# Why Rational People Make Panics *The Coleman Course · Part 9 · [[00-social-dynamics|all 24 laws]]* > **Law 9 — Panics are made of rational people. Fix the exit, not the crowd.** Depositors empty a bank that was solvent this morning. Millions buy an asset nobody can value. A false rumour convinces a country in an afternoon. The comfortable explanation is that crowds make people stupid — a "group mind" switches rationality off. Coleman's explanation is harder and far more useful: **everyone in a panic is behaving sensibly.** The panic is what sensible people produce when they hand control of their actions to each other. Why hand it over? Because under uncertainty, other people's behaviour is the cheapest information there is. When you can't verify what's true, watching what others do is a *good strategy* — most of the time. Coleman is precise about what makes collective behaviour special: the transfer is **unilateral** (nobody pays you for it) and **broad** (to whoever's around, not to a chosen authority — contrast [[04-why-you-do-what-youre-told|Law 4]]). The trouble starts when everyone does it at once. Then the crowd is watching itself, nobody holds new information — the information just *circulates* — and a system where everyone copies everyone has no brakes. At the core sits a structure worth having a name for — **double contingency**: ```mermaid flowchart TD A["A's best move depends on<br/>what B is about to do"] --> B["B's best move depends on<br/>what A is about to do"] B --> A A -.-> C["Neither can wait to find out"] B -.-> C C --> D["A few early movers →<br/>'evidence' manufactured →<br/>everyone moves at once"] ``` **The bank run is the cleanest specimen.** If nobody withdraws, the bank is fine. If everyone does, the last in the queue lose everything. So the moment you believe others will run, running is *correct* — even if you know the bank is healthy. The pathology lives in the structure, not in any head. All that's changed since Coleman wrote: clock speed. Silicon Valley Bank, 2023 — the queue was a group chat, and tens of billions moved in hours. The signal now outruns the facts; you don't need to see a crowd anymore, only a screenshot claiming one exists. Deposit insurance is the structural answer, and notice what it does *not* do: persuade anyone. It changes the payoff so that being last costs nothing — and a queue nobody needs to win stops forming. Every central-bank "whatever it takes" works identically: aimed at the contingency, not the balance sheet, and often effective without a cent being spent. ## One machine, many costumes Coleman's chapter is really a taxonomy — the same transfer of control pointed in different directions, and it's worth seeing the family resemblance: **The craze** is the panic in reverse — everyone rushing in instead of out. Tulips, railways, dot-coms, crypto, and the largest capital-spending wave in tech history. One question cuts every boom, including this one: **are people buying because of what they think it's worth, or because of what they think other buyers will do next?** The first is a market; the second is a craze — and real technologies host crazes happily. Railways and the internet were both genuinely transformative *and* crashed spectacularly. Being right about the technology does not protect you from the structure. **The rumour cascade** is the same transfer in the currency of belief: unable to verify, people count how many others repeat a claim — but the others did the same, so repetition gets mistaken for evidence. Recommendation algorithms are machines that measure the transfer of control and amplify it; generative AI removed "I saw the video," the last cheap check most people had. **The hostile crowd** adds Coleman's most unsettling detail: people in mobs do things they'd never do alone, not mainly from anonymity, but because each member has *deposited their right to judge* into the crowd. The first stone is expensive; the tenth is free, because the crowd has by then issued permission. That's why escalation is stepwise, why one unpunished act flips a demonstration, and why an online pile-on feels acceptable from inside — everyone is drawing on a permission the group already granted, and [[07-just-doing-my-job|responsibility was deposited along with the judgement]]. ## The five-condition dashboard Coleman closes with the conditions that make all of this likely — readable as a risk checklist for any system you're in: **high uncertainty** (nobody can verify) · **strong interdependence** (your outcome hangs on others' moves) · **communication faster than verification** · **no authority people actually believe** · **cheap to join, expensive to be last.** Several lit at once → expect a run, a craze, or a pile-on — and expect the trigger to look ridiculously small afterwards. It always does. Triggers don't cause cascades; structures do. The [[01-why-society-did-it-explains-nothing|Ceuta surge]] lit all five simultaneously, which is why sixty thousand people moved on a rumour that was never even true. ## Where people misapply the law Don't conclude that crowds are always wrong — the same copying machinery that makes panics also makes markets informative and conventions possible; most days, following the crowd *is* the rational play, which is exactly why the failure mode is so hard to resist. And don't over-diagnose: a rush isn't automatically a cascade. Sometimes everyone moves at once because everyone received the same *real* news. The test is the source of the signal: are people reacting to the world, or to each other? Only the second has no brakes. ## Use it this week Next time you see people doing something en masse, don't ask what's wrong with them. Ask: **what does each of them think everyone else is about to do?** Answer that and you can usually see the whole thing coming — and you'll know that arguing with individuals is useless (they're not making a mistake), while changing the payoff for being last is everything. If you *run* something — a team, a product, a market — the checklist runs in reverse: which of the five conditions could you switch off before you need to? **Remember it as:** *don't lecture the crowd. Widen the exit, insure the last in line.* --- *Next: [[10-where-rules-come-from|Law 10 — rules grow where complaints can't become deals]] · Previously: [[08-how-trust-collapses|Law 8]] · Companion: [[19-who-edits-your-wants|Law 19 — why joining feels good]] · [[00-social-dynamics|Index]] · Coleman, ch. 9.*