.png)
Tigray War Death Toll: The Range That Became a Count (Cold Files 017)
At 1 a.m. on 4 November 2020, minutes before Abiy Ahmed announced a law enforcement operation in Tigray, the internet and the telephones went out. NetBlocks logged it the same day. Over the next two years a great many people died in northern Ethiopia, and every outside body with the standing to count them was kept out. What the world got instead was a number. It has been printed as 600,000, as 385,000 to 600,000, as half a million. None of those figures is a count. Two came from one researcher's worst case, hedged in the original and unhedged everywhere after. The third came from a mediator who gave no method at all.
Part of the Cold Files series.
The region was closed before anyone started counting
A death toll is a rate applied to a population over a period. Tigray's war destroyed all three inputs.
The population was guesswork. Ethiopia's last census was taken in 2007, when Tigray held 4,314,456 people. Wartime estimates of the region's population ran from roughly 5.69 million, used by OCHA in 2021, to 7,070,260, used by Belgian country-of-origin research in 2024. That spread of about 24 percent moves any rate-based total by nearly a quarter before a single death is counted.
The baseline rate did not exist. A pre-crisis review by ACAPS in February 2021 found no recent Tigray-specific information on mortality or life expectancy at all. The only crude death rate available was national, about 6.5 per thousand per year. There was nothing to measure a war against.
Then the region was sealed. The New York Times correspondent Simon Marks was driven to the airport and flown out on 20 May 2021 with no explanation given. On 30 September 2021 the federal government declared seven UN officials persona non grata and gave them 72 hours to leave: five from OCHA, the UNICEF representative in Ethiopia, and one from the UN human rights office. The foreign ministry's stated reason was meddling in the internal affairs of the country.
Aid went last. On 2 September 2021 the acting humanitarian coordinator for Ethiopia, Grant Leaity, published the arithmetic of the blockade in two consecutive sentences: "A minimum of 100 trucks of food, non-food items, and fuel must enter Tigray every day to sustain an adequate response. To date, and since 12 July, only 335 trucks have entered the region, or about 9 percent of the required 3,900 trucks." No trucks at all had entered since 22 August.
The counts that came from inside
The counting that did happen was done by Tigrayans, on paper, in the dark.
In January 2022 the head of the Tigray Regional Health Bureau, Hagos Godefay, published the results of a mortality assessment covering July to October 2021. It reached 333 tabias, about 40 percent of all tabias in the region, across 53 districts. It found 5,421 deaths. Infectious disease accounted for 32 percent, non-communicable disease 27.5 percent, and malnutrition or starvation 27.3 percent. Seventy-seven percent of the deaths had happened at home. Western Zone was wholly inaccessible and parts of three other zones were partly so. Godefay wrote that the survey "is bound to underreport the real extent of the crisis." He was describing his own data.
Why the deaths were at home is the second finding, and it is the more important one. A census of Tigray's health system published in BMJ Global Health in April 2022 assessed 1,007 facilities as of June 2021. Thirty-six, 3.6 percent, were fully functional. Ninety-one were recorded as not functional. And 820, 81.4 percent, were logged as "no communication," the paper's term for a facility under occupation by Eritrean or Amhara forces. All 741 of Tigray's health posts fell into that category. A place with no functioning health post generates no death certificate.
The machinery that remained ran on couriers. A 2025 study in Population Health Metrics reconstructed every airstrike casualty report filed in six accessible zones between June 2021 and October 2022, and noted that the regional health bureau's forms had to be carried between hospitals by hand because the telecommunications were down. It found 80 incidents, 1,143 casualties and 385 deaths, and warned the blackout risked underreporting in its own dataset.
The most careful count came after the shooting stopped. A verbal autopsy study published in BMC Public Health in December 2024 took nine randomly selected woredas and 53 displacement sites, had three physicians assign causes of death by consensus, and found starvation to be the leading cause: 1,329 deaths, 49.3 percent of the total. Nine woredas out of 76 accessible ones, sampled for budgetary reasons, projected to about 16,342 deaths.
Three categories, and no survey
Into that vacuum stepped a geographer.
