The anatomy of a bubble — what all of them share
The answer
Every episode in this cluster follows the same 5-stage shape: a real change, easy money, a widening of who participates, a justification for why the old rules no longer apply, and a stop that comes from inside rather than from news. Recognising the shape does not let you predict the date. It lets you decide how much money to have in it.
Why this costs you money
Most people who read bubble history come away with pattern recognition and no instrument. They can describe 1929. They can name Pets.com. And they still hold too much of the thing that has risen most, because knowing the story does not change a position size.
The cost is concentrated in 1 behaviour: letting the winner grow.
You bought 6 holdings in equal amounts. One rose 5 times while the others did nothing. It is now 45% of your portfolio, and you never decided that. The price decided it, in the direction of the thing that has already risen most.
That is not diversification failing. It is diversification quietly reversing itself, and it happens in every mania, to careful people, without a single bad decision being taken. The bubble does not need you to buy at the top. It only needs you to hold while the arithmetic concentrates you.
How it works
Charles Kindleberger, building on Hyman Minsky, set out a sequence. Stated plainly, and checked against the episodes in this cluster, it has 5 stages.
Stage 1: Displacement. Something real changes. A new trade route, a new technology, a new law, a new source of demand. It is genuine, and this is why bubbles are hard. Tulips from the Ottoman Empire. British government debt becoming tradeable. Electrification in the 1920s. The internet in the 1990s. Indian corporate profits after the 1991 reforms. Securitised lending in the 2000s.
Stage 2: Easy money. Credit becomes available on new terms, or a new mechanism lets people control more than they own. This stage turns a re-rating into a bubble, and it is the most reliably observable of the 5.
Look at how the money got in each time:
- 1637: forward contracts requiring no payment until the summer.
- 1720: the South Sea Company lending money to buy its own shares.
- 1929: margin loans at 10% to 20% down.
- 1992 India: bank funds moving through the settlement gap.
- 2000: initial public offerings with tiny free floats and no profits required.
- 2008: mortgages with no income documentation.
- 2021: zero-commission trading and cheap short-dated options.
Stage 3: Broadening. New participants enter, and they enter because prices have already risen. The composition of buyers changes: from people who understand the asset, to people who understand that it has been going up. This is where the sharpest signal lives, and it is behavioural rather than financial.
Stage 4: The justification. A framework appears explaining why the usual measures no longer apply. Not a lie — usually a real method, applied where it does not belong. Replacement cost in 1992 India. "Eyeballs" in 1999. Correlation assumptions in 2006. "Permanently high plateau" in 1929. The justification always arrives after the price, which is the tell.
Stage 5: The stop. Buying slows. It rarely stops because of news. It stops because the pool of new buyers is exhausted, or because the credit from stage 2 is withdrawn. Then the leverage runs in reverse, and the selling is forced.
One more property, and it is the most useful thing here. Stages 2 and 3 are measurable in advance. Stage 5 is not. You can find out how much credit is behind an asset and who is buying it. You cannot find out when it ends. So every practical conclusion from bubble history is about position size, never timing.
What it tells you, and what it does not
What it tells you. A checklist. If you can answer 5 questions about an asset, you know roughly where you are standing, which is enough to decide how much of your money belongs there.
It also tells you the strongest signal is compositional, not numerical. A high valuation is weak evidence, because valuations can stay high for years and the correct multiple genuinely varies. A change in who is buying and why is stronger evidence, and it is the thing nobody records.
What it does not tell you. It does not tell you when. This must be said plainly, because the whole appeal of bubble history is the implicit promise that you could get out in time. Nobody does this reliably. The people remembered for calling a top were usually early by 1 to 3 years, which for most investors is indistinguishable from being wrong.
It also does not tell you that every large rise is a bubble. Prices rise for good reasons constantly. A framework that labels every rise a bubble has no information in it, and using this checklist to avoid all gains would cost you more than the crashes would.
And it does not tell you that bubbles are always destructive. Several produced infrastructure that outlasted them: railways in the 1840s, fibre optic cable in the late 1990s. The capital was destroyed and the cable stayed. That explains why bubbles keep happening in genuinely important technologies.
The decision rule
Bubble history is a position-sizing instrument, not a timing instrument. Use it to decide how much, never when.
The conditional form. If an asset scores high on the checklist below and it has grown to a large share of your portfolio through price rather than through your decision, reduce it to the weight you would choose today with fresh money. That is the only action supported by the evidence.
If it scores high and it is a small position you can lose entirely, you may keep it, and you should write down that you are keeping it on those terms.
What is not supported: selling everything, predicting the top, or shorting. This cluster contains many people who were right about value and ruined by timing.
Try this now
Fifteen minutes. This is the exercise the whole cluster was building toward.
