195 countries ship dangerous goods. Two of them count.

This is the first edition of an annual review. We checked the official statistical output of every sovereign country for a published, verifiable count of dangerous goods incidents in transport. We did not estimate, model or fill gaps. Where a country publishes nothing, the row is blank, and blank means unavailable, never zero.

The two countries with national multimodal totals are not comparable with each other. The United States recorded 27,200 incidents in 2024 and Canada recorded 299 (preliminary). That gap is a difference in what each country requires to be reported, not a difference in safety. Reading it any other way is the single most common mistake made with this data.

Only the European rail figures are genuinely comparable between countries, because Eurostat and the European Union Agency for Railways apply one shared definition of a significant accident. Across that group, accidents involving dangerous goods fell from 55 in 2010 to 29 in 2024, though not every country in the group reported in both years, so the change may partly reflect coverage rather than accidents.

Reported incidents rose 84%, from 14,755 to 27,200. Over the same fifteen years deaths fell from 8 to 1, injuries from 167 to 100, and incidents classed as serious from 483 to 247. Deaths are small enough to bounce year to year, so the steadier evidence is the serious incident count, which is built on hundreds rather than single figures and has roughly halved.

A shipment that is undeclared or misdeclared defeats every control downstream of the paperwork. The packaging, the segregation, the stowage and the emergency response all follow the declaration, so when the declaration is wrong the whole chain is working on the wrong information. This is the failure that documentation checks exist to catch, and it is the one getting worse.

Highway carried 24,439 of 27,200 reports, 90% of the total. Corrosives and flammable liquids are the two classes that appear most often. Because a single report can name more than one class, those figures describe how often a class turns up, not a count of separate incidents.

Method, and what this data cannot tell you

Every figure was taken from an official statistical publication and checked against the publishing authority. Nothing was estimated, modelled or interpolated. Where a country publishes no verified figure, the cell is empty.

Four rules we ask you to keep

• Blank means unavailable, never zero. A country with no published figure is a country with no published figure.
• Never compare counts between countries. Thresholds and definitions differ so widely that a higher count usually means broader reporting, not worse safety.
• Never add hazard classes together. One report can carry several, so the class figures overlap.
• A rising count is not a rising danger. Reporting behaviour and freight volume move the number as much as safety does.

Known limits of this edition

For the United States, 2024 is the latest complete year we could verify: the accessible primary record held only part of 2025 when we checked. A widely repeated figure of 27,355 incidents for 2025 could not be reproduced from a complete primary export, so we do not use it. Canada changed its reporting threshold in December 2016 and its most recent year, 2024, is preliminary, so its series is not suitable for trend analysis. Marking a country as having no data records what this review found in the sources it checked; it does not prove that no national or unpublished source exists.

Questions this data answers

Sources

Every figure traces to one of these. Licensed publications such as the IATA Dangerous Goods Regulations and the IMDG Code are named where relevant but never reproduced.

Last verified 2026-09-03. Reviewed by DGM One, part of the DGM network of more than 50 dangerous goods offices in more than 30 countries. Corrections are welcome: please contact your nearest office.