Jan Nyssen, at the time a professor of geography at Ghent University, had worked in Tigray for decades. From March 2022 he and colleagues produced the only sustained attempt to put a number on the whole war. The method had three categories. Deaths from starvation were estimated by taking the IPC's count of people in famine conditions, about 400,000 in Phase 5 in mid-2021, and applying a mortality share. Deaths from violence were extrapolated upward from a documented floor. Deaths from healthcare collapse were calculated by raising the baseline crude death rate from about 6 per thousand to 20, and then to 32 under total blockade, and applying the result not to the population but to the 3.9 million people OCHA described as in need of medical care.
The March 2022 version gave half a million as a worst case. Nyssen noted the categories might double-count and declined to sum them. By a webinar in January 2023 the figure had a shape: an average of 518,000 civilian deaths for 4 November 2020 to 31 December 2022, a low of 311,000 and a high of 808,000, with roughly 60 percent attributed to famine, 30 percent to healthcare collapse and 10 percent to violence. Battlefield deaths were excluded entirely.
None of the totals was peer reviewed. The Ghent mortality work exists as blog posts, conference slides and repository deposits. His peer-reviewed Tigray work is on soils, erosion and landscape change.
Then, in May 2023, something the headlines never followed. Led by Emnet Negash, the same group revised its own total downward, to 162,000 to 378,000. The reason was specific. A study by the London School of Hygiene and Tropical Medicine with WHO and UNICEF had produced an excess-mortality figure for Somalia of roughly 43,000, against the group's own IPC-based projection for Somalia of at least 115,000. The ratio was about 38 percent. They applied it to their famine component, halved the healthcare one, and cut their per-location mortality assumption from 5 to 10 percent of local population down to 2 to 5 percent. A published model that revised itself downward still beats the silence of every body that could have counted.
The number that left its caveats
Because by then the figure had left its authors.
On 21 November 2022, USA TODAY published "383,000 to 600,000 civilians" killed, attributed to Nyssen's Ghent team. The range was Nyssen's own. An NRC interview he deposited on 22 October 2022 gives at least 384,000 and at most 600,000 civilian victims. It has circulated ever since, usually as 385,000 to 600,000, and usually footnoted to a newspaper.
Then came the mediator. In the Financial Times on 14 January 2023, Olusegun Obasanjo, the African Union's High Representative for the Horn of Africa and the man who had brokered the Pretoria agreement ten weeks earlier, said of the war: "The number of people killed was about 600,000." He gave no method and named no source. In the same article Tim Vanden Bempt of the Ghent group thought the figure roughly correct, and separately offered 300,000 to 400,000 civilians. Ethiopian officials in the same piece put the toll at 80,000 to 100,000. And Daniel Bekele, chief commissioner of Ethiopia's own Human Rights Commission, said: "We need to be cautious about overly exaggerated death toll estimates by all sides."
ACLED, which maintains the event-level record, wrote in February 2023 that "it is unknown which methods were used to produce the estimates," and, in the next sentence, that no known surveys or samples were provided as evidence for them.
The gap between the documented and the modelled is the whole problem. The group's own event-based count of the dead, the figure nearest to documentation, was 7,060 to 12,330, against ACLED's 5,325. Its headline estimate was 42 to 73 times larger.
One peer-reviewed estimate of the whole toll now exists. Published in Population Health Metrics in May 2025 by a team including the LSHTM crisis epidemiologists Francesco Checchi and Abdihamid Warsame, it surveyed the global Tigrayan diaspora, analysed 791 responses covering 10,725 individuals, and estimated 102,466 excess deaths among people aged 15 and over, with a 95 percent confidence interval running from 17,455 to 159,382. The authors call their own figure crude and think it may be too low. They also describe Ghent's multiplicative modelling of cause-specific projections as methodologically problematic.
And their composition differs sharply. Of 310 wartime deaths in their sample, 224, or 72.3 percent, were from war-related injury. Ghent put violence at 10 percent and famine at 60. The verbal autopsy study inside Tigray found starvation leading at 49.3 percent. Three datasets, three definitions: Ghent excludes battlefield deaths, McGowan does not, and the autopsy study counts 2023. The composition depends on who and when you count: a diaspora respondent remembers the cousin who was shot, and may never learn that a grandmother died of untreated hypertension in a village with no health post.
