Two scientifically audited numbers today constitute the best available and most cited evidence quantifying Iraqi civilian deaths directly associated with the war in that country which began in March 2003. Each is generated by a credible and independent source, though their conclusions vary widely: one gives a running total of 48,783 (as of 18 October 2006), the other gives 654,965 for the period March 2003 to July 2006.

At this stage in the Iraq war, these different orders of magnitude for civilian casualties are too often relayed by a number-loving (and sensation-hungry) media in ways that both reflect and serve the preordained views of those in favour of or against the war. A statistical language about Iraqi casualties that is able to bring numbers and words, tallies and stories, into a coherent relationship requires understanding of what "48,783" and "654,965" are really measuring, how they were respectively computed, and what they reveal.

At a glance and for a lay audience, the contrast in these figures suggests that science must have gone wrong. The reaction of an intelligent sceptic might be to dismiss both figures and ask for additional homework. Moreover, any attempt to reconcile these sharply opposing tallies would only add to the statistical confusion and shift the political debate from the legitimacy of a conflict to the legitimacy of its metrics.

However, as the power of numbers rarely escapes political influence in the context of human conflict and suffering, it is necessary to elaborate a coherent statistical language that might help to answer the two essential questions raised by war: that of the jus ad bellum (whether entering into a war is justifiable), and the jus in bello (whether a war itself is conducted justly).

When thousands have already perished, it could be argued that it is in principle too late for the former (even though retrospective studies can still be done) and too technically challenging for the latter; and that as a result, to invoke numbers to support a case is morally irrelevant. Paul Farmer expresses it well in his 2005 Tanner lecture: "Where, in the midst of all of these numbers, is the human face of suffering? Can the reader discern the human faces in these reports?"

Indeed, the death toll that most makes sense to a Bosnian, a Rwandan Tutsi, or an Iraqi survivor is the "one" death of their beloved, not the 100,000, 800,000, or 45,000-650,000 deaths recorded in conflict statistical databases. Conversely, the heartrending experience of individual human beings offers no rationale to dismiss the value on mortality statistics for entire populations. True, when deaths begin to be numbered in the thousands there is a growing danger that they will be "anonymised"; yet large numbers tell a different kind of tragic story, one that belongs to humankind. It is critical to encounter and learn from these stories, as they illuminate the meaning and the unacceptability of all the individual deaths that they subsume.

Michel Thieren is a Belgian physician specialising in humanitarian affairs and human rights. He has spent more than a decade managing emergency operations in non-governmental organisations and the United Nations, and was head of office in northern Bosnia for the World Health Organisation in 1995-6. He works as a medical scientist at the statistical department of the World Health Organisation in Geneva

Michel Thieren writes here in his personal capacity only

Also by Michel Thieren in openDemocracy:

"There was genocide in Srebrenica…"
(11 July 2005)

"Katrina's triple failure" (7 September 2005)

"Kashmir: brothers in aid" (17 October 2005)

"Dayton plus ten: Europe interrogated" (24 November 2006) – with Louise Lambrichs

The Iraq Body Count tally

The Iraq Body Count (IBC) database is an ongoing human-security project that seeks to record "all civilian deaths in Iraq caused by military actions perpetrated by the coalition forces and by military or paramilitary responses to them" since coalition forces began their assault on 20 March 2003. More straightforwardly, IBC measures civilian deaths caused by the war.

At the time of writing, the total is calculated as somewhere between 43,937 and 48,783. A regularly updated spreadsheet tabulates the events one by one with time and geographical information indicating also which weapon caused the death (gunfire, roadside bomb, explosives, among others). As presented, the database allows for further computations, such as geographical distribution of deaths over time as well as their breakdown by cause. Although more calculation is needed, a quick review of IBC's statistical tables indicates a sharp escalation in civil casualties as a result of the war in recent months.

The IBC numbers derive from online media reports and eyewitness accounts. Each event is closely screened and verified. The project acknowledges possible "on-the-ground uncertainties and potential political bias" in the identification of fatal events, and in light of such uncertainties provides a maximum and a minimum count for each incident.

The design of the IBC project is admittedly imperfect. It is criticised above all for the fact that it does not attempt to itemise all civilian deaths directly associated with military or paramilitary action. The estimated current death toll is, then, almost certain to be lower than the total actual figure because the quality criteria applied to the IBC's reporting process may disqualify real events that caused civilian deaths. Additionally, in a situation of escalating violence, sporadic and single-death incidents tend not to be reported by the media, further affecting the true measure of bodies.

From a purely statistical perspective, however, actually counting deaths - painstakingly, one by one - constitutes the gold-standard for measuring mortality. The requirement for this to be done well is that no death escapes the statistical count and that all morgues and hospitals of Iraq communicate all fatal events that occur in their vicinity. The IBC initiative does not have that ambition and has never claimed it. Rather, Hamit Dardagan, its co-founder and principal researcher, readily concedes that many Iraqi civilian casualties are not included in the IBC tally.

The Johns Hopkins / Lancet tally

An eminent research team from the Johns Hopkins Bloomberg School of Public Health in Baltimore, United States estimates that as of July 2006, 654,965 Iraqis have died as a consequence of the war. Of these, 601,027 deaths are due to violence. The study specifies that people in Iraq are dying at a rate two-and-a-half times greater since the invasion occurred, and that violent deaths are primarily responsible for the increase.

