C2
Work
The second of four kinds of harm, and the one where injuries are counted rather than modelled. Deaths and injuries at work, by country, year and industry, as the national systems that record them report them.
This is what makes injuries a real measure. For war, nobody counts the wounded and the figure has to be estimated; here a labour inspector, an insurer or an employer recorded each case, so the number is reported, not estimated. What it is NOT is complete: it covers only the workers a country’s reporting system reaches, which in much of the world means formal jobs and nothing more.
Is something on this page wrong? Report it →
The products this kind of harm reaches
Each product whose figure includes Work, with its deaths per item and the part of that figure this kind of harm accounts for.
Scenario Economic-embedded, using FIGARO 2023.
| Product | Deaths per item |
|---|---|
| Glass Part from here: 91% | 2.2 × 10⁻¹⁰ per item(0.93–4.5 × 10⁻¹⁰)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Aluminium Part from here: 46% | 4.8 × 10⁻¹⁰ per item(0.19–1.2 × 10⁻⁹)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Pack of cigarettes Part from here: 20% | 3.6 × 10⁻¹⁰ per item(1.4–8.8 × 10⁻¹⁰)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Butter Part from here: 11% | 3.2 × 10⁻⁷ per item(1.3–7.2 × 10⁻⁷)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Croissant Part from here: 11% | 0.42–2.4 × 10⁻⁸ per itemINDICATIVES-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Wheat flour Part from here: 11% | 1.1 × 10⁻⁸ per item(0.44–2.6 × 10⁻⁸)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Refined sugar Part from here: 10% | 6.1 × 10⁻⁹ per item(0.24–1.5 × 10⁻⁸)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Cotton fabric Part from here: 8.7% | 3.9 × 10⁻⁸ per item(1.5–9.5 × 10⁻⁸)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Cotton T-shirt Part from here: 8.7% | 6.4 × 10⁻⁹ per item(0.25–1.6 × 10⁻⁸)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Cotton yarn Part from here: 8.7% | 3.8 × 10⁻⁸ per item(1.5–9.3 × 10⁻⁸)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Denim fabric Part from here: 8.7% | 3.8 × 10⁻⁸ per item(1.5–9.8 × 10⁻⁸)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Jeans Part from here: 8.7% | 3.2 × 10⁻⁸ per item(1.2–8.0 × 10⁻⁸)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| NMC cathode powder Part from here: 7.4% | 1.4 × 10⁻⁷ per item(0.71–3.2 × 10⁻⁷)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Lithium-ion battery pack Part from here: 6.7% | 6.9 × 10⁻⁸ per item(0.36–1.4 × 10⁻⁷)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Lithium-ion cell Part from here: 6.7% | 8.4 × 10⁻⁸ per item(0.50–1.7 × 10⁻⁷)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Electric car Part from here: 5.6% | 3.0 × 10⁻⁵ per item(1.8–5.5 × 10⁻⁵)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Steel Part from here: 4.9% | 3.1 × 10⁻⁹ per item(1.7–6.7 × 10⁻⁹)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Passenger car tyre Part from here: 4.9% | 5.2 × 10⁻¹⁰ per item(0.27–1.1 × 10⁻⁹)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Coffee, one cup Part from here: 3.7% | 4.7 × 10⁻¹⁰ per item(0.19–1.2 × 10⁻⁹)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Roasted coffee Part from here: 3.7% | 6.7 × 10⁻⁸ per item(0.27–1.7 × 10⁻⁷)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Petrol car Part from here: 3.1% | 9.5 × 10⁻⁶ per item(0.55–1.7 × 10⁻⁵)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Chocolate bar Part from here: 2.1% | 2.5 × 10⁻⁹ per item(1.2–5.8 × 10⁻⁹)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Diamond ring Part from here: 2.0% | 9.2 × 10⁻⁶ per item(0.44–2.0 × 10⁻⁵)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Copper Part from here: 1.7% | 2.0 × 10⁻⁷ per item(0.81–4.2 × 10⁻⁷)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Laptop Part from here: 1.6% | 5.4 × 10⁻⁷ per item(0.21–1.2 × 10⁻⁶)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Smartphone Part from here: 1.6% | 3.0 × 10⁻⁸ per item(1.2–6.3 × 10⁻⁸)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Printed circuit board assembly Part from here: 1.5% | 1.6 × 10⁻⁶ per item(0.60–3.4 × 10⁻⁶)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Cocoa butter Part from here: 1.1% | 4.1 × 10⁻⁸ per item(0.16–1.1 × 10⁻⁷)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Cocoa mass Part from here: 1.1% | 4.1 × 10⁻⁸ per item(0.16–1.0 × 10⁻⁷)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
| Bag of charcoal Part from here: 0.48% | 1.5 × 10⁻⁷ per item(0.61–3.5 × 10⁻⁷)LOWS-MAX · run 2259 · methodology v1.0 · data dv-20261007-44r |
The figure is the product’s own, with its range. The share beside it is a middle estimate only: the calculation stores a range for each figure, not for each part of one.
