Jana Asher

As most statisticians and survey methodologists know, social data collected under any circumstance can have problems associated with them. Missing data, incorrect codes, implausible responses, and other issues appear in data collected using even the best survey methods. However, data on casualties caused by political violence or war are collected under some of the most trying circumstances. The best-conceived sampling plan might fall prey to an instantly obsolete sample frame or sampled clusters that have become unstable and are unsafe to enter. The population to be surveyed is almost always traumatized and fearful. Often, such data must be collected in developing or transitional countries, and the population of interest could represent multiple languages or culture identities.

After the violence has ceased, gathering data on what happened—even pre-existing data—is almost as problematic. There might have been good media coverage (e.g., television and newspaper) of local events and episodes of violence, or there might have been no media access. Administrative records might or might not be available, or might be in paper format and require long hours of data entry. Qualitative or quantitative data collected by nongovernmental organizations might or might not be of sufficient quality to be useful for statistical analysis. If new data collection is required, it will have many of the same issues as data collection efforts would have had during the actual period of violence.

Clearly, recording and estimating casualties during political violence is not easy. Many have attempted such estimation using methods known to them through their academic and professional fields. Until recently, those diverse methods for recording and estimating casualties have not been subject to much scrutiny. That changed with the release of two studies published in The Lancet: “Mortality Before and After the 2003 Invasion of Iraq: Cluster Sample Survey,” by Les Roberts, Riyadh Lafta, Richard Garfield, Jamal Khudhairi, and Gilbert Burnham, and “Mortality After the 2003 Invasion of Iraq: A Cross-Sectional Cluster Sample Survey,” by Gilbert Burnham, Riyadh Lafta, Shannon Doocy, and Les Roberts.

After those studies became publicly available, a political and scientific controversy that echoed across the scientific community ensued. Advocates of projects such as the Iraq Body Count and studies such as the World Health Organization Iraq Family Health Survey (PDF) raised questions about the validity of the methods employed and suggested different approaches.

Partly in response to that controversy and partly to understand and develop a scientific consensus on the best approaches for recording and estimating casualties, a group of researchers at Carnegie Mellon University and the University of Pittsburgh hosted a conference on casualty recording and estimation in times of conflict. The 30 experts on war-related casualties who participated represented an array of academic fields, including statistics, demography, epidemiology, political science, history, sociology, psychology, and survey methodology.

Over two days, several established methods for recording, estimating, and describing war-related deaths were discussed, including demographic approaches, epidemiological approaches, multiple systems estimation, random sample surveys, repeat interviews, and complete enumerations through newspaper reports or other qualitative records. Candid discussions about each method demonstrated both strengths and weaknesses. The following three themes seemed to emerge:

• No one method is best for recording and/or estimating casualties. In fact, because of the myriad situations within which casualty estimation is sought, a ‘toolbox’ of methods is required.
• Although casualty estimation has drawn upon many academic fields, the science of that estimation is complex enough to warrant the development of a unique academic field focused on its development.
• The science of casualty estimation requires far more development.

The conference organizers plan to produce a book based on the presentations given to capture the methodologies for casualty estimation as they exist now, but they also want to answer the call for developing better casualty estimation science in the future.

The convening of this conference represents a strong first step toward developing a science for casualty recording and estimation; however, much work remains and statisticians and survey methodologists will play a vital role.