University of Southern California Dissertations and Theses

Evaluating spatial changes in the rate of insurgency‐violence in Central Africa: the Lord's Resistance Army 2008-2012

(USC Thesis Other)

Evaluating spatial changes in the rate of insurgency‐violence in Central Africa: the Lord's Resistance Army 2008-2012

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

EVALUATING SPATIAL CHANGES IN THE RATE OF
INSURGENCY-VIOLENCE IN CENTRAL AFRICA:
THE LORD’S RESISTANCE ARMY 2008-2012

by

Robert J. Franssen Jr.

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

May 2014

Copyright 2014 Robert J. Franssen Jr.
i

TABLE OF CONTENTS
LIST OF TABLES iv
LIST OF FIGURES v
LIST OF EQUATIONS vii
LIST OF ABBREVIATIONS viii
ABTRACT ix
CHAPTER 1: INTRODUCTION 1
1.1 African Militancy and Counter-insurgency Directives 1
1.2 Existing Research Gaps 4
1.3 Using Geographic Information Systems to Monitor Insurgency Violence 5
CHAPTER 2: BACKGROUND AND LITERATURE REVIEW 7
2.1 The LRA Militancy: From Uganda to Modern-day DRC, CAR, and South Sudan 7
2.2 Area of Operations: People and Topography 9
2.3 LRA Strategy 13
2.4 Notorious LRA Massacres 14
2.5 Counter-insurgency Programs 18
2.6 Conflict Geographies and Theories of Communication and Insurgencies 22
2.7 Use of GIS in Insurgency Studies 26
CHAPTER 3: METHODOLOGY 29
3.1 Study Area and Areal Units of Analysis 29
3.1.1 Modifiable Areal Unit Problem (MAUP) 31
3.2 Data Sources, Variables, and Metrics 33
ii

3.2.1 Armed-Conflict Event Data 33
3.2.2 Civilian Protection Program Data: High-frequency Radio Locations 35
3.2.3 Data Strengths, Assumptions, and Limitations 36
3.3 Methodology-framework 37
3.3.1 Data preparation within ArcGIS 38
3.4 Explored methods 40
3.4.1 Clustering and Heat Map Analysis 40
3.4.2 Polynomial Regression and Density Score 44
3.4.3 Spatial Distribution of Violence: Summaries of Violence Aggregated by Country,
Rates of Change by Territory, and Comparative Changes using Difference-in-Differences 46
3.5 Summaries of Violence Aggregated by Country 48
3.6 Annual Rates of Change in the Spatial Distribution of Violence by Territory 48
3.7 Comparative Changes Represented by a Difference-in-Differences Descriptive Analysis 51
3.8 Expected Outcome 53
CHAPTER 4: RESULTS 55
4.1 Summaries of Violence Aggregated by Country 55
4.2 Annual Rates of Change in the Spatial Distribution of Violence by Territory 58
4.2.1 Baseline Distributions by Category of Violence 60
4.2.2 Interval Rate Changes by Category of Violence 63
4.3 Comparative Changes Using a Difference-in-Differences Descriptive Analysis 75
CHAPTER 5: DISCUSSION AND CONCLUSIONS 80
iii

5.1 Key Observations 80
5.2 Contrast with Previous Studies 82
5.3 Recommendations for Future Research Directions 83
5.3.1 Comparative Analysis for High-frequency Radio Distribution 84
REFERENCES 86
APPENDIX A: ARMED-CONFLICT DATA COLLECTION METHODS 94
APPENDIX B: RAW COUNTS BY TERRITORY 95

iv

LIST OF TABLES
Table 1: Territory Areas and Population Estimates 31
Table 2: Territory Areas and Population: Mean, Max, and Min. 32
Table 3: Armed-Conflicts by Territory, 2008-2012: Normalized by Population 49
Table 4: Civilian Murders, 2008-2012: Normalized by Population 49
Table 5: Civilian Abductions, 2008-2012: Normalized by Population 50
Table 6: Variables referenced in the analysis of comparative changes 53
Table 7: Data collection codebook methodology 94
Table 8: Armed-Conflicts - Raw Counts by Territory 95
Table 9: Civilian Murders - Raw Counts by Territory 96
Table 10: Civilian Abductions - Raw Counts by Territory 96

v

LIST OF FIGURES
Figure 1: Dem. Rep. of Congo, Central African Republic, and South Sudan 2
Figure 2: GlobCover 2009: 300m Mosaic Land Cover Classification 12
Figure 3: Locations of Christmas Massacres (HRW 2009) 15
Figure 4: Map of Makomba Massacre (HRW 2010) 16
Figure 5: Generalized HF Radios 2010-2012 indicating approximate location 21
Figure 6: Map representing Administrative Boundaries in the DRC, CAR, and South Sudan 30
Figure 7: Map of LRA Armed-Conflicts 2008-2012 35
Figure 8: High-level Methodology-Framework 38
Figure 9: Armed-conflicts Heat Map: 2008 41
Figure 10: Armed-conflicts Heat Map: 2009 41
Figure 11: Armed-conflicts Heat Map: 2010 42
Figure 12: Armed-conflicts Heat Map: 2011 42
Figure 13: Armed-conflicts Heat Map: 2012 43
Figure 14: Armed-conflicts Heat Map: 2008-2012 43
Figure 15: Aggregating categories of violence by Country and exporting line graphs 48
Figure 16: Calculating rates of change by territory and comparative changes from Difference-in-
Differences 51
Figure 17: Conceptualizing comparative changes between territory groups 52
Figure 18: Trends of Armed-conflicts by Country (raw counts) 56
Figure 19: Trends of Civilian Murders by Country (raw counts) 57
Figure 20: Trends of Civilian Abductions by Country (raw counts) 58
Figure 21: The Rate of Armed-conflicts in 2008 60
vi

Figure 22: The Rate of Civilian Murders in 2008 61
Figure 23: The Rate of Civilian Abductions in 2008 62
Figure 24: Change in the Rate of Armed-conflicts from 2008 to 2009 63
Figure 25: Change in the Rate of Civilian Murders from 2008 to 2009 64
Figure 26: Change in the Rate of Civilian Abductions from 2008 to 2009 65
Figure 27: Change in the Rate of Armed-conflicts from 2009 to 2010 66
Figure 28: Change in the Rate of Civilian Murders from 2009 to 2010 67
Figure 29: Change in the Rate of Civilian Abductions from 2009 to 2010 68
Figure 30: Change in the Rate of Armed-conflicts from 2010 to 2011 69
Figure 31: Change in the Rate of Civilian Murders from 2010 to 2011 70
Figure 32: Change in the Rate of Civilian Abductions from 2010 to 2011 71
Figure 33: Change in the Rate of Armed-conflicts from 2011 to 2012 72
Figure 34: Change in the Rate of Civilian Murders from 2011 to 2012 73
Figure 35: Change in the Rate of Civilian Abductions from 2011 to 2012 74
Figure 36: Comparative Changes in Violence 2008 to 2009 (no HF radios exist) 76
Figure 37: Comparative Changes in Violence 2009 to 2010 77
Figure 38: Comparative Changes in Violence 2010 to 2011 78
Figure 39: Comparative Changes in Violence 2011 to 2012 79

vii

LIST OF EQUATIONS
Equation 1: Comparative Changes ( x1 - x2) 53

viii

LIST OF ABBREVIATIONS

CAR Central African Republic
DRC Democratic Republic of the Congo
FARDC Armed Forces of the Democratic Republic of the Congo
HRW Human Rights Watch
HUMINT Human Intelligence - Covert intelligence gathered from human sources
IDP Internally displaced person
LRA Lord’s Resistance Army
MONUC United Nations Organization Mission in the Democratic Republic of Congo (later
renamed MONUSCO)
MONUSCO United Nations Organization Stabilization Mission in the Democratic Republic of
Congo (formerly MONUC)
NGO Non-Government Organization
NRA National Resistance Army
SAF Sudanese Armed Forces
SIGINT Signals Intelligence - Intelligence gathered from communications sources
SPLA Southern People’s Liberation Army
SSD South Sudan
UN United Nations
UPDF Ugandan People’s Defense Force (formerly the NRA)

