Manias ↔ technology diffusion ↔ forecasting
Every mania feels unprecedented from the inside. From the outside they rhyme — tulips, railways, dot-com, crypto, maybe AI. There's a recurring shape to financial euphoria, a recurring function to technology bubbles, and a rigorous discipline for forecasting them. None of it gives you the one thing you actually want: when.
How to read this. Four stages, in order — each opens with a real puzzle, builds the model under it, lets the strongest objections fight, and hands you to the next. Stage 3 has a bubble curve you can drag a marker along. The aim isn't to call the top; it's to see exactly which part of the question is answerable and which part isn't.
"This time is different." — the four most dangerous words in finance, per Carmen Reinhart & Kenneth Rogoff, whose book takes them as its title
The phrase is what every bubble tells itself at the top. New technology, new monetary regime, new kind of asset — the particulars are always genuinely new, which is exactly why the warning is so easy to dismiss. Reinhart and Rogoff's point, drawn from eight centuries of defaults and crashes, is that the feeling of novelty is the most reliable signal you have. The story is new; the shape isn't.
Charles Kindleberger, in Manias, Panics, and Crashes, argued that bubbles follow a recurring arc rather than appearing as one-off accidents. He built the spine of that arc on Hyman Minsky's stage model: displacement — a new opportunity, a real one, that resets expectations — then boom, then euphoria, then profit-taking as the early money quietly leaves, then panic.
Minsky's deeper claim is why the arc keeps recurring. His financial-instability hypothesis says stability itself breeds instability. A long calm stretch makes leverage look safe, so financing drifts through three regimes: hedge finance, where income covers both interest and principal; speculative finance, where income covers only the interest and the principal must be rolled over; and Ponzi finance, where income covers neither and only a rising asset price can service the debt. The further the system slides toward Ponzi, the smaller the shock needed to tip it — until a Minsky moment, when asset prices stop rising and the whole structure has to unwind at once.
Efficient markets
"Bubbles are hard to identify in advance. Prices already reflect available information; if a 'bubble' were obvious, traders would act and it would deflate. The arc is a story we tell after the fact, fitting a curve to whichever rises happened to reverse."
Behavioral / Minsky
"The arc is recognizable and recurrent across centuries and asset classes. Displacement, euphoric leverage, the drift to Ponzi finance, the snap — the sequence repeats whether the asset is a tulip, a railway, or a token. The pattern is the data, not the narrative."
Where this leaves us
The shape repeats even though the story is always new. That's the unsettling part of "this time is different": the sentence is literally true about the technology and reliably false about the trajectory. Knowing the arc tells you a great deal about the structure of what you're inside. It tells you nothing yet about whether the structure is doing any good.
Because some bubbles, it turns out, leave behind the rails the future runs on.
A crash looks like pure destruction — fortunes erased, companies gone, savings wiped out. But ask a different question: after the dust settles, what's still standing? Often it's the infrastructure the mania paid to over-build — track, fiber, compute — sitting on the ground at a fraction of what it cost to lay down.
Carlota Perez, in Technological Revolutions and Financial Capital, mapped a recurring surge that each technological revolution drives through five phases: irruption, when the new technology appears; frenzy, the bubble, when financial capital pours in and over-invests; crash; synergy, a deployment "golden age" when the technology spreads through the real economy; and maturity. The frenzy over-builds the new infrastructure — canals, then railways, then fiber-optic cable in the dot-com era, perhaps AI compute now. The crash wipes out the speculators. But the cheap, over-built infrastructure they financed is what powers the golden age that follows.
Paul David's essay The Dynamo and the Computer supplies the timeline correction. His subject is the productivity paradox: a general-purpose technology can be installed for decades before it shows up in the productivity statistics, because the gains require reorganizing everything around it. Electric motors took roughly forty years to reorganize the factory — moving from one central steam shaft to a motor on every machine, which let factories be laid out around workflow instead of around the drive train — before they lifted measured productivity. That lag is exactly the question hanging over AI's timeline: a real revolution and a slow payoff can look, from inside the frenzy, identical to a fake one.
Bubbles are pure waste
"Capital that could have funded sound investment is incinerated chasing a mania. The survivors would have built the infrastructure anyway, more cheaply, without the boom-bust whiplash. Calling the waste a 'subsidy' is just romanticizing a disaster after the fact."
Perez: bubbles install infrastructure
"This is how capitalism funds the unfundable. No rational investor pays to lay a continent of track or fiber on day one; only a frenzy, with its suspended disbelief, mobilizes capital at that scale. The crash transfers the asset to the economy at a discount. The deployment golden age runs on it."
Where this leaves us
A bubble can be both a financial disaster and an infrastructure subsidy at the same time. The crash isn't the end of the technology's story — it's frequently the start of its useful one. So "is this a bubble?" turns out to be the less interesting question. The interesting one is which phase you're standing in, and David's lag warns you that even a real revolution will spend years looking like a bust.
Which means the honest move is to stop arguing from inside this case — and enlarge the sample.
The cure for "this time is different" is not better data about this time. It's to enlarge your sample — to stop treating the present mania as a unique event and start treating it as one more draw from a reference class that has played out many times before.
Daniel Kahneman, in Thinking, Fast and Slow, draws the distinction this stage turns on: the inside view versus the outside view. The inside view builds a forecast from the specifics of this case — this team, this technology, this set of reasons it'll work — and is systematically overconfident, because the specifics are vivid and the ways things go wrong are not yet visible. The outside view ignores the specifics first and starts from the base rate of the reference class: of all the situations that looked like this one, how did they actually turn out? Reference-class forecasting — anchoring on that base rate, then adjusting — is the rigorous core of "read history to enlarge your sample."
