A brief history of getting carried away

Agree to buy a bulb before it's delivered, and the flower can get cheaper while your agreement keeps the old price.

Anne Goldgar follows those speculative contracts and the sharp rise in some bulb prices before the market broke in early February 1637. Her research describes a much smaller trade than the familiar national-disaster story.

No numeric price benchmark is approved for this episode, and none is shown: the archival work cited here supplies no market-price series, and it rejects the familiar story of universal participation and national ruin. A promise to pay for a bulb is not the same thing as the bulb's resale price, and neither is a modern bank loan.

Sell the coin and the lender still wants paying. If the sale brings less than you owe, you have a loan and no coin. Buying outright means no lender, and the price can still drop.

The rounded prices come from the ECB's November 2022 article. Its negative outlook is the authors' view, not proof of permanent collapse. The BIS finding about limited wider financial spillovers also concerns 2022.

Price per bitcoin, at its peak

about $69,000

the November 2021 peak

Price per bitcoin

about $17,000

mid-June 2022

Implied fall

about 75%

between those two observations · our arithmetic

1 − 17,000 ÷ 69,000 = 75.4%

Rounded narrative observations, not one exchange's exact daily close, and not the whole 2022 peak-to-trough history. The first figure is a peak the source names, not a representative price for that November, and the fall between the two is our own arithmetic on the source's two rounded figures rather than a decline the source states. The cited article's broader interpretation is strongly negative; it is used here for these two dated figures only, not as proof of a permanent collapse.

A government-security holder in 1720 could hand over that certificate for South Sea Company shares. The new certificate went in the same drawer. The holder now had an ownership interest instead of the same debt claim.

When the shares crashed, the holder lost money, and a falling share price is not the government paying anyone back.

The company also participated in the trade in enslaved people. Its financing can't be separated from that business.

No share-price endpoints and no present-day equivalent are approved for this episode, so none appear. The South Sea Company also held a right granted in 1713 to supply enslaved people to Spanish colonies; that is recorded here and in the source text, and it is not drawn. The engraving reference in the source list is a late-18th-century print after Hogarth, not an image made during the 1720 event and not evidence for any 1720 price.

The prospectus is on a London desk and the borrower is overseas. One is easier to inspect.

The New York Fed describes speculation and poor information in the lending boom. It reports Latin American bond prices roughly halving by summer 1825. Those are resale prices. The principal owed didn't halve, and it isn't a formal index of every London bond.

Poyais took the paperwork further. The land and its inhabitants were real and the issuing state was made up. It was one fraud within the wider panic.

Latin American bond prices, in this historical account

Roughly halved

by summer 1825

A bond is a written promise to pay that can itself be bought and sold, which is the idea this card introduces. The halving is an approximate historical generalization from the cited account, not a defined index for all London bonds and not a dated series; no numeric index observations exist for it here. It is not a principal write-off and not a recovery rate. Poyais is one swindle among many and is not the explanation of the whole crisis; the invented state is the fraud, not the land or the people said to live there.

A website can be busy and broke at the same time. Investors who paid for years of expected growth still need the business to earn it, and a working website doesn't tell them whether it will.

The Nasdaq comparison uses closing index levels over more than two years. It measures a price decline, not debt paid off.

Nasdaq Composite, closing level

5,048.62 points

10 March 2000

Nasdaq Composite, closing level

1,114.11 points

9 October 2002

Decline between those two closes

77.9%

between these two dated closes · our arithmetic

(5,048.62 − 1,114.11) ÷ 5,048.62 = 77.9%

Two daily closing levels of an index, measured in points. The closes are the dataset's; the percentage between them is ours, worked out on this card rather than quoted from anyone. Index points are not dollars owed, and no total dollar loss can be inferred from them. This was a decline over more than two and a half years, not a one-day crash. The series is a daily closing index, not a total-return series. Nothing here reclassifies any company's reported debt elsewhere on this site as a dot-com loss.

The racks need paying for before anyone knows what they'll earn. A company can use its cash or borrow, including through private credit. Either way, the equipment arrives before the answer.

The BIS research asks whether earnings can sustain the build-out. Its September 10, 2026 speech describes anticipated AI-related capital expenditure by the five largest big-tech companies across 2025 and 2026.

That spending plan isn't an account of OpenAI's or Anthropic's current debt, and it doesn't say how the boom ends.

Anticipated AI-related capital expenditure, five largest big-tech companies

more than $1 trillion

across 2025 and 2026 · BIS speech, 10 September 2026 · partly forecast

This entry is unresolved and stays unresolved. The spending figure is an attributed, forward-looking two-year plan stated in a speech dated 10 September 2026; it is not an outstanding debt balance, not a valuation, and not a claim that the money has already been spent. The separate January 2026 bulletin's risk assessment is dated to January and is not an updated September measurement or a guarantee of any outcome. No crash is predicted here, and neither is a payoff. Reviewed 13 September 2026.

Imagine selling the house and handing the proceeds to the lender. In our example, the sale brings $70, but the loan is $80. There's $10 left to find.

