Here is where the story is supposed to widen into farce: the market as a lecture hall of rereaders, the mania as fluency at the scale of nations, tulip bulbs trading for houses. The research files refuse the script, and their refusal is the lesson.

Take the tulips. When the economist Peter Garber went to the seventeenth-century price records, the archetype of collective madness came apart in his hands: after the 1637 peak, prices of the celebrated rare bulbs fell about 32 percent a year, a pace he found not far from the ordinary depreciation of prized bulbs in later, calmer decades, as novelties became commonplace. What looked like civilization losing its mind was, on the bulb ledgers, roughly a flower behaving like a flower. Four centuries of writers have retold tulipmania without checking it, which means the most famous story about market delusion is itself a fluency artifact: a tale so retellable that nobody audits it.

Or take the folk rule, sell whatever has soared. When William Goetzmann assembled a century of national stock markets, he found that after a market doubled in real terms, it went on to double again over the next five years nearly twice as often as it went on to halve, roughly 26 percent of cases against 15. Booms, at the level of whole markets, are mostly not ticking bombs. A young investor who learns only this has already avoided the genre's oldest mistake.

So what does survive? Something narrower, stranger, and more useful, and it is best introduced through a country that did nothing wrong. In 2007, Germany's own credit gauges were quiet. No boom, no overheating; by its domestic numbers it deserved calm. But a third of the countries around it were deep inside what one team of economists calls the red zone, the condition where rapid credit growth and rapid asset-price growth arrive together, and Germany's banks were wired into all of them. The model gave Germany elevated odds anyway, on contagion alone. In 2008, its crisis arrived on schedule.

That red-zone result, built on 42 countries across seven decades, is the sturdiest thing the mania literature owns, and what makes it sturdy is not its power. It is that it publishes its own failures. A country in the zone faces roughly a 45 percent chance of financial crisis within three years for business credit, 37 for household, against a background rate of about 4 percent in a given year. Now read the failure rates in the same breath, because the authors do: most alarms are followed by no crisis at all, and roughly a third of crises arrive with no warning from the indicator, error rates the research file computes from the paper's own figures. The signal even survives the hindsight objection, the one that flattens most of this genre, because it still works when estimated only on data available at the time, ending before 2008. A warning system wrong more often than right, blind to a third of what it watches for, and still lifting three-year odds toward a coin flip: that is what the best macro-forecasting instrument in existence actually looks like. A regulator, whose errors are priced asymmetrically, should act on it. An investor who exits markets on it will be wrong more often than right and will pay in forgone returns, so for an individual it is a dial for attention and position size, never a date.

Inside a single sector the evidence sharpens, and it comes with a concession that should raise your trust rather than lower it. Researchers who tracked American industries back to the 1920s began by conceding, in as many words, that the great skeptic of bubbles, Eugene Fama, is right about averages: a sector that has soared does not, on average, go on to deliver unusually poor returns. What soaring changes is the tail. Define a crash as a 40 percent collapse within two years, and the background odds for any industry run about 14 percent; after a sector doubles against the market, those odds reach 53 percent; among run-ups half again as large, four of every five ended in a crash, though that last figure rests on just fifteen episodes in a century of data, small enough to treat as a direction rather than a law. Both findings are true of the same distribution: the average holds while the left tail fattens. Which dictates the only coherent response: reduce your size, never your direction; betting against the boom is how the boom's skeptics go broke. The practitioners' version is a dated note, written before the run-up matures, stating what you will do at each threshold. And the evidence grades the practice with unusual severity: if you quietly amend the note as the number approaches, the amendment is not a footnote to your experiment. It is the result.

One market instrument remains, and it is the one for stories. Every boom arrives narrated, and the narratives have no predictive power at all; new-era stories attend the booms that break and the booms that build the future, which is why their presence proves nothing. But a good story always rests its full weight on one number, and that number can be audited. In the railway mania of the 1840s, the number was the traffic forecasts, produced by surveyors called traffic takers, and the historian who studied the episode concluded that investors were operating under a serious delusion about those estimates' accuracy: checkable, and almost never checked. In the telecom boom of 2000, the number was the claim, repeated by a major executive, that internet traffic was doubling every hundred days, while measured traffic doubled about once a year. The story is not the signal. The story is the map to the audit: write the number your favourite narrative depends on beside the narrative, with the date you checked it.

Today

Log

Drills

Weekly Review

Reading Room