No one argues the toll was trivial. Ethiopian officials' 80,000 to 100,000 is still a catastrophe. The dispute is about whether the circulating numbers are knowable.
No outside body ever counted
No body with full access ever produced an estimate, which is why a geographer's spreadsheet became the world's figure. The International Commission of Human Rights Experts on Ethiopia, created by Human Rights Council resolution S-33/1 in December 2021, published no death toll at all. It could not: its findings paper records that it was "not able to visit Ethiopia for the purpose of in-country fact-finding," that it was unable to interview a single serving government official or senior military officer, and that the vast majority of its interviews were conducted remotely. Its mandate expired in December 2023 because no renewal resolution was tabled. No court or tribunal has ruled on the number of dead.
And then the instrument came back. Tigray's interim administration collapsed in May 2026 when the TPLF reinstated the pre-war regional council, and Chatham House wrote on 1 October 2026 that the Pretoria agreement had effectively unravelled. Fighting resumed on 23 September 2026, with the airports at Mekelle, Axum and Shire seized; Mekelle's was retaken. On the night of 24 to 25 September, Ethio Telecom services went down across northern Tigray while the smaller private network stayed up, and NetBlocks reported a sub-national shutdown. Sources told Reuters that this time the TPLF had ordered the blackout, possibly to disrupt government drones. The TPLF has not commented. Mekelle spent ten days cut off from the outside world, with each side blaming the other.
Before any of that, Tigray's own disaster commission reported in September 2026 that 3.9 million people, more than half the region's population, needed immediate emergency food assistance. That is an interested party's figure, and the only one published.
The first war produced a range that became a count because nobody could get in to do better. The second began the same way, the switch in the other hand.
Sources consulted
NetBlocks, Tigray internet and telephony disruption, 4 November 2020
ACAPS, pre-crisis situation in Tigray, secondary data review, 23 February 2021
Committee to Protect Journalists, Ethiopia press-freedom reporting, 2021 to 2022
UN News, seven UN officials declared persona non grata, 30 September 2021
Grant Leaity, OCHA, statement on the blockade of Tigray, 2 September 2021
IPC, Ethiopia acute food insecurity analysis, 22 July 2021
Hagos Godefay, Tigray health bureau mortality assessment, Ethiopia Insight, 26 January 2022
Gebregziabher et al., health system functionality, BMJ Global Health 7, 2022
Gebregziabher et al., airstrike death and injury, Population Health Metrics 23, 2025
Haileselassie et al., verbal autopsy study, BMC Public Health 24, December 2024
Jan Nyssen and Tim Vanden Bempt, mortality estimates, March 2022 and January 2023
Nyssen interview, NRC Handelsblad, October 2022, deposited 22 October 2022
Emnet Negash, Nyssen and Vanden Bempt, revised starvation assessment, May 2023
Kim Hjelmgaard et al., USA TODAY, 21 November 2022
David Pilling and Andres Schipani, Obasanjo interview, Financial Times, 14 January 2023
ACLED, Ethiopia Peace Observatory analysis, February 2023
McGowan, Checchi, Warsame et al., excess mortality, Population Health Metrics 23, May 2025
Commission of Human Rights Experts on Ethiopia, A/HRC/54/55 and A/HRC/54/CRP.3, 2023
Tigray Disaster Risk Management Commission, food security overview, September 2026
Soliman and Abate Demissie, Chatham House, 1 October 2026
BBC and Reuters, northern Tigray network shutdown, 25 to 26 September 2026
More Cold Files: Biafra War Death Toll: The Forecast That Became a Count and Okinawa Civilian Death Toll: The Four Counts That Do Not Add Up.
Next in Cold Files: East Timor, 1975 to 1999. A truth commission counted the dead properly, published a number, and was ignored by the people who had asked for it.
Project Leaflet publishes Cold Files twice a week, alongside the daily situation reports. Follow along at t.me/TheLeaflet.
Comments