ix

ABTRACT

Understanding the geographic distribution of insurgency violence is critical for assessing where
counter-insurgency and civilian protections operations are effective. It allows researchers and
policymakers to detect trends in violence and propose local programs designed to quell
insurgency aggression in vulnerable areas. This thesis examines the spatial distribution of armed-
conflicts in Central Africa committed by the Lord’s Resistance Army from 2008 to 2012 and
offers a descriptive evaluation regarding the geographic fluctuation of violence throughout the
region. Existing counter-insurgency programs are discussed, and additional analysis is performed
on the development of a high-frequency radio network designed to facilitate information sharing
between communities. Resulting geographic representations indicate a steady decline in armed-
conflicts in the Democratic Republic of the Congo and South Sudan with violence becoming
more prevalent in the Central African Republic. The revealed fragmentation and variance in the
LRA’s operations supplement a growing body of research that seeks to better understand the
geographic evolution of conflict, identify why violence may increase or decrease in certain areas,
and assess the capacity for civilian protection initiatives in regions afflicted with insurgency
(Buhaug and Lujala 2005; Flint et. al. 2009; Kobayashi 2009).

1

CHAPTER 1: INTRODUCTION
Since 2002, the world has witnessed ongoing ethnic and sectarian violence across the African
continent. The civil war in Sierra Leone, the mass killings in Darfur, and ongoing paramilitary
violence in Democratic Republic of Congo (DRC) are three examples of African insurgencies
where civilians have been specifically targeted (Johnston 2008). Although these conflicts span
politically and environmentally diverse regions, there are recurring themes of civilian
vulnerability. The most heavily impacted communities are often those with underdeveloped civil
infrastructures including information technology and communications. This has inspired
nongovernmental organizations (NGOs) and state leaders to develop initiatives meant to enhance
civilian security and combat insurgency operations (Berman et. al. 2013; Mukeba 2013).
1.1 African Militancy and Counter-insurgency Directives
This thesis uses the term “Central Africa” in reference to the DRC, the Central African
Republic (CAR), and South Sudan as depicted in Figure 1. Highlighted are the 24 territories that
compose the study area applied in this thesis. In each territory, at least one act of violence
committed by the Lord’s Resistance Army (LRA) occurred between 2008 and 2012.

2

Figure 1: Dem. Rep. of Congo, Central African Republic, and South Sudan
Basemap Source: National Geographic, Esri, DeLorme, HERE, UNEP-WCMC, USGS, NASA,
ESA, METI, NRCAN, GEBCO, NOAA, Increment P Corp.

While this region has a rich history of diverse tribal cultures, deep-rooted ethnic and
ideological differences have grown into decades of war and insurgency -- particularly in the
DRC, which is the largest country on the African continent.
Since the 1990s, conflict has escalated within several Congolese regions. It is estimated
that four million people have died there since 1998 as a result of insurrection and violence (BBC
2014). Ethnic clashing in the Northeastern Ituri Region, and fighting between the Armed Forces
of the Democratic Republic of the Congo (FARDC) and foreign militias in the Eastern Kivu
regions, has prompted the United Nations Security Council to establish the UN Organization
Peacekeeping Mission in the Congo (MONUC). This organization was renamed the United
3

Nations Organization Stabilization Peacekeeping Mission of the Congo (MONUSCO) in 2010
after adopting a revised mandate authorizing the use of force to ensure civilian protection
(Copeland 2012).
Moctar Aboubacar, a policy writer for the Africa Center for Strategic Studies in
Washington D.C., questions whether MONUSCO has been successful in stabilizing the region.
In The MONUSCO Contradiction, he describes the mission’s obligatory cooperation with the
Congolese FARDC as a hindrance to success due to the military’s ineffectiveness and
disorganization adding that they have achieved neither cooperation nor stability in the region
(Aboubacar 2013).
Copeland (2012) presents similar conclusions, carefully defining civilian protection tasks
and describing three metrics by which the effectiveness of MONUSCO’s mission in protecting
Congolese civilians could be measured. The metrics include the rate of civilian deaths, the rate of
internally displaced persons (IDPs), and the occurrence of sexual violence. Copeland offers a
critical assessment of MONUSCO’s peacekeeping operations citing consistently high rates of
sexual violence, murders, and displacement, as evidence that the operation is failing to enforce
security in the Congo (Copeland 2012). However, these assessments are based on macro-level
analyses that do not recognize the variance of insurgency operations within a study area nor
provide consistent metrics by which trends in violence can be accurately measured over time and
across conflicts.
Research, like Copeland’s and Aboubacar’s, speak of African insurgency broadly and
group together insurgency organizations, which in terms of religion, ethnicity, and ideology, are
unrelated. Studies often fail to distinguish between the motivations behind armed-conflicts and
the counter-insurgency operations tailored to combat them. In aggregating armed violence along
4

the Eastern and Northeastern Congolese border, such studies combine violence perpetrated by
four distinct insurgency groups: the FARDC, the Democratic Forces for the Liberation of
Rwanda (FDLR), the National Congress for the Defense of the People (CNDP), and the LRA.
Because rates of violence persisted in these regions, despite new UN Security Council
resolutions tasked with ensuring civilian protection, researchers conclude that the mission has
fallen short of its goals (Copeland 2012).
1.2 Existing Research Gaps
There are recurring themes in African insurgencies including the conscription of children
and the deliberate targeting of civilian populations. These have been thoroughly explored in
several case studies (Aning and McIntyre 2004; Beber and Blattman 2011) with resounding calls
for programs addressing civilian protection, child education, and former soldier rehabilitation.
But, how do we begin assessing whether the LRA insurgency is changing; i.e., how do we model
spatial and temporal changes in violence at a scale that allows us to see variance between
adjacent territories?
Existing studies evaluating the impact of programs, like those headed by MONUSCO,
often assess the presence of violence rather than examining which groups are committing which
types of violence and where. We know that violence is ongoing along the Eastern and
Northeastern Congolese border, but this border is over 2,000km long and is shared with six other
countries. We do not know whether individual groups are becoming stronger or weaker, or if
they are shifting their area of operations in response to political allegiances or counter-
insurgency operations. Insurgents are treated as bands of rebels and little regard is paid to their
ethnic, political, or ideological aspirations. Studies do not consider the difference between each
group’s capacity for violence and cohesion of their leadership, or their susceptibility to
5

dissuasion. Assessments rely on macro-level analyses, which describe violence in general.
Rarely the metrics for violence extend beyond estimates of civilian deaths and IDPs.
This thesis addresses the flaws of what Fluri (2011) terms a “one-size-fits-all” approach
common in African insurgency studies. By executing a spatial analysis, specific to the LRA, in
greater resolution than what has been done previously, and by applying a yearly interval to the
spatial observations, patterns and trends in violence are revealed indicating flux in the LRA’s
area of operations. This suggests external variables may be responsible for the change. Resulting
observations allow for aid-workers and policy-makers to make informed decisions surrounding
the logistics of humanitarian and counter-LRA specific programs, which is critical for
strategically allocating military and financial resources. With quantifiable variables, the format
of the study allows for researchers to make unbiased comparisons between the LRA insurgency
and other regional conflicts that may also require military or UN assets.
1.3 Using Geographic Information Systems to Monitor Insurgency Violence
We know that the strength and orientation of the LRA changed from 2008 to 2012, but
we struggle to determine where changes occurred, the extent of the changes, and whether
particular territories improved or worsened outside of the overall trend. Also, while we know that
some territories employed specific counter-insurgency and civilian protection initiatives, we do
not know whether the rate of violence within those territories changed at a rate faster or slower
than territories which did not employ the same initiatives.
The objective of this thesis is to leverage geographic information systems (GIS) to
demonstrate how data pertaining to the spatiality of armed-conflict can be used to better
understand the non-static distribution of insurgency. Using our knowledge of other forces
engaged in the region, we can position better arguments on what factors might influence
6

