Iraq War Logs: What the numbers reveal

First published 23 Oct 2010

• The Iraq War Logs contain an estimated 15,000 previously unknown civilian deaths.
• The majority of these new deaths come from small incidents of one to three deaths.
• Additionally, IBC calculates that over 150,000 violent deaths have been recorded since March 2003, with more than 122,000 (80%) of them civilian.

The Iraq War Logs released by WikiLeaks on 22 October 2010, contain 54,910 records compiled by the United States military whose numerical fields register 109,032 violent deaths between January 2004 and December 2009. These casualty records contain four categories of casualties, ‘Civilian’ (66,081 deaths), ‘Host Nation’ (15,196 deaths),‘Enemy’ (23,984 deaths), and ‘Friendly’ (3,771 deaths).

But how much of this is new knowledge? How many of the 109,032 deaths documented in the logs are previously unpublished? What, in other words, do these records add to the already known death toll? A precise and final answer to this question can only be gained after a detailed reading of each individual log to discover whether or not what it contains has already been documented and reported elsewhere.

Such analysis will take many months yet to complete. Nonetheless, IBC has been able to provide a preliminary assessment of how many deaths the new data adds by analysing a carefully chosen sample of these logs. On the basis of these analyses IBC is able to conclude that some 15,000 hitherto unrecorded civilian deaths will be added to the public record from the Iraq War Logs, and that these, together with new information on combatant deaths contained in the logs, will bring the recorded death toll since March 2003 to over 150,000, roughly 80% of whom were civilians.

Full details of the IBC analyses are given below and in the technical appendix.1

Our analysis in detail

We constructed the sample for our preliminary analysis from all death categories in the Iraq War Logs except for 'Friendly' (defined as US and other allied, non-Iraqi military forces). Such military deaths are never included in the IBC database and are, in any case, very well documented in other sources.2 The remaining three categories will overlap with the IBC database to one degree or another. For each of these categories we assessed how and where the two data-sets overlap and, as a result, what the new logs would add to the IBC database.

In the three categories (‘Civilian’, ‘Host Nation’, ‘Enemy’), we assessed every log that records 20 or more deaths. We chose to look at each of these logs in detail, since the presence of previously unknown incidents of such a large size would be particularly important to know, and because their total number (360 logs) was small enough to be manageable within the limited time available. For logs containing death totals below 20 (52,024 logs), random samples were drawn from each of the three categories above.

First, we have compared these logs with similarly detailed information on civilian deaths in the IBC database to provide a projected total of previously unknown civilian deaths that will be added. Then we have evaluated the extent to which the US military classification of deaths as 'Enemy' are inconsistent with IBC-recorded data on these same events. Next, we have performed a like with like comparison of the same types of deaths in both datasets to assess the number of IBC-recorded deaths that are missing from the logs. And finally, we have derived an overall figure for total violent deaths in Iraq (both civilian and combatant) by integrating the estimated log numbers with IBC and other relevant totals already in the public domain.

We stress that the following assessments are preliminary and are in no way our final conclusions about what these logs contribute to existing knowledge. A full and final accounting will require many months of dedicated work evaluating all of the logs in detail, one by one.

Deaths in the 'Civilian' category

• More than 12,000 previously unknown deaths are contained in the ‘Civilian’ category of the logs.
  
• The vast majority of previously unreported deaths arise from small incidents of 1-3 deaths.

Civilians are defined in the Iraq War Logs as Iraqi and other civilians, including foreign contractors. The logs record 66,081 deaths in this category. Since IBC is itself a civilian deaths database, this is the most clearly relevant category for IBC. One exception is that the logs include foreign security contractors in this category, while IBC does not include them in its count. However, the vast majority of 'Civilians' in the logs are Iraqis, so matching this category against IBC is fairly straightforward.

War logs Category	Total deaths	Number of deaths sampled	Percent matched	Estimated new deaths
Civilian	66,081	11,177	81.2%	12,438 (95% CI: 10,999-13,877)

The 'Civilian' category contains 239 logs recording 20 or more deaths. We assessed each of these logs individually. From the remaining logs recording between 1 and 19 deaths, a sample of 300 logs was drawn. We matched this sample to the IBC database to determine which deaths were not already recorded in IBC and would, therefore, provide new deaths to add to the database.

We estimate that when all of the work is complete, 81.2% of the deaths in the logs will match those that were in the IBC database prior to the release of the logs. This means that a significant majority of deaths recorded in the logs had already been put into the public domain, mainly via press and media reports contemporary with the events being recorded. On the other hand, 18.8% of deaths in the sample were not previously known.

