Every crisis in this Part, from tulips to Credit Suisse, had the same ingredients: a trigger, leverage or borrowed funding, a liquidity mismatch, a regulatory gap and someone who was surprised. The mismatch decides how fast a fall becomes a fire sale. When forced selling approaches a day’s market volume, a central bank or a group of banks has had to step in.
Why it matters: Leverage and a liquidity mismatch together can turn a small trigger into a big event.
Summary: Every crisis in this Part, from tulips to Credit Suisse, had a trigger, leverage or borrowed funding, a liquidity mismatch, a regulatory gap and someone who was surprised. The mismatch decides how fast a fall becomes a fire sale: when forced selling approaches a day’s market volume, a central bank or bank group has had to step in.
- Bubbles need a detached narrative; runs such as UK LDI and Credit Suisse do not.
- Measure the mismatch as forced selling ÷ normal daily volume: 38% of a day in 1987, 4.2 days for gilts in 2022.
- Solvency tests are not enough; run a one-week cash test.
- Rescues can rewrite the order of losses, as AT1 holders learned.
- Treat leverage above about 10× or a one-week cash ratio above 1 as fragile (a heuristic for funds and leveraged positions, not regulated banks).
Section 10.7: The Repeating Pattern — What Every Bubble and Crisis Shares‘s stages describe bubbles. This table uses five features that fit every case in this Part: the trigger that started the fall, the leverage that turned price losses into forced selling, the liquidity mismatch (claims that could be pulled faster than assets could be sold), the regulatory gap the case exposed, and who was surprised. The 2008 row summarizes Volume I’s Part 3.
| Case | Trigger | Leverage | Liquidity mismatch | Regulatory gap | Who was surprised |
|---|---|---|---|---|---|
| Tulips, 1637 (10.1) | Buyers stopped appearing, February 1637 | Contracts for future delivery with little cash down | Promises to pay with no cash behind them | Informal contracts; Holland suspended them in April 1637 | First sellers in each chain |
| South Sea, 1720 (10.2) | Price slide from the summer peak | Company loans on its own shares; installment subscriptions | Loans secured on falling shares | Insider promotion and bribery; Bubble Act aimed at rivals | Late buyers, Isaac Newton among them |
| 1929 (10.8) | October selling after the September peak | Margin of about 10% down | Call loans withdrawn: $2.4 billion in October | No federal margin rules or securities regulator | Margin buyers and non-bank call lenders |
| 1987 (10.8) | Falls in the days before October 19 | Index futures | Hedges that needed buyers at every price | No cross-market halts or coordination | Portfolio insurers relying on continuous trading |
| Harshad Mehta, 1992 (10.8) | Exposure of the scam, April 1992 | Bank money diverted to brokers’ stock buying | Repo collateral that did not exist | Paper bank receipts; weak bank controls | Banks holding receipts |
| Asia, 1997 (10.3) | Baht floated, July 2, 1997 | Dollar debt on local revenue | Short-term foreign borrowing against long local assets | Supervision “not up to” global capital flows (IMF); pegs treated as permanent | Borrowers who never hedged |
| LTCM, 1998 (10.4) | Russia’s default and devaluation, August 1998 | Over 25 to 1 on balance sheet; $1.4 trillion notional | Collateral calls on crowded trades | Leverage limited only by lenders; little disclosure of risk | Lenders relying on the partners’ reputation |
| Japan, 1990 (10.5) | Rate rises ending the land and stock boom | Property-backed corporate and bank lending | Banks rolling over bad loans | Slow recognition of bank losses | Firms and banks that assumed land never fell |
| Dot-com, 2000 (10.6) | Peak in March 2000; lockup expiries | Firms funded mainly by selling shares; investors’ margin debt | Loss-making firms needing new funding | Conflicted analyst research | Buyers paying for growth that never came |
| 2008 (Volume I) | Mortgage losses; Lehman’s failure | Securities firms and off-balance-sheet vehicles | Mortgage securities funded by repo and commercial paper | Shadow banking outside bank rules | Investors relying on ratings |
| IL&FS, 2018 (10.9) | Defaults, July to September 2018 | Group debt above ₹91,000 crore | Long projects on short borrowing | No liquidity ratio for non-bank lenders | Mutual funds holding its paper |
| Yes Bank, 2020 (10.9) | Moratorium, March 5, 2020 | Bad loans against thin capital | Deposit withdrawals | AT1 sold to individuals | Retail AT1 buyers |
| Archegos, 2021 (10.9) | Falls in its concentrated stocks | $160 billion on $36 billion capital, via swaps | Margin calls on unsellable blocks | Swap positions invisible in aggregate | Prime brokers that each saw only their own slice |
| UK LDI, 2022 (10.9) | Growth Plan, September 23, 2022 | Leveraged gilt hedges | Collateral due within days; schemes’ cash slower to raise | Stress tests sized at about 100 basis points | Pension schemes that believed they were hedged |
| Credit Suisse, 2023 (10.9) | Outflows and the March 2023 US bank failures | Bank balance sheet, with AT1 counted as capital | Client money that left faster than assets could be sold | Backstop and write-down rested on an emergency ordinance | AT1 holders who ranked themselves above equity |
Read across the columns. The dot-com row has the least borrowing inside the firms themselves, and its losses fell mainly on shareholders. Every row has a liquidity mismatch, including the five that were not bubbles. The trigger column is the least predictable: a missed auction, a budget, a default abroad.