insurgency behavior such as the utilization of a high-frequency radio network designed to
facilitate communication between municipalities.
The remainder of this thesis is divided into four chapters. Chapter Two reviews the origin
of the LRA including historical contexts and political dynamics that influence this study as well
as a discussion regarding the motives behind insurgency actors and an assessment as to why
isolation contributes to civilian vulnerability. This chapter describes the types of counter-
insurgency programs operating and examines past research on conflict geographies. Chapter
Three describes the study area, data sources, and the various metrics that are calculated and
mapped. The conceptual framework is explained along with the expected outcomes. Chapter
Four documents and interprets the study outcomes. Chapter Five discusses the impact of the
findings, their contribution to existing research on conflict geographies, and their relevance to
ongoing discussions regarding the development of counter-insurgency efforts in Central Africa.
This thesis concludes by describing limitations of the methods used and by identifying future
research directions in conflict geography and insurgency.

7

CHAPTER 2: BACKGROUND AND LITERATURE REVIEW
To understand the relationships and dynamics between the LRA and the states in which they
operate, it is important to first discuss their history and the social variables that have come to
influence this study. This chapter examines the Ugandan Alcholi origin of the LRA in the
context of a post-colonial Central Africa inundated with violence. The factors contributing to
civilian vulnerability are examined and existing counter-insurgency and civilian protection
initiatives are reviewed along with past research on conflict geographies.
2.1 The LRA Militancy: From Uganda to Modern-day DRC, CAR, and South Sudan
Prime Minister Milton Obote, a descendent of the Alcholi people from Northern Uganda,
suspended the constitution and declared himself President of Uganda in 1966. In 1971, he was
ousted from power by a military coup led by his commander, Idi Amin. During the eight years of
Amin’s presidency, substantial human rights abuses were committed against the Acholi people.
While the number of Ugandans murdered during these exchanges is unknown, estimates of
casualties range between 100,000 and 500,000 people (LOC 1990).
In 1979, Milton Obote and other Ugandan exiles fought back into Uganda and ousted Idi
Amin with the help of the Tanzanian army. Elections were held, and Obote became president for
the second time in 1980. Alleging electoral fraud, the National Resistance Army (NRA), led by
Yoweri Musevini, gained popularity as a resistance group and began launching attacks against
the government (LOC 1990). Retaliation from Obote’s security forces decimated areas that were
either controlled by or sympathetic to the NRA movement. This was the beginning of the
Ugandan Bush War, also known as the Ugandan Civil War, which lasted from 1981 to 1986. The
war concluded in 1986 with the NRA taking control of the Ugandan government. Yoweri
8

Musevini was named President of Uganda, and he remains President as of this writing. Those left
in the Acholi-controlled regime fled to the Sudan.
Since 1986, a number of Acholi-based rebel movements have come out of Northern
Uganda to resist the Ugandan government citing mistreatment of the Acholi people by both
Amin’s regime and the NRA. Most notable of the resistance groups was the Holy Spirit
Movement established by a medium named Alice Lakwena. The militant arm of the Holy Spirit
Movement is now known as the Lord’s Resistance Army (Bunting 2011). Joseph Kony assumed
power of the Lord’s Resistance Army in 1997, and he remains the leader of the rebel group. Like
Lakwena, Kony claims to receive messages from God and is said to be fighting the Ugandan
government in order to establish a new government and constitution in line with his religious
philosophy (Bunting 2011; HRW 2012).
Over the next decade, the LRA continued a war of attrition against the Ugandan
government. They launched attacks against the Ugandan army and even local communities in the
Acholi region. The LRA quickly lost favor with the northern communities because of their
forced-conscription of child soldiers and for attacking communities they accused of cooperating
with the Ugandan government (Bunting 2011; Cakaj 2007). Indiscriminate killings were carried
out to demonstrate to the northern populations that the government was unable to protect them if
they chose to betray the LRA.
The LRA found a brief ally in the Sudanese government in Khartoum. The LRA was in a
strategic, offensive position near the Southern People’s Liberation Army (SPLA) -- a Sudanese
rebel group based out of Southern Sudan (this region later became South Sudan following an
independence referendum in 2011). The SPLA was a rival group to the Sudanese Armed Forces
(SAF) who represented government interests in the north. The SAF provided the LRA with a
9

temporary base of operations in the south, near Juba, and the LRA received weapons, supplies,
and training in exchange for guerilla-style attacks against the SPLA (Cakaj 2007). Because the
Ugandan government supported the SPLA, the LRA was even more amendable to the
arrangement (Cakaj 2007).
Like other insurgency groups, there are items needed to ensure the organization’s
longevity. These items include food, supplies, and members to fill the ranks. With Sudanese
backing having deteriorated after a comprehensive peace treaty was signed between Sudan and
South Sudan in 2005, the LRA increasingly relied on looting civilian-communities for food and
supplies (Cakaj 2007). Weakened, the LRA signed a truce with the Ugandan government in
2006. The LRA agreed to leave Uganda and move to an area near Garamba National Park in the
northeastern region of the DRC. In 2008, ongoing violence perpetrated by the LRA provoked a
unified attack from the South Sudanese, Ugandan, and Congolese armies which, in turn,
prompted new waves of retaliatory violence against civilians including the abduction and forced
recruitment of children. It is believed that today there are between 200 and 300 members of the
LRA remaining (Cakaj 2007). This is a significant decline from the 3,000+ members the army
once comprised. Today, many people question whether the group still has centralized or political
motives. While they have been effectively eradicated from Uganda, they remain active in the
DRC, CAR, and South Sudan where looting, murder, rape, and kidnapping are still prevalent.
2.2 Area of Operations: People and Topography
The LRA’s area of operations surrounds the borders between the DRC, CAR, and South
Sudan. Poverty and malnutrition are prevalent within these states whose public health metrics
rank among the most severe in the world. The average infant mortality rate in the region is 80
deaths per 1,000 live births. The combined literacy rate is just over 50% and the median age falls
10

under 20 years in all three countries. In comparison, the infant mortality rate in the United States
is under 6 deaths per 1,000 live births, the literacy rate is 99%, and the median age is 37.2 years
(CIA 2014). A lack of public health facilities compounded by an enormously disproportionate
number of youth lacking a formal education contributes to a state of instability and civilian
insecurity. Most LRA victims, whom vary in ethnicity and political alignments, share the
struggles. Victims include civilians from the northern, Alcholi region in Uganda as well as the
Lugbara people who live on both sides of Uganda’s western border with the DRC. Christian,
Islamic, and Animist tribes have also been targeted in the CAR and South Sudan. This includes
members of the SPLA whose ranks contain the Sudanese Dinka.
The LRA traditionally targets rural, isolated communities along sparse countryside.
These agriculturally dependent villages and hamlets are particularly vulnerable due to their lack
of connectivity to other communities. Without telecommunications infrastructure, detached
communities must rely on “runners” and ground transportation in order to relay messages
between communities. Not only might those carrying messages be targeted by the LRA, but the
slow reporting time results in a delayed military response in the event of an attack.
The region’s topography consists primarily of evergreen and deciduous forests in the
Northeastern DRC, and shrub-lands, grasslands and croplands in the northern area of the study in
Southeastern CAR and Southwestern South Sudan. Heavily vegetated and unsettled areas are
colloquially referred to as “the bush.” Figure 2 is a 300m mosaic prepared by the ESA
GlobCover 2009 Project Team which documents land cover types using the UN Land Cover
Classification System (LCCS) (Bontemps et. al. 2011). The map provides additional context to
the environment and topographical composition which influences modern development and
insurgency behavior. The particularly dense forests in the DRC have hindered civilian protection
11

initiatives including the construction of communication infrastructure. The terrain conversely
aids the LRA by providing cover and place to retreat from military forces while remaining
hidden from aerial surveillance.