A quick compression of some figures provided by the study (published in the London-based medical journal The Lancet) reports that the conflict is taking a dramatic turn. During 2003-04, 200 people were dying from violence every day; in 2004-05, the number rose to 500 per day, and in 2005-06, it reached the alarming rate of 900 violent casualties per day. The report indicates too that of the 601,027 violent deaths, 56% resulted from gunshots; this means that almost 500 people are executed every day in Iraq.

The study draws on a methodology (a good explication of which is here) that is commonly applied to measure mortality in areas of conflicts. In May-July 2006, two survey teams visited 1,849 households (covering about 12,000 people) spread across the various governorates of Iraq. In each household, a team gathered verbal information about pre-invasion and post-invasion deaths.

Two questions immediately arise in relation to the validity of the results of an approach like this: is the sample selected representative of the entire population of Iraq, and do the interviews omit any deaths that have happened or invent those that have not?

A blurry answer applies in both cases. To collect information on a door-to-door basis is difficult in highly violent settings. Insecurity interferes with the mobility and timing of the teams, reducing their capacity to visit a large number of households and the willingness of residents to open doors. As a result, the survey team's sample frame was the best possible compromise between the ability to represent every Iraqi and the feasibility of enforcing that statistical obligation on the ground.

It is hard to assess how objectively the interviewers were able to carry out their work. No one would expect a fully neutral interaction between two Iraqis, one answering the enquiries of a peer about the human cost of a conflict they both share. Under these circumstances, the teams certainly deserve some benefit of trust. (In this context, personal survey experiences in Bosnia and the former Zaire come as a genuine reminder: after stepping into a family home and being welcomed, mutual empathy and willingness to help prevail over suspicion and deception.)

The survey teams, anticipating possible bias, were requested to ask for death certificates. The proof of death corroborated 80% of the deaths accounted verbally: a quite convincing proportion. Overall, no major methodological flaw exists that could simply invalidate the study results.

Unlike mathematics, statistics is not an exact science. Even more, it is the science of error. "All surveys have potential for error and bias," admits the study report. In their final comments, the research authors address many possible biases one by one, acknowledging possible resulting fluctuations in the findings. That is why each total is reported within a margin called the "confidence interval". 601,027 excess violent deaths is called the "best guess" or the "best estimate" because it is the closest to the actual number of deaths ranging from 426,369 and 793,663. The same reasoning applies to 654,965, the estimated total number of Iraqis (within a range from about 400,000 to 950,000) who would not have died from any cause if the invasion had not occurred in March 2003.

The reconciled story

These numbers differ in absolute terms largely because they are generated through two very different approaches, with their respective flaws. Methodologically, a passive compilation of fatal events sporadically communicated by the press has very little in common with a data-gathering method that is equivalent to the one used for public-opinion polls.

The larger tally is a snapshot estimate of casualties in the whole Iraqi population. It arrived as a flash in the dark, electrifying everyone. Immediately after its publication, 654,965 triggered both the dismissal chorus of the men in charge in Washington, London and Baghdad and the encouraging verdict of leading statistical experts (one of whom told ABC News: "I'll vouch for it 95%, which is as good as it gets in survey research").

By contrast, the 48,783 appears steadily but silently, making its no less tragic case more visibly each time it passes a rounded benchmark. The unfortunate element of the smaller tally is that its figures quickly pale when suddenly confronted with a figure thirteen times bigger.

The story these two numbers convey does not lie and should not be understood to lie in their respective size. Regardless of their absolute divergence and methodological approximations, each figure, when properly analysed and broken down, reveals the inhumane and intolerable story of approximately 27 million people trapped in escalating violence and revenge killings.

More work is required to build the statistical language needed to fully analyse the ongoing jus in bello - how the war is conducted - in Iraq. In the absence of humanitarian organisations and independent civil-society observers, the military forces could play a critical role in assisting Iraqi officials' collectio of more mortality information. But (as Sarah Sewall has said) the military is concerned with what it values the most: military effectiveness. By disregarding the jus in bello, military forces tend to neglect looking at ways to improve their profession. This is especially regrettable considering that the jus ad bellum - why the war was fought - of Iraq was given a "humanitarian" label.

If launching a war in the name of helping people can ever be a legitimate claim, the proponents of the Iraq war have the duty to show more evidence that the war is actually helping the people of Iraq.

"Our key message is not to produce numbers but to ask in conflict situations, how can we think more effectively about protecting (a) population ... The general pattern ... in conflicts (is) that the host population takes more and more of the hits, which is what we are seeing here." With these words, Gilbert Burnham, the principal investigator of the mortality survey, gave his "numbers" the last word, and rendered them human.

For the Johns Hopkins study, a civilian dies in Iraq every three minutes from a war that has long perverted the meaning of both the jus ad bellum and the jus in bello. That civilian would not have died if the coalition had not come. For Iraq Body Count, the same story is told almost twice per hour. How much of a difference does it make? That is the political question these two statistics convey to the world, even if statistics themselves cannot answer it.