AMERICAN EMPIRE CRISIS ASSESSMENT: CORRECTED EDITION

Supersedes: American Empire Systemic Collapse: Multi-Vector Crisis Intelligence Assessment, 13 January 2026. Status of this document: correction and re-calibration. The earlier text remains published unaltered. Corrections are worth nothing if the original is quietly edited.

The January assessment made a claim of systemic imperial stress across eight vectors and attached to it a confidence of 85 to 95 percent. On review, the direction of that assessment is defensible and the confidence figure is not. The error is not political and it is not a matter of taste. It is that a conjunction of eight claims was assigned the confidence appropriate to a single claim, and the arithmetic correcting this takes one line.

That is the substance of this correction. What follows also fixes a second, quieter defect: the original document contained numbers that a reader cannot reconstruct or check, presented in the same typeface as numbers that can be. Those are now marked.

Correction 1: compound claims do not inherit component confidence

The January core judgment was that eight named failures constitute a single pattern. Treated honestly, that is a conjunction: the pattern claim holds only if the component claims hold together.

If eight independent sub-claims each hold with probability p, the conjunction holds with probability p to the power 8.

Read the third row. To justify the published 85% floor, every one of the eight indicators must independently hold at 98%. Not one of the eight in the January document was sourced to that standard, and three were not sourced at all.

The immediate objection to this table is correct and I will make it myself, because a correction that only strengthens its author’s next position is not a correction. The eight indicators are not independent. Fiscal strain, alliance strain and currency reserve behaviour share drivers, and under perfect correlation the conjunction collapses to the confidence of a single claim, 90%, not 43%. The true value sits between 43.0% and 90%, and where it sits depends on a correlation structure the original document never specified and this one cannot supply either.

That is the honest end state: the correct answer is a range whose width is set by an unmeasured quantity. It is not 85 to 95 percent. Publishing a narrow band across a compound claim with unmodelled dependence is false precision, and false precision in an assessment is worse than an admitted range, because it survives being quoted.

Corrected confidence in the January core judgment: 55 to 80 percent, dependence structure unspecified.

Correction 2: the indicator audit

The January executive summary listed eight indicators, four of which carried specific figures. Those figures were not accompanied by an accessed source. They are marked below at the level they actually stand at.

Five of eight indicators carry figures or forecasts that a reader cannot check from the document. That is the actual finding of this review, and it is a finding about the analyst, not about the subject.

Correction 3: a forecast is not an indicator

Indicator 6 projected 15 to 20 percent unemployment by Q4 2027 and was then counted among the present-tense evidence for the pattern. This is circular: the forecast was produced by the model of decline, and was then used to support the model of decline. Remove it and the evidence base drops from eight items to seven, which by the table in Correction 1 moves the 98%-each conjunction from 85.1% to 0.98^7 = 86.8%. The direction of that shift, upward, is itself the tell. Removing a bad input improved the number, which means the input was doing work it had not earned.

Indicator 6 is withdrawn from the evidence set and retained only as a forecast, where it belongs, and where it is subject to the tests in the next section.

Correction 4: the scenario space was not exhaustive

The January document assigned 65% to chaotic stumbling decline, 20% to managed transition, and 15% to catastrophic collapse. Those sum to 100%, which asserts that the three named scenarios exhaust the possibilities. They do not. Absent from the set: a period of consolidation and renewed capability, which has occurred in several of the historical cases the original cited as precedents, and outcomes that do not fit any of the three descriptions closely enough to be scored.

Corrected distribution, with the missing bucket restored:

The 20 percentage points in the fourth row did not come from new evidence. They came from the three named scenarios, which had been assigned mass that belonged to a possibility the author had not written down. This is the most common failure in scenario work and it is invisible unless someone insists the rows be exhaustive before they are scored.

Correction 5: falsification criteria, which were missing entirely

The January assessment stated an accuracy record of 73% within confidence intervals but specified no way for a reader to score the present forecast. A forecast without a named observable and a date is not a forecast. The following are the tests this corrected assessment stands or falls on.

• If, by 31 December 2027, measured unemployment attributable to automation displacement is below 8%, indicator 6 is falsified and the decline model loses its labour-market leg.
• If allied defence procurement decisions through 2027 continue to be routed primarily through US-integrated programmes rather than substituted, the alliance-fracture reading is falsified.
• If reserve-currency composition shows no continued directional shift over a rolling eight-quarter window ending Q4 2027, indicator 4 is falsified.
• If two or more of the above fail, the corrected 52% modal case should be reduced below 40% and this document should itself be superseded.

Earliest observable signal that this corrected assessment is wrong: allied procurement substitution stalling while reserve composition flattens, both visible in routine quarterly data well before 2027, and both checkable by a reader without special access.

Consequence

Three things follow, and only the third is about the United States.

First, the January assessment should continue to be read, with its confidence figure struck through and this document attached. Withdrawing it would be tidier and would destroy the only useful artefact a forecaster produces, which is a record of being wrong in a specific way on a specific date.

Second, the defect corrected here is generic. Any assessment that lists n indicators, calls them a pattern, and attaches a single high confidence to the pattern is making the same error, and the reader can test it in one step by asking what p to the power n would have to be.

Third, the substantive judgment survives in weakened form: the modal expectation remains prolonged, uneven decline rather than either collapse or restoration, at 52% rather than 65%, over a horizon of 5 to 15 years, with a fifth of the probability mass now explicitly reserved for outcomes nobody in this document has described.

An analyst’s confidence is not evidence. It is the last thing that should be published and the first thing that should be corrected.

Related Documents

• American Empire Systemic Collapse, original assessment
• USA Recovery Policy Framework, 9SENS
• American Imperial Trajectory Assessment