In 1994, in response to a request by senior U.S. policy makers, the State Failure Task Force was established to design and carry out a data-driven study of the preconditions of state failure, defined to include ethnic and revolutionary wars, adverse or disruptive regime transitions, and genocides and politicides. In 1998, in response to President’s Clinton’s policy initiative on genocide early warning and prevention, Barbara Harff, a senior consultant of the Task Force, was asked to design and carry out a study that would use her and other data to establish a workable and theoretically sound data-based system for risk assessment and early warning of genocidal violence. The results of this effort are published in a recent issue of the American Political Science Review (see below) and the data are available for replication via a University of Maryland website. The latest structural model, summarized in Table 1, identifies six causal factors that jointly differentiate with 74% accuracy the 36 serious civil conflicts that led to episodes of genocidal violence between 1955 and 2002 and 93 others that did not. Accuracy increases to nearly 90% when temporal inconsistencies in the data are taken into account. These results are very good given the limits of social science indicators and analytical techniques.

Interpretation: When the model is applied to current data, it provides the basis for a watch list of countries at risk of future genocidal violence. The author has constructed such a watch list by applying the model to information on current conflicts and groups at risk of victimization. The results are shown in Table 2. The highest-risk countries for the foreseeable future are Sudan – despite the ongoing peace process – plus Burma, Burundi, Rwanda, and Congo. Each has five of six risk factors. Next are four countries with four each: China, Algeria, Uganda, and Somalia. The conflict-challenged countries with three risk factors include Afghanistan and Iraq – despite US-led regime changes – along with Pakistan and Ethiopia.

Risk assessment of this quality and generality does not exist elsewhere so far as we know. Since the data and model are publicly available, the study can easily be updated yearly and the watch list replicated by social scientists familiar with quantitative social science techniques, for example by NGO’s concerned with anticipating humanitarian crises.

Problems and Issues: Some contend that there is no shortage of early warnings, and that the obstacle to early action is the lack of political will. We agree with the latter statement but not the former. There have been many warnings by area specialists and human rights groups, but none early enough and none that routinely scans the globe for signs of impending disaster. By early we mean months in advance, not after killings are underway.

Risk Assessment, Early Warning, and Early Response: Whereas risk assessment is better than what we had before, it is not enough to tell us more precisely WHEN genocidal violence is likely to begin. What it tells us is that a country is in the latter stages of upheaval that may result in genocide. This alone can be enough to prompt preventive action. In other words it is then that less costly approaches, i.e. financial, humanitarian or rescue operations combined with subtle or not so subtle political pressures, could work to prevent genocidal escalation. To bridge the gap between risk assessment and the onset of genocidal violence, the Clinton Administration (under the leadership of then Vice President Gore), supported a pilot study, also designed by Harff, to monitor on a daily basis countries identified at high risk. The theoretical underpinnings of this study were published in 1998 in the Journal of Peace Research (a Norwegian journal; see below). The theoretical base is extremely complex using 10 factors and triggers that are measured by observing political events. It requires us to track roughly 70 indicators on a daily basis. Initial results suggest that we can narrow the time frame and identify warning flags that a genocide is in the making a few months prior to its onset. The encouraging results have led to efforts to automate procedures for coding news reports. Automated coding would allow us to process thousands of pieces of information in
minutes. Unfortunately the information gathered thus fare is not publicly available. But the author of this paper plans to work with NGO’s to develop a simplified tracking device can be provided that will help policy makers analyze and provide early warnings in specific situations.

The theoretical basis of the early warning system described above is anchored in the genocide literature. The definition of genocide and politicide used closely resembles those employed by other scholars. Thus discussion can focus on prevention rather than debating etiology or definitions. Moreover the risk assessment and early warning systems are easy to understand and can be simplified for use by policy makers and observers in the field who are not necessarily familiar with social science techniques and jargon. I believe what is most needed now are preventive tools that are tailored to the specific needs of particular communities at a particular time. The next big challenge for early warning research is to learn more about what works to prevent genocidal violence in which kind of situations and at which time.

Note:
This paper reflects the views of the author and not those of the US government. Although the research has been funded in part by the U.S. government,Task Force analysis and modelling results do not represent the views of any branch of the US government.

References:
Barbara Harff, “No Lessons Learned from the Holocaust? Assessing Risks of Genocide and Political Mass Murder since 1955,” American Political Science Review vol. 97 (February 2003), pp. 57-73.
Barbara Harff and T. R. Gurr, “Systematic Early Warning of Humanitarian Emergencies,” Journal of Peace Research vol. 35 (No. 5,1998), pp. 551-579.
Website: http://www.cidcm.umd.edu/inscr/stfail