12

Figure 2: GlobCover 2009: 300m Mosaic Land Cover Classification
Source: Bontemps, S., Defourny, P., Van Bogaert, E., Arino, O., Kalogirou, V., Perez, J.
"Globcover 2009: Land Cover, Africa, and the Arabian." UCLouvain & ESA Team, 2011.

13

Combined with a scattered population and an agriculturally-based economy where the
demand for consumer goods and mobile services is low, there has been little commercial
incentives to venture into these regional markets. In 2010, the Invisible Children NGO, with the
Resolve NGO, attempted to help close the communication gap by expanding a network of high-
frequency, short-wave radios (HF radios) with the objective of providing communities with
intelligence on LRA movements. The nature of the HF radio transmission allows for broadcasts
to be sent up to 3,000 kilometers leveraging the signal’s ability to bounce between the
ionosphere and the forest canopies around the curvature of the Earth (Ionosphere 1999). This
essentially enables a radio in the town hub of Dungu to communicate with every other
community in the LRA’s area of operations -- a feat otherwise unachievable with existing
infrastructure.
2.3 LRA Strategy
LRA members are constantly on the move as they are pursued by Ugandan and
Congolese troops. It is rumored that some LRA commanders have access to satellite phones;
however, they have reduced their use of satellite phones for fear of being tracked by government
and military forces. The groups tend to move alongside rivers, secondary roads, and cattle paths,
and they use runners to relay messages from one group to another. LRA commanders have
access to GPS devices used for navigation, and they have been known to code GPS coordinates
in order to organize meetings with partner groups (Cakaj 2007).
Prior to attacking a community, LRA members will often perform reconnaissance. They
have been known to kidnap locals and force captives to spy on their behalf. They are most
interested in whether there are Congolese or Ugandan military forces stationed near their target.
If there is a nearby threat, they will typically refrain from attacking (Cakaj 2007). However, if
14

they are to engage a military force, they are more likely to attack the Congolese army (FARDC)
than the Ugandan army (UPDF) (Cakaj 2007). While the group is averse to risk, they have on
multiple occasions pretended to be opposing military groups to infiltrate communities prior to
attacking (Cakaj 2007).
2.4 Notorious LRA Massacres
While LRA violence has been ongoing for over 20 years, two massacres in the DRC
stand out as particularly violent and illustrative of vulnerabilities faced by at-risk communities.
The importance of emphasizing these events is to illustrate the brutality of the LRA and their
irregular military tactics, which include the callous and systematic execution of civilian
populations. Both of these events fall within the study area and the associated casualties are
included in the quantitative analysis described in Chapter Three.
The first series of events described are known as the Christmas Massacres. This series of
killings occurred from December 24
th
, 2008 to January 17, 2009. In 24 days, 865 civilians were
murdered and 160 children were abducted (HRW 2009). The LRA had coordinated attacks
across three major communities in the Haut-Uele district located in the Orientale Province in
Northeastern DRC. The communities were Doruma, Duru, and Faradje.
One of the first villages attacked was Batande -- a small hamlet in Doruma with a
population of approximately 100 people. Ugandan soldiers, responding to the LRA attack,
arrived at the village three days after the violence began. They found that the LRA had murdered
82 people leaving less than 20 survivors. Soldiers learned that residents had gathered for a
Christmas celebration before the LRA attacked them by surprise. LRA members gathered
residents, marched them into the woods, and then killed them (HRW 2009). LRA members ate
the food that had been prepared for the celebration and stayed overnight in the village.
15

The next day, LRA members traveled to the nearby town of Nagengwa and killed 30
people. Then, they traveled to Mabando where they killed another 50. The LRA traveled to nine
more villages killing people as they went (HRW 2009). Meanwhile, similar attacks were
happening in Duru and Faradje, yet no villages were equipped with a way to request assistance
or alert adjacent communities. The map in Figure 3 depicts Northeastern DRC around the
Garamba National Park, which borders modern-day South Sudan. Shown in the map are the three
major communities, which sustained iterative LRA attacks over the 24-day period.

Figure 3: Locations of Christmas Massacres (HRW 2009)
Source: Human Rights Watch. "The Christmas Massacres: LRA Attacks on Civilians in Northern
Congo." (16 Feb 2009). Map inset added for context.

16

The second series of events are known as the Makomba Massacres. These occurred
between December 14
th
, 2009 and December 17
th
, 2009. The LRA attacked 10 villages in the
Makomba area again located in the Haut-Uele district, DRC. They killed 321 civilians and
abducted 250 others, including 80 children (HRW 2010). The map shown in Figure 4 traces the
LRA’s four-day route making note of the villages they attacked.

Figure 4: Map of Makomba Massacre (HRW 2010)
Source: Human Rights Watch. "Trail of Death: LRA Atrocities in Northeastern Congo." (28
March 2010.) Map inset added for context.

Abducted civilians were ordered to march 20km a day through the bush carrying looted
items and supplies. Civilians who could not keep up were killed. Congolese and Ugandan
17

soldiers stationed in Haut-Uele were alerted to the incident on December 16
th
– two days after
the violence started. When soldiers arrived on December 18
th
, the LRA was gone (HRW 2010).
MONUSCO was tipped off to the event in late December; however, it took two months for
MONUSCO and the UN to obtain the complete details of the attack. When the UN did receive
preliminary reports of the massacre, they assigned the intelligence a grade of “D” on a scale of
“A” to “F” indicating a low confidence in credibility (Poole 2013).
The Human Rights Watch (HRW), an international NGO based out of New York City,
estimates that, since the LRA’s inception in 1987, over 20,000 civilians have been abducted, and
tens of thousands have been murdered (HRW 2012. According to records compiled by two
NGOs based in the United States, the Invisible Children and Resolve, the LRA are responsible
for 2,681 murders between 2008 and 2012. These killings demonstrate the enormity of the
LRA’s crimes and a serious lack of counter-insurgency intelligence.
An aggravating factor in these events is the remoteness of communities and the distance
to counter-insurgency operations headed by state military forces. When the LRA attacks a small
village or hamlet, it takes several days or weeks for the army, state, or media to be made aware.
This has prompted organizations and NGOs to explore ways to strategically intervene by
tracking the movements of LRA members, developing local programs to convince LRA
members to leave the group, and by networking at-risk communities using radio and cellular
communications. Together, these efforts should allow counter-insurgency actors to respond to
LRA attacks quickly, alert nearby communities to LRA operations, and slowly break down the
LRA disposition by reducing the success of their operations while increasing the risk.

