On Sept. 1, the bullet-riddled bodies of four Iraqi men were found on a Baghdad street. Two days later, a single dead man, with one bullet in his head, was found on a different street. According to the U.S. military in Iraq, the solitary man was a victim of sectarian violence. The first four were not.
Such determinations are the building blocks for what the Bush administration has declared a downward trend in sectarian deaths and a sign that its war strategy is working. They are made by a specialized team of soldiers who spend their nights at computer terminals, sifting through data on the day's civilian victims for clues to the motivations of killers.
The soldiers have a manual telling them what to look for. Signs of torture or a single shot to the head, corpses left in a "known body dump" -- as the body of the Sunni man found on Sept. 3 was -- spell sectarian violence, said Chief Warrant Officer 3 Dan Macomber, the team leader. Macomber, who has been at his job in Baghdad since February, rarely has to look it up anymore.
"If you were just a criminal and you just wanted to take somebody's money, just wanted to discipline them, you're not going to take the time to bind them up, burn their bodies, cut their arms off, cut their head off," he explained. "You're just going to shoot them in the body and get it over with." That, the team judged, is what happened to the four Shiite men, sprayed with gunfire and left where they dropped.
In the Iraq conflict, traditional military measures of achievement -- troops deployed, enemy dead, territory won -- are challenged by the chaos of counterinsurgency warfare. But Congress, the public and the military itself demand an accounting. Far from the battlefield, platoons of soldiers in Iraq and at the Pentagon are assigned to crunch numbers -- sectarian killings, roadside bombs, Iraqi forces trained, weapons caches discovered and others -- in a constant effort to gauge how the war is going.
In recent months, most of the military's indicators have pointed in a favorable direction. As with all statistics, however, their meaning depends on how they are gathered and analyzed. "Everybody has their own way of doing it," Macomber said of his sectarian analyses. "If you and I . . . pulled from the same database, and I pulled one day and you pulled the next, we would have totally different numbers."
Apparent contradictions are relatively easy to find in the flood of bar charts and trend lines the military produces. Civilian casualty numbers in the Pentagon's latest quarterly report on Iraq last week, for example, differ significantly from those presented by the top commander in Iraq, Gen. David H. Petraeus, in his recent congressional testimony. Petraeus's chart was limited to numbers of dead, while the Pentagon combined the numbers of dead and wounded -- a figure that should be greater. Yet Petraeus's numbers were higher than the Pentagon's for the months preceding this year's increase of U.S. troops to Iraq, and lower since U.S. operations escalated this summer.
The charts are difficult to compare: Petraeus used monthly figures on a line graph, while the Pentagon computed "Average Daily Casualties" on a bar chart, and neither included actual numbers. But the numerical differences are still stark, and the reasons offered can be hard to parse. The Pentagon, in a written clarification, said that "Gen. Petraeus reported civilian deaths based on incidents reported by Coalition forces plus Iraqi government data. The [Pentagon] report only includes incidents reported by Coalition forces for civilian causality data."
"There is a current effort to consolidate multiple databases in theater," a Multi-National Force-Iraq headquarters spokesman said in an e-mail.
The number of sectarian killings in 2006 -- a key reference point in measuring improvement this year -- has changed considerably in the line graphs used in the Pentagon's past three quarterly reports, increasing between the March and June assessments this year and again in last week's report. Macomber, the analyst in Baghdad, said the first jump occurred when his office realized after the March version was published that a backlog of Iraqi government data had not been included in the 2006 figures.
The most recent increase came when the Pentagon decided to include Iraqis killed in vehicle and suicide bombings, the most obvious forms of sectarian violence. Baghdad had always tallied those numbers along with other killings, Macomber said, but the Pentagon had always taken them out in compiling its own graphs. Asked about the change, a Pentagon spokesman e-mailed that "We regularly review our metrics to determine the most informative way to report what is happening in Iraq."
In an Iraq assessment released this month, the Government Accountability Office said it "could not determine if sectarian violence had declined" since the U.S. troop buildup began in the spring and saw no decrease in overall attacks against civilians as of the end of July. The GAO recommended that the administration expand its statistical sources to include "all relevant U.S. agencies" and that it use "broader measures of population security" to establish trends. An unpublished, classified annex to its report listed the sources of differing agency opinions and provided more detail on the kinds of measurements the GAO thought should be included.
The U.S. intelligence community considers more than numbers in making its war assessments. "What the Iraqis perceive" about their country and their daily lives "may be more important than what the numbers are," said a senior intelligence official, who discussed the subject on the condition of anonymity. Even so, he said, intelligence officials found contradictions in the available statistics as they wrote last month's National Intelligence Estimate on Iraq, whose conclusions were somewhat less optimistic than the military's.
"It's not anybody trying to make it come out one way or another way," said the official, who sympathized with the military's need to quantify. But it is important, he said, to determine "what the numbers meant. Who collected them? Why do numbers that come in from this piece of the U.S. government differ from those coming in from another part of the government?"
While both Petraeus and the recent Pentagon report emphasized improved statistics over the past three months, the intelligence community generally declines to declare trends based on data measured in periods shorter than six months to a year. Several senior intelligence officials said last week that most numerical indicators appear to be moving in a uniformly positive direction in the nearly two months since the intelligence estimate's data cutoff -- although they said it is too early to determine definitive trends.
As questions have been raised about its statistics, the military has tried to make them more transparent. After his congressional testimony, Petraeus released an unclassified version of a Multi-National Force-Iraq document titled "Ethno-Sectarian Violence Methodology," and the Baghdad command last week provided a telephone interview with Macomber, the man directly in charge of implementing it.
Macomber, an 18-year Army veteran, said that he is a "senior all-source intelligence analyst" and that the mission of his six-person team is "to compile [data] and track trends and analysis for General Petraeus." Daily data on civilian killings are compiled in a database called the Combined Information Data Network Exchange. The source of the information "could be a coalition force out on patrol," Macomber said. "It could be police, or somebody who called and said they found a body."
'We look at every single record'
"We look at every single record and de-conflict between coalition and host nation [information] to ensure that nothing is duplicate or erroneous," he said. "Then we look at every record and apply our methodology and criteria to it and assess whether it's ethno-sectarian."
Their written definition of that term is: "An event and any associated civilian deaths caused by or during murders/executions, kidnappings, direct fire, indirect fire, and all types of explosive devices identified as being conducted by one ethnic/religious person/group directed at a different ethnic/religious person/group, where the primary motivation for the event is based on ethnicity or religious sect."
The process to determine whether a body is that of a Shiite, a Sunni or a member of any one of a welter of minority sects in Iraq is imperfect, Macomber said. "Sometimes they know by any type of identification," he said. "There are times when they don't know. . . . A lot of times it comes down to, a body was found in a Shiite area, it wasn't moved anywhere, and we'll make that call that it was likely a Shiite person."
Recent sectarian fighting in an area is another clue. "It's not perfect to be able to identify every single person," Macomber said. "But there are pieces out there that we can use to help us." At the end of the day, he said, "it's an analyst making an analyst's call."
The killing of seven Iraqis on Aug. 25 in the predominately Shiite Baghdad neighborhood of Kadhimiya was judged sectarian. The victims were Shiites, and the method and location -- a car bomb in a marketplace -- pointed to Sunnis.
Two Iraqis killed by a car bomb on Sept. 3 were not included in the sectarian database, however. The attack occurred on a road near Ramadi, not far from where President Bush was meeting with government officials that day. But the victims, regardless of ethnicity or sect, were Iraqi policemen. They were counted elsewhere.