The matching rates varied considerably according to the size of the incident. Deaths in logs recording 10 or more civilians killed had a matching rate of over 95%, while deaths in logs where only one or two people were killed had lower matching rates of around 73%. This means that most of the new deaths revealed by the Iraq War Logs are not from big bombings and explosions, but from those multiple small tragedies (targeted assassinations, executions, drive-by shootings, checkpoint killings) that the Iraqi people have endured on a daily basis throughout the entire period.

Taken together, our sample of 300 logs across the Civilian category, along with the 239 logs reviewed containing 20 or more deaths each, allow us to estimate that 12,438 (10,999-13,877) previously unrecorded deaths will be added to the IBC database from the ‘Civilian’ category of the logs.

Deaths in the 'Host Nation' category

• 2,700 previously unreported deaths are of Iraqi Police and other Iraqi security forces killed after capture.
  
• More than 15,000 previously unknown civilian deaths are revealed once the ‘Civilian’ and ‘Host Nation’ categories are combined.
  
• This analysis confirms that police have been the most targeted profession.
  

‘Host Nation’ is defined in the Iraq War Logs as Iraqi security forces, including all official Iraqi government forces from local police to National Guard and Iraqi Army. The logs record 15,196 deaths in this category. Some of these types of deaths are included in the IBC civilian deaths database, and others are excluded. IBC includes deaths of local police, but only includes other Iraqi security force members, such as Iraqi Army, National Guard or National Police, when they have been executed after capture.3 This means there is only partial overlap between the inclusion criteria of IBC and the ‘Host Nation’ category of the logs.

Below is our estimate of how many new 'Host Nation' deaths are recorded in the logs that are compatible with the criteria used by IBC.

War logs category	Total deaths	Number of deaths sampled	Percent matched or not compatible	Estimated new deaths
Host Nation	15,196	889	82.2%	2,706 (95% CI: 1,965-3,446)

The ‘Host Nation’ category contains 14 logs that record 20 or more deaths. Each of these 14 logs were assessed individually, and a sample of 100 logs was drawn from the remaining logs that recorded between 1 and 19 deaths. These 114 logs allowed us to estimate that 2,706 (1,965-3,446) IBC-compatible deaths (e.g., Iraqi Police) will be added to the IBC database from the ‘Host Nation’ category.

As in the ‘Civilian’ category, IBC’s coverage of ‘Host Nation’ deaths varied considerably depending on incident size: in incidents where 4 or more Iraqis were killed in this category, 96% were either already in IBC's civilian database or incompatible with it. For logs recording 1-3 deaths, this dropped to around 75%. So it is again deaths in the smallest incidents which tend to have been previously unreported.

Combining the ‘Civilian’ and ‘Host Nation’ results allows us to estimate that 15,144 (13,525-16,762) new deaths will be added to the IBC database from the Iraq War Logs. This suggests that the current IBC total of over 107,000 civilian deaths will increase to over 122,000 as a result of the release of these documents by WikiLeaks.

Deaths in the 'Enemy' category

• The US military's classification of over 3,000 deaths as ‘Enemy’ in the logs is disputed by previous reports collected by IBC
  
• Over 20,000 deaths of insurgents or anti-Coalition forces are recorded in the logs
  

‘Enemy’ are defined in the Iraq War Logs as Anti-occupation forces, 'insurgents' or militia who opposed the US and Coalition Forces. The logs record 23,984 deaths defined as ‘Enemy’. 107 logs recorded 20 or more deaths in the ‘Enemy’ category and each of these was assessed individually. A sample of 100 logs was drawn from the remaining logs recording between 1 and 19 deaths.

The IBC database does not include 'Enemy' fighters (i.e., 'insurgents', ‘anti-occupation forces’), so the 'Enemy' category is, in theory, entirely non-overlapping with IBC. However, there is often disagreement over such classifications across different sources. Many will question whether every designation of a death as 'Enemy' in these US military logs is reliable. Therefore we assessed this category against IBC to determine what proportion of these deaths are classified differently in IBC. Specifically, to what extent does IBC record civilian deaths where the logs record these deaths as 'Enemy'?

War logs category	Total deaths	Number of deaths sampled	Percent matched / Classification disputed	Estimated death classifications disputed
Enemy	23,984	4,950	14%	3,485 (95% CI: 2,992-3,978)

We found that the US military's classification is disputed by existing press reports and other public sources for roughly 14% of deaths in their ‘Enemy’ category, or an estimated 3,485 deaths (2,992-3,978). We noted that in some of these cases the IBC database recorded a number for civilian deaths given as a range beginning with zero, such as 0-22 (e.g. k15375), which usually indicates that there already was some disagreement over their classification in our previously collected sources.

