return to ICG Spaces home    ICG Risk Blog    discussions    newsletters    login    

ICG Risk Blog - [ Tangible statistical evidence of the long denied civil war in Iraq ]

Tangible statistical evidence of the long denied civil war in Iraq


Numbers of dead - even directional if not exact, who caused those deaths - and who did not, and the means by which death was inflicted are excellent measures to evaluate causal conditions of war, i.e., while they are necessary but not sufficient indicators of root cause they can, by process of elimination, point us to root cause. For example, deaths in Iraq due to, say, aerial attack would point to coalition forces as its cause while deaths due to thermobaric weapons (offering near-nuclear effects in confined spaces) could either be coalition forces or insurgents employing say, Russian, Ukrainian or Czech weapons flowing through friendly states or purchased through the global arms market. Deaths due to car bombs or IEDs point to insurgent/jihadist sources while deaths due to small arms fire could be any participant.

The recently released, and incandescently received, 2006 survey, Mortality after the 2003 invasion of Iraq: a cross-sectional cluster sample survey and the earlier 2004 Mortality before and after the 2003 invasion of Iraq: cluster sample survey , both in The Lancet, has offered this observer the first concrete statistical corroboration of an inter-Iraqi civil war, long denied by the US and Iraqi administrations, that is distinct from coalition activity to suppress an admittedly robust and rising insurgency

Absolute numbers aside, directional differences between the two surveys show that causes of Iraqi deaths shifted from primarily coalition aerial action, with victims among adult males, children and women, to primarily non-coalition-caused gunshots, IEDs and explosions, with an overwhelming number of adult males with few children and women. While deaths due to coalition action continue to rise in absolute terms, those deaths are dwarfed by deaths due to actions by Iraqi nationals. Iraqi-on-Iraqi, Sunni on Shia, and Shia upon Sunni predation have become the major causes of death, and have ramped significantly in 2005-2006.

The singular implication of that analysis in concert with years of obvious, rising anecdotal violent death data is that Iraq is firmly in a state of civil war.

As this note was going to release, there are waves of retaliatory Reprisals between Shias and Sunnis, calls to halt the sectarian government, and a "bloc of Sunni insurgent groups marked the [one-year anniversary of a referendum on a U.S.-backed constitution] by declaring a separate Islamic republic in Iraq, stretching from the western province of Anbar to Baghdad, Kirkuk and other parts of the north" as it noted the "creation of a separate Kurdish republic in northern Iraq and a push by some Shiite parties for a separate republic in the south":

Anthony Cordesman, an analyst for the Center for Strategic and International Studies, said by telephone: "If you total up the number of people that are being killed, that are being wounded, that are being displaced and are being forced to leave the country, and the zones in which there is major civil conflict . . . trying to declare there isn't a civil war borders on the absurd."

What the data tells me about civil war in Iraq

2004: Cause of violent death was overwhelmingly due to coalition action.

Results were calculated with and without data from Falluja, ultimately excluding Falluja data given the intense combat operations there. From cluster data, 46 deaths were reported prior to March 2003 and 142 deaths post invasion, Of the 73 violent deaths, 84% were due to coallition action and of those, 95% were due to helicopter gunships, aerial bombardment, artillery and other forms of aerial weaponry. 46% of violent deaths involving coalition forces were men ages 15 to 60, but 46% were children younger than 15, and 7% were women.

2006: Cause of violent death overwhelmingly shifted to non-coalition action.

Coalition caused deaths now drops to 26% for June 2005 to June 2006, with 30% by other fighters, and 44% undetermined. Causes of death from 302 deaths show 56% gunshot, 14% ordnance explosion, 13% car bomb, 13% aerial attack, 2% unknown violent and 2% accident. Most of the deaths are of military-age men, undifferentiated between civilians, insurgents and Iraqi security forces:

Of the deaths reported by the study population since 2003:

  • 92% were from violent causes, more than half from gunshots.
  • Only about a third of violent deaths were attributed to actions by coalition forces.
  • The percentage of violent deaths attributed to coalition action has fallen, though the absolute number of deaths per year from that cause has climbed.
  • Men and boys aged between 15 and 44 accounted for 59% of violent deaths, despite making up just 24% of the population.
  • Despite disruptions in sanitation and health care, death from non-violent causes do not appear to have climbed significantly.