Philip Tetlock's Superforecasting is the empirical payoff. Tracking thousands of forecasts, he found that ordinary people using disciplined methods — base rates first, frequent small updates, calibrated probabilities instead of verdicts — reliably beat credentialed experts working from the inside view. Forecasting turns out to be a trainable skill, and the trainable part is precisely the discipline of the outside view.
Anatomy of a bubble
Drag the marker to move the "present" along the canonical curve. Read the phase — but watch the x-axis.
Stealth Smart money quietly accumulates
It rhymed with…
You can read the shape — which phase you're in. You cannot read the scale of the x-axis — how long each phase lasts. That's the timing, and it's unknowable. "Markets can stay irrational longer than you can stay solvent."
Where this leaves us
You can attach calibrated probabilities to the phase you're in — and you can update them as evidence arrives. That is a real, trainable skill. It is a different skill from naming the date. The outside view sharpens your read of where on the curve you stand; it stays deliberately silent on when the next point arrives.
And that silence isn't a gap in the method. It's a property of the thing being forecast.
Suppose you read the shape perfectly. You know the arc, you know the function, you've done the outside view and you're calibrated. You still don't know when it breaks — and the reasons you don't are not fixable with more work.
First, reflexivity. A forecast about a market is not a measurement from outside it; it's an input into it. A widely believed crash forecast changes behavior — it can delay the break, as believers prepare and sell early into strength, or trigger it, as the same selling becomes the shock. The thing being predicted moves in response to the prediction, so there's no fixed target to aim at.
Second, Frank Knight's distinction from Risk, Uncertainty and Profit between risk and uncertainty. Risk is quantifiable: known outcomes, known odds, a roulette wheel. True uncertainty is not: you cannot assign a meaningful probability distribution at all, because the reference class is too thin or the future too open. The timing of a crash lives in the uncertainty column. You can price the phase; you can't price the day.
Third, fat tails. Even when the setup is clear, the size and timing of the break are radically uncertain, because market moves don't follow a tidy bell curve — the extreme events that end manias are far more common and far larger than a normal distribution would allow. That's why Keynes's warning, commonly rendered as "markets can stay irrational longer than you can stay solvent," is the practical last word: being right about the shape and wrong about the timing is indistinguishable, on your balance sheet, from being wrong.
The deepest objection comes from efficient-market theory and cuts at whether bubbles were ever knowable in advance. If a bubble were reliably identifiable while it inflated, rational traders would sell or short it, and their trades would arbitrage it away before it grew large. So — the efficient-market defenders argue — what looks like an obvious bubble in hindsight wasn't actually exploitable in advance; the very fact that it kept inflating is evidence that informed money couldn't profitably bet against it.
Behavioral economists counter on two fronts: the documented recurrence of the arc across centuries and asset classes is hard to wave off as coincidence, and there are well-understood limits to arbitrage — short-selling is costly and dangerous, a bubble can stay inflated longer than a short-seller can fund the position (Keynes again), and "smart money" can rationally ride a bubble rather than fight it. Both sides have a real case. The "were they knowable?" question stays genuinely contested — which is itself the honest answer to the title.
The judgment
Yes, you can see the shape: Minsky and Perez give you the phase, the outside view gives you calibrated odds, and the signatures of euphoria — "new paradigm," Ponzi-stage leverage, "this time is different" — are recognizable from inside. No, you cannot reliably see the timing: that's set by reflexive crowd behavior and fat-tailed shocks, and it lives in Knight's uncertainty, not his risk. The honest skill is probabilistic phase-reading. It is not date-calling, and anyone selling you the date is selling the one part of this that can't be sold.
The short answer
The arc of a mania is real and recurrent — Minsky's slide from hedge to Ponzi finance, Perez's frenzy that over-builds the infrastructure the golden age will run on. Enlarge your sample with the outside view and you can attach calibrated odds to the phase you're in. What you cannot get is the date. Timing is set by reflexive crowds reacting to the forecasts themselves and by fat-tailed shocks no distribution captures — Knightian uncertainty, not risk. So the answer is half yes and half no, and the honest half is the limit: you can see a bubble's shape before it bursts. You cannot see when.
C. Kindleberger, Manias, Panics, and Crashes (orig. 1978) — the canonical history of financial bubbles as a recurring arc; builds its spine on Minsky's stages. The scholarly entry point.
H. Minsky, the financial-instability hypothesis — stability breeds instability; the drift from hedge to speculative to Ponzi finance, and the "Minsky moment" of collapse.
C. Reinhart & K. Rogoff, This Time Is Different (2009) — eight centuries of financial folly; the title is the warning. The data behind "the story is new, the shape isn't."
C. Perez, Technological Revolutions and Financial Capital (2002) — the surge model: irruption → frenzy → crash → synergy → maturity; the bubble as how capitalism installs new infrastructure.
P. David, "The Dynamo and the Computer" (1990) — the productivity paradox; general-purpose technologies take decades to reorganize production and show up in the statistics. The timeline check for AI.
P. Tetlock, Superforecasting (2015) — disciplined ordinary forecasters beat experts; base rates, frequent updates, calibrated probabilities. The empirical case for the outside view.
D. Kahneman, Thinking, Fast and Slow (2011) — the inside view vs the outside view; why the specifics of this case breed overconfidence and the base rate corrects it.
F. Knight, Risk, Uncertainty and Profit (1921) — the distinction between quantifiable risk and true uncertainty; why the timing of a crash can't be priced.