Scenarios
The model does not have one answer. It has five, and the difference between them is the sensitivity of the result to the assumption — which is more informative than any single figure.
A scenario is a row in the database, not a comment in the code. Every published number names the scenario that produced it, and every parameter that could change a number is stored as data so that a model run can be reproduced exactly and audited by someone who disagrees with it.
What scenarios may not do
No scenario deletes harm and no scenario renormalises. Every one of them reserves a sink sector — “Other” — and the shares sum to one including it. A scenario that declines to route a share moves it to the sink, where it stays counted and stays visible.
This matters more than it sounds. If 40% of a conflict’s attribution is evidenced and the rest were discarded and rescaled, a 30% oil share would become 75% — the model would make its strongest claim about oil exactly where its evidence is weakest. Holding the remainder in “Other” keeps every evidenced share at its true size and turns the gap into a number you can see.
renormalise: false — Stored on every scenario as renormalise: false, and asserted by the seed and by the test suite — so it is a fact about the data, not a promise in prose.
Economic-embedded S-MAX
The default. It is the project’s founding hypothesis, and the site leads with it rather than hedging about it.
Routes the entire attributable toll through the economic system, including the share no specific sector can be evidenced for. Quantifies the economic activity associated with conflict rather than the motivation of anyone fighting it.
| Propagates the “Other” sink into product footprints | yes |
|---|---|
| Requires an evidence link before a share reaches a named sector | no |
| Minimum confidence for a link to be propagated | LOW |
| Sectors eligible to receive a share | all |
| Attributes deprivation and exposure harm | yes |
| Applies the occupational under-reporting correction | 19.5–24.1 (×22.7) |
| Indirect conflict deaths, as a multiple of direct | 3–15 (×8) * |
| Sink sector | OTHER |
| Renormalises after withholding a share | no |
Evidence-weighted S-EVID
Only shares with a documented evidence link reach a named sector. The remainder is held in Other and is not propagated into products — it is still counted, still displayed, and never renormalised away.
| Propagates the “Other” sink into product footprints | no |
|---|---|
| Requires an evidence link before a share reaches a named sector | yes |
| Minimum confidence for a link to be propagated | LOW |
| Sectors eligible to receive a share | all |
| Attributes deprivation and exposure harm | yes |
| Applies the occupational under-reporting correction | 19.5–24.1 (×22.7) |
| Indirect conflict deaths, as a multiple of direct | 3–12 (×5) * |
| Sink sector | OTHER |
| Renormalises after withholding a share | no |
Resource-only S-RES
Only energy, minerals, agriculture, land, water and forestry are propagated. Shares belonging to any other sector move to Other rather than being deleted.
| Propagates the “Other” sink into product footprints | no |
|---|---|
| Requires an evidence link before a share reaches a named sector | yes |
| Minimum confidence for a link to be propagated | LOW |
| Sectors eligible to receive a share | ENERGY, MINING, AGRICULTURE, LAND, WATER, FORESTRY |
| Attributes deprivation and exposure harm | yes |
| Applies the occupational under-reporting correction | 19.5–24.1 (×22.7) |
| Indirect conflict deaths, as a multiple of direct | 3–12 (×5) * |
| Sink sector | OTHER |
| Renormalises after withholding a share | no |
Arms and resources S-ARMS
Resource-only plus the military-industrial sectors. Tests the founding structural point: that arms are a product inside the model, not an externality outside it.
| Propagates the “Other” sink into product footprints | no |
|---|---|
| Requires an evidence link before a share reaches a named sector | yes |
| Minimum confidence for a link to be propagated | LOW |
| Sectors eligible to receive a share | ENERGY, MINING, AGRICULTURE, LAND, WATER, FORESTRY, ARMS, MILITARY_PROCUREMENT |
| Attributes deprivation and exposure harm | yes |
| Applies the occupational under-reporting correction | 19.5–24.1 (×22.7) |
| Indirect conflict deaths, as a multiple of direct | 3–12 (×5) * |
| Sink sector | OTHER |
| Renormalises after withholding a share | no |
Conservative S-CONS
Only high-confidence causal links are propagated, deprivation and exposure harm is not attributed at all, and reported occupational figures are used without an under-reporting correction. Exists so that a reader who rejects the model's more contestable steps still has a number.
| Propagates the “Other” sink into product footprints | no |
|---|---|
| Requires an evidence link before a share reaches a named sector | yes |
| Minimum confidence for a link to be propagated | HIGH |
| Sectors eligible to receive a share | all |
| Attributes deprivation and exposure harm | no |
| Applies the occupational under-reporting correction | no |
| Indirect conflict deaths, as a multiple of direct | 1–3 (×1) * |
| Sink sector | OTHER |
| Renormalises after withholding a share | no |
Provisional parameters *
Some parameters are not yet evidence-backed and are marked as such on the scenario. Indirect conflict deaths are the largest example: published estimates range from three to fifteen times direct deaths and every point in that range has defensible support. It is carried as a scenario parameter with a wide interval rather than as a constant, and it moves the headline more than anything else in the model.
Questions
Why is the maximal scenario the default?
Because the project exists to quantify a hypothesis, not to hedge about it: that organised violence is embedded in economic activity. The model does not claim to establish anyone’s motivation — it quantifies the economic activity associated with conflict. A conflict that began in religious hatred still buys rifles, burns fuel, seizes land and sells what it digs up.
What if I reject that assumption?
Then read the conservative scenario, which is why it exists. It propagates only high-confidence links, does not attribute deprivation or exposure harm at all, and uses reported occupational figures without correction. It will give a much smaller number, and the difference between the two is the most honest thing the model can show you.
Can I compare figures from two different scenarios?
Compare them to each other, yes — that comparison is the point. Do not mix them in one total, and do not compare a figure from one scenario with a figure from another methodology version; the version is part of what the number means.