Every material Warconomy value carries two independent judgments: an epistemic class (what kind of value this is — observed, reconstructed, nowcast, forecast, scenario, claim, or unknown) and an evidence status (how well-established it is — verified, corroborated, reported, provisional, disputed, unverified-lead, or unknown). These are independent: an observed value can be disputed, a reconstructed value can be corroborated, and a forecast can never be 'verified' because it hasn't happened yet, but its methodology can be well-validated. This page explains each class; see /forecasts for real examples.
- A reconstructed value is never made visually indistinguishable from an observation.
- A forecast always shows its issue date and target horizon, and is preserved once issued rather than silently overwritten.
- A scenario is never presented as the expected outcome and carries no probability.
- An unknown value never renders as zero.
The seven epistemic classes
Directly reported, measured, published, or documented. An observed value can still be disputed by credible sources — the two are independent.
A historical estimate Warconomy produced to fill a missing period, geography, metric, or category — never made visually indistinguishable from an observation, and always documenting why the gap exists, why the method fits, and why alternatives were rejected.
An estimate of the present or very recent past where official data has not yet been released. Never labeled 'current official data.'
A model-based prediction of a defined future variable and horizon. Issued once and preserved — a later forecast supersedes an older one, but the original stays available so accuracy can be evaluated honestly.
A conditional projection based on explicitly stated assumptions — never presented as the most likely outcome, and never assigned a probability (that's what separates it from a forecast).
A statement or number attributed to a named party that Warconomy has not independently established. May later become corroborated and re-classified as observed once independently confirmed.
No responsible value or estimate is currently available. Never rendered as zero.
What this does not do
- It does not fabricate a missing historical value merely to complete a chart.
- It does not imply statistical precision the underlying data cannot support — see the uncertainty-interval methodology.
- It does not use a combatant's unsupported claim as a verified forecast input.
- It does not describe correlation as causation, or a scenario as a prediction.
- It does not claim real-time or live information anywhere on the site.
See it in use
The forecasts hubapplies this model to a real example of each modeled type, built from IOM DTM's published Sudan displacement figures. The human cost of conflict uses the related (domain-specific) evidence-status model for casualty and displacement figures.