Executive Summary
Contents:
1. Context: Current
Conflict Trends in Sub-Saharan Africa
2. Objectives and Value
of Conflict Early Warning Systems
3. Stages of Conflict or
Crisis Development
4. Information Sources
and Types of Early Warning Analysis
5. Risk Assessment and
Early Warning Models: Empirical Findings
6. Linking Early Warning
to Collaborative Early Response and Preventive Action
7. Implications for
Practice
I. Context: Current Conflict Trends in Sub-Saharan Africa
The number and magnitude of wars and related disasters increased
globally throughout the Cold War, peaked in 1990-1993, and have fallen sharply since then,
with both a marked decline in the onset of new wars and an unprecedented success rate in
containing or ending existing wars (Marshall, 1998; Ayers, 2000; Gurr, 2000a & b;
Davies and Gurr, forthcoming). This success reflects a strengthening of norms, capacity
and political will in favor of protecting at-risk minority groups and preventing the
escalation and spread of political violence.
There remain significant levels of instability, with new democratic
norms and institutions under challenge in many parts of the developing world, and it is
these conflicts that continue to dominate the news, rather than successes in conflict
management and prevention. Positive trends in Sub-Saharan Africa are in any case more
modest, with six high intensity wars still current in early 2000, twenty more minority
groups substantially at risk of future wars (Gurr, 2000c), and with the Central
African/Greater Horn region yet to share the positive trends experienced elsewhere.
Nevertheless, these trends point to an opportunity to provide international support for
strengthening capacity for conflict management and prevention throughout this troubled
region.
II. Objectives and Value of Conflict Early Warning Systems
The goal of early warning systems may be conceived as avoiding or
minimizing violence, deprivation or humanitarian crises that threaten the sustainability
of human development. Human development implies improving quality of life through
expanding human and social capital, as required to adapt to change and better satisfy
human needs such s those for security, identity/valued relationships and effective
participation.
Reliable early warnings buy time not only to prepare for short term
containment and relief strategies, but also to design, build support for and implement
longer term proactive strategies and development programs that can reduce the likelihood
of future disasters. An early warning system is more than the flow of information and
reports from those on the ground regarding highly visible or rapidly escalating crises. It
should also provide reliable analyses that identify still-latent or low-level conflicts or
instabilities before they attract CNN and international headlines, when there is still
time for investing in appropriate structural development and building conflict management
capacity.
More than that, early warning systems can generate analyses that
identify key factors driving the instability, providing a basis from which to assess
likely future scenarios, and recommend appropriate options for local and international
policy makers oriented toward preventive action. They also should keep track of what
preventive strategies have been used in what contexts in the past, to what effect and at
what cost.
Such analyses improve the reliability of conflict early warnings
(current model-based warnings have been found accurate close to 80% of the time—see
below), and can be used to build political will and coalitions among potential responders
for appropriately designed preventive action, at a fraction of the cost of reactive
humanitarian and peace-keeping initiatives. In addition, these coalitions should include
where possible, domestic as well as international, private as well as public sector
partners, thereby significantly adding to the strength of civil societies and democratic
institutions, and their capacity to manage their own conflicts in the longer term.
Along with disease (AIDS) and “natural” disasters such as
drought, ethnic wars and repression within states constitute an enormous threat to human
security in Sub-Saharan Africa. Yet much more needs to be done to develop systems for
providing early warning of violent civil conflict and of related phenomena, such as mass
killings (genocide or politicide), gross human rights violations, regime failure, refugee
and IDP flows, environmental degradation and food shortages.
III. Stages of Conflict or Crisis Development
Three stages of conflict or crisis development will need to be
distinguished for purposes of building early warning systems (see e.g., Lund. 1996):
i. Structural
tensions or instability: Potential crises at this stage may be identified through
tracking background or structural conditions that constitute root causes of tension and
crises. Examples include a history of state repression, exclusionary ideologies, lack of
democratic experience, increasing gaps in income and economic opportunity, high cohesion
and external support for aggrieved groups, land desertification, and increasing population
pressures. These structural conditions tend to change slowly, and so form the basis for
long-term risk assessments: the probability that a crisis will occur at some time
in the next several years or decade, and are needed for preventive peace-building or
peace-making initiatives and long-term planning.
ii. Escalation:
Potential crises at this critical stage may be identified through tracking dynamic factors
or “accelerators,” which may exacerbate the underlying conditions, driving up
tensions. Examples include arms or resource acquisitions, incidents of aggressive
posturing or low-intensity violence, new discriminatory and repressive policies, crop
failures, and major currency devaluations. These are events that unfold rapidly and
provide the basis for dynamic early warnings: indicating the probability of a
crisis within the coming months or weeks, as needed for last minute preventive diplomacy
and containment efforts, or at least in planning for humanitarian aid and peace-keeping.
iii. Crisis/War: The
transition to a full-blown humanitarian crisis or war may be marked by trigger incidents
such as a coup attempt, assassination, or declared state of emergency that may act as a
match to ignite the flammable mix of dry trees or logs (structural factors) with dry brush
or lighter fluid (accelerators). At this stage the opportunity for prevention has passed,
but the need for relevant information and (comparatively high-risk, high-cost) reaction is
no less urgent.
