Civilian Deaths by Conflict: Which Wars Kill the Most Civilians?
Across this site's 29 theaters, UCDP recorded 187,033 deaths from organized violence in the twelve months to July 2026. About 30,000 were recorded as civilians. Where they fall is not where the headlines are: the highest civilian shares are in eastern DR Congo, Myanmar, Syria and Sudan, and the deadliest war by far, Ukraine, has the lowest.
Open the live conflict map →The short answer
Measured as a share of reported deaths, the wars hardest on civilians are eastern DR Congo, Myanmar, Syria and Sudan, where deaths recorded as civilian make up two-thirds or more of the toll. Measured in absolute numbers, the largest civilian counts in this window are Iran's 2026 war (7,182) and Sudan (6,802). The Russia–Ukraine war accounts for more than half of all reported deaths across the 29 theaters, 99,431 of 187,033, and has the lowest civilian share of any of them, under 3 percent.
Two caveats govern everything below. First, "civilian" here means recorded as civilian by UCDP: another 34,896 deaths, 19 percent of the total, could not be attributed to a side or to civilians at all, and in some theaters that unattributed share is so large that the civilian share is unknowable from this data. Second, these are direct deaths from organized violence, preliminary, and about a month behind. They are a floor, not a total.
How this was measured
Every figure comes from one source, the UCDP Candidate Events Dataset, the Uppsala Conflict Data Program's monthly first pass at coding each violent event. For each event UCDP records a best estimate of deaths, how many of those were civilians, and how many it could not attribute. We summed the twelve months to July 2026 for each theater's countries, using the same public theater-to-country mapping the homepage globe uses (see how the globe counts deaths). Two things follow from that mapping: UCDP codes Gaza and the West Bank under Israel, so that row is the Israel–Gaza theater as a whole; and the Sahel row sums Mali, Burkina Faso and Niger.
The civilian share is shown as a range, not a point. The low end is deaths recorded as civilian divided by all reported deaths. The high end adds every unattributed death, as if all of them were civilians. The truth is somewhere in between, and the width of the bar is itself information: it is a direct measure of how much a dataset built from press and NGO reporting could actually see.
Theaters with fewer than 100 reported deaths are left out of the ranking, because a share of a few dozen deaths is noise: Turkey (16 reported deaths) and Libya (34). Armenia–Azerbaijan, the South China Sea and the Taiwan Strait recorded no deaths from organized violence in the window, which is what "tensions" means on this site.
The ranking
| # | Theater | Reported deaths | Recorded civilian | Civilian share (range) | Unattributed |
|---|---|---|---|---|---|
| 1 | Eastern DR Congo (M23) | 3,799 | 2,765 | 73–82% | 9% |
| 2 | Myanmar civil war | 1,962 | 1,385 | 71–90% | 20% |
| 3 | Syria | 1,142 | 776 | 68–77% | 9% |
| 4 | Sudan civil war | 10,248 | 6,802 | 66–87% | 21% |
| 5 | Afghanistan | 772 | 473 | 61–72% | 10% |
| 6 | Mozambique (Cabo Delgado) | 431 | 216 | 50–66% | 16% |
| 7 | Haiti gang violence | 624 | 299 | 48–67% | 20% |
| 8 | Central African Republic | 277 | 92 | 33–69% | 36% |
| 9 | Iran (2026 war) | 23,660 | 7,182 | 30–95% | 64% |
| 10 | Nigeria (Boko Haram / ISWAP) | 8,393 | 2,457 | 29–47% | 18% |
| 11 | Yemen | 1,093 | 310 | 28–37% | 9% |
| 12 | South Sudan | 643 | 182 | 28–79% | 51% |
| 13 | Colombia | 1,875 | 448 | 24–72% | 48% |
| 14 | Sahel insurgency (Mali, Burkina Faso, Niger) | 5,819 | 1,180 | 20–29% | 8% |
| 15 | Kashmir (India) | 375 | 68 | 18–30% | 12% |
| 16 | Israel–Gaza (Israel, Gaza, West Bank) | 5,766 | 906 | 16–96% | 80% |
| 17 | Ecuador | 290 | 45 | 16–95% | 80% |
| 18 | Ethiopia | 4,294 | 581 | 14–26% | 12% |
| 19 | Pakistan (TTP / Baloch) | 4,564 | 504 | 11–14% | 3% |
| 20 | Mexico (cartel violence) | 2,944 | 216 | 7–97% | 89% |
| 21 | Iraq | 254 | 18 | 7–29% | 22% |
| 22 | Lebanon | 4,539 | 308 | 7–82% | 75% |
| 23 | Somalia (al-Shabaab) | 3,788 | 116 | 3–39% | 36% |
| 24 | Russia–Ukraine war | 99,431 | 2,763 | 3% | 0% |
Four patterns in the table
1. The wars where civilians are the target
Five theaters have a recorded civilian share above 60 percent even before any unattributed death is counted: eastern DR Congo (73 percent), Myanmar (71), Syria (68), Sudan (66) and Afghanistan (61). These are conflicts in which massacres, reprisals and attacks on villages, camps and minorities do most of the killing, not battles between armed formations. Sudan is the clearest case: 6,802 recorded civilian deaths in twelve months, more than the entire toll of any theater outside the top five. It is also not a coincidence that Sudan and Myanmar are the two theaters on the globe carrying a formal genocide determination (see the atrocity indicator). The two lists overlap, but they are not the same list, and the next pattern shows why.