The mismatch decides speed. Measure it as forced selling divided by what the market normally absorbs.
| Case | Forced selling or withdrawal | Market capacity | Ratio |
|---|---|---|---|
| 1987 | $8 billion of model-required selling still unexecuted entering October 19 ($12 billion called for less $4 billion sold) | About $21 billion of NYSE sales on October 19 | 8 ÷ 21 = 38% of that day’s sales |
| UK LDI, 2022 | At least £50 billion of gilt sales expected | About £12 billion of normal daily trading | 50 ÷ 12 = 4.2 days of all trading |
Near or above one day’s volume, the forced seller sets the price, and a central bank or group of banks steps in: New York banks in 1929, the Federal Reserve in 1987, the Bank of England in 2022.
The table is a checklist, not a forecast.
| Situation | What changes | Why |
|---|---|---|
| A bubble in equity-funded firms | Expect a large price fall with smaller credit losses | With little debt inside the firms, losses stay mostly with shareholders (dot-com, 2000) |
| Leverage without a bubble | Solvency tests pass; run a cash test instead | LDI liabilities fell with gilt prices, yet funds still ran out of cash |
| Fraud at the core | Reported numbers cannot be used; verify collateral and exposure yourself | Fake receipts (1992) and misstatements to banks (Archegos) hid the leverage |
| A rescue that rewrites the order of losses | Legal terms decide outcomes, not the ranking chart | AT1 holders lost everything at Yes Bank and Credit Suisse, then challenged it in court |
| Your own government is the trigger | Stress tests must include domestic policy moves | The UK Growth Plan and Thailand’s float both came from home |
| A fast crash versus a slow one | Judge recovery by private balance sheets, not by the size of the fall | The Dow regained its 1987 level within two years; the Nikkei took 34 years |
Score any institution or position on the five columns. Trigger: name two events, including your own government’s actions, that would force a sale. Leverage: compute the fall that wipes out capital, 1 ÷ leverage. Mismatch: divide the cash that could be demanded within a week by the cash and sellable assets on hand. Gap: ask which rule covers this structure. Surprise: ask who believes they are protected, and why. Treat leverage above about 10× or a one-week ratio above 1 as fragile whatever the valuation; these are heuristics for funds and leveraged positions (judge regulated banks, which run higher leverage, on their capital and liquidity ratios), tighter for illiquid or concentrated assets.
Testing solvency and skipping liquidity. Most UK pension schemes in 2022 were solvent and Credit Suisse met its capital requirements, yet one market needed £19.3 billion of central bank buying and the other bank a weekend takeover. Expected LDI forced sales equaled 50 ÷ 12 = 4.2 days of all gilt trading. Run a cash test: the largest collateral or withdrawal call in a stressed week against the cash you can raise without fire sales.
What do all financial crises have in common?
In every case in this Part, from tulips in 1637 to Credit Suisse in 2023, someone held assets that could not be sold as fast as the money behind them could leave, and almost always borrowed money turned price falls into forced selling. Bubbles add a story that detaches prices from value; runs do not need one.
What is a liquidity mismatch?
The gap between how fast funding can leave and how fast assets can become cash. Banks lending deposits, LDI funds owing daily collateral on long gilts and IL&FS funding 20-year projects with shorter borrowing all had one. It is harmless until funders stop rolling over at the same time.
