The Biggest Collateral Killings in History

Lists like this usually fail in the same way. They put a deliberate massacre and a mistaken airstrike in the same column, sort by body count, and leave the reader with the impression that these are all one thing. They are not, and the difference is the most useful thing an index of them can give you.

The distinction that has to come first

Collateral damage is a legal category, not a synonym for atrocity. It means harm to civilians that was not the object of an attack. A massacre is the deliberate killing of people known to be civilians. And a third kind sits between them, where an area rather than an object was the target. All three are on this page, and every entry says which it is.

Why keeping them apart matters

Collapsing the two does two kinds of damage at once. It flatters the deliberate killings, by wrapping them in a term that implies accident. And it libels the genuine mistakes, by implying intent nobody has shown. The full explanation of the term is here, and the three rules that decide lawfulness are here.

Collapsing the two flatters the deliberate killings and libels the genuine mistakes.Why every entry here says which category it belongs to

Deliberate killings of civilians

Attacks in which the city itself was the target

This column exists because the other two cannot hold these. Civilians were not incidental, so it is not collateral damage. Nobody was killed face to face, so it is not a massacre in the sense the first table uses. An area was the objective, and the people in it died because they were in it.

These are the largest entries in this index by a wide margin, and they are the hardest to judge by its standards, because the rules on distinction and proportionality were codified after them and partly in response to them.

Mistaken, negligent or misidentified strikes

The 2007 Baghdad airstrike this archive is built around sits between the two tables, which is the honest place for it. The first burst reads as a mistaken identification; the attack on the van does not.

The pattern across all of them

Three things recur in almost every case

1. The first account is wrong. My Lai was reported as a battle. The Kabul strike was called righteous. Haditha was reported as deaths from a roadside bomb.
2. It is outsiders who correct it. A reporter, a survivor, a news organisation, a forensic team. Rarely the institution involved.
3. The correction does not produce accountability. Establishing what happened and answering for it turn out to be separate problems.

What this index leaves out, and one thing it used to

An earlier version of this page excluded the atomic bombings and the firebombing raids, on the reasoning that they were policies rather than discrete incidents. That was wrong, and the correction is worth recording rather than quietly making.

Each of them has a date, an hour, an aircraft, an aiming point and a decision, which is the test applied to everything else here. They are also the largest civilian death tolls any single acts of war have produced. An index of the biggest that omitted the biggest was not defensible. They now have the third column.

What remains absent is whole campaigns rather than raids, sieges and blockades, and incidents where the record is too thin to write responsibly. There are many in the last group, and their absence is a fact about documentation rather than about what happened.

Why the death tolls are ranges

Almost every figure in the tables above is contested, and for reasons that repeat.

The party responsible counts one way, often classifying the dead by assessment at the time. Survivors and local records count another. An official inquiry may certify only those it can individually evidence, which is always fewer than died. At My Lai the US Army figure and the Vietnamese memorial list differ by more than 150 people; at No Gun Ri the certified count and the survivors' estimate differ by a factor of more than two.

None of that means a number is unknowable. It means a single confident figure is usually the least trustworthy thing on offer, and a range with its sources attached is the honest presentation. The mechanics of casualty counting are set out here.

• My Lai: 347 by US Army count, 504 on the memorial list.
• No Gun Ri: about 150 certified, around 400 by survivors' accounts.
• Minab: 156 named and identified, from an attendance estimate between 170 and 264.

Reading these collateral killings against one another

The value is in the comparison, not in any single entry. Read My Lai next to Kunduz and the difference between intent and error becomes concrete. Read Nisour Square next to Haditha and you can see what it takes to get a conviction, and how little it takes to undo one.

Read all ten together and the finding is the one this archive keeps arriving at: the evidence usually surfaces eventually, and it almost never reaches a courtroom. The names, where names exist, are kept here.