What this kind of harm counts
This counts injuries caused by work — the fall, the machine, the collapse, the vehicle — as each country’s own labour statistics record them. Like war deaths, and unlike harm from poverty or pollution, each case is recorded when it happens, so nothing has to be estimated from a cause.
DISEASES caused by work are a different matter and are not in these figures. A cancer thirty years after working with a harmful substance has to be estimated, not counted, and that comes later. International estimates put work-related disease at about four times the deaths from injuries, so what is shown here is only part of a floor.
Reported by each country’s labour statistics. Left out: accidents on the way to work, where the country reports them separately, and any death already counted as a war death.
What the records hold today
23,730
Events in the records
1,348,184
Deaths recorded by the source
144
Countries
0
Wars
1970–2024
Years covered
These figures are the publisher’s own, shown with the credit its licence asks for. Nothing on this page has been linked to a product, a company or a material.
| Confidence of the death records | Records |
|---|---|
| HIGH | 53,863 |
| MEDIUM | 1,191 |
The causes it includes
| Cause | Estimated from its cause | Records |
|---|---|---|
| Occupational injury C2-INJURY |
no | 23,730 |
| Occupational disease C2-DISEASE |
yes | not counted here |
Which sectors carry this harm
Which industries this kind of harm is linked to, weighted by how much harm lies behind each share rather than averaged over countries.
100.0% of the deaths recorded for this kind of harm belong to a subject that has been attributed at all — that is, to a war or a report that has been linked to industries — 144 of 144.
| sector | share | Grade |
|---|---|---|
| Construction F |
22.3% | HIGH |
| Manufacturing C |
17.7% | HIGH |
| Transportation and storage H |
10.6% | HIGH |
| Agriculture, forestry and fishing A |
10.3% | HIGH |
| Other and unattributed economic activity OTHER |
9.4% | INDICATIVE |
| Wholesale and retail trade; repair of motor vehicles and motorcycles G |
6.7% | HIGH |
| Mining and quarrying B |
4.7% | HIGH |
| Public administration and defence; compulsory social security O |
2.9% | HIGH |
| Administrative and support service activities N |
2.8% | HIGH |
| Other service activities S |
2.0% | HIGH |
| Accommodation and food service activities I |
1.5% | HIGH |
| Electricity, gas, steam and air conditioning supply D |
1.3% | HIGH |
These are shares, not counts of people. The number of deaths each industry stands for is calculated but not published: it has no range yet and has not passed the test on past years.
Occupational attribution is a CONCORDANCE, not a judgement: the International Labour Organization reports deaths and injuries at work by industry, so the shares are the publisher’s own numbers divided by their total — nothing is chosen by us. What sits in "Other" is harm reported under a label that covers several industries at once — "market services" covers eight — which can only be split using how many people work in each.