These results show that the US military are more likely to classify deaths as ‘Enemy’ in the logs than are other sources used by IBC, such as media reports. Setting aside these disagreements, the Iraq War Logs still record 20,499 (20,006-20,992) ‘Enemy’ deaths whose classification as such is not disputed by other existing data.

Deaths in Iraq War Logs & IBC compared

• 64,000 deaths are recorded in both IBC and the logs
  
• 15,000 recorded in the logs are unrecorded in IBC
  
• 27,000 recorded in IBC are unrecorded in the logs
  
• Both datasets include some deaths that are absent in the other
  

Both IBC and the Iraq War Logs have omissions that can be revealed by a detailed like with like comparison. As discussed above, our analysis suggests there are roughly 15,000 new civilian deaths in the logs which were previously unknown and can be added to the IBC database. Since we know the number of deaths reported in the two datasets over the same periods, we can also calculate the likely number of omissions in the logs.

IBC has recorded 91,055 civilian deaths for the period covered by the logs. Our analyses of the logs suggest that they record 79,187 (78,191-80,183) deaths of the same types recorded by IBC, with the remainder of roughly 30,000 being deaths which IBC defines as 'combatant' and does not include (e.g., Iraqi Army, US military, insurgents).

Our prior analysis above concluded that the logs are likely to contain roughly 15,000 comparable deaths that were not recorded by IBC. When this like with like comparison is performed, it suggests that the two datasets record roughly 64,000 of the same deaths (79,000 minus 15,000). As such, there must remain roughly 27,000 civilian deaths recorded by IBC that are unaccounted for in the logs (91,000 minus 64,000). These relationships are illustrated by the following Venn diagram:

At this early stage it is not possible to give a full account of the IBC-recorded deaths that are missing from the logs. We have, however, noticed some problems with the logs that may be relevant. One such issue is the presence of numerous simple coding errors. For instance, there are many records where the numerical columns register zero deaths yet the verbal summary descriptions mention specific deaths. This is another reason why every log requires detailed human examination before the casualty information it contains can be fully extracted. Such coding errors make it impossible to accept simple summations of the numerical columns in the logs as providing a fully accurate representation of the fatal incidents that these logs contain.4

In addition to such technical errors, some examples of major omissions have already been found, such as the complete absence of any substantive reports of civilian casualties in the two major US military operations in Fallujah in April and November 2004. Corroborated sources that have been incorporated into the IBC database provide evidence of at least 1,226 civilian deaths in these two assaults. The logs do record almost 800 deaths classified as ‘Enemy’ during these assaults, but appear to register no more than a few deaths classified as ‘Civilian’. The reasons for the almost total absence of civilian deaths recorded in these cases are not clear, but this does show two instances where many civilian casualties already established in the public record are not recorded in the logs. This and other such instances will likely contribute to explaining the more than 25,000 civilian deaths recorded in IBC that appear to be missing from the logs.

Total deaths 2003-2010

• More than 150,000 people have been recorded killed in the Iraq war to date.
  
• 80% of those killed were civilians.
  

The Iraq War Logs record deaths of all types, including combatant deaths that fall outside the scope of IBC’s civilian deaths database. This means that they can also contribute to a broader accounting of the total number of persons killed in the war, civilian and combatant alike. Such a figure can be derived by combining the IBC database, the new logs (2004-2009), and other official information available on combatant deaths in 2003, 2010 and the two months missing from the logs (May 2004 and March 2009).

Combining these sources, the detailed calculation below provides a figure of total Iraqi deaths, both civilian and combatant, of 150,726. Adding figures on Coalition military deaths, which now stand at 4,744, brings the number up to 155,470. That is, given our analysis of the new logs, as combined with other previously reported deaths, we are now able to say that more than 150,000 people have been recorded killed in the Iraq war since 2003, of which around 80% were civilian.

Iraq Body Count 2003-10 October 2010	107,369
Iraq War Logs new 'Civilian' and comparable 'Host Nation' - central estimate	15,114
Iraq War Logs ‘Host Nation’ combatant - central estimate	5,575
Iraq War Logs ‘Enemy’ (minus IBC overlaps) - central estimate	20,499
Insurgents killed June-December 2003	597
Insurgents & Iraqi soldiers killed January-September 2010	861
Insurgents killed May 2004	652
Insurgents & Iraqi soldiers killed March 2009	59
TOTAL IRAQI	150,726
US & Coalition military killed	4,744
TOTAL	155,470