The chart accompanying this NYT piece offers a useful graphic rendering of the shift between 2004 and 2006. The 2006 study indicates that civilian death estimates are in line with the Vietnam War and other conflicts. The interpretation of the 2006 report states:

The number of people dying in Iraq has continued to escalate. The proportion of deaths ascribed to coalition forces has diminished in 2006, although the actual numbers have increased every year. Gunfire remains the most common cause of death, although deaths from car bombing have increased.

I take the next step in interpreting this pattern as civil war divorced from coalition action.

Shooting the messenger (again): attacking the survey method to blunt the message

Cluster sampling is the 'gold standard' of epidemiological research and of determining mortality rates in areas where infrastructure collapse degrades surveyor access or where local security cannot protect the surveyors.

The same cluster sampling survey methodology used in the Johns Hopkins 2004 and 2006 surveys has been used in the Democratic Republic of Congo (DRC) (here, here and here), Kosovo (here and here) and in Darfur (here and here) without dispute. Even when Burnham and others were investigating Iraqi small arms usage there had been no complaint. The only occasions where the methodology has been attacked are the two Iraqi casualty surveys where parochial and political interests have an intense interest in discrediting its results.

Both surveys were criticized for predicting higher casualty figures than that tabulated from bodies reported, but such tabulation is difficult in a culture that requires burial in a day, many areas are remote to authority and hostile to investigators, and Iraqi government has closed public posting of morgue lists.

The 2004 survey by Johns Hopkins and the Al Mustansiriya University estimated more 100,000 Iraqi deaths from all causes, excluding Falluja households to gauge mortality rates:

[Roberts and Burnham's] regular technique is to estimate total mortality by personal surveys of a sample of the households in the area under study; this method being chosen in order to avoid the undercounting inherent in using only reported deaths in areas so chaotic that many deaths are unreported, and to include those deaths not directly attributable to violence but nevetheless the result of the conflict through indirect means, such as contamination of water supply or unavailability of medical care. The baseline mortality rate calculated from the interviewees' reports for the period prior to the conflict is subtracted from that reported during the conflict, to estimate the excess mortality which may be attributed to the presence of the conflict, directly or indirectly. This technique has been accepted uncritically in the previous mortality surveys [as noted above].

Because of the impracticality of carrying out an evenly distributed survey, particularly during a war, Roberts' surveys use "cluster sampling", dividing the area into a number of randomly-selected, approximately equally-populated regions; a random point is chosen within each region, and a fixed number of the households closest to that point are surveyed as a "cluster". While not as accurate as an evenly distributed survey of the same number of households, this technique is more accurate than merely surveying one household for each selected point.

Interestingly, Burnham is the lead course faculty for Johns Hopkins' Quality Assurance Management Methods for Developing Countries program, has substantive experience in humanitarian and health development programs in what I would call challenged areas of the globe, and is one that would not be expected to use dodgy methods.

The only 'criticism' of the 2004 report that I recommend to readers is that of Jack Straw, then UK Secretary of State for Foreign and Commonwealth Affairs. While it also relies on the other lower numbers as its principle defense, it was the only one that addressed the uncertainties in the survey and, in the process, offers information about the Iraqi condition. It shames all commentary from the US administration on either the 2004 or 2006 surveys.

The OECD defines Cluster sampling (Échantillonnage par conglomerats) as:

When the basic sampling unit in the population is to be found in groups or clusters, e.g. human beings in households, the sampling is sometimes carried out by selecting a sample of clusters and observing all the members of each selected cluster.

Readers can get a quick Wiki intro but are recommended to Description and comparison of the methods of cluster sampling and lot quality assurance sampling to assess immunization coverage:

In a cluster sample, the population is divided into non-overlapping subpopulations usually based on geographic or political boundaries. For a simple cluster sample, a random sample of subpopulations (clusters) is obtained and, within each selected cluster, each subject is sampled. More often, a two-stage cluster sample design is used where a random sample of clusters is selected and, within each cluster, a random sample of subjects. The two-stage design can be expanded into a multi-stage one, in which samples of clusters are selected within previously selected clusters. A benefit of this type of cluster sample is that a list of the units in the population is only needed for those clusters that are selected…

Statisticians go here.

