A War's Death Toll Is Three Different Numbers Wearing One Name
A list of named dead, a ministry's running total and a statistical estimate answer different questions, and each can be wrong in its own way. Here is how casualty figures are made, who checks them, and how to weigh the next one you see.
Every war ends up with a death toll, but the number in the headline is rarely one thing. It may be a list of named people, an official running total, or a statistical estimate, and each answers a different question.
The habit worth building is to ask, before anything else, what the figure counts and who produced it. A precise-looking total is not the same as a complete one. Everything below follows from that.
What is the difference between a count and an estimate?
Think of a hotel. You can count the guests who signed the register, or you can estimate how many people stayed last month from the water bill and the laundry. The first gives you names and rooms and misses anyone who slept in without signing.
The second reaches people the register never saw, but rests on assumptions about how much water a guest uses.
War tolls work the same way. An incident-based count records deaths that have been found and documented under defined rules. A population mortality estimate asks how many people died in a given population over a given period, including deaths that were never individually reported. Neither is automatically the definitive answer. They are built to answer different questions.
How is a documented list built?
The clearest public example of the first kind is Iraq Body Count, a project that records violent deaths in Iraq. Its detailed database is limited to civilians under its own stated definition. It draws on cross-checked media reporting and reviews hospital, morgue, NGO and official records. [1]
Its figures, it says, are documented deaths, not statistical estimates, and it revises them when new information arrives. [1]
For each incident, the project extracts the date, location, number killed and injured, the weapons used, the sources and, when available, the victims' names and demographic details. It says those details help prevent double counting and let it tell one incident from another. [2]
When reports of the same incident conflict and the gap cannot be resolved, it does not pick a winner. It publishes a low-high range based on figures from independent sources, and it uses ranges for some possible overlaps between entries and for disputes over whether the dead were civilians. [2]
Every new entry is checked by at least two additional volunteer data checkers before it is published, and old entries can be revised later. [2]
A range or a revision can look like weakness. It is the opposite. A recorder that shows its uncertainty is telling you where the edges are.
How do official running totals work?
The second kind is the running total, kept by a ministry or an international body. The United Nations Office for the Coordination of Humanitarian Affairs, known as OCHA, keeps a Protection of Civilians database for the occupied Palestinian territory. Its field staff collect the casualty information, which is then reviewed and verified. The general rule is to validate an incident with at least two independent and reliable sources, with stated exceptions. [4]
The database also has inclusion rules. It covers specified casualties linked to the occupation and conflict and excludes some kinds of incidents. It defines civilian status by whether a person is a member of the security forces or has a combat function within an armed group. OCHA says that classification does not itself imply whether a killing was lawful. [4]
Note what that means. Who counts as a civilian is a definition, written down in advance, and a different definition produces a different total from the same bodies.
OCHA also draws a line that readers rarely see. Casualties from the hostilities in Gaza and Israel that began on October 7, 2023, enter its database only after independent verification. Until then, the reported figures appear in its humanitarian updates attributed to whoever supplied them. [4] That is the right instinct. A number with its author's name attached is a claim, and the label tells you so.
What happens when nobody can count?
The third kind is the estimate, and it exists because counting breaks down exactly where wars are worst.
The cleanest mortality data usually comes from complete civil registration, the system by which births and deaths are recorded as they happen. But many countries lack complete registration even in peacetime, and conflict can disrupt whatever exists. [5]
The fallbacks each have a flaw. Facility-based records, from hospitals and clinics, miss deaths that occur outside health facilities. Media-based counts tend to undercount and to favour deaths that are especially newsworthy. [5]
Researchers can fall back on retrospective household surveys or censuses, which ask people about deaths in their families.