Can regulation prevent the next crisis?
It closes the gap each crisis exposed: margin rules in 1934, circuit breakers after 1987, liquidity ratios for Indian non-bank lenders in 2019, LDI buffers in 2023. The next failure usually appears where no rule has yet been written, which is why the five-column check matters more than any single rule.
India supplies three of this Part’s cases, one for each of the columns that regulators can most easily fix. Volume I’s Part 10 covers the 1991 balance-of-payments crisis, the fourth great Indian episode.
| Column | Indian case | Rule now in force |
|---|---|---|
| Collateral and settlement | Harshad Mehta, 1992: bank receipts with no bonds behind them; ₹3,018.63 crore of hard-to-recover bank exposures | Special court and attachment of property (1992); securities held in dematerialized form, as Volume I’s Part 1 and Part 11 describe |
| Liquidity mismatch | IL&FS, 2018: group debt above ₹91,000 crore funding long projects | RBI liquidity coverage ratio for large non-bank lenders, 100% from December 1, 2024; scale-based regulation from October 1, 2022; SEBI segregated portfolios for mutual funds (December 2018) |
| Who was surprised | Yes Bank, 2020: ₹8,415 crore of AT1 written off, ₹679 crore of it bought by 1,346 individuals | New AT1 issues only for qualified institutional buyers, minimum ₹1 crore (SEBI, October 2020) |
| Trigger and market halts | Any sharp fall | Market-wide circuit breakers at 10%, 15% and 20% moves in the Sensex or Nifty, whichever comes first (since July 2, 2001), against the US 7%, 13% and 20% of the S&P 500 |
Two Indian features stand out. Courts, not only regulators, decide who bears losses: the Special Court Act was still producing Supreme Court rulings in 2024, and Yes Bank’s AT1 write-down awaited the Supreme Court’s judgment after hearings closed on February 26, 2026. And India’s first halt needs a 10% index move, against 7% in the US, so Indian markets keep trading through falls that would pause New York. For an Indian reader the practical rule is the same as Section 10.9‘s: before buying any instrument sold as “like a fixed deposit,” find the clause that can write it to zero.
Across all fifteen cases the repeating features are a trigger, leverage or borrowed funding, a liquidity mismatch, a regulatory gap and someone who was surprised. Measuring the mismatch as forced selling against normal daily volume shows why some crises needed a central bank within days, and why a cash test matters as much as a solvency test.
Six questions on this chapter. Decide on your answer first, then click “Reveal Answer.”
1. Forced sales of £50 billion are expected in a market that normally trades £12 billion a day. How many days of normal volume is that?
- About 6.0
- About 0.24
- About 2.1
- About 4.2
Reveal Answer
Answer: D. 50 ÷ 12 = 4.2 days of all trading.
2. What does the section’s mistake box say is skipped too often when testing a leveraged institution?
- A review of the board’s composition
- A cash test for a stressed week
- A test of reported return on equity
- A comparison with peer valuations
Reveal Answer
Answer: B. Solvent institutions such as UK pension schemes and Credit Suisse still failed for lack of cash.
3. Which case in the table had the least borrowing inside the companies themselves?
- The dot-com crash
- The 1929 crash
- LTCM’s 1998 collapse
- Archegos in 2021
Reveal Answer
Answer: A. Internet firms were funded mainly by selling shares, so losses fell mainly on shareholders.
4. Under the section’s heuristic, which position should be treated as fragile?
- Leverage of 3× with a one-week cash ratio of 0.50
- Leverage of 6× with a one-week cash ratio of 0.90
- Leverage of 12× with a one-week cash ratio of 0.8
- Leverage of 1× with a one-week cash ratio of 0.20
Reveal Answer
Answer: C. The heuristic flags leverage above about 10× or a one-week cash ratio above 1; 12× breaches the first.
5. Worked problem: Forced selling is $30bn and normal daily volume is $15bn. What is the forced-selling ratio?
Reveal Answer
Answer: Ratio = $30bn ÷ $15bn = 2.0 days of normal volume.
6. Worked problem: Another episode forces $5.7bn of selling against $15bn of daily volume. How does it compare?
Reveal Answer
Answer: Ratio = 0.38 of a day (38%), far gentler than two days.