We should be counting the dead in Iraq, but let's not get the figures out of proportion like this

There are good reasons for thinking that the true figure is towards the lower end of 'The Lancet's' range

Have we really killed 100,000 people in Iraq? The short answer is No. Even The Lancet study, highlighted by The Independent, which produced the 100,000 figure, suggests that the number of Iraqis, military and civilian, killed by coalition forces is about 20,000. And, partly because that figure is an extrapolation from nine cases, it is subject to a large degree of uncertainty.

Not many people know that the estimate of the deaths directly attributable to coalition forces is one-fifth of the total "excess deaths" in post-invasion Iraq, or that it is based on just nine cases. Yet The Lancet's headline figure has been accepted as fact by those who opposed the war in Iraq, and shrugged off as unconvincing by those who supported it. Neither response is justified. Those who opposed the war ought to be much clearer about what the evidence suggests. Those who supported it need to be more critical of the failure by the US and British governments to afford the Iraqi dead the minimal respect of counting them.

It is therefore worth getting the facts about The Lancet article straight. Whether 194,000 have died as a result of the invasion, or 8,000, is terribly important. Some people were absolutely opposed to the invasion of Iraq on principle, the mere fact of Saddam Hussein's being a tyrant not being enough to justify for them the violation of the sovereignty of the Iraqi nation. Others - a rather smaller number - were prepared to pay any price to remove a dangerous regime. For most people, however, it must be suspected that where the real figure is within that range would make quite a difference to their view of the morality of invading Iraq.

What, then, does The Lancet article show? It is an attempt to compare death rates in pre-invasion and post-invasion Iraq, by carrying out a survey of random households across the country. From the deaths reported by interviewees, the researchers estimated that there were 98,000 more deaths in the 18 months during and after the war than would have been expected from the pre-war death rate. Hence the 100,000 figure.

However, this number is only the central point of a range that extends from 8,000 to 194,000. This huge disparity was mocked ignorantly by one American commentator as "not an estimate, it's a dartboard". It was also defended, equally ignorantly, by the editor of The Lancet, who said: "It's highly probable the figure is 98,000. Anything more or less is much less probable." Both wrong. What the figures say is that there is a 95 per cent chance that the true figure lies between 8,000 and 194,000. There is a 5 per cent chance that it is even lower or even higher. And the probability that any number is right gets lower the further away from 98,000 it is, up or down, but even so the spread of possibilities is unusually wide.

The reason for this is that the survey was carried out at only 32 locations across Iraq. Instead of choosing 990 households fully at random, interviews were carried out in clusters at 32 points chosen at random in different provinces. That was the only way such a survey was possible using seven interviewers. It is statistically respectable, which is why The Lancet article passed its peer reviews, but it produces estimates hedged about with great uncertainty.

And there are good reasons for thinking that the true figure is towards the lower end of The Lancet's range. The 98,000 figure is extrapolated from an excess of 44 deaths reported since the invasion. Of these, most - 23 - were not caused by violence. They were attributed to heart attacks, strokes, accidents, "neonatal and unexplained infant deaths", and infectious diseases. This is curious, considering the poor state of Iraqi health services before the invasion. Indeed, opponents of US policy before the war blamed sanctions for the suffering of the Iraqi people, and a figure of 500,000 avoidable infant deaths was often cited. Now, many of the same people imply that the citizens of Iraq were in rude health under Saddam, until the Americans came in and destroyed all their hospitals. Given the efforts made by the coalition to provide basic services - opinion polls in Iraq produce approval ratings for the state of the health and education services that Tony Blair could do with in this country - this seems unlikely.

More plausible, I would have thought, is that The Lancet study suffers from recall bias, a well-known phenomenon in such surveys where interviewees find it harder to remember accurately events that took place long ago. This would reduce the estimated pre-war death rate, and make it seem that there were more "excess" deaths after the invasion.

However, sticking for the moment with the 44 "excess" deaths reported in The Lancet study, just under half - 21 - were caused by violence. But most of these were not attributed to coalition forces. Seven were criminal, two were anti-coalition fighters, two were unknown and one was caused by Saddam's forces in the early days of the war. Which leaves the nine for which US, British or other allied forces were blamed.

This is where we get to the issue of responsibility. One of Tony Blair's less attractive evasions has been his pretence that the murderous activities of the insurgents in Iraq are nothing to do with him. It ought to be obvious that the suicide bombers and murderers of Iraqi police trainees would not be killing people in Iraq if the US and Britain had not invaded last year. The insurgency was not wholly foreseeable, but some people did foresee it, including the US National Intelligence Council which reported to the CIA in January 2003. It suggested that rogue elements from the overthrown regime could work with existing terrorist groups.

The Prime Minister and the US President ought to accept some responsibility for the murderous state of the country. Equally, however, their critics ought to accept a moral distinction between casualties inflicted by coalition troops and those inflicted by the insurgents. That is one of the weaknesses of the tally kept by Iraqbodycount.net, currently running at 15,000 to 17,000. Its policy is that "we record all civilian deaths attributed to our military intervention in Iraq". That means it does not include Iraqi military deaths during the invasion, which The Lancet study does. But it also means that it does not distinguish between deaths caused directly by coalition forces and those attributed to insurgents.

So far, however, it seems to me that the burden of explanation falls more heavily on the US and British governments. Jack Straw, the Foreign Secretary, published a polite response to The Lancet article nearly three weeks after it appeared. He refused to accept any estimates of Iraqi death tolls, but avoided putting forward any of his own. His real constraint is probably the fact that Britain's role in Iraq is subsidiary to that of the US, and the Bush administration is ruthless not just in refusing to make its own estimates but in preventing others from doing so.

For a coalition claiming to fight a just war, pledged to minimise civilian casualties, the minimum requirement for fulfilling that pledge is surely to know how many there have been.

j.rentoul@independent.co.uk

The writer is chief political commentator for 'The Independent on Sunday'