In conflict zones, surveys face sampling problems, because it is hard to reach a fair cross-section of a population that is displaced or under fire. They can also be skewed by three kinds of bias. Recall bias means people misremember when or how someone died. Reporting bias means people leave out or shade what they say. Survival bias means the survey hears only from people who lived to be asked, so a household wiped out entirely leaves no one to report its dead. Censuses may be delayed or disrupted, and migration complicates them. [5]
There is also the question of which deaths belong in the total. Conflict mortality includes direct deaths associated with combat and indirect deaths that follow damage to health care, transport and other infrastructure. [5] A figure for violent deaths and a figure for all excess deaths from a war, meaning deaths above what would normally be expected in that population and period, including those from disease or hunger, can differ widely, and both can be honest. They are answering different questions. Measuring the whole picture takes multiple sources and complementary methods. [5]
Who has a reason to shade the number?
Everyone with a stake. Parties to a war have reasons to emphasize, minimize or reclassify deaths in ways that serve their case. That is a reason to attribute and test their numbers, not to wave away every figure from a government or an institution by reflex.
A figure can come from an interested party and still be accurate. The question is whether the method behind it can be seen and checked.
The record is not neutral at the other end either. A media count rewards the dramatic incident. A hospital list rewards the body that reaches a hospital. Even a careful volunteer project works from what reporters and records happened to capture.
What does a contested toll look like in practice?
Gaza is the live case. A BBC report published on April 22, 2025, laid out how the territory's health ministry builds its lists. The method initially relied on bodies brought to hospitals and later added online forms submitted by relatives. [3]
The ministry says thousands of bodies remain under rubble. It also said about 900 bodies were unidentified and not included in its list at that time. [3] Those people are dead, and they do not appear in the total.
The lists have also been revised. According to Gaza health official Zaher al-Wahidi, the ministry revised and verified its lists in what he called a "revision and verification process." The BBC reported that, according to the official, more than 3,000 names were removed from the lists in August and October 2024 while checks continued. [3]
Israeli officials have challenged the accuracy of the fatality figures. The Gaza health ministry denies manipulating them. [3] The BBC said it could not independently verify either side's figures, because Israel blocked international journalists from entering Gaza independently. [3]
So here is what that one example shows. The reporting system changed under pressure. Online reports entered the process. Names were removed pending checks. Bodies remain unidentified.
Two parties dispute the result, and the outside observers who would normally test it have been kept out. Each point is a reason to read the figure as a claim with a method, not as a settled fact. None of them, taken alone, shows the number is wrong.
Why do the missing matter more than the argument?
The debate over a toll can sound like a contest of statistics. For a family, it is a question about one person.
The International Committee of the Red Cross says people go missing in armed conflicts, and that parties have obligations to take measures to establish the fate of missing people and inform their families. It says deaths should be registered and information kept about burials and human remains. [6]
That is the real purpose of a careful record. A total is a summary, and the summary exists because someone, somewhere, is owed an answer about a name.
How should you read the next number?
Five questions will carry you through most headlines.
First, who made the number? A named ministry, a UN agency, a volunteer project and a survey team have different incentives and different access.
Second, what exactly does it include? Civilians only or everyone, direct violence only or indirect deaths too, and what date does the count close?
Third, is it a list, a confirmed count or an estimate? A list has names, a confirmed count has a verification rule, and an estimate has assumptions. The wording in the headline rarely tells you which one you have.
Fourth, what checks stand behind it? Two independent sources, a second pair of reviewers, a published range, or nothing you can see?
Fifth, what is left out, and what is still unknown? The bodies under rubble, the unidentified, the district no one could reach.
A number that survives those questions is not necessarily the final one. Good counts get revised, and the ones that never change should worry you more than the ones that do. The honest figure is the one that tells you how it was made, and the best ones tell you where it might be wrong.
Every edition in brief, three times a day, on our Telegram channel, on Bluesky and on Threads.
Spotted an error? Tell the editors
- About IBC :: Iraq Body Count iraqbodycount.org
- Methods :: Iraq Body Count iraqbodycount.org
- Gaza health ministry denies manipulating death toll figures bbc.com
- Data on casualties | United Nations Office for the Coordination of Humanitarian Affairs - Occupied Palestinian Territory ochaopt.org
- Global Health Data Methods: Systems to estimate mortality in conflict globalhealthdata.org
- Protected persons: Missing persons | ICRC icrc.org




