Death tolls in war-torn regions such as Beirut may have been even more grisly than thought. Credit: SAS / Alamy

A new study on violent deaths shows that we may have seriously underestimated the number of war deaths in the past half-century.

There have been 5.4 million violent war deaths between 1955 and 2002, says the study, a figure three times higher than previous estimates. This latest entry into the controversial area of war death tallies also contradicts the view that there has been a recent decline in the number of casualties. If anything, say the study's authors, the results are a conservative estimate.

The reason for the huge discrepancy, says study author Christopher Murray, is that the main source of information on war deaths has generally been ‘passive estimates’ from observers of the conflict. Writing in the BMJ1, Murray and his colleagues present a new estimate based on answers to surveys run by the World Health Organization (WHO) in 2002 and 2003.

“If you think of the nature of modern warfare, there’s an awful lot of dispersed conflict,” says Murray, director of the Institute for Health Metrics and Evaluation at the University of Washington in Seattle. “There isn’t a big set-piece battle where you would go and have someone count the bodies.”

The hidden dead

The surveys, carried out in 45 countries, asked one random person in a number of households about deaths among their siblings, and whether such deaths were due to war. When analysed by Murray and his colleagues, survey data combined with known population data indicates 5.4 million violent war deaths in 13 countries.

By far the biggest killer was the Vietnam War, one of the late twentieth century's defining conflicts, estimated to be responsible for 3.8 million deaths.

However, Murray says much of the previously unidentified mortality probably comes from smaller, underreported conflicts such as the 30,000 deaths estimated from the Philippines.

This is currently the only option available for reasonably estimating violent war deaths. Christopher Murray , University of Washington

Only a third of the total deaths picked up by the survey questions are indicated by previous methods relying on passive observations. Data from the Uppsala/PRIO database (produced by a collaboration between Uppsala University in Sweden and the Peace Research Institute in Oslo, Norway), which compiles passive reports on violent death, puts annual war deaths globally at 137,000 between 1985 and 1994, while Murray’s paper puts the number at nearly 378,000.

“There’s almost no reason to believe the passive surveillance strategies should work," says Murray, who describes his method as "a real, direct measurement" and "probably currently the only option available for reasonably estimating violent war deaths”.

Fog of war

There has been great controversy about war deaths, especially studies over the toll from the current war in Iraq (see: New estimate of Iraq death toll.)

Household surveys in Iraq have been criticized for potentially being open to manipulation, such as people reporting more deaths than actually occurred to score political points.

However, Murray says that his study shows household surveys can be used retrospectively, and appear to be reasonably accurate. The deaths described in the surveys tied up well with historical accounts.

In an editorial running alongside the paper, Richard Garfield, of Columbia University in New York, says there are some limitations with the new work, especially the small number of surveys analysed and the small samples in some of those surveys2.

However he also says, “This method eliminates some of the ambiguity rampant in this highly politicised field.”

Ensuring household surveys take a representative sample of the local population is a major challenge, agrees Gilbert Burnham, co-director of the Center for Refugee and Disaster Response at Johns Hopkins University in Baltimore, Maryland.

Still too low

Burnham, author of one of the Iraq mortality surveys3, says, “In general the passive reporting of war deaths clearly greatly underestimates the [deaths from] conflicts. Population based surveys are clearly an enormous advance. Population based isn’t a gold standard, but it’s what we need to use.”

Murray’s team and Garfield agree that, sadly, even Murray’s new estimate is likely to be lower than the actual number of deaths that occurred.

This underestimation is likely because households may choose not to report deaths, and if an entire family is killed there is no one to report. The surveys also account only for violent deaths; questions such as those asked by the WHO cannot account for all the indirect deaths that wars can cause, for example, by disrupting healthcare, or water and power supplies.