The usual account of a town that accepts a bad deal is about judgement. The council was naive, or captured, or dazzled by a jobs number. That account locates the pathology in the decider.
WP-32 made the opposite move about households: prospect theory locates the pathology of selling in the seller's value function; this paper locates it in the seller's balance sheet. A household does not sell at the bottom because it misjudges the bottom. It sells because rent is due and there is nothing else to sell. The preference governs the quiet regime; the constraint governs the loud one.
Transpose that. A county has non-deferrable obligations too — debt service, pension contributions, a school roof that is already failing, a water system under consent order. It has a tax base it does not control and a revenue calendar it cannot move. When those bind, the county is not choosing between a data centre and no data centre. It is choosing between a data centre and a millage increase in an election year. That is not a preference over outcomes. It is a shadow price on cash, and it is the same λ WP-32 names.
If the problem is judgement, the remedy is information: better studies, expert testimony, a citizens' guide. If the problem is a constraint, information changes nothing, because the council already knows. Every campaign that arrives with a slide deck about water consumption is treating a loud regime as a quiet one.
WP-56's Theorem 1 is about what an outside observer can and cannot see. The regulator observes trades and prices; it does not observe λ. Without the latent state, “loss aversion” and “rational forced-sale behaviour” are mathematically indistinguishable — and that indistinguishability is not an inconvenience, it is the cover story. “The lender's defence — this is risk-based pricing — is precisely the observational equivalence of Theorem 1.”
In siting, the asymmetry is wider, because the constraint is published. A developer selecting a county reads its bond rating, its comprehensive annual financial report, its deferred-maintenance schedule, its school funding gap, its unfunded pension liability, the minutes in which it last discussed a levy. All of it is disclosed by law. The county's λ is a matter of public record, assembled by people whose profession is assembling it, and read by nobody on the other side of the table.
What the state sees afterwards is the agreement: an abatement schedule, a jobs commitment, a payment in lieu of taxes. It does not see the distress the terms were priced against, and it has no instrument that would make that distress an input. So the defence writes itself, and it is the same sentence with one noun changed: this is competitive siting. Every county is free to decline. That is true, and it is exactly as true as a borrower being free to decline a loan on the day the rent is due.
A forced sale requires a clock. In markets the clock is a margin call. In siting it is built by hand, and the components are standard:
None of these changes what the county wants. Each one shortens the interval between first hearing of it and having to answer. A public is not a standing object; it is something that has to assemble, and assembly takes weeks that are mostly spent finding out who else is worried. Compress the interval below that and the vote happens in a jurisdiction that contains no public — only a council and a proposal. That is a forced sale, executed on a rezoning.
WP-32's amplification is A = 1/(1−ρ): constrained sellers depress the price, the lower price tightens everyone's collateral, more sellers are forced. The crisis is the amplification; the shortfall is only its seed. And the unconstrained buyer accumulates at each turn — the concentration ratchet.
Regionally the coupling is more direct than in markets, because it runs through physical stock rather than through sentiment. Each completed project consumes interconnection capacity, so the next county's grid argument is weaker. Each water allocation sets an administrative precedent, so the next county's hydrologist is arguing against a decision already made elsewhere. And each abatement schedule becomes the comparable the next developer cites — this is the going rate. The county in round n+1 is negotiating against the outcome of round n, having had no standing in it.
Campaigns are organised county by county because that is where the vote is. But ρ is regional and the vote is local, so each campaign fights the seed and none fights the amplification. Winning in one county and losing in four leaves the fifth county worse off than before anyone organised at all.
The catastrophe reading in WP-44 and Disaster Theory says the useful thing about a fold is that it separates two regimes: before it, a perturbation decays back and small interventions have proportionate effects; after it, the system has branched and outcomes stop being a function of effort.
Siting has that structure and it is procedural rather than metaphorical. The sequence is a comprehensive-plan amendment, then a rezoning or special-use permit, then an interconnection agreement, then construction. Discretion is broad at the first step, narrow at the second, effectively spent by the third: once a use is permitted and a queue position is held, the remaining questions are conditions, not whether.
Organised opposition almost always forms after the second step, because the second step is the first one that is legible from outside — it has a hearing, a sign on the property, a date. Everything before it looks like planning staff doing planning. So the movement reliably arrives in the regime where its own effort no longer predicts the outcome, and then experiences exactly what that regime feels like: enormous energy, litigation, referenda, years, and no relation between input and result. The frustration is not a failure of will. It is a correct report of which side of the fold the work is being done on.
A municipality's λ is not a household's. Rent is due on the first; a council can defer, and sometimes does. The non-deferrability here is political and budgetary rather than contractual, and it is softer for that. The mechanism survives the weakening — a softer constraint still binds when the alternative is a levy in an election year — but anyone arguing this should concede the difference before it is pointed out.
No number crosses over. ε₀ = 1/3 is a calibrated quantity in the dm³ setting and WP-44 marks its calibration OPEN in three separate places. Nothing here borrows it. The fold in this paper is identified procedurally — a named vote on a published calendar — which is a weaker and checkable claim.
No case is counted. This is a mechanism, offered because it predicts things a judgement account does not: that better information will not change a constrained council's vote, that campaigns will feel disproportionate after the rezoning and not before, and that a regional win rate below one leaves the last county worse than no campaign at all. Those are falsifiable. None of them is tested here.
The remedy, and why it cannot be installed from outside the thing it repairs, is WP-118.