Before the price drop, the home was worth $100 and the owner had $20 in equity. After it, equity is negative $10. These are hypothetical figures, with payments, interest, and loan changes held still.

The actual mortgage crisis also involved lending, securitization, collateral, and wider financial stress.

The $100 / $80 / $70 figures are an arithmetic illustration written for this card, not observations and not a figure from the cited source. There is no claim that U.S. homes fell exactly 30%, and this is not a foreclosure or recovery model. Not all mortgage lending was subprime, and the crisis had no single cause. The loan is held fixed by construction.

The train arrives full. Whether the shares were worth buying still depends on the ticket money, the business's costs, and what you paid for them.

The study follows fourteen large railways continuously listed from 1840 to 1860. Its 1844–1845 return belongs to that portfolio, not every railway proposal or the whole boom and bust.

Cumulative return, a portfolio drawn from 14 large railways continuously listed 1840–1860

63.5%

1 January 1844 to 9 August 1845

A reported return for a selected portfolio over a selected window: 14 large railways that stayed continuously listed from 1840 to 1860, not every railway proposed in the 1840s. It is not a total-return index for the period and no chart has been digitised from the study. No numerical bust endpoint is supplied here, and no railway mileage or reconstructed loss total is claimed. This is an analogy for later technology booms, not a rule that technology investment ends badly.

A building can back a loan. Raise its valuation and it may back a bigger one. Use that loan to buy more property, and borrowing feeds the market that supplied the valuation. Falling prices can reverse the process.

Ito and Iwaisako examine Japan's stock and land prices alongside lending and collateral, including partial explanations based on fundamentals. Stocks and land followed different timelines and don't fit one neat rise and fall.

A qualitative, research-attributed mechanism. No land-price index, stock-price level or headline monetary total is approved for this entry, and none is shown. The cited findings are the authors' own, not necessarily the Bank of Japan's official position. Late-1989 stock-price growth is distinct from land-price increases in 1990; the two markets are not given a shared peak date. No claim is made that later Japanese economic outcomes all followed from this episode.

Part of the share purchase is your money; the rest is borrowed. A price drop eats into your smaller stake while the lender's claim remains. That's how a modest movement can become a large loss relative to the cash you put in. Gains are magnified too.

Not every investor used a margin loan. The Dow Jones figures run from September 1929 to July 1932, almost three years. Black Tuesday is one date inside that decline.

Dow Jones Industrial Average, closing level

381.17 points

3 September 1929

Dow Jones Industrial Average, closing level

41.22 points

8 July 1932

Decline between those two closes

89.2%

over almost three years, between these two dated closes · our arithmetic

(381.17 − 41.22) ÷ 381.17 = 89.2%

These are index points, not dollars, even though the source's chart description uses a currency symbol in places. The two closes are the publisher's; the 89.2% between them is ours — the essay itself says “89 percent” — and it is marked as our arithmetic rather than as a figure anyone published. They cannot be converted into a total of debt or of investor cash losses. This is not a one-day Black Tuesday return. Borrowing to buy assets magnifies what a price change does to an investor's own stake, which is a separate mechanism from the index level; no share of investors who borrowed is claimed here. One stock market does not by itself explain the Great Depression.

Cage disclosed approximately $14 million owed to the IRS in January 2010. Later reporting described that tax bill and separate creditors. It's a large pile even before the other paperwork arrives.

His roughly $6 million repayment recollection in 2023 is a separate account, not a replacement for the earlier disclosure.

Owed to the IRS, as stated

about $14,000,000

disclosed January 2010

A historical, approximate, self-reported figure, attributed to a January 2010 statement and reaching us through a blog post quoting a magazine. It is not his current debt, not a verified lifetime maximum and not a record of any kind. No court filing or tax record states $14 million; the largest amount tied to actual recorded documents is about $13.3 million in federal tax liens for 2002, 2003, 2004, 2007 and 2008, which is a different measure and a different set of years. In a 2023 interview he said he had paid it all back and put the total at about $6 million — his own later account of the same debt, not a second, separate one. Both numbers are his; they are thirteen years and one repayment apart, and neither is audited.

Bowie's 1997 bonds raised $55 million against royalties from his earlier music catalogue. Prudential bought them.

The songs could keep earning; the deal let him receive money up front instead of waiting for those royalties. The bondholder had a reason to care what was playing.

Raised by the bond issuance

$55,000,000

1997

A 1997 financing issuance, not an overdue personal bill and not a sign of financial distress. The citation is a 2002 entry on the artist's official site, which does not describe the issuance in its own words — it quotes a Q magazine piece. That is one hop further from the event than a citation normally implies, and the entry's comparison of the bonds' performance with government stock is its own claim, not adopted here. What happened next is settled and better documented than the entry: Moody's rated the issue A3 at the outset and cut it to Baa3 in March 2004, one grade above speculative, citing weaker recorded-music sales and a downgrade of the guarantor; the bonds were repaid in full at maturity in 2007. The drawn royalty notes illustrate the arrangement, not a measured cash-flow rate.