The Methods described here go into greater detail, but remain consistent with, the original Iraq Body Count Methodology published in 2003, and reflect the experience IBC has gained over the intervening years.
Note that these Methods apply only to IBC’s formal output, and not to other material such as the Recent Events interim updates to the database, or to comment pieces – except where these are directly informed by, or reference, the formal work.
3. Data extraction
3.1 Standardization and reliability
IBC’s data arrives in a non-standardised format: that is, it doesn’t come to us in the convenient form of a standard questionnaire designed according to our needs and filled out in accordance with a protocol of our choosing. Nonetheless, experience of the nature of reporting as it obtains in the Iraq conflict has allowed us to identify data points that are consistently, if not perfectly, present across a wide range and large number of reports. Key and most consistent among them are the basic data gathered of date, location, the number of civilian deaths, and the weapons they were killed by. In all, IBC systematically extracts 18 pieces of information in relation to each incident and/or person killed:
1A standardised place-names transliteration system from the Arabic into English, based on UN data, is employed to avoid conflicting spelling conventions for Iraqi locations.
2Where individuals' names are transliterated from the Arabic using different spelling in different English reports, all variants are entered in the database entry for that individual.
| For each incident | For each person | |||
|---|---|---|---|---|
| 1 | Date | 13 | Name2 | |
| 2 | Time | 14 | Age | |
| 3 | Place1 | 15 | Gender | |
| 4 | Target | 16 | Marital status | |
| 5 | Minimum deaths | 17 | Parental status | |
| 6 | Maximum deaths | 18 | Occupation | |
| 7 | Minimum injuries | |||
| 8 | Maximum injuries | |||
| 9 | Weapons | |||
| 10 | Killers | |||
| 11 | Media sources | |||
| 12 | Primary witnesses | |||
The choice of this specific data-set is motivated by two aims central to the project: (a) to ensure that as much detail as is feasible and practicable is recorded in a standard format for every person killed; (b) to ensure against double counting and reliably differentiate between similar incidents which might be confused.
Data extraction and coding is subject to strict internal reliability checks, with particular emphasis on the number killed. Each new data-entry is reviewed and signed off by at least two additional volunteer data checkers prior to publication. All past data entries are kept under review, and if necessary modified when new information becomes available.
3.2 Treatment of inconsistencies in reporting
Reports of numbers killed often vary across sources. On-the-ground uncertainties and potential political bias can result in a range of figures reported for the same incident. When such variation exists and cannot be reliably reconciled, each incident is associated with a range for the number of deaths in that incident.
In such cases, the high end of the range (the ‘maximum’ number) is the highest number of civilian deaths published in at least two independent sources. The low end of the range (the ‘minimum’) is the lowest number of civilian deaths published by two of our sources. The lower IBC number can receive an entry of zero if two sources explicitly state that “no was one killed,” despite two other reports stating that there were deaths. However, “unable to confirm any deaths,” and similar forms of wording reflecting lack of knowledge, are not considered to conflict with reports that do confirm a number of deaths and do not lead to an entry of zero or use of the range.
Similarly, the wording “at least” preceding a death toll is not considered to contradict a higher number, since it allows for that higher number. Further, updates to a story, whether correcting its death toll higher or lower, supersede numbers in early-breaking reports collected by IBC.
A range is also employed to handle uncertainties regarding potential overlaps between separate data entries (the potential for double counting). When there is clear or particularly likely overlap between two separate IBC entries, the overlapping number is subtracted, requiring no range. However, in cases where overlap is plausible but neither certain or likely, the potential overlap is subtracted from the lower IBC number (which allows for the possibility that an overlap does exist), but zero is subtracted from the upper IBC number (allowing for the possibility that no overlap exists) .
Additionally, there can be uncertainty in different accounts about the civilian status of those killed. Analysis of all available sources can often reconcile such conflicting reports, but where these cannot be reconciled the high-low range is again employed to reflect both possibilities, with deaths added to the higher IBC number but not the lower one.
Where conflicting data arises from differing accounts by primary sources (e.g., eyewitnesses, police, medics, government and military officials), IBC adopts a pragmatic process in choosing between them. This only assumes that people can convey facts no more accurately than they are able to discern them.
Instead of attempting to consistently divine the ‘political’ position of any given primary source in relation to the realities of an incident, IBC instead takes careful account of their physical position in relation to the pertinent facts. In short, no assumptions are made about the reliability of primary sources, beyond that some are better placed to know the facts of the incident than others.
Eyewitnesses may, for example, relate a bombing event vividly and exactly as they experienced it. Nonetheless they may have seen only a limited proportion of its victims compared to emergency medics, and consequently understate its dead and wounded. At other times eyewitnesses may mistake the unconscious wounded as among the dead, leading to an overstatement of the death toll. Conversely, eyewitnesses may have more direct knowledge than medics about whether an incident was a vehicle-borne suicide attack or a bomb planted in a parked car, making them better placed to know those facts.
3.3 Inclusions and exclusions
The range given in the IBC count refers to civilian deaths. Use of the term “civilian” by definition involves making a distinction between some people and others. We determine this distinction on a case-by-case basis through careful and systematic scrutiny of the data sources we consult.
The boundary between civilians and others is not always clear-cut. Analysis and adjudication of this boundary can raise deep legal, moral, and philosophical issues which are far from fully resolved. The distinctions which follow are those we routinely apply for the purposes of the IBC project.
Excluded from IBC are those aged 18 and over who, at the point of death, were reported as initiating deadly violence or being active members of a military or paramilitary organisation. We also exclude overseas ‘contractors’ providing security and other private services related to the occupation of Iraq.
Included are all others killed violently, including regular local police forces.
As every society, at war or at peace, has police forces who live and work among the civilian population, we consider such police forces to be a customary part of civil society, and therefore include them in our civilian count. However we do not include police ‘commando’ units who work under the Interior Ministry and are best described as paramilitary.
Under one special circumstance we also include members of Iraqi military or paramilitary/militia forces in our database, namely when they are killed — i.e. summarily executed — after capture. Under those conditions even military personnel automatically acquire ‘protected person’ (effectively, POW) status under International Humanitarian Law, and this distinction is respected by IBC.