18

2.5 Counter-insurgency Programs
Since 2008, there have been numerous efforts to curb LRA violence. Under U.S.
President George Bush, the U.S. agreed to provide financial and logistical support to the
Ugandan government in their pursuit of the LRA. In 2010, under President Obama, 100 military
advisors were sent to region in order to assist Ugandan and Congolese military forces in tracking
the LRA. Other efforts include an arrest warrant placed on Joseph Kony by the International
Criminal Court (ICC), and a five-million dollar bounty for Kony’s arrest offered by the U.S.
(Baguma 2012). Included in the ICC’s charges against Kony is the extensive use of children in
warfare – a trend that is particularly prevalent in Sub-Saharan Africa where over half of the
population is under the age of 18 (Anning and McIntyre 2004).
In From Youth Rebellion to Child Abduction: The Anatomy of Recruitment in Sierra
Leone, researchers Prof. Emmanuel Aning and Angela McIntyre analyzed child soldiering in
Sierra Leone. They described a 10-year civil war where both military and non-state actors had
compelling incentives to develop strategies of mobilizing youth for their causes (Aning and
McIntyre 2004). In the case of Sierra Leone, the civil war was preceded with economic despair,
lack of economic opportunities, and a failing education system. The disadvantaged and
uneducated have become prime targets for recruitment due to the ease at which they can be
manipulated (Aning and McIntyre 2004). Their study draws many parallels with the LRA’s
recruiting and forced-conscription of children in Central Africa. Aning and McIntyre suggest that
in order to combat the recruitment of children in warfare, youth-oriented initiatives must be
considered an active component of the peace process, and access to education must be
emphasized when working with former combatants. (Aning and McIntyre 2004; Beber and
Plattman 2011).
19

In their 2011 study, The Logic of Child Soldiering and Coercion, Beber and Plattman
argue that education-based, intervention programs that target youth susceptible to LRA influence
can teach youth about the LRA including how to escape if captured. (Beber and Plattman 2011).
Education is critical considering the methods by which children become soldiers in the first
place. While many are abducted and forced to commit violent acts against their own community
or family, others are tricked or convinced that life as soldier is prestigious, prosperous, and easy
(Riddell 2009).
With LRA members dispersed and without a centralized command point, there has been a
surge in information campaigns to reach and educate LRA members and to convince them to
leave the organization. The Enough Project, an American NGO based out of Washington D.C.,
promotes the use of FM radio broadcasts encouraging LRA defection. They suggest that these
messages should target junior- and mid-level officers (Cakaj 2010). LRA members should be
made to feel that opportunities, such as resuming studies in vocational training, exist elsewhere
(Cakaj 2010).
Beber and Plattman (2011) cite evidence that the LRA have been known to halt
abductions in some areas after local children were exposed to FM radiobroadcasts offering
amnesty to LRA members. Senior LRA leaders feared that these community broadcasts would
catch on with LRA youth and trigger mass desertion among lower-ranking members that do not
otherwise have access to FM radios (Beber and Plattman 2011; Cakaj 2010). Leaflet distribution
and helicopter-mounted speakers flown over LRA areas have also been used in encouraging
defection among LRA members. In 2012, 25 Ugandan members defected from the LRA with 21
of them stating they had seen or heard these messages (Poole 2013; Vanvider 2013). With a
declining roster of active LRA members, this represents a substantial reduction.
20

The HRW recommends continuing to improve communication infrastructure and
building cellular phone towers as a means of combating the LRA citing both physical and
ideological benefits. More towers imply an expansion of mobile phone coverage in LRA-
affected areas, which may result in more intelligence sharing and increased coordination between
communities (HRW 2012). The Pyramid Research group, a consulting firm specializing in
telecommunications and emerging markets, released a study in 2009 analyzing changes in
mobile phone access in neighboring Uganda. They estimate that, in 2002, the percentage of
Ugandans with access to mobile communications services was 1.9%. In 2009, that number had
risen to 39%. The firm predicts that 70.7% of Ugandans will be using mobile services in 2014
(Baker 2009). Nearby countries like the DRC, South Sudan, and CAR are expected to experience
similar growth (Estefan 2012). In fact, at the time of writing, some communities in the CAR,
including Obo, Mboki, and Rafai, as well as some communities in the DRC, including Dungu
and Faradje, have already constructed cellular towers. However, existing service is limited and
does not yet extend to the many villages and hamlets, which surround the more populated
community-centers (Poole 2013).
While the construction of a cellular communication network is slowly expanding, the
Invisible Children and Resolve have taken measures to repurpose an existing network of high-
frequency (HF), short-wave radios maintained by local Catholic churches. The churches have
long used the HF radios for logistical purposes, although they are increasingly being used to
defend against the LRA. In an interview, Sean Poole, the Counter-LRA Program Director at the
Invisible Children, explained how the radio network functions as a complex security reporting
apparatus. There are currently 38 radios composing an “early warning” network with 17
additional radios slated for installation by the first quarter of 2014.
21

The purpose of the HF radios is to report LRA sightings and attacks to UN and military
forces as well as alert nearby communities of the LRA’s presence. This allows citizens there to
take precautionary measures. With the aid of radios, communication can occur faster than other
methods such as sending runners or vehicles that risk being targeted by LRA members (Poole
2013).
The map in Figure 5 represents the approximate location of HF radios in use between
2008 and 2012. Points on the map have been inflated and staggered in order to shield radios’
actual location. The geocoordinates of the HF radios are considered confidential due to their
sensitive nature and public concerns that having an HF radio could result in a community being
specifically targeted by the LRA. The HF radios and the territories in which they are located are
discussed further in Chapter 3, Methodology.

Figure 5: Generalized HF Radios 2010-2012 indicating approximate location
22

The Human Rights Watch supports efforts to ensure HF radios are available for all
communities within the LRA’s area of operations (HRW 2012). For isolated communities, these
radios may serve as the only means of telecommunication. LRA encampments can be as close as
10-12km away. Even after the LRA leaves the vicinity, there is still a chance that they will
return. If they do, the radios offer an avenue by which that information can be relayed to
government, military organization, or UN personnel (Poole 2013).
With the assistance of the radios, LRA attacks are geographically identified with GPS
coordinates and logged in a central database known as the LRA Crisis Tracker. Leveraging this
geographically referenced information, the LRA area of operations is more clearly delineated
which allows for education-based programs to be tailored for specific regions. Affected regions
have also become candidates for economic and commercial investments, including
telecommunications, due to the LRA’s preference in attacking isolated communities. Overall, the
information gathered in the HF radio communication network offers a picture not previously
available. It remains the primary means of intelligence gathering information used by the African
Unions (AU), the United Nations, Ugandan military forces, and the United States (Poole 2013).
2.6 Conflict Geographies and Theories of Communication and Insurgencies
There have been numerous studies discussing conflict geographies involving various
types of African insurgency over the last 10 years. In a 2008 paper on insurgent violence in
Liberia and Sierra Leone, Patrick Johnston argued that a military’s effectiveness is a direct result
of their hierarchical organization, which is determined by geographic and technological factors
(Johnston 2008). His conclusions provide insights on the tactics armed groups will use given a
defined set of circumstances and technological limitations. He argues that a military member’s
preferences and strategies will vary relative to the cost of defection (Johnston 2008). Classifying
23

insurgency groups as unitary or multidivisional, he concludes that unitary insurgencies are more
effective on the battlefield, and more reliable in negotiations due to their centralization and
ability to control their members. Conversely, multi-divisional insurgencies are more fragmented
with members dispersed with many small sub-divisions of authority, although a centralized
leadership does exist (Johnston 2008).
There is a close parallel with Johnston’s description of multi-divisional insurgencies and
the modern-day LRA. Since leaving Uganda, the group is widely dispersed with operations
across three countries. They are scattered more today than ever before, and they currently have
the fewest number of members in their history. The fragmented nature of the group negotiations
difficult, and a peace treaty is nearly impossible to broker as many groups are scattered,
undisciplined, and unregulated (Johnston 2008). However, it also renders them less effective in
their overall threat to the Ugandan government whose overthrow had been their original
mandate.
The lack of available communication has an effect on the LRA’s ability to regulate and
coordinate amongst their members. While this thesis advocates for the development of
communication infrastructure as a civilian protection initiative, there are those who suggest that
introducing communication technology, like cellular communications, may empower and allow
new organizational models for insurgency organizations, i.e. a better functioning “networked
insurgency” (Arquilla et. al. 1999; Andreas 2002; Muckian 2006). For insurgencies and non-
state guerilla groups, mobile communications provide a platform by which attacks can be
executed in an organized fashion while remaining flexible and reactive to counterinsurgency
efforts (Blakely 2005; Cordesman 2005; Leahy 2005)
24