2. The war that kills the most, and the fewest civilians
Ukraine is a war between armies along a front, and the data shows it: 99,431 reported deaths, 53 percent of everything in the table, and a civilian share of 2.8 percent that barely moves even if every unattributed death is added, because UCDP could attribute almost all of them. The absolute civilian toll, 2,763, is still the fourth-largest in the table, roughly equal to eastern DR Congo's. A low share is not a small number.
3. The theaters where the data cannot say
In five theaters the unattributed share is so large that the civilian share is effectively unmeasured. Mexico is the extreme: 90 percent of 2,944 deaths could not be attributed, so the civilian share is anywhere from 7 to 97 percent. Gaza is the case that matters most: 80 percent of 5,766 deaths are unattributed, largely because Gaza's health authorities report deaths without distinguishing combatants from civilians, so this data supports a range of 16 to 96 percent and nothing narrower. Ecuador (80 percent unattributed), Lebanon (75) and Iran (64) are the same problem. Note also that Gaza's twelve months run mostly under the ceasefire in place since October 2025, so 5,766 is a fraction of the war's toll, not a measure of it.
4. The theaters where even the total is a guess
UCDP publishes a high and a low estimate alongside its best one, and the spread between them is a second honesty signal, independent of attribution. For Ethiopia, the high estimate is 3.7 times the best estimate (15,825 against 4,294); for Afghanistan it is 3.4 times; for eastern DR Congo 2.9 times. Ukraine's high estimate is 2 percent above its best. Where the bracket is wide, reporting is thin, and thin reporting undercounts civilians before it undercounts soldiers, because soldiers have institutions that count them. Ethiopia's 13.5 percent civilian share should be read with that in mind.
What this does not show
- Indirect deaths. Hunger, disease and the collapse of health care are outside UCDP's definition entirely. In Sudan and Gaza, researchers' estimates of indirect deaths exceed the direct toll. Every share above is a share of direct deaths only.
- Revision. Candidate data is UCDP's first pass, revised in its annual release each June. A month's total, and its civilian count, can change after publication. This piece is dated for that reason and will not be silently updated.
- Who counts. A civilian death is recorded when a source reports one. Where the press cannot go, the recorded share falls even if the real one does not. The unattributed column and the high–low bracket are the best available measures of that gap, which is why both are shown rather than a single tidy number.
- The window. Twelve months is long enough to smooth a bad month and short enough to describe the current phase of a war. It is not a war's total: Gaza's and Syria's rows describe a year of ceasefire and transition, not the wars that preceded it.
How to cite this
Cite the dataset, the window and the measure: "UCDP Candidate Events, best estimate, twelve months to July 2026, via armedconflicts.org." Quote the range, not one end of it, for any theater with a large unattributed share. The per-theater files behind this piece are published as JSON at /data/fatalities/, and the live version of the same twelve-month window is on every theater's panel on the globe, where it rolls forward each time UCDP publishes a month.
Sources & further reading
- Uppsala Conflict Data Program, UCDP Candidate Events Dataset, monthly files for August 2025 to July 2026 (accessed 2026-09-02). Best estimate, civilian deaths, unattributed deaths, and high/low estimates per event.
- UCDP definitions of organized violence and battle-related deaths; the annual release that supersedes candidate data is described in Where ACLED and UCDP disagree.
- On this site: how the globe counts deaths, the atrocity indicator, and the deadliest conflicts of 2026 by total toll and displacement.
Related reading
Frequently asked questions
Which war kills the most civilians?
By share of reported deaths, eastern DR Congo (at least 73 percent civilian), Myanmar (71), Syria (68) and Sudan (66) in the twelve months to July 2026, according to UCDP candidate data. By absolute number in that window, Iran's 2026 war (7,182 recorded civilian deaths) and Sudan (6,802). The two measures answer different questions.
Why is the civilian share in Ukraine so low?
Because the Russia–Ukraine war is overwhelmingly fought between armies along a front. UCDP recorded 99,431 deaths in the twelve months to July 2026, of which 2,763 were civilians, about 2.8 percent. That is the lowest share of any theater, but the fourth-largest civilian count.
Why does Gaza show a range instead of a number?
Because 80 percent of the deaths UCDP recorded for Israel, Gaza and the West Bank in the window could not be attributed to combatants or civilians, largely because Gaza's health authorities report deaths without that distinction. The data supports a civilian share anywhere from 16 to 96 percent. The window also runs mostly under the ceasefire in place since October 2025, so the twelve-month total is a fraction of the war's toll.
Where does the civilian-deaths data come from?
From the UCDP Candidate Events Dataset, the Uppsala Conflict Data Program's preliminary monthly coding of violent events, summed over twelve months per theater using the same country mapping as this site's globe. It counts direct deaths from organized violence only, not hunger or disease, and is revised in UCDP's annual release.