact balance-sheet figures without a source. Reports differ by date and by what they count.
In practice
A stress table for any levered or concentrated book lists the capital lost for a given move, the gain to recover, and the size of forced selling. Research on FinUties is where studies built on methods like these are published, with their method and limits stated.
Share of capital lost as asset values fall, by leverage
- no leverage
- $10 of debt per $1 of capital
- $30 of debt per $1 of capital
Worked numbers
Each result is recomputed from the formula shown.
| What | Formula | Result |
|---|---|---|
| assets per $1 of capital at $30 of debt | 30 + 1 | $31 |
| capital lost when assets fall 1% | assets_per_capital * 0.01 | 31% |
| a second, larger fall in asset values | 0.015 | 1.5% |
| capital lost when assets fall 1.5% | assets_per_capital * move15 | 46.5% |
| fall in assets that removes all capital | 1 / assets_per_capital | 3.2% |
| gain needed to recover from a 90% loss | 1 / (1 - 0.9) - 1 | 900% |
| implied capital after the injection if $3.625 billion bought 90% | 3.625 / 0.9 | 4.03 |
| implied capital left before the injection | after_total - 3.625 | 0.40 |
Check yourself
At about $30 of debt per $1 of capital, how much capital is lost when asset values fall 1%?
Assets are about $31 per $1 of capital. A 1% fall is about $0.31, which is 31% of the capital.
At that leverage, what fall in asset values removes all the capital?
1 / 31 = 3.2%. A move that is ordinary for many markets ends the fund.
After losing 90% of capital, what gain is needed to get back to the starting level?
1 / (1 - 0.9) - 1 = 900%.
Why is a convergence trade not safe because the prices should meet eventually?
Prices can move further apart first. A leveraged holder faces margin calls and forced sales on the way, so it may not survive until convergence.
Name two reasons a model built on a short history can mislead.
The sample may miss rare stress, so volatility and correlations look low. And relationships that held in calm periods can break together when many holders sell at once.
Key terms
Checked facts
Every figure below carries how it was checked: Primary checked (read in the cited primary, official, academic or data source), Two sources agree, or Approximate (an order of magnitude only).
- Primary checked
In 1998 Long-Term Capital Management lost more than 90% of its capital.
- Primary checked
At the end of 1997 LTCM held roughly $30 of debt for every $1 of capital.
- Primary checked
On 23 September 1998 fourteen banks and brokerage firms agreed to put $3.625 billion into LTCM in exchange for 90% of the fund.
- Primary checked
The Federal Reserve Bank of New York convened the meetings, and the money came from the private firms, not from the government.
- Primary checked
Jorion (2000) concludes that LTCM underestimated its risk through reliance on a short history and through concentrated exposures.
- Primary checked
GAO found that weak risk management at creditors and counterparties let LTCM's size and leverage grow unchecked.
Sources
- Hedge Funds, Leverage, and the Lessons of Long-Term Capital Management. Report of the President's Working Group on Financial Markets, April 1999, 1999. Accessed 9 Nov 2026
- Long-Term Capital Management: Regulators Need to Focus Greater Attention on Systemic Risk (GAO/GGD-00-3). U.S. General Accounting Office, October 1999, 1999. Accessed 9 Nov 2026
- Responses to Questions Concerning Long-Term Capital Management and Related Events (GAO/GGD-00-67R). U.S. General Accounting Office, 2000, 2000. Accessed 9 Nov 2026
- Near Failure of Long-Term Capital Management. Federal Reserve History, Board of Governors of the Federal Reserve System, 2013. Accessed 9 Nov 2026
- Statement on Long-Term Capital Management (William J. McDonough, 1 October 1998). Federal Reserve Bank of New York, 1998. Accessed 9 Nov 2026
- Risk Management Lessons from Long-Term Capital Management. European Financial Management 6(3), 277-300 (Philippe Jorion), 2000. Accessed 9 Nov 2026