ViEWS: A political violence early-warning system
| Published date | 01 March 2019 |
| DOI | 10.1177/0022343319823860 |
| Date | 01 March 2019 |
| Subject Matter | Research Articles |
Research Articles
ViEWS: A political violence
early-warning system
Håvard Hegre
a,b
, Marie Allansson
a
, Matthias Basedau
a,d
,
Michael Colaresi
a,c
, Mihai Croicu
a
, Hanne Fjelde
a
, Frederick Hoyles
a
, Lisa Hultman
a
,
Stina Ho
¨gbladh
a
, Remco Jansen
a
, Naima Mouhleb
a
, Sayyed Auwn Muhammad
a
, Desire
´e Nilsson
a
,
Håvard Mokleiv Nygård
a,b
, Gudlaug Olafsdottir
a
, Kristina Petrova
a
, David Randahl
a
,
Espen Geelmuyden Rød
a
, Gerald Schneider
a,e
, Nina von Uexkull
a
, and Jonas Vestby
b
Abstract
This article presents ViEWS – a political violenceearly-warning system that seeks to be maximally transparent, publicly
available, andhave uniform coverage, and sketches the methodological innovationsrequired to achieve these objectives.
ViEWS produces monthly forecasts at the country and subnational level for 36 months into the future and all three
UCDP types of organized violence: state-based conflict, non-state conflict, and one-sided violence in Africa. The article
presents the methodology and data behind these forecasts, evaluates their predictive performance, provides selected
forecasts for October 2018 through October 2021, and indicates future extensions. ViEWS is built as an ensemble of
constituent models designed to optimize its predictions. Each of these represents a theme that the conflict research
literaturesuggests is relevant, or implements a specificstatistical/machine-learning approach.Current forecasts indicatea
persistenceof conflict in regions in Africawith a recent history of politicalviolence but also alertto new conflicts such as in
Southern Cameroon and Northern Mozambique. The subsequent evaluation additionally shows that ViEWSis able to
accuratelycapture the long-term behaviorof established politicalviolence, as well as diffusionprocesses such as the spread
of violencein Cameroon. The performancedemonstratedhere indicates that ViEWScan be a useful complementto non-
public conflict-warning systems, and also serves asa reference against which future improvementscan be evaluated.
Keywords
Africa, armed conflict, forecasting
ViEWS: Guiding principles
Large-scale political violence kills thousands every month
across the globe and forces many more to relocate within
countries and across borders. Armed conflicts have dis-
astrous economic consequences, undermine the func-
tioning of political systems, prevent countries from
escaping dire poverty, and hinder humanitarian assis-
tance where most needed.
The challenges of preventing, mitigating, and adapt-
ing to large-scale political violence are particularly daunt-
ing when it escalates in locations and at times where it is
not expected. Policymakers and first responders would
benefit greatly from a system that systematically moni-
tors all locations at risk of conflict and assesses the prob-
ability of conflict onset, escalation, continuation, and
a
Department of Peace and Conflict Research, Uppsala University
b
Peace Research Institute Oslo (PRIO)
c
University of Pittsburgh
d
German Institute of Global and Area Studies
e
University of Konstanz
Corresponding author:
havard.hegre@pcr.uu.se
Journal of Peace Research
2019, Vol. 56(2) 155–174
ªThe Author(s) 2019
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/0022343319823860
journals.sagepub.com/home/jpr
geographic diffusion. This article presents ViEWS – a
political violence early-warning system — which seeks to
address this need. We outline the methodological frame-
work and evaluate the predictive performance of the
system as of 1 October 2018.
The forecasting task ViEWS has set out is multidi-
mensional. ViEWS provides forecasts 36 months into
the future for three types of political violence: armed
conflict involving states and rebel groups, armed conflict
between non-state actors, and violence against civilians
(Pettersson & Eck, 2018). The probability that political
violence occurs in a given month is forecasted for both
countries and subnational geographical units. This
means that ViEWS provides forecasts for continued con-
flict as well as new conflicts. To be useful as an early-
warning system, ViEWS has since June 2018 published
monthly updated forecasts for Africa at http://views.pcr.
uu.se/.