The 2006 survey claims a 95% certainty in its prediction that "the number of Iraqis killed violently during the war ranges from 426,369 to 793,663" with a probable value of 601,027 having "suffered violent deaths since the March 2003 invasion." Worse, their data indicates that Iraq had become more violent in 2005-2006.

Criticism devolves, again, to the fact that its figures are "out of whack with every other estimate" rather than a realistic examination of the survey methodology. The comparison figures are based on notoriously unreliable data from the Iraqi Ministry of Health and the Baghdad morgue:

The Johns Hopkins researchers said methods that relied on morgue data or news reports may be inaccurate and undercount the deaths. Paul Bolton, an associate professor at the Boston University School of Public Health who reviewed the study at the behest of the authors, said official statistics were often less reliable than surveys. "Most of the time people would trust a random sample more than they would any kind of official reporting or official statistic," Bolton said. "Often we use this type of study to check if the official statistics are accurate."

Regarding the 2006 survey, Human Rights Watch (which had initially criticized the 2004 report, saying that its casualty projections were high and that HRW data indicated that fatalities were due more to ground action than aerial attack) said that its "group had no reason to question the accuracy of the new survey" and "If there is surprise about the size of the figure, it has more to do with our existing death tolls… The conventional wisdom is based on shoddy information." Just as in the response to the 2004 survey, a Pentagon spokesman said that "the U.S. took care to avoid civilian casualties, whereas insurgents deliberately targeted civilians" and "referred questions on civilian deaths to the Iraqi Health Ministry."

Burnham notes that "This clearly is a much higher number than many people have been thinking about… It shows the violence has spread across the country." Burnham stated that "the researchers' intent wasn't "to try to find a precise number" and, in a conference call with reporters, added, "If it turns out that we were able to do a much bigger study and we found that the death rate was 580,000 that would be essentially the same magnitude as what we're turning up here."

Burnham says that he expected criticism due to the implicit indictment of US and UK claims of war-related deaths and the political implications of coalition actions or inactions in the rising death toll. Lost in the criticism of both the 2004 and 2006 surveys is the fact that Johns Hopkins had originally expected to find that health and environmental issues were leading causes of death, not hostile action.

Dozens Of Iraqis Killed in Reprisals
River Towns Trade Sectarian Strikes As Militias Move In
By Ellen Knickmeyer and Muhanned Saif Aldin
Washington Post
October 16, 2006

A deadly Iraqi numbers game
By Sanjay Suri
Asia Times
Oct 13, 2006

Survey says 600,000 have died in Iraq war
By Clive Cookson, Science Editor, and Steve Negus, Iraq Correspondent
Financial Times
Published: October 11 2006 17:37 | Last updated: October 11 2006 17:37

Study Puts War's Iraqi Death Tally at More Than 600,000
The research team was criticized for its high estimates two years ago, and it again draws flak.
By Julian E. Barnes
LA Times
October 11, 2006

Iraqi Dead May Total 600,000, Study Says
New York Times
October 11, 2006

Death Toll of Iraqis Exceeds 600,000, Report Says (Update2)
By Nadine Elsibai
Oct. 11, 2006 - Last Updated: October 11, 2006 12:54 EDT

Enormous death toll of Iraq invasion revealed
Debora MacKenzie news service
13:32 11 October 2006

Huge Iraqi death estimate sparks controversy
Authors of study deny accusations of political bias.
Jim Giles
Published online: 11 October 2006

Mortality after the 2003 invasion of Iraq: a cross-sectional cluster sample survey
Gilbert Burnham, Riyadh Lafta, Shannon Doocy, Les Roberts
The Lancet
Published Online October 11, 2006

Mortality in the Democratic Republic of Congo: a nationwide survey
Benjamin Coghlan, Richard J Brennan, Pascal Ngoy, David Dofara, Brad Otto, Mark Clements, Tony Stewart
Lancet 2006; Vol 367: 44–51
January 7, 2006