- Open your holdings. Identify the single position that has risen the most since you bought it, and note what percentage of your total portfolio it is now.
- Note what percentage it was when you bought it. If you cannot find that, work it out roughly from the amount you invested.
- Now score that holding out of 5. Answer each question yes or no, in writing.
Q1. Displacement. Is there a genuine change behind this — a real technology, policy, or demand shift? A yes here is not reassuring. Every bubble has one.
Q2. Easy money. Is borrowed money, or an instrument that lets people control more than they own, involved in the buying? Check: margin lending in the stock, promoter pledge levels, high options volume relative to the shares, or a lending boom in the industry itself.
Q3. Broadening. In the last 12 months, have you seen this asset discussed by people who do not normally discuss investments? Count actual instances, not impressions.
Q4. Justification. Is the price defended using a measure that was not used for this asset 3 years ago? Revenue multiples replacing profit multiples, "total addressable market", users instead of customers, adjusted earnings far above reported earnings.
Q5. Your own reason. Write the reason you are still holding. Does it contain a number about the business, or does it contain a price?
- Add up the yes answers. Write the number and today's date.
- Then the action that matters. Ask 1 question: if I had this money in cash today, would I put this percentage of it into this holding? If no, sell the difference between the weight you have and the weight you would choose. Not the whole position. The difference.
What you should see. Most readers score between 2 and 4 on their biggest winner. Very few things score 0, and a score of 5 is rare.
The number is not a prediction. Its purpose is step 5. Almost everyone finds that their largest holding sits at a weight they would never choose deliberately, and that they have carried that concentration without ever agreeing to it.
Keep the dated sheet. Together with the other dated notes this cluster asks for, it is the only record you will have of what you believed before the outcome was known.
Three real cases
1. Railways in Britain, 1844 to 1846 — the bubble that built something Parliament authorised a very large number of railway companies. Share prices rose sharply, collapsed from 1845, and many companies never built anything. What it turned out to be: enormous capital destruction and a national rail network that was genuinely built and used for the next 150 years. Investors lost; the country gained.
2. Japan, 1986 to 1990 — the one that did not come back Japanese share and land prices rose enormously in the late 1980s, supported by loose monetary policy and bank lending against land. The Nikkei 225 peaked in December 1989 and took more than 30 years to regain that level. What it turned out to be: the counter-example to every "markets always recover quickly" argument. The displacement was real — Japan genuinely was the world's most successful manufacturing economy. The credit in stage 2 was the difference.
3. Indian small and micro-cap stocks, 2021 to 2025 — the live one to practise on Indian small and mid-sized company indices rose very substantially from the March 2020 low. From early 2024 SEBI required mutual funds to publish stress test results for small-cap and mid-cap schemes, and SEBI and the exchanges tightened requirements for SME platform listings after very heavy oversubscription. What it turned out to be: still resolving as you read this, which is why it is the right one to run the checklist on. Every episode in this cluster was unresolved once.
The question that resolves it
A novice asks: is this a bubble?
An expert asks: what would I own if I am wrong in either direction?
The second question is answerable today and the first is not. If the asset keeps rising for 3 more years and you own a small position, you make money and you are not ruined by having missed it. If it falls 80% next year and you own a small position, you are annoyed and you are fine.
The question "is this a bubble" has only ever been answered afterwards. The question of what you can survive can be answered before lunch.
What would make this wrong
The falsification condition. The claim is that the 5 stages recur, and in particular that stage 2, easy money, is present in the large episodes. If a significant historical bubble occurred with no credit expansion and no new mechanism allowing control beyond capital, the framework would be incomplete. There are candidates: tulip mania itself may be one, since the forward contracts created obligation but there was little actual bank lending. That is a genuine weakness and it should be stated rather than hidden.
The honest limits, and there are 3.
First, survivorship in the evidence. We study the episodes that ended badly. Many periods with all 5 stages present did not end in a crash. Because the checklist is built from the failures, it will produce false positives, and there is no way to know how many.
Second, you cannot score yourself honestly. You will judge the things you own more kindly than the things you do not. The only defence is to write the answers down before you look at the total.
Third, and most important: this framework has no timing content at all. Every stage can persist for years. A person who applies it and then sits in cash for 5 years has used it wrongly, and will lose more to inflation and missed compounding than the crash would have cost.
The hindsight trap, which is the whole subject of this article.
Every one of the 9 episodes in this cluster looks obvious in retrospect and was not obvious at the time. The reasons were different each time and are worth collecting in one place, because the collection is the actual lesson.
- In 1636, the scarcity of the rarest bulbs was real, and the Dutch economy was genuinely the strongest in Europe.
- In 1720, the scheme had been approved by Parliament, the King was the company's governor, and a smaller version had already worked.
- In 1929, the productivity gains of the decade were real and profits had genuinely risen. The most respected economist in America said prices were reasonable.
