IBC begins to include credible single-sourced reports

The new procedures summarised

From 21st December, IBC will begin to include single-sourced incidents in its public database. The most important effect of this is to ensure that trend analyses derived from IBC will be more accurate for recent periods, including the month just elapsed.

These single-sourced incidents comprise a small proportion of overall incidents and an even smaller proportion of deaths (since these incidents mainly involve smaller numbers killed - two, on average). Such small incidents are rarely misreported: inconsistent reporting mostly applies to very large incidents where the exact death toll is difficult to determine. Further, these single-source reports stem from the same reputable media and primary sources which provide most of IBC's fully-corroborated data, and many of them are subsequently corroborated through later-released official cumulative totals.

Single-sourced incidents do not differ in any significant way from those we routinely cross-corroborate, beyond that they tend to describe smaller-scale incidents which failed to attract the attention of other media sources.

The detailed case for including single-sourced reports

The data entries for deaths in the IBC database, while methodically collected and compiled by IBC, do not originate from us. The vast majority are based on media reports of violent events which led to civilian deaths, along with cumulative information produced by various Iraqi official sources (e.g., Iraq’s Health Ministry, regional morgues or local police records). Most of the latter data, usually released after the end of calendar months, but in some cases many months after the fact, are themselves reported by the media in addition to their reporting of individual incidents. After careful analysis by IBC to avoid double counting and other possible errors, these are also integrated into the media-based IBC database in whole or in part.

Since misreporting of even relatively straightforward events does occur, IBC has operated a system of corroboration to provide an acceptable level of confidence in the veracity of the reports used. In fact it is IBC’s policy to gather every reputable report about each incident (or death), in a more or less open-ended process, with a minimum of two sources required for each death added to the database.

After having compiled and cross-corroborated every incident possible in a given month, as described above, there is always a remainder of incidents with only one press source reporting them. In the past, these reports have simply remained in our off-line archives, awaiting a second corroborating source. The most problematic aspect of this practice is that it can introduce distortions when analysing trends using IBC data.

When most months are completely processed, but other months are left only ‘mostly’ completed, this can create the appearance of fluctuations in violence which did not actually occur. IBC-derived trends will be more reliable if every month’s data is processed in the same manner, and within the same time-frame, so that different months’ data are not left for long periods at varying stages of completeness.

It has been our experience that, after some time has elapsed, almost all single-sourced reports are either directly corroborated by other emerging reports, or indirectly corroborated by figures gathered by various local or national institutions in Iraq. This means, effectively, that the publication of single-sourced events are simply delayed for varying lengths of time until an official figure is released of which the events they describe would have constituted a part.

For instance, almost all but a handful of genuinely questionable single-sourced reports during 2006 have now been integrated into the database through later corroboration, although this process took well into 2007 to complete. These single-sourced reports ultimately added some 3,500 violent civilian deaths out of a total of nearly 28,000 in that year:

• The current total for 2006 is 5,176 incidents involving 27,665 violent civilian deaths
• This includes 1,630 later corroborated incidents involving 3,497 deaths

In 2007, most such single-sourced deaths have already been corroborated by official figures, so that there are fewer outstanding single-sourced incidents to add, and these are isolated to particular months during the year. Nonetheless, we are confident that these remaining figures can and should be released ahead of their probable corroboration in future.

In the cases where single-sourced incidents remain uncorroborated this is less likely to reflect unreliability in the published record than it is to reflect the ups and downs of the complex and politically-charged context for official decision-making in regard to the release of ministry-held data.1

Beginning on 21st December, 2007, IBC will add to its database single-sourced deaths without waiting a month or more for official or other corroborating figures to emerge. Far from leading to a reduction in accuracy and reliability of the database, we believe that this will make it more accurately reflect recent trends, and obviate the need for complex caveats which we have in the past had to provide to account for the comparative patchiness of recent data.

In brief, the relevant issues to be considered are:

1. All else being equal, incidents and deaths reported in multiple media sources provide a higher degree of certainty than those reported in only one source.
2. Nearly five years of IBC’s experience in collecting and analyzing this data suggests that misreporting of specific deaths is scarce, or minor, and readily identified by our routine controls.
3. Single-sourced incidents almost invariably involve small numbers killed, on average two, so that their contribution to the database is always likely to be small.
4. These reports are as a rule otherwise impeccable – originating from the same news agencies and the same types of primary sources as those which are reported by multiple sources.
5. Like all IBC entries, these remain ‘provisional’ in the sense that further information on them will constantly be sought.
6. Ultimately, the lack of information here is one of official recording. Early publication of these single-sourced media reports bridges the gap in the publication of comprehensive official figures (as well as providing, as the IBC database does in its entirety, a detailed record of deaths which is absent from official cumulative figures).
7. Since corroborating official figures can take time to emerge, recent periods often appear less violent than they should relative to preceding months, which has required us to issue complicated caveats for recent trends. Including single-sourced incidents earlier rather than later goes some way towards alleviating this.
8. The number and proportion of single-sourced deaths can now be displayed on the database home page, keeping their contribution open and transparent.

It is our firm view that points 2–7 far outweigh the concerns raised in point 1, which is why we are introducing single-sourced entries from today.

Including single-sourced data, in a transparent way, has been made possible by improvements to the database which allow these incidents to be tracked and their present contribution reported. They are also clearly marked on an incident’s own page: if you know of a corroborating source for any of these incidents, please let us know.