Data Shows Faster-Rising Death Toll Among Iraqi Civilians
New York Times
July 14, 2005

Darfur: Counting the Deaths
Mortality Estimates from Multiple Survey Data
Debarati Guha-Sapir, Olivier Degomme with Mark Phelan
Centre for Research on the Epidemiology of Disasters (CRED)
University of Louvain, School of Public Health
May 26, 2005

Iraq Living Conditions Survey 2004
Volume II: Analytical Report
Central Organization for Statistics and Information Technology
Ministry of Planning and Development Cooperation, Baghdad, Iraq
UNDP, First published in 2005

The Role of Small Arms during the 2003–2004 Conflict in Iraq
By Riyadh Lafta, Les Roberts, Richard Garfield, Gilbert Burnham
A Working Paper of the Small Arms Survey

Chapter 5 Cluster Sampling
Lecture Notes, Semester 2, 2005
620-374: Sample Survey Section
University of Melbourne

Disasters: Introduction and State of the Art
Eric K. Noji
From the Centers for Disease Control and Prevention, Washington, DC
Epidemiologic Reviews
Volume 27, Number 1. pp 3-8, doi:10.1093/epirev/mxi007
accepted for publication March 18, 2005

Mortality in the Democratic Republic of Congo: Results from a Nationwide Survey
Conducted April - July 2004
Reported by Ben Coghlan, Rick Brennan Pascal Ngoy, David Dofara, Brad Otto, Tony Stewart
International Rescue Committee
December 2004

Mortality before and after the 2003 invasion of Iraq: cluster sample survey
Roberts L, Lafta R, Garfield R, Khudhairi J, Burnham G
The Lancet - Vol. 364, Issue 9448, 20 November 2004, Pages 1857-1864
PMID: 15555665 [PubMed - indexed for MEDLINE]
Note: This abstract also points to a series of comments and author replies
Full Text Mirror
2nd Full Text Mirror

Iraq (Casualty Estimates)
The Secretary of State for Foreign and Commonwealth Affairs (Mr. Jack Straw)
17 November, 2004

100,000 Civilian Deaths Estimated in Iraq
By Rob Stein
Washington Post
October 29, 2004

Iraqi Civilian Deaths Increase Dramatically After Invasion
Johns Hopkins Bloomberg School of Public Health
October 28, 2004

Horrific death toll in Darfur revealed
12:41 01 October 2004 news service
Katharine Davis

Retrospective Mortality Survey Among the Internally Displaced Population, Greater Darfur, Sudan, August 2004
World Health Organization
European Programme for Intervention Epidemiology Training
15 September 2004

Mortality in the Democratic Republic of Congo: Results from a Nationwide Survey
Conducted September November 2002
by Les Roberts, Pascal Ngoy, Colleen Mone, Charles Lubula, Luc Mwezse, Mariana Zantop, Michael Despines
International Rescue Committee
Reported April 2003

Kosovo Living Standards Measurement Study Survey 2000
Poverty and Human Resources
Development Research Group
The World Bank
October 2001

Description and comparison of the methods of cluster sampling and lot quality assurance sampling to assess immunization coverage
by Stacy Hoshaw-Woodard, Ph.D
Center for Biostatistics, The Ohio State University
World Health Organization, Geneva
WHO/V&B/01.26, 2001

War and mortality in Kosovo, 199899: an epidemiological testimony
Paul Spiegel, Peter Salama
International Emergency and Refugee Health Branch, National Center for Environmental Health, Centers for Disease Control and Prevention
The Lancet, Volume 355, Issue 9222, Pages 2204-2209
PMID: 10881894 [PubMed - indexed for MEDLINE]

The Rapid Assessment of Health Problems in Refugee and Displaced Populations
Michael J. Toole, M.D.
Medicine & Global Survival (M&GS) 1994;1:200-207

Gordon Housworth

InfoT Public  Strategic Risk Public  Terrorism Public  


  discuss this article

<<  |  September 2019  |  >>
view our rss feed