‘Capital punishment’
| DOI | 10.1177/0022343314552717 |
| Date | 01 March 2015 |
| Author | Charles Butcher |
| Published date | 01 March 2015 |
| Subject Matter | Research Articles |
‘Capital punishment’: Bargaining
and the geography of civil war
Charles Butcher
National Centre for Peace and Conflict Studies, University of Otago
Abstract
Civil wars show substantial variation in where they are fought. One dimension of this variation is the proximity of
fighting to the capital city. While some wars are fought in the periphery, others devastate capital cities, often for
months, or years, on end. What explains this? This article approaches the puzzle from a bargaining perspective and
argues that wars with evenly balanced belligerents (bipolar conflicts) should be less likely to see fighting in the capital
while wars with multiple, evenly matched belligerents (multipolar conflicts) should be more likely to do so. Empirical
analysis of new conflict-year data on the location of fighting in civil war and measures of conflict actor ‘fractionaliza-
tion’ and ‘polarization’ from 1975 to 2011 support these claims. Highly asymmetric conflicts are fought furthest
from the capital city. Bipolar conflicts are fought closer to the capital, but only modestly so. The transition from
a bipolar to a multipolar conflict sharply increases the risk of fighting within 10 km of the capital and decreases the
expected distance of conflict from the capital. In general, this article points to the utility of bargaining theory to help
explain spatial patterns in violent conflict, in addition to questions of onset, duration, and termination, to which this
theory has been traditionally applied.
Keywords
civil war, bargaining, geography
Introduction
Civil wars exhibit variation in where they are fought,
especially in relation to the capital city. Fighting in
Afghanistan occurred primarily in the countryside dur-
ing the 1980s before Kabul was attacked repeatedly from
1992 to 1996. Some capitals, such as Bamako or Nia-
mey, never experience civil war related violence, and
some, such as Luanda in Angola, experience compara-
tively low levels. Some capitals, such as Monrovia and
Freetown, are subject to repeated episodes. What
explains this variation? Understanding the spatial dimen-
sions of civil war is important. Capital cities are often a
country’s most economically productive and populous
location and the destruction and displacement visited
upon these cities sets recovery and post-conflict peace-
building back many years. Moreover, the ‘battle for the
capital’ is often the most destructive phase of these con-
flicts. Kabul was nearly destroyed in the period 1992–96.
Perhaps 10,000 people were killed in the battle for Braz-
zaville in 1997 and 14,000 in the 1991 battle for
Mogadishu. Thousands more were killed in fighting over
Monrovia, and the horrific images of ‘Operation No Liv-
ing Thing’ in Freetown in 1999 endure in documen-
taries such as Cry Freetown and in popular culture
(Blood Diamond). Anecdotal and statistical evidence sug-
gests that the onset of fighting in capital cities attracts
foreign military intervention (Butcher, 2011: 217),
deters mediation efforts, and makes them more likely
to fail when they occur (Greig, 2014). Understanding
the characteristics of these conflicts that attract the atten-
tion and resources of the international community is an
important first step to designing policy responses that
can effectively stop them.
