Recorded deaths in armed conflicts across 185 countries, from 1989 to 2025, by where the fighting occurred.
Source: Uppsala Conflict Data Program, via Our World in Data. Deaths are recorded by where the fighting occurred, not by the nationality of those killed. UCDP counts only deaths it can verify from reported sources, so these are documented deaths rather than total deaths — the true figures are higher, often substantially. The low and high estimates are UCDP's own uncertainty range, and the distance between them is a useful indicator of how contested the reporting is. Countries shown in the palest band recorded no conflict deaths that year, which is different from having no data.
This is the distinction that governs how the whole dataset should be read.
The Uppsala Conflict Data Program counts deaths it can verify from reported sources — news reports, NGO documentation, official records. Where reporting is thin, dangerous or actively suppressed, deaths go uncounted. The series is conservative by design, and the true figures are higher, sometimes by a wide margin.
Excess-mortality studies, which estimate deaths by comparing observed to expected mortality, routinely produce figures an order of magnitude above UCDP’s for the same conflict. Neither approach is wrong. They answer different questions: UCDP asks how many deaths can be documented, excess mortality asks how many people died who otherwise would not have.
For cross-country comparison over time, a consistent counting rule matters more than completeness, which is why UCDP is the academic standard. But a headline figure drawn from it should be described as recorded deaths, not deaths.
UCDP publishes three figures for each country and year: a best estimate, and low and high bounds. The hover panel shows all three.
The distance between the bounds is informative in its own right. Where sources agree, the range is narrow. Where the fighting is remote, the reporting contested, or the parties have reason to misstate casualties, it widens sharply.
Ukraine in 2024 records a best estimate of 91,779 deaths, within a range of 72,626 to 119,391 — a spread of nearly 47,000. Syria in the same year records 3,587 within a much tighter range of 3,539 to 3,633. Both are documented conflicts; the confidence in the counting differs substantially.
Reporting only the central figure discards that information.
Zero is not the same as no data. Countries recording no conflict deaths are shown in a distinct pale colour rather than the grey used for missing data. Most countries, in most years, record zero — that is the substantive finding, and greying it out would hide it.
Deaths are attributed by location, not nationality. A country hosting a conflict carries its deaths regardless of who is fighting or where the combatants came from. Foreign intervention, proxy conflict and cross-border fighting all register against the territory rather than the intervening power.
The rate view corrects for population. A thousand deaths means something different in a country of one million than in one of a hundred million. Switching to deaths per 100,000 reorders the map considerably, and is the fairer comparison for assessing the burden a society is carrying.
Annual figures fragment long conflicts. A war spanning a decade appears as ten separate years, none of which conveys its scale. Use the year buttons to trace a conflict rather than reading any single year as its total.
SDG 16 commits to peace, justice and strong institutions, and it is the Goal most often treated as separate from the environmental agenda.
It is not separate. Conflict destroys the institutional capacity that every other Goal depends on. A state fighting a war does not enforce emissions standards, maintain water infrastructure, or run a certification regime. The governance indicators on this site measure that capacity directly; this map shows what happens when it collapses.
The relationship runs the other way too. Resource scarcity, water stress and climate displacement are all documented contributors to conflict onset, though the causal pathways are contested and the literature cautions against simple environmental determinism.
Data from the Uppsala Conflict Data Program, via Our World in Data.