How to read these numbers
A figure without a range is a claim. A figure with one shows how much is actually known — and on this site the range is not decoration: it decides whether a number may be published at all.
Every published figure comes with a range that the true value very probably lies in (nine times out of ten), and a grade saying how sure we are. Both are worked out by the calculation, never chosen. The software that shows a figure refuses to work unless it also has the calculation number, the scenario, the low and high ends of the range, the grade and the versions of the method and the data — so no page on this site can print a bare number, even by mistake.
The rule, and why it is built into the software
A result with no interval is not published.
The software that prints figures enforces that sentence: it stops with an error if any of the six things a reader needs is missing. A page cannot get around it, because there is no other way to print a number. A figure graded INDICATIVE is shown as a range with no central value at all — the model can say roughly how big it is, but has no best guess worth showing.
Is anything published yet? yes — every figure on the site comes from a published calculation.
What the latest calculation actually did
Run 2259, scenario S-MAX, year 2023, engine 0.30.0. The numbers below are that calculation's own settings, read from it — not a description of them.
- Repetitions — 2,000 repetitions per figure (monte_carlo_pcg64, starting value 20260921). Using the same starting value gives exactly the same range every time.
- The records of harm, deaths — spread 0.115 — measured, not assumed: the records of harm give a low and a high estimate for every event, and added up over a year the low and high totals are 1.46 times apart.
- The records of harm, injuries — spread 0.039 — measured, not assumed: the records of harm give a low and a high estimate for every event, and added up over a year the low and high totals are 1.14 times apart.
- How the economy is connected — spread 0.23 — estimated from published studies that compare different tables of how industries buy from each other, because this table cannot be checked against a second one here. See below.
What is varied, and where each uncertainty comes from
There are three kinds of uncertainty, and the difference matters. One is read from data that states its own uncertainty. One is estimated from published research. And one is a rule of thumb tied to a confidence grade, because nobody has measured it. The third kind is the first that should be replaced.
| Step | Kind | Where the uncertainty comes from |
|---|---|---|
| The records of harm | measured | War data gives a low, a best and a high estimate for every event; health data gives its own ranges. Added up over the year being calculated, that is the records' own range. |
| Which industry the harm belongs to | graded | Each industry's share of a war has a confidence grade, and the grade sets how far that share may move. |
| Which material an industry's harm belongs to | graded | How good the production data is. The materials of one industry are varied together, because they share one amount of harm between them. |
| How much of a material a product needs | graded | The grade on each line of the product's list of materials — a measured composition moves less than a national average applied to a typical product. |
| World production of a material | graded | Whether its publisher reported the figure, estimated it, or filled a gap. |
| How the world economy is connected | estimated from research | Kept as published, with one adjustment standing in for its uncertainty, estimated from studies that compare different tables of trade between industries. |
| The scenario | not varied | On purpose. The assumptions that change a figure most — how many deaths a war causes indirectly, whether harm is passed through the economy at all — are choices a reader can switch, not error margins. Five scenarios show how much they matter better than one wider range would. |
Inputs that rise and fall together are varied together. The overall level of harm in the records is varied once and shared by every material in every product, so a long supply chain cannot look more precise than it is by averaging. And the materials of one industry move against each other, because when one share rises another must fall. Varying them separately would make every range narrower than it should be — the dangerous direction to be wrong in.
The one honest shortcut
The table of how the world economy is connected is held fixed, and its uncertainty is added as a single adjustment.
Re-varying the whole table for every repetition is possible but very slow, and its effect would be small next to the uncertainty in the harm data. So the table is used as published, and one adjustment, estimated from research, stands in for how much a different but equally reasonable table would have changed the answer.
The check that would keep this honest is to run the same product through a second, independently built table and publish the difference. This site cannot do it: the licence of the obvious second table does not allow this project to use it. So the adjustment comes from published research alone, and this paragraph exists so that no reader assumes a cross-check that was never done.
Estimated from: Giljum et al. (2019), Journal of Industrial Ecology 23(4) 946-958, doi:10.1111/jiec.12833 — a 15% corridor holds 46% of countries; carbon-footprint convergence (Moran and Wood 2014) is tighter and is NOT used
What a grade means
A grade is the worst of three readings: how wide the range is, how many steps stand between a recorded death and this figure, and how good the inputs themselves were judged to be. A figure is only as good as its weakest part.
| Grade | Means | Figures in the latest calculation |
|---|---|---|
| HIGH | Main sources, few steps, and a range no wider than about 40% either side of the central figure. | 0 |
| MEDIUM | Linked by the model, three or four steps, and a range of up to about three times between its low and high ends. | 0 |
| LOW | Little data, or estimates standing in for data, more than four steps, and a range that spans ten times or more. | 0 |
| INDICATIVE | Shown as a range only. There is no central figure worth printing. | 0 |
No grade on this site is typed in by hand. Each is worked out from the figure's own range, number of steps and inputs — which is why most are LOW: the model knows how much it does not know.
What no range on this site covers
These are not inside any range here, because nobody has measured them well enough to put a number on them. They are listed because the uncertainties nobody lists are the ones that get found out later.
- Deaths a war causes indirectly — The largest single unknown. Estimates from three to fifteen times the deaths in fighting can all be defended, and the choice changes the headline more than anything else in the model — so it is a scenario you can switch, not a number hidden in an average.
- The main scenario itself — Passing all harm that can be linked to the economy through the economy is a position, not a measurement. The other scenarios show what changes if you take a different position.
- Informal, small-scale mining — Almost absent from official statistics, and exactly where war and work harm are worst. It is undercounted; we know in which direction, not by how much.
- Work injuries that go unreported — A serious problem, and worse in poorer countries. A scenario allows for it, rather than a made-up correction factor.
- Everything after a product is made — These figures cover production only — making things, not using them. Using a car — road deaths, exhaust — probably costs more lives than making it, and is not counted anywhere on this site yet. Every page that shows a figure says so.
- What the supply chains do not reach — A product's list of materials traces what it names. Where the chain stops at an input the model has no data for, that part carries no harm at all, and each product page says how much of the product it could trace.
Questions readers ask first
What does the range beside a figure mean?
The whole calculation is run 2,000 times, each time with the uncertain inputs varied within what is known about them. The range holds the middle nine in ten of those results. It shows what the model knows, not every way it could be wrong: harm it cannot link to a product is not in it at all, which is why every figure is also a floor.
Why only two meaningful digits?
Because the model is not more precise than that. A third digit would claim a precision the data cannot support, so the software that prints figures rounds them once, to two meaningful digits, in every language.
What does the confidence grade say?
How far to trust the figure, judged by the weakest of three things: how wide its range is, how many steps stand between the recorded harm and the product, and how good its inputs are. It is worked out, never chosen.
Why do some figures show only a range?
A figure graded INDICATIVE has a believable range but no central value worth quoting, so the site shows the range alone and will not print a single number for it.
Can two products be compared by their ranges?
Not by checking whether the ranges overlap: both come from the same records of harm, so they rise and fall together. The comparison page runs the two together and reports how many times larger one is than the other, with its own range, instead.