In the context of the wars in Iraq and Afghanistan, reactions by Al-Qaida and the Taliban
to the development of communication infrastructure have been observed. There are reports of the
groups attacking or threatening cell phone towers in some areas, while demanding that services
are improved in other areas. These seemingly contradictive behaviors reflect their perceived
security in their surroundings, i.e. whether they felt locals would share intelligence or provide
tips to coalition forces or police (Shapiro and Weidman 2012).
Wireless communication channels enable the general population to easily share
information about insurgent activity -- oftentimes anonymously thereby reducing the fear of
reprisal attacks (Shapiro and Weidman 2012; Trofimov 2010). Instead of having a single
intelligence, counter-insurgency effort, the entire population can be employed in collective
intelligence gathering – a concept for which Emeritus Professor of UC Santa Barbara, Michael
Goodchild, coined the term “citizens as sensors” (Goodchild 2007). In The New Digital Age, Eric
Schmidt, the executive chairman of Google, discusses the future of multi-dimensional conflict in
an era where information is distributed and accessed freely. He argues that a connected and
informed public has a greater potential to mobilize against injustice (Schmidt and Cohen 2013).
To engage civilians in Iraq, the Coalition Provisional Authority, the interim transitional
government from 2003 to 2004, budgeted $10 million USD for billboard, print, radio, and
television advertising instructing civilians how to leverage the National Tips Hotline to report
information and expand the intelligence profile on local insurgencies (Semble 2006). These
efforts have multiplied the channels by which civilians can share information and participate in
security building.
There are few works that quantitatively assess the effects of introducing civilian
protection initiatives, including communications development, in war-affected areas. However,
25

in Talking About Killing: Cell Phones, Collective Action, and Insurgent Violence in Iraq,
published by Profs. Jacob Shapiro and Nils Weidman, improvements in communication are
explored to assess their impact on the production of violence in Iraq (Shapiro and Weidman
2012). While acknowledging that introducing cell phone coverage can make it easier for
insurgents to coordinate attacks, their findings suggest that new coverage in areas where Iraqi
insurgencies are ongoing creates more opportunities for “passive signals intelligence”. This
includes information gathered from people and telecommunications that can be shared with
counter-insurgency forces (Shapiro and Weidman 2012). One of their key conclusions suggests
that the “provision of information” by non-combatants is the most critical element. In an open-
communication environment, civilians experience only a minimal risk to safety for cooperating
with government and counter-insurgency forces. These information-sharing efforts have a
substantial effect on local conflict (Shapiro and Weidman 2012).
In a working paper titled Modest, Secure and Informed: Successful Development in
Conflict Zones, the garnering of civilian support in information gathering is referred to as part of
a “hearts and minds” model (Berman et al 2013). This model describes a three-sided, game
theoretical decision making process between non-combatants, governments, and rebel forces.
Berman et al. (2013) model is searching for an optimum ratio of conditions that encourages
information-sharing by non-combatants without fear of retribution or of being identified. Similar
models have been in place in Iraq where coalition forces appeal to civilians to “fight the war in
secret” and provide information on insurgents (Berman et. al. 2013; Miles 2004). The model
predicts that with the expansion of mobile technology, and a tilt of the equilibrium by which
civilians are willing to participate, there should be a reduction in violence in rebel-active areas
(Berman et. al. 2013). This “hearts and minds” method corresponds well with the LRA’s strategy
26

in Central Africa where tactics of intimidation and fear are used to dissuade locals from
participating in anti-LRA efforts. Logic presented in these studies suggests that, given the proper
conditions, civilians will participate in information sharing.
Research conducted by Shapiro and Weidman in Is the Phone Mightier than the Sword?
Cell Phones and Insurgent Violence in Iraq suggests that the introduction of new mobile
coverage indeed reduces the rate of violence relative to overall trends (Shapiro and Weidman
2011). Using a 50% significant threshold, they found that activating a new cell phone tower
predicted .896 fewer attacks per 100,000 people, per district, per month. Conceding that they
could only speculate whether the decrease was because new coverage allowed civilians to inform
on insurgents, or if the new coverage enabled enhanced coalition intelligence gathering efforts,
the decline in the rate of insurgent attacks was clear (Shapiro and Weidman 2012).
2.7 Use of GIS in Insurgency Studies
GIS is increasingly being used to explain the spatial dynamics of insurgency and civil
war by identifying variables that vary in space and at scales consistent with armed-conflicts and
conflict-zones (Buhaug and Lujala 2005). Buhaug and Lujala (2005) argue that conflict-specific
variables are dependent on the scale of measurement and that country-level geographical
variables often omit specific features of terrain, natural resources, and subsets of ethnic diversity,
which might otherwise correlate with increased risks of conflict. This holds especially true in
regions experiencing simultaneous conflicts that may be ethnically unrelated, yet motivated by
distinct geographic variables.
Most civil wars and insurgencies are isolated to certain parts of the country due to either a
concentrated group of people identifying with a particular ethnicity, religion, or ideology, and
conflicts often extend from territorial disputes or ownership of natural resources within their
27

region (Buhaug and Lujala 2005). Typically, commonly used metrics are measured at the
national level despite substantial deviations between statistics aggregated by country versus
those that are specific to particular conflict zones. This is often due to a lack of available data at
the zonal level.
In 2005, Buhaug and Lujala examined the suitability of country-level statistics in place of
data collected at levels specific to conflict zones. They classified individual conflicts as
stationary points with a fixed radius accounting for a total of 252 civil conflicts occurring
between 1946 and 2001. Next, they separately coded individual countries and individual conflict
zones by land-cover types (e.g. mountains or forest) and by whether they contained “lootable”
items such as gemstones, coca, cannabis, or opium. After conducting multivariate regression
analysis on the geographic variables, they reaffirmed their notion that the scale of measurement
does affect the significance levels associated with the presence of land-cover types and lootable
items specific to individual conflict-zones. The correlations were less significant when using
country-level aggregates (Buahug and Lujala 2005).
Other GIS studies (Cederman et. al. 2007) focus solely on the sub-national divisions of
ethnicity and whether these are a driving force behind certain conflicts. Cederman et al. (2007)
essentially lay the foundation for country-level analysis by applying a spatially enabled
disaggregated approach using polygonal feature classes to correspond with ethnic sub-divisions
(i.e. boundaries). Arguing that this can then be used to better identify whether ethnic groups are
participating as collective actors in various conflicts and civil wars, Cederman et. al. underwent
the arduous process of digitizing a series Soviet-era ethnic classification maps known as the
Atlas Narodov Mira (Cederman et. al. 2007). By layering the digitized maps with rasterized
representations of population densities, they propose creating a series of ethno-linguistic
28

fractionalization indices (ELFs) to be used in future GIS projects. Completed, the digitization
project yields greater quantitative facts regarding the “micro-level” mechanisms that influence
the emergency and evolution of ethnic conflicts and insurgency (Cederman et. al. 2007).
Knowing that environmental features and international relationships are influencing
individual conflicts, Flint et. al. (2009) argue that focused spatial considerations are needed to
understand how actors behave. This can be achieved by documenting each actor’s physical
position and by defining their role in geographic and community contexts while recognizing
where individual groups lie in the “social terrains” of conflict and in their relationships to other
entities (Flint et. al. 2009; Tuathail 2010). These fundamentals must be incorporated in future
studies to include disaggregated spatial analysis at a scale that lends specific insights to these
theories, which is necessary to avoid losing the particularities of specific conflicts (Flint et. al.
2009; Fluri 2011; Tuathail 2010).
In line with the suggestions outlined in this section, this thesis offers an integrated
examination of the LRA insurgency in a geographic framework that acknowledges external
influences and derives patterns of insurgency behavior (Chapter 3, Section 3.3.3). Spatial
observations address disjuncture in examining insurgencies as static operations. Accordingly, all
findings will be provided to the Invisible Children NGO and made available to other
organizations working to assess the possible impacts of counter-insurgency initiatives and
improve civilian security.

