- The six questions organise evidence about cash flows, financing and forced selling; they do not measure a crash probability.
- Historical bubbles are selected examples. Published industry research also includes price booms that did not crash.
- A market gauge needs a date, denominator and coverage definition; nominal margin debt alone is not a timing signal.
- In the hypothetical example, a 20% asset decline becomes a 40% equity loss when half the purchase is financed with debt.
A compelling new technology can be real while the price of an investment built around it is difficult to justify. That is the useful starting point for a bubble checklist: examine assumptions and financing, rather than trying to name the day a market will turn. This is general education. The six questions below are an editorial checklist, not a validated crash indicator or an instruction to buy or sell.
What makes a boom financially fragile?
In his 1992 financial instability hypothesis, Hyman Minsky distinguishes financing that can meet its obligations from cash flow, financing that needs debt to be rolled over, and financing that requires new borrowing or asset sales even to meet interest. Prolonged prosperity can encourage a move towards more fragile financing. The vulnerability is not simply enthusiasm: it is dependence on continued refinancing or favourable asset prices. Minsky, original working paper, pp. 6–8.
Minsky’s term “Ponzi finance” describes a cash-flow position in this model; it does not by itself allege a fraudulent Ponzi scheme. Nor does his framework establish that every boom follows the same timetable.
The six-point bubble checklist
Use the questions to collect evidence. Do not add ticks into a supposed probability of a crash.
| Question | Evidence to look for | What it cannot establish |
|---|---|---|
| 1. What changed? | A technology, business model or policy that changes plausible future cash flows. | A real innovation does not establish that any price is justified. |
| 2. What does the price require? | Revenue, margins, rents or other cash flows needed to support the valuation. | A high valuation alone does not date a fall. |
| 3. How is the buying financed? | Borrowing, collateral requirements, repayment dates and refinancing dependence. | A headline debt total does not measure every investor’s leverage. |
| 4. How strong is the evidence behind the enthusiasm? | Disclosures, issuance, trading activity and assumptions that can be checked independently. | Popularity or a memorable slogan does not prove mispricing. |
| 5. Who can sell, and why? | Disclosed ownership, lockups and the purpose of insider transactions. | Insider selling can have ordinary reasons; it is not automatically an exit signal. |
| 6. What could force selling? | Margin calls, redemptions, debt maturities or cash-flow shortfalls. | Visible stress may arrive after prices have already fallen. |
For each question, record the date, geography, source and uncertainty. “Unknown” is a valid answer. A story about what investors must be feeling is weaker evidence than a documented financing condition.
Historical examples: useful, but easy to overstate
Tulips, Netherlands, 1630s. Historian Anne Goldgar’s archival work challenges familiar tales of universal participation and nationwide ruin. Her account describes contracts, high prices and broken promises, while showing how later retellings relied on repeated propaganda. A memorable historical analogy needs source criticism, not just repetition. Goldgar, author’s book excerpt.
US stocks, 1929. Federal Reserve History records the Dow falling 89% from its 1929 peak to its 1932 low and not reaching the old level again until November 1954. It also describes extensive margin buying. These are nominal price-index levels, not an investor’s inflation-adjusted total return with dividends reinvested. Federal Reserve History.
Neither episode supplies a modern crash date. Selecting only famous collapses leaves out booms that did not end in the same way.
What systematic research can tell us
Greenwood, Shleifer and You studied US industry returns from 1926–2014 and international sector returns from 1985–2014. Their published paper identified 40 US episodes with two-year gains above 100% in both raw and market-adjusted returns, also requiring at least a 50% raw gain over five years. A crash was defined as a 40% drawdown within two years. Twenty-one of those episodes crashed; the rest did not. Among the eventual crashes, prices peaked about six months after identification on average. Sharp increases raised crash risk without, by themselves, predicting unusually low average future returns. Bubbles for Fama, published paper, pp. 20–21.
These are historical sample results from industries, not a 53% probability that today’s whole market will crash. The small number of episodes matters. A sample average is not a reliable countdown, and our six questions are not the model estimated in that paper.
Reading a current gauge without turning it into a forecast
One directly verified example: FINRA reported 1,453,832 million US dollars of debit balances in customers’ securities margin accounts for August 2026, versus 1,502,072 million for June. The unit is millions, so the August figure is approximately 1.454 trillion, not 1.454 billion. These are balances reported by FINRA member firms. FINRA margin statistics, checked October 4, 2026.
A nominal total needs context: market size, who borrowed and what other borrowing is missing. Its recent movement is not a stand-alone measure of excess leverage. Similarly, a valuation ratio based on realised earnings and one based on analysts’ forecasts answer different questions. Specify the earnings period, index and data date before comparing them. This page does not claim a current valuation extreme that has not been independently checked.
Worked example: leverage changes the loss
Assume an asset costs 10,000 neutral currency units. Ignore interest, fees and taxes for this example.
| Position | Asset falls to 8,000 | Equity after repaying principal | Loss on initial equity |
|---|---|---|---|
| Pay 10,000 entirely from own funds | 8,000 | 8,000 | 20% |
| Pay 5,000 and borrow 5,000 | 8,000 | 3,000 | 40% |
The asset falls 20% in both cases. Borrowing magnifies the equity loss because the debt principal has not fallen. Real lending terms may force a sale earlier, and interest or fees increase the loss. The example illustrates financing risk; it does not estimate the chance of a crash.
What to do with the checklist
Review whether a loss would interfere with essential spending, whether one sector dominates your portfolio, and whether borrowing creates obligations you cannot meet from other resources. A risk decision can be based on those constraints without a prediction about tomorrow’s price.
Diversification explains concentration risk, while investment fees covers costs. The compound interest calculator lets you explore explicit return assumptions; smooth model paths are not forecasts of market returns. The investing vs. gambling reflection helps examine your decision process without certifying an investment.
Sources you can check
- Minsky — The Financial Instability Hypothesis (1992), original paper
- Greenwood, Shleifer and You — Bubbles for Fama, published paper (2019)
- Anne Goldgar — Tulipmania, author excerpt
- Federal Reserve History — Stock Market Crash of 1929
- FINRA — Margin Statistics, August 2026 data