Early warning and early response systems will require both structural
risk assessment processes to point to opportunities for appropriate and well-planned
preventive action to address structural problems, and linked dynamic early warnings to
flag the need for more immediate containment efforts.
IV. Information Sources and Approaches to Early Warning
The more conventional approach to early warning among operational
agencies has involved the preparation of internal, interpretive reports for policy
makers, focusing on the substance of the assessment and the need to package the report to
bring out available policy options within the constraints on the agency’s capacity to
respond. These are typically based on field monitoring and in-country situation studies,
supplemented by limited but increasingly systematized sharing of field reports and
indicator monitoring, through mechanisms such as UNOCHA’s Humanitarian Early Warning
System (HEWS) for UN agencies, or its publicly available, internet-based ReliefWeb and
African IRIN reports. UNHCR’s Center for Documentation and Research similarly has
begun to make versions of its more structured country situation assessments for UN
interagency meetings also publicly accessible via the web.
A complementary approach is the generation of analytical reports
employing explicit structural or dynamic models of crisis phenomena. Models provide an
essential complement to field monitoring and indicator monitoring through specifying the
combinations of risk factors and sequence of events that are likely to lead to different
types of crises, distinguishing between remote and proximate conditions. This approach has
been favored by academic researchers, and its advantages include being accessible for
systematic testing and progressive refinement and adaptation based on cumulative results,
and providing a means for identifying the key factors to be addressed for effective
prevention.
These model-based analyses may draw on country or group profiles or
databases of structural indicators (e.g., Gurr’s, 2000a, Minorities at Risk data),
coded assessments by country experts (e.g., Scarborough, 1998; Schmid, 1998), and/or coded
analyses of multiple filtered, publicly accessible news and information sources (e.g.,
Harff and Gurr, 1998). Analysis of data from both risk assessment and dynamic early
warning models can make use of powerful statistical tools for complex pattern recognition,
such as neural net analysis or time-series impact assessment analysis, to distinguish
where and when escalation and crises are likely.
Another rich information source is to be found in episodic databases
which profile past conflicts and crises, including information on crisis development and
the effectiveness of attempts at crisis prevention or management (e.g., Brecher and
Wilkenfeld, 1988; Bercovitch, 1996; Bloomfield and Moulton, 1997). This can provide an
empirical basis for recommending early response options which are likely to be more
effective in current potential crises.
V. Risk Assessment and Early Warning Models: Empirical Findings
The most extensive empirical research evaluation of risk assessment and
early warning models has been conducted as part of the ongoing White House-initiated State
Failure Project (e.g., Esty et al., 1998) and the derivative Genocide Early Warning Center
in the Department of State. State failures (including partial failures, with 243 cases
between 1955 and 1994) were divided into four categories: revolutionary wars, ethnic wars,
genocides and politicides, and adverse or disruptive regime transitions.
Phase 1 of the project compared these with 339 matched control cases,
617 variables were evaluated as potential structural indicators, leading to identification
of 31 demographic/social, political, and economic/environmental variables that were
significant discriminators of impending state failure at least two years in advance. The
best overall risk assessment model (using regression and neural net analyses, and about
70% accurate) combined three of these variables:
i. Openness to
international trade (best measure of a country’s integration into the global economy,
requiring respect for international norms and rule of law);
ii. Infant mortality (best
measure of quality of life); and
iii. Democracy (measuring the
conflict inhibiting effects of democratic governance, requiring responsiveness to popular
discontents).
More specific models were developed and tested for the three more common
forms of state failure: ethnic war, regime collapse and genocide/politicide. Inclusion of
variables for the ethnic character of the ruling elite and for the existence of a
demographic “youth bulge” improved accuracy of the ethnic war model to 72-78%,
and similar levels of accuracy have since been achieved using more detailed and
theoretically grounded models for ethnic war, regime crises and genocide/politicide.
Phase 2 of the project focused on the development and testing of dynamic
early warning models to supplement the structural models. These “accelerator”
models are designed for use with cases that have been identified as high risk using the
structural models, to track events in order to anticipate when an unstable situation is
likely to escalate to full-blown crisis (war, genocide or adverse regime transition) in
the following months. These models each allow accurate anticipation of when a crisis will
erupt six months in advance, about 78% of the time. The genocide/politicide model is now
being tested real time as a project of the Genocide Early Warning Center.
More detailed versions of the ethnic war and genocide/politicide models,
both structural and dynamic, are available (Harff and Gurr, 1998; Gurr, 2000a), with
guidelines for measuring each variable. More recent research findings will be published
shortly, including details of the statistical procedures that are required to take
advantage of the power of these models. These procedures and findings, and details of the
as yet unpublished regime transition model, should also be available internally to USAID
from the State Failure Project.