- In 1987, the strategy that amplified the crash rested on published mathematics that had already won broad academic acceptance.
- In 1992 India, the liberalisation was real and a large re-rating of Indian equities was genuinely justified.
- In 1999, the internet was more transformative than the optimists claimed.
- In 2006, US house prices had not fallen nationally in 70 years, so every model said what the data said.
- In March 2020, no one knew whether a vaccine was possible, and serious people doubted it.
- In January 2021, the original GameStop analysis was researched and correct, and it worked.
There is a pattern in that list, and it is not that people were stupid. In every case the central factual claim was true, and the price was still wrong. The error was never in the story. It was in treating a true story as though it settled the question of price.
What a careful person could have seen, in every case, is 2 things that were public at the time: how much of the buying was financed, and what the price required to be true about the future. Both are available today, for whatever you own, in about 15 minutes.
What nobody could see, in any case, is when.
In India
The Indian record is short compared with the Western one, and it is denser.
1865, Bombay. The American Civil War cut off American cotton, Indian cotton boomed, and money went into bank and land reclamation shares bought on time bargains with little money down. The war ended in April 1865 and the market collapsed. Every stage is present, including the credit mechanism.
1992, the securities scam. Real reform, bank money entering through a settlement gap, a replacement-cost justification, and a collapse triggered by exposure rather than by news.
2001, Ketan Parekh. Real global technology enthusiasm, cooperative bank funding and badla carry-forward, a concentrated set of stocks, and a collapse after the February 2001 budget.
2008. No Indian credit failure, but a fall from over 20,000 on the Sensex in January 2008 to about 8,160 in March 2009, driven by foreign selling and a genuine global shock.
2021 to 2025. Small-cap and micro-cap strength, heavy SME IPO oversubscription , and enormous growth in individual participation in index options, where SEBI's studies have repeatedly found that a large majority of individual traders lose money.
What came out of those episodes. SEBI's statutory powers in 1992. The NSE in
- Dematerialisation from 1996. Rolling settlement completed in 2002. The ASM
and GSM surveillance frameworks. Mutual fund stress test disclosure from 2024. Each was a response, and each is now free public information you can use before the next one.
In the United States
The American record is longer, and it produced the disclosure system most of the world copied.
The recurring episodes: the 1790s panic, the 1830s land boom, the 1869 gold corner, the 1907 panic, 1929, the 1960s conglomerate boom, the 1980s savings and loan crisis, 2000, 2008, and 2021.
The responses, each following a specific failure: the Federal Reserve in 1913, after 1907; the SEC and the Securities Acts of 1933 and 1934, after 1929; margin rules under Regulation T; circuit breakers in 1988, after 1987; Regulation FD in 2000 and the research settlement in 2003, after the dot-com era; Dodd-Frank in 2010, after 2008; and the move to T+1 settlement in 2024, partly informed by January 2021.
The public tools an American investor has today, all free: quarterly institutional holdings through Form 13F, insider transactions through Form 4, twice-monthly short interest from FINRA, and full company filings through the SEC's EDGAR system.
Where they differ, and what that tells you
The difference is what each system tries to control.
The United States controls information: require disclosure, punish falsehood, and let prices do what they do. There is no daily limit on how far most US stocks can move.
India controls the mechanism: daily price bands, restrictions on short selling, surveillance frameworks that change margins and trading modes, upfront margin, and settlement rules. Disclosure exists too, but the primary instrument is structural.
What that tells you is which failure each market will produce.
An American bubble can go further and faster, because nothing stops it mechanically. The compensation is that you can see the positioning and the insiders while it happens.
An Indian bubble is capped in daily speed, but the information about who is buying and why is thinner, and the biggest recent Indian episodes happened where the mechanism controls were weakest — cooperative bank funding in 2001, the SME platform in 2023 and 2024, index options for individual traders.
So the instruction differs. In the United States, use the disclosures: 13F, Form 4, short interest. In India, use the structural signals: ASM and GSM lists, delivery percentages, promoter pledge levels, mutual fund stress test results, and the monthly statement from the depository.
Both sets are free. Both are ignored by almost everyone. That is the honest reason the same shape keeps repeating: not because the information is unavailable, but because reading it is dull and the story is exciting.
Carry this
- 5 stages: real change, easy money, new buyers, new justification, a stop from inside.
- Stage 2 and stage 3 are measurable today. Stage 5 is never predictable.
- This is a position-sizing instrument. Ask what weight you would choose with fresh money, and go to that weight.
Knowledge check
Related
- Tulip Mania, 1637
- The South Sea Bubble, 1720
- 1929 and the Great Depression
- The dot-com bubble, 2000
- GameStop and the meme-stock mania, 2021
- ← Market history: manias, crashes and manipulation