1
This is made possible by the monthly release of
candidate events from the Uppsala Conflict Data Pro-
gram (UCDP; see Hegre et al., 2018).
The ViEWS forecasts build on a number of constitu-
ent models drawing on insights from decades of quanti-
tative peace and conflict research. Some of the models are
thematic, concentrating on topics such as conflict his-
tory, the economy, political institutions, and geography.
Others are more general, combining multiple themes or
using information at the country and the subnational
level to generate forecasts. We subsequently combine the
forecasts from these individual models to ensembles. Our
evaluation shows that the ViEWS ensembles improve
forecasting of political violence at both the country and
the subnational level compared to multiple tough base-
line models.
The forecasts from October 2018 indicate that con-
flict will persist up to and beyond 2021 in several coun-
tries that have a recent history of political violence, such
as Burundi, Nigeria, and DR Congo. The system also
alerts to new conflicts in Southern Cameroon and
Northern Mozambique.
The aims of ViEWS are maximal transparency, uni-
form coverage, and public availability. Transparency
requires that the risk assessments can be traced back to
a fully specified argument and accessible information,
allowing readers and potential users to evaluate what lies
behind the forecasts. ViEWS is therefore exclusively
based on publicly available data. Moreover, its results,
input data, and procedures are available to researchers
and the international community. Uniform coverage of
the regions at risk helps to alert observers to locations
that receive insufficient attention. In principle, ViEWS
seeks to be able to issue a warning with equal probability
for any location independent of its geo-strategic impor-
tance, past conflict history, or current humanitarian sit-
uation. Public availability of the results is useful for
domestic actors and small international NGOs, and
essential to ensure transparency regarding decisions they
might make based on these results.
In the following, we first briefly review the litera-
ture that has informed ViEWS. Second, we outline
themethodologicalframework.Third,weevaluatethe
performance of the system and present the ViEWS
forecasts from October 2018 to October 2021.
Finally, we discuss future extensions of the system
and conclude.
Literature review
Prediction has long been considered a core task for peace
research (Singer, 1973), and a comprehensive literature
review is beyond the scope of this article (rather, see
Schneider, Gleditsch & Carey, 2010; Hegre et al.,
2017). Conflict forecasting has taken a number of meth-
odological approaches, for example game theory (Bueno
de Mesquita, 2010), machine-learning tools such as
neural networks (Schrodt, 1991), and algorithms for
automatic coding of event data (Schrodt, Davis & Wed-
dle, 1994). Ward, Greenhill & Bakke (2010) arguably
represents a turning point, bringing prediction into the
mainstream of peace research.
ViEWS builds on innovations in the academic early-
warning systems for conflict that have been proposed
since the 1970s (Andriole & Young, 1977). The State
Failure/Political Instability Task Force (PITF) aimed to
forecast political crises two years in advance (Esty et al.,
1995; Gurr et al., 1999; Goldstone et al., 2010). A key
insight from PITF is that simplistic models with a few
powerful variables performed as well as complex models,
at least at the country-year level. The Integrated Crisis
Early Warning System (ICEWS) focused on a range of
domestic and international crises (O’Brien, 2010). Valu-
able insights from ICEWS include separate modelling of
conflict phases (onset, continuation, termination) and
the utility of a multimethod approach to forecasting.
As the literature has matured, real-time forecasts have
become increasingly common (Brandt, Freeman &
Schrodt, 2011; Ward & Beger, 2017). Some of these
are publicly available. For instance, the US Holocaust
Memorial Museum has a regular updated early-warning
1
Given sufficient funding to cover the required data-collection
needs, these ambitions will be scaled up to a wider geographic scale.
156 journal of PEACE RESEARCH 56(2)
To continue reading
Request your trial