Data

Why airline accident death counts differ between sources

Ask three sources how many people died in the same accident and you can get three answers. Most of the time nobody is wrong: they are counting different things. Here is how it happens, with examples from official reports.

Updated 31 October 2026. Figures come from our accident statistics register of 171 accidents.

Five questions to ask about any figure

1. Does it count people on the ground?
2. Does it include people who died days or weeks later?
3. Is it one aircraft or two?
4. Does it count crew, or only passengers?
5. Does it come from the investigation report, or from a copy of a copy?

1. People on the ground

Investigation reports count people killed on the ground as deaths from the accident, but news summaries often quote only the people aboard. 17 of the accidents in our register include ground deaths.

2. Deaths after the day of the accident, and injury categories

Some people die of their injuries days or weeks later, and reports differ on whether to include them in the headline figure. Many apply a time limit or a separate injury category. United Airlines Flight 232 shows a different kind of gap: the account counts 112 deaths overall, while the report classifies 111 as fatal injuries. At Delta Flight 191 the total of 137 includes a motorist and two passengers who died later. In UPS Airlines Flight 2976 the NTSB main tables list 14 deaths, while including a victim who died 51 days later gives 15.

3. One accident, more than one aircraft

In a collision, the death count covers every aircraft involved, and the number of people aboard is a sum.

4. Who counts as an occupant

Crew, jump-seat occupants, security officers, infants and company engineers are all counted differently in different summaries. Our register uses the report's total for everyone aboard. A death can also be counted when the certified cause is medical: Singapore Airlines 321 recorded one death after severe turbulence, with the cause of death certified as heart failure and lung edema.

5. Copies of copies

Public data sets often copy figures without checking the report. We compared our register against the "number of deaths" field in Wikidata, the free public database that feeds many websites and assistants, on 16 October 2026. Two of its figures plainly disagree with the official report:

Wikidata is edited by volunteers and may be corrected after this page is published. A third difference is one of counting rules rather than an error: for Japan Airlines Flight 516 the database shows no deaths, which is true of the 379 people aboard the airliner, while the report's five deaths were the crew of the Coast Guard aircraft it collided with.

What we do

Every death count in the register comes from the investigation authority's own report and links to it. When the report itself counts something in a way readers might not expect, the row carries a note. The site build checks each figure against the account and against a second classification table, and stops if they disagree.

Open the full accident register or download it as CSV (credit AirlineIncidents.com, CC BY 4.0). Think a figure is wrong? Tell us with the page reference.