A damned statistic: how a Johns Hopkins team got to 654,965.

THE respected statisticians at the Johns Hopkins Bloomberg School of Public Health have once again proved the truth of the old aphorism "There are lies, damned lies, and statistics." In their second study estimating the death rate in Iraq, recently published in the British medical journal Lancet, they say that 654,965 (give or take a quarter million) deaths since 2003 can be directly attributed to the coalition invasion. This is a whopping fivefold increase over the estimate in their first study, released in 2004.

If anything in this world is certain, it is that from now until the end of time every critic of American policy in Iraq will repeat this number. But it simply cannot be true. After the current report's predecessor was released in 2004, Fred Kaplan at the online magazine Slate made a solitary attempt to expose its problems. I will not belabor his arguments concerning its methodological flaws, which were repeated in the more recent study. Rather, I will address only a few of the Lancet report's largest problems.

First, its results are implausibly out of line with every other estimate of the number of deaths associated with the invasion, occupation, and insurgency. Of these estimates, that of the Iraqi Body Count Project (IBCP) is the most comprehensive, and is generally considered the most accurate. It places the war's death toll around 48,800, or about one fifteenth of what the Lancet study reports. The Lancet authors dismiss this discrepancy by blithely claiming that the media (the main source of the IBCP numbers) have historically underreported deaths in violent areas. But is it really credible that the most intensely focused media gaze in the history of warfare overlooked more than 600,000 deaths? These supposed deaths would have been concentrated overwhelmingly among men of military age, amounting to the loss of an entire generation of Iraqi manhood--carnage on a scale not seen since World War I. But maybe thousands of journalists just missed it.

A good rule of thumb is that for every person killed in a conflict, another three or four are wounded. If the Lancet estimate is right, where are these millions of people? Are they secretly bleeding to death in hidden corners? How come the interviewers who gathered data for the study did not notice the wounded and maimed populating the households they surveyed?

Which leads to another question: Even if we grant that the directors of the study are honest and reliable, what reason do we have to believe the same of the Iraqis they employed to conduct their surveys? It's far from inconceivable that some would lie to make the U.S. look bad. Remember that this is the Middle East, where media outlets try to outdo one another in reporting the highest number of deaths, and are expert at detecting Western and Israeli malfeasance where none exists. This is what produced reports of the Jenin "massacre" that never happened.

But even if the media have underreported the number of Iraqi deaths, as the Lancet study claims, they would still pick up an increase in this number. And if the extent to which the media underreport deaths is consistent over time--which it presumably would be--then their reports should provide an accurate gauge of the rate of increase. Accordingly, the ratio of deaths estimated by the IBCP (which bases its numbers on media reports) to deaths estimated by the Lancet team should have remained approximately the same over the past two years, if indeed the Lancet numbers are accurate. Yet the IBCP has deaths rising by a bit over a thousand each year over the past two years--a growth rate of less than 10 percent--while the authors of the Lancet study report an increase of 550,000, or 500 percent. There is no plausible way of accounting for this discrepancy without concluding that the Lancet methodology is seriously flawed.

Then there is a problem with the pre-invasion Iraqi death rate that the Lancet study assumes. Its estimate of the total number of deaths depends on comparing the supposed pre-war Iraqi death rate with the supposed post-war Iraqi death rate; get the pre-war rate too low, and you are overestimating the war's victims. Yet this is just what the study does. It puts the pre-war death rate at 5.5 out of every 1,000 Iraqis. That number is almost certainly wrong. Incredibly, it is considerably lower than the death rate in the United States. Death rates, particularly infant mortality, correlate closely with wealth. The richer a society, the lower its infant-mortality rate and the longer its average life expectancy. International sanctions on Iraq prior to the coalition invasion had reduced the country's per capita GDP to the level of sub-Saharan Africa's. Even if the accompanying degradation of Iraq's public-health infrastructure was gradual, the pre-invasion death rate must have been well in excess of twelve per thousand. Furthermore, current Iraqi per capita GDP is about $3,500, and the average death rate for other countries with similar amounts of wealth is nine out of a thousand.

If Iraq's GDP is used to provide a more realistic estimate of the pre-war death rate, 600,000 of the study's estimated deaths are erased. The number of deaths left over is close to the number given by the IBCP, whose estimate is looking more reliable all the time. And if we apply to pre-war Iraq the death rates of Pakistan and Syria--two countries in the same region, with socioeconomic conditions comparable to those of Iraq--we find that more than 100,000 additional predicted Iraqi deaths disappear.

The World Health Organization (WHO) says that Iraq's pre-war child-mortality rate was over 130 per 1,000 births, and that the average life expectancy was 51 years. These numbers are more than twice the pre-war death rate assumed in the Lancet study. Today, the WHO puts the child-mortality rate at 35 in 1,000 and the life expectancy at 69 years. This improvement is hardly surprising, given that the United States has pumped billions of dollars into Iraq's health system, and that these funds have helped build or repair over 60 hospitals and clinics throughout the country. If the WTO's numbers are right, then the invasion has actually saved lives--about 2 million of them. The statisticians who compiled the Lancet study should be pressed either to prove that the WHO's numbers are wrong, or to explain how a nation's death rate can double while its life expectancy grows by 18 years.

The reality of the situation is that the Iraqi death rate, far from doubling since the invasion, has probably been cut in half. To believe that the Lancet's estimate of the pre-invasion death rate is correct requires you also to believe that the sanctions imposed on Iraq following the Gulf War had no effect on public health. Of course, that requires throwing out a much-hyped pre-war U.N. study that said sanctions were ruining the health of Iraqis and had killed over 500,000 children.

The biggest problem with the Lancet's pre-invasion death rate, however, is the time it covers. It looks only at the 14 months preceding the coalition invasion, a period when Saddam was on his best behavior in order to get U.N. sanctions lifted. Consequently, it ignores much of the death and misery that were a staple of his regime. It fails to record, for example, the 300,000 Iraqis (or, in some estimates, up to a million) who were murdered by the regime and dumped in mass graves. Even if the study had looked back ten years, though, it still would have missed many of these victims. The study's research teams visited households and asked about deaths in the family; but since Saddam regularly wiped out entire families--and even villages and towns--there would have been no survivors for the teams to interview.

The study also fails to account for lives the coalition has saved by preventing the wars that Saddam inevitably would have started with his neighbors if his regime had survived. His unprovoked invasions of Iran and Kuwait brought hundreds of thousands of Iraqi deaths in direct military combat, and many more were killed indirectly. How many hundreds of thousands would have died in future conflicts with a post-sanctions, WMD-armed Iraq?

Finally, the study does not tell us that the vast majority of the deaths it attributes to the post-invasion conditions in Iraq were men of military age killed by gunshots. A large part of these men were likely insurgents who, if not killed, would have fought either to return Saddam to power or to plunge Iraq into a new dark age.

Every death of an innocent Iraqi should be lamented, but we should not succumb to the idea that wars only kill. They can also save lives--millions of them--and there is reason to believe that this is precisely what has happened in Iraq. Two fundamental points must be remembered. First, people continue to die in Iraq because a certain segment of its population will do anything to gain power and return the country to barbarism. Their victory would come at a catastrophic human cost. Second, and more important, as we feel sorrow for those who have died, we should also celebrate the millions who would already be dead or would eventually die--needlessly--if America had not removed Saddam's evil regime from power.

Mr. Lacey is a military analyst based in Washington, D.C. His next book, Takedown, will be published in March 2007.