There is an emerging literature on the spatial dimen-
sions of civil war (Buhaug & Rød, 2006; Buhaug, 2010;
Raleigh & Hegre, 2009; Hegre, Østby & Raleigh, 2009;
Buhaug, Gates & Lujala, 2009; Rustad et al., 2008;
Corresponding author:
charles.butcher@otago.ac.nz
Journal of Peace Research
2015, Vol. 52(2) 171–186
ªThe Author(s) 2014
Reprints and permission:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0022343314552717
jpr.sagepub.com
Buhaug, Cederman & Rød, 2008; Cederman, Buhaug &
Rød, 2009; Wucherpfennig et al., 2011; Weidmann &
Ward, 2010) including a handful of studiesexamining the
proximityof fighting to the capital cityacross different civil
wars. Quantitative studies show that separatist conflicts
start, and are fought, further from capital cities than con-
flicts over government (Buhaug & Gates, 2002; Buhaug
& Rød, 2006; Wucherpfennig et al., 2011; Hallberg,
2012) and studies that either treat conflict geography as
an independent variable (Buhaug, Gates & Lujala, 2009)
or examineregionally restricted samples(Raleigh & Hegre,
2009; Hegre, Østby & Raleigh, 2009) make findings sug-
gesting that stronger rebels fight closer to the capital. This
literature le aves much variation u nexplained, however . The
finding that separatist conflicts are fought furtherfrom the
capital is important but not especially surprising given
that separatist rebels generally do not aim to take the cap-
ital (Buhaug, 2006). In addition, the most interesting
variation (and the most variation) occurs withinthe sam-
ple of governmental conflicts (Hallberg, 2012). That
weak rebels fight further from the capital is also fairly
uncontroversial as they lack the capability to strike the
center of power. Moreover, some strong rebel groups
attack the capital while others do not.W hy,for example,
was Luanda not the location of intense fighting from
1992 to 2002 while Monrovia, Mogadishu, Brazzaville,
and Kabul were? These conflicts all involved belligerents
that were fairly evenly matched. This article argues that
fighting in the capital city becomes more likely when
structural conditions make it difficult for belligerents
to make or accept mutually agreeable bargains. Specifi-
cally, fighting in capital cities should be more likely
when there are three or more evenly matched belliger-
ents (multipolar conflicts) and uncommon when there
are two evenly matched belligerents (bipolar conflicts)
or the conflict is highly asymmetric.
These propositions are tested with a new database of
the minimum distance of fighting from the capital for
the period 1975–2011 based on the UCDP/PRIO
Armed Conflict Dataset (ACD; Themne
´r & Wallens-
teen, 2012) and new measures of conflict actor fractiona-
lization and polarization based on the Non-State Actor
Dataset (NSA; Cunningham, Gleditsch & Salehyan,
2013) and the UCDP Conflict Encyclopedia. Polariza-
tion approximates the extent to which the civil war is
characterized by bipolarity, and fractionalization
approximates the extent to which it is characterized
by multipolarity. The hypotheses are also retested at the
conflict-month using the UCDP Georeferenced Event
Dataset (GED; Sundberg & Melander, 2013). Empiri-
cal results show support for these hypotheses. Conflicts
with high levels of fractionalization are fought substan-
tially and significantly closer to the capital city than
conflicts with low levels of fractionalization. Polarized
conflicts are fought closer to the capital than asym-
metric conflicts but significantly further from the capi-
tal city than fractionalized conflicts. These effects are
independent of other plausible correlates such as the
size of the government army, state capacity, and the
incompatibility. The article also finds evidence to sug-
gest that commitment problems help explain conflict
geography. Internally divided rebel groups appear to
fight closer to the capital city. This likely reflects the
inability of victorious rebel groups to credibly commit
to a post-conflict distribution of state resources or the
willingness of government actors to gamble on weak
rebels splintering before they reach the capital.
The article proceeds as follows. The first section dis-
cusses a theoretical framework linking the proximity of
fighting to the capital city to the likelihood of
information-rich fighting. This is the basis upon which
hypotheses regarding the role of fractionalization and
polarization and their impact on the location of fighting
in civil wars are derived. Descriptions of the data and
methods are then outlined followed by a discussion of
the results. The article concludes with an appraisal of the
hypotheses in light of the evidence, problems with the
evidence, and where this study points to future direc-
tions for research.
Information and the spatial dimensions of
civil war
The article approaches the question of civil war geogra-
phy from a bargaining perspective.
1
This approach
emphasizes that violent conflict is the inefficient product
of commitment problems and information asymmetries
(Fearon, 1995; Powell, 2002, 2006; Reiter, 2003).
2
War
is inefficient because it is costly and belligerents have
incentives to distribute resources based upon the fore-
casted outcome of a war. But war may result if neither
side trusts the other to uphold the terms (commitment
problems) or when incentives to misrepresent military
capabilities, endemic to armed conflicts, obscure credible
communication and create uncertainty about the likely
1
Other studies assume rationality, but do not specifically incorporate
bargaining (Buhaug, Gates & Lujala, 2009; Buhaug, 2010).
2
Bargaining theories also include issue indivisibility but since the
divisibility of the state is assumed to be constant across countries,
or at least across regions (Herbst, 2000), this cannot explain the
variation I am most interested in here.
172 journal of PEACE RESEARCH 52(2)
To continue reading
Request your trial