What it does not capture
- A country that reports nothing is recorded as reporting nothing. It is never entered as a zero, because a zero would put the countries with the weakest labour inspection at the top of the safety table — the exact opposite of the truth.
- Reports arrive two to three years late almost everywhere, so the latest years are thin and will fill in later. A figure worked out today from the last reported year is not a figure about today.
- Countries define things differently. Some count accidents on the way to work and some do not; an injury must be reported after three days off work in one country and four in another. Each record keeps the definition it came with, and the differences are recorded rather than smoothed away.
- Only people inside the reporting system are covered, not everyone who works. Informal and small-scale work — where conditions are most dangerous — is mostly outside these statistics, exactly where it matters most.
- European figures appear twice, from two publishers, and are counted once. The international series takes Europe’s numbers from the European one, so the European records are loaded only for comparison and are never added to the total.
- A death here may also be counted under another kind of harm: a miner killed in an armed attack is both a death at work and a war death. No total for one kind of harm is added to another until the overlap is measured.
Sources and credit
| Version of the data | Downloaded | File fingerprint (SHA-256) |
|---|---|---|
| ilostat · fatal-20261006 | 2026-10-06 | 700fa580091e41e7… |
| ilostat · non_fatal-20261006 | 2026-10-06 | a5573994bf32a56a… |
| eurostat · esaw-fatal-20261006 | 2026-10-06 | 892fa9c813eed547… |
| eurostat · esaw-non_fatal-20261006 | 2026-10-06 | 934c233125af525b… |
Eurostat, European Commission — esaw-fatal-20261006, retrieved 2026-10-06 from https://ec.europa.eu/eurostat/api/dissemination/sdmx/2.1/data/hsw_n2_02?format=SDMX-CSV. Licensed under Creative Commons Attribution 4.0 International (https://creativecommons.org/licenses/by/4.0/). Data modified by HumanToll.Eurostat, European Commission — esaw-non_fatal-20261006, retrieved 2026-10-06 from https://ec.europa.eu/eurostat/api/dissemination/sdmx/2.1/data/hsw_n2_01?format=SDMX-CSV. Licensed under Creative Commons Attribution 4.0 International (https://creativecommons.org/licenses/by/4.0/). Data modified by HumanToll.International Labour Organization, ILOSTAT. The ILO did not prepare, review or endorse these derived figures; responsibility for them rests solely with Quiescence — fatal-20261006, retrieved 2026-10-06 from https://sdmx.ilo.org/rest/data/ILO,DF_INJ_FATL_ECO_NB,1.0/all. Licensed under Creative Commons Attribution 4.0 International (https://creativecommons.org/licenses/by/4.0/). Data modified by HumanToll.International Labour Organization, ILOSTAT. The ILO did not prepare, review or endorse these derived figures; responsibility for them rests solely with Quiescence — non_fatal-20261006, retrieved 2026-10-06 from https://sdmx.ilo.org/rest/data/ILO,DF_INJ_NFTL_ECO_NB,1.0/all. Licensed under Creative Commons Attribution 4.0 International (https://creativecommons.org/licenses/by/4.0/). Data modified by HumanToll.
Questions
Are these HumanToll’s numbers?
No. They are the publishers’ own, shown with the credit their licences ask for and with the exact version of the data file they came from. Nothing here has been linked to a product or a company.
Why are the two publishers not simply added together?
Because one copies the other. Europe’s figures reach the international series from the European one, so adding both would count every European injury twice. The European records are kept as a second opinion, not an addition — which also means any disagreement between them is visible.
Why is there no range on these numbers?
Because the publisher reports a count. The uncertainty in work statistics is not a matter of sampling; it is cases that were never reported, and allowing for those is a choice made by the model, not something in the record. It is a scenario setting, so you can see it and switch it off.
Does a work injury here mean a product is responsible for it?
Not yet, and not by itself. This page records that the injury happened, in which country, year and industry. Linking an industry to the things it makes is a separate step that needs evidence.