29

CHAPTER 3: METHODOLOGY
This chapter will review LRA-active areas, discuss the data sources being used to capture
metrics of violence, and it will provide a conceptual framework for conducting a descriptive
analysis of change in the rates of violence given the introduction of radios over time.
3.1 Study Area and Areal Units of Analysis
This study takes place north of the Great Lakes region in Central Africa and encompasses
the borders between the DRC, CAR, and South Sudan. The level of analysis selected in this
study represents a “territory-level” analysis because this was the most granular level of
administration for which boundary and population data were available for all three countries.
Each country applies different terminology in referring to this level of administrative
area. These areas are known as “counties” in South Sudan, “territories” in the DRC, and “sub-
prefectures” in the CAR. This study uses the term “territories” in reference to all three.
Territories were included in the study if they sustained one or more LRA armed-conflicts
between 2008 and 2012. Territories without any conflicts were considered outside of the area of
operations and were excluded from the analysis. Administrative boundaries for these areas were
downloaded from www.diva-gis.org in shapefile format. The map shown in Figure 6 outlines
each of the 24 territories demonstrating the LRA area of operations from 2008 to 2012.
30

Figure 6: Map representing Administrative Boundaries in the DRC, CAR, and South
Sudan comprising the LRA’s area of operations 2008-2012.

Population data were acquired from www.geohive.com, www.citypopulation.de, and the
World Gazetteer at www.archive.js. While there are many villages and hamlets of varying
populations within each territory, data at these resolutions were unavailable. The territory-level
population estimates were used to normalize each category of violence. Table 1 describes the
total area and population estimates for each territory.

31

Table 1: Territory Areas and Population Estimates
Country Territory Area (km
2
) Population
SSD Raja 56,643 54,340
Wau 28,853 151,320
Nahr Yei 12,955 201,443
Meridi 19,523 82,461
Mundri 16,361 82,290
Tombura 27,592 55,365
Yambio 15,671 152,257
DRC Ango 34,303 8,675
Bambesa 9,750 15,483
Bondo 37,139 18,576
Poko 22,711 9,592
Dungu 33,804 23,726
Faradje 13,303 25,000
Niangara 8,750 13,977
Watsa 16,317 33,385
CAR Djemah 33,923 1,835
Obo 14,391 36,029
Zémio 8,039 16,812
Yalinga 43,856 5,175
Bakouma 16,348 20,975
Bangassou 8,019 66,515
Rafaï 28,714 13,962
Birao 40,349 48,367
Ouanda Djallé 6,595 3,888

3.1.1 Modifiable Areal Unit Problem (MAUP)
Because this analysis is performed at a territory level delineated by administrative
boundaries whose area and population vary widely, there is a clear risk of the areal units
introducing bias in the results. Table 2 shows the variance between territories.

32

Table 2: Territory Areas and Population: Mean, Max, and Min.
Area (km
2)
Population
Mean 23,080 47,560
Max 56,643 201,443
Min 6,595 1,835

The MAUP problem will potentially impact the rate changes calculated by territory along
with the comparative changes as represented by the Difference-in-Differences statistical method
(Abadie 2005) descriptive analysis (Section 3.7). Territories are a composition of many small
villages and hamlets – many of which lie beyond the LRA’s area of operations. However,
because at least one or more armed-conflict occurred within the jurisdictional area, the entire
territory is included in the analysis. This means a territory could have a small community
impacted by the LRA while every other municipality in the region is unaffected. To reduce the
MAUP associated with the spatial distribution, more analysis would need to be completed at
additional scales. Preferably, a provincial, municipal or other type of zonal level analysis would
be used to offer more general or granular pictures of insurgency variance to validate any
findings. However, for this thesis administrative and population data at other resolutions were
unavailable. This would constitute a next step for future research.
In the Difference-in-Differences descriptive analysis, only four distinct territories are
identified as having HF radios in 2012 even though a total of 28 HF radios existed by that year.
At other scales, the number of distinct areas containing radios could increase or decrease.
Aggregating radios by territory implies that other municipalities within the same territory are
affected by the HF radios’ presence. Similarly, there may be municipalities in adjacent territories
that are impacted by an HF radio across the border despite there being no HF radio within their
33

own jurisdiction. These are examples of errors that may be attributable to the delineation of
territories without accounting for the extent of an HF radio’s expected reach.
3.2 Data Sources, Variables, and Metrics
Two datasets are applied in this thesis: armed-conflict event data and HF radio locations.
3.2.1 Armed-Conflict Event Data
The Invisible Children and Resolve are two NGOs that maintain an online database of
LRA attacks and sightings known as the LRA Crisis Tracker. The Crisis Tracker is a browser-
based GIS developed by the sales marketing firm, Salesforce, and it allows LRA conflict data to
be available to users through a geographically-enabled web interface (Ungeleider 2011).
Incidents involving the LRA, including attacks, kidnappings, lootings, sightings, or even
member-defections, are logged in the database. While some information is redacted from the
public version of the database (e.g. victim names or other sensitive data pertaining to the
movement of high-profile LRA members), the Crisis Tracker database is readily available for
download in .XLS format.
The information included in the armed-conflict database is collected from a variety of
sources. Sources include HF radio transmissions, UN and media outlets, military forces, NGOs,
and other first-hand accounts and interviews such as those conducted by the Human Rights
Watch. The complete data sourcing process by which the information is collected, verified, and
catalogued in the Crisis Tracker database, can be found in Table 7 in Appendix A. This table was
extracted from a codebook prepared by the Invisible Children and the Resolve.
In its raw form, the dataset contains 95 columns that describe the conflict data. These
fields include the date of the event, the event’s GPS coordinates, description, community,
34

verification level, and a detailed disaggregation of casualties and victims including violence
indicators like fatalities, rapes, and abductions as well as victim sex and age.
At the time of download (02 June 2013), 2,392 records were in the table. Records missing
GPS coordinates (23) were removed from the dataset. Of the records that remained, SQL was
used to retain only those events that included violence against civilians. Violence against
civilians is defined in this study as any LRA attack that resulted in injury to a civilian. Injury
includes physical harm, death, sexual assault, and abduction. LRA sightings and looting were not
included in this study unless the event included direct harm to a civilian.
Of the 2,369 records obtained, 1,132 of them met these criteria, i.e. each of the 1,132
records constitutes an “armed-conflict”. The number of civilian murders and civilian abductions
were then logged in separate tables. For example, if a single armed-conflict resulted in five
murders and seven abductions, the record in the Armed-conflict table would receive a value of
one, the record in the Civilian Murders table would receive a value of five, and the record in the
Civilian Abductions table would receive a value of seven. If an armed-conflict resulted in only
injury to a civilian, and no one was killed or abducted, then the Armed-Conflict table would
receive a value of 1, and the other tables would receive a value of 0.
The map shown in Figure 7 shows the location of all 1,132 armed-conflicts committed by
the LRA between 2008 and 2012.
35