The structural model for ethnic war, for example, includes factors
reflecting group incentives for collective action (lost political autonomy, active
political, economic or cultural discrimination); group capacity for collective action
(strength of identity, militant mobilization); and opportunities for collective action
(recent regime transition, support from kindred groups). In contrast, the structural model
for genocide/politicide includes factors such as: history of the regime’s reliance on
coercion, international economic status, exclusionary ideologies, lack of constraints on
security agencies, elite (and target group) fragmentation, and economic hardship that
leads to differential treatment of disadvantaged groups.
Research has also been conducted to test risk assessment models for
other conflict-related crises such as refugee and IDP flows (e.g., Schmeidl and Jenkins,
1998; Onishi, 1998) and environmental conflicts (e.g., Baechler, 1998). The more
specifically targeted models should allow greater accuracy, but require greater investment
to the degree that multiple models are needed to anticipate the full range potential
crises.
VI. Linking Early Warning to Collaborative Early Response and
Preventive Action
Some success is being achieved through the development of collaborative
partnerships or networks that bring together government or IO agencies with NGO’s and
academic research institutions concerned with conflict and crisis early warning. Examples
include the EU-supported Conflict Prevention Network, the OSCE’s Conflict Prevention
Center, the London-based Forum on Early Warning and Early Response (FEWER), the
Utrecht-based European Platform for Conflict Prevention and Transformation, and Center for
Preventive Action supported by the New York based Council on Foreign Relations.
FEWER, for example, aims to inform and promote coordinated and early
responses to violent conflict through the provision of response-oriented analyses of
conflict areas to key actors with relevant conflict prevention capacities. FEWER is a
global consortium of NGO’s and academic research centers which is building global and
regional networks and preventive action partnerships with governmental and IO agencies.
Emphasis is placed on utilizing and enhancing the capacities and skills of local member
organizations, supported by international members, in developing early warning reports up
to international standards, in developing regional conflict prevention tool-boxes
appropriate to local conditions, convening strategic roundtables for the development of
action plans for peace, convening governmental/inter-governmental contact groups where
policy makers are convened to discuss responses to conflict generating factors (as
recently for the Djavakheti region of southern Georgia, on the border with Armenia), and
pooling non-governmental conflict prevention capacities where needed to respond to urgent
escalating situations.
VII. Implications for Practice
The early successes of such collaborative initiatives, and the empirical
research findings reviewed above, point to certain principles and priorities for the
development of conflict early warning and response capacity in USAID’s Africa Bureau.
These include (see Gurr, 2000c):
i. High priority should be given to identifying and
monitoring latent and emerging crises well in advance of the outbreak of war or related
forms of humanitarian crisis.
ii. To this
end, an inventory of ethnic, religious, national and political groups at risk from adverse
discrimination or potential conflict in Sub-Saharan Africa should be compiled. A starting
point for this is available from the Minorities at Risk data.
iii. Explicit
models specifying what risk factors are known or thought to be associated with what types
of conflict or crisis should be used to aid interpretation of available information.
iv. Long-term
risk assessments from structural factors should be supplemented by near-real-time tracking
of events in unstable (high-risk) areas to anticipate escalation of crises.
v. Monitoring and analyses should be done outside the
reach of political control and considerations, to ensure that results are, and are seen to
be, objective. This can be accomplished by engaging non-government and academic
participants in an early warning network, as well as through ensuring operational
independence of analysts within the Africa Bureau.
vi. Use of
standard protocols for generating risk assessments and early warnings helps ensure both
comprehensiveness and precision in the reports, and allows comparison of assessments both
across different locations, and form different sources, as needed for deciding priorities
in the allocation of limited resources for response.
vii. It is
essential to develop close and regular communications between early warning analysts and
officials with operational responsibility for preventive action and humanitarian response.
This is required to ensure that early warning reports contain information useful for
informed decision making, including drawing out the implications of structural and dynamic
analysis into likely future scenarios and identification of a range of early response
options which are appropriate to the situation and within operational and policy
constraints of potential intervenors. It will also give the early warnings a greater level
of credibility, as needed for an early warning system to be effective.
viii. Such contact
groups should also be regularly engaged in constructing and reviewing tool boxes of
regionally appropriate preventive or early response options, drawing lessons from past
applications within the region and elsewhere.
ix. Early
warnings and risk assessments should be widely circulated, most urgently to those in a
position to contribute to collaborative preventive initiatives, and also to other
NGO’s, scholars, activists and journalists within and beyond the region as far as is
possible without risking the safety of those directly involved. This will help to promote
public discussion of the issues, build political support for preventive action, and
generate new ideas and action responses.
x. USAID
has a responsibility to support the building of local capacity for conflict prevention, as
needed for strengthening civil society, democratic norms and sustainable peace and
development in the long term. In addition, by engaging local professionals in the early
warning process, AID will avoid the resentment and suspicions of bias that would follow
release of comprehensive in-house assessments, while still being able to use the reports
to mobilize cooperative efforts with broad public support.