Figure 7: Map of LRA Armed-Conflicts 2008-2012

3.2.2 Civilian Protection Program Data: High-frequency Radio Locations
One objective of this thesis is to identify variance in rate changes between territories,
which either do or do not employ HF radios. Data describing the locations of the HF radio
installations were provided by the Invisible Children in table format. Data fields for the 38 radios
included country, community, GPS coordinates, and the date of installation. Of the 38 radios,
two radios were expired, and either were installed after 2012 or were still undergoing
construction. The non-operational radios’ expiration dates were unclear, so they were excluded
from the analysis leaving 28 radios accounted for in this thesis. Due to the sensitivity
surrounding the radios’ geographic coordinates, the exact locations of radios have been omitted
36

from any maps included in this thesis. The radio data are instead aggregated in the attribute
values for the territories in which they are located. As a result, two categories of territories were
identified: a sample group that included those territories that contained HF radios, and a control
group that included those territories that did not.
3.2.3 Data Strengths, Assumptions, and Limitations
Data included in this study were obtained from disparate sources using different
collection methods with varying levels of confidence and accuracy. The armed-conflict data is
derived from first-hand accounts and other primary sources. This presents two challenges. First,
GPS coordinates were used to georeference the data, but the coordinates themselves carry an
inherent source of error. When recorded, many of the GPS coordinates assigned to the conflict
reflect the coordinates of the nearest municipality or other verified landmark. It is not feasible for
all counter-insurgency forces to rely on handheld receivers to log each location as would
otherwise be expected in higher-resolution applications. However, this does not present a
substantial concern in this thesis due to the territory-level scale being used. Second, it is possible
that witnesses or reporters of LRA violence have exaggerated certain accounts. It is difficult to
corroborate every attack, and duplicate or inaccurate reporting of events has been known to
occur. Conversely, there may be additional acts of violence that are unaccounted for just as there
may be acts of violence included in this study wrongly attributed to the LRA. This is especially
feasible due to numerous reports of rogue military units and other paramilitary organizations
committing crimes against civilians (Menondji 2013). Still, the LRA Crisis Tracker database
remains the best source of primary conflict data pertaining to the LRA, and this thesis is assured
in the collection, entry, and validation of data as outlined in Appendix A, Table 7, Armed
Conflict Data Collection Method (Resolve and the Invisible Children 2012).
37

Regional population counts are difficult to estimate and may be another source of error in
this thesis. The incorporated population data were obtained from three different sources and the
same population counts were used for all five years. In reality, there are a large number of IDPs
whose location may vary year to year according to political climates and their proximity to other
communities or even insurgencies. While estimates of IDPs are a common metric in describing
the impact of military conflict, they are extremely difficult to quantify and therefore have not
been incorporated into this thesis. Instead, the population estimates acquired for each territory
remained unchanged so the rate of change could be calculated while maintaining a consistent
level of precision across each year-to-year transition.
3.3 Methodology-framework
A high-level methodology-framework followed in this study is shown in Figure 8. The
methodology involved a series of steps for data acquisition, preparation, and processing that are
discussed in detail in Section 3.3.1. The flowchart also outlines the executed exploratory
methods (Section 3.4) before identifying suitable approaches for qualitatively assessing the
spatial distribution of violence. These approaches include aggregating summaries of violence by
country (Section 3.5), calculating annual rates of change by territory (Section 3.6), and
calculating comparative changes as represented by a Difference-in-Differences descriptive
analysis at the territory-level (Section 3.7)(Abadie 2005).
38

Figure 8: High-level Methodology-Framework.

3.3.1 Data preparation within ArcGIS
Data used in this study were imported and geo-referenced using ArcCatalog. All data
were projected to the WGS 1984 UTM Zone 35N coordinate system (SRID 32635) as this
encompasses the overwhelming majority of the study area and allowed SQL calculations to be
executed in metric units. ArcMap Modelbuilder was used to align feature topology, execute
spatial joins, and export cartographic representations.
Each datasets was imported to a single file geodatabase as feature classes using
ArcCatalog. Individual feature classes representing the country-level administrative areas of the
DRC, CAR, and South Sudan were joined to create a single layer of countries. Similarly, the
individual feature classes representing the territory-level administrative areas were joined to
create a single layer of territories. The process of acquiring and preparing the data is described in
Figure 9.
39

Figure 9. Data acquisition, preparation, and integration workflow
Armed conflicts
Armed-conflicts were imported to the file geodatabase, after the acquired GPS
coordinates were geocoded. These data were spatially joined with the “Territories” and
“Countries” feature classes and data for each category of violence (i.e. number of armed-
conflicts, civilian murders, and civilian abductions) were aggregated by region (i.e. territory or
country). Having executed the spatial joins, the annual raw counts of violence were known for
each category of violence for each territory and each country from 2008 to 2012.
HF Radio Locations
HF radio location data were imported to the file geodatabase, after the acquired GPS
coordinates were geocoded. While the actual geocoordinates of the HF radios are not indicated in
this thesis, their use was considered in conjunction with a polynomial regression analysis
described in Section 3.4.2. However, as indicated in the workflow, the HF radio locations were
40

limited to a spatial join with the “Territories” feature class where they were absorbed into the
attribute values of the territories in which they were located.
Population
Population data were imported to the file geodatabase and joined with the “Territories”
feature class. The ArcMap Field Calculator was used to normalize the conflict data by
calculating the rate of violence per 1,000 people in each territory. Conflict data were normalized
for the three categories of violence in each year.
3.4 Explored methods
Three methodological directions were initially explored: Clustering and Heat Map
analysis, Polynomial Regression and Density Score, and Annual Rates of Change and
Comparative Changes using Difference-in-Differences. Attempts and limitations of the
clustering and polynomial regression methods are presented before introducing, in detail, the
qualitative analysis of the spatial distribution of violence, which is represented as summaries of
violence by country, annual rates of change by territory, and as comparative changes represented
by Difference-in-Differences.
3.4.1 Clustering and Heat Map Analysis
Clustering methods were explored to show where armed-conflicts are concentrated and
whether there is evidence of heavy clustering. In a heavily clustered environment, there is the
potential to identify predictor variables that can be used to explain the highs and lows of the data.
Six kernel density maps were created to produce rasterized surfaces indicating concentrations of
LRA activity. Figures 9-13 reflects the kernel density of armed-conflicts in 2008, 2009, 2010,
2011, and 2012 respectively. Figure 14 reflects the kernel density with all years considered.
41

Figure 9: Armed-conflicts Heat Map: 2008

Figure 10: Armed-conflicts Heat Map: 2009
42

Figure 11: Armed-conflicts Heat Map: 2010

Figure 12: Armed-conflicts Heat Map: 2011

43

Figure 13: Armed-conflicts Heat Map: 2012

Figure 14: Armed-conflicts Heat Map: 2008-2012
While the maps demonstrate how the concentration and extent of the LRA’s operations
has changed with time, the changes are relative and are dependent upon the total number of
44

armed-conflicts each year. This complicates our ability to quantify changes in density for each
year-to-year transition. This method would be increasingly viable in the presence of a more
robust collection of predictor variables enabling other regression techniques.
3.4.2 Polynomial Regression and Density Score
This approach began with coding each armed-conflict event with the ID of the nearest HF
radio installation. Armed-conflicts were only coded to a nearby HF radio if that radio was in
existence at the time the armed-conflict occurred. Then, the distance between the armed conflict
event and the nearest HF radio was calculated in meters. Armed-conflicts events were classified
as being in Tier 1 or Tier 2 of the nearest radio. An event was classified as Tier 1 if it occurred
within 40 km of a HF radio and classified as Tier 2 if its distance was greater than 40km. The
furthest any event was to a radio was 954km. Because the year of the earliest HF radio
installation date is 2010, only armed-conflicts occurring after that point in time could be included
in this analysis. Of the 547 armed-conflicts meeting this criterion, 270 were in Tier 1, and 277
were in Tier 2.
A value, termed in this thesis as “Density Score”, was then calculated for each armed-
conflict using spatial SQL. The variable was designed to be indicative of the density of violence
surrounding each individual point feature (i.e. armed-conflict event) within a temporal window.
A radius of 15 km and a temporal window of +/- 15 days were selected. Once calculated, the
Density Score value for each armed-conflict represented how many other conflicts occurred
within 15 km and within 15 days. Armed-conflicts that occurred nearby and in close succession
with other conflicts were assigned the highest scores. Calculating the Density Score in this
manner allowed for testing at different radii (e.g. 20km or 50km) and varying windows of time
45

(e.g. +/- 30 days, or +/- 45 days) while maintaining the same units of reference (events per km
2

per day).
The goal was to aggregate all armed-conflicts by the two tiers of the HF radio installation
nearest to them and assign each Tier a mean density score value. Then, the mean density score
for both tiers of the HF radio installation would be plotted over the course of three years: 2010,
2011, and 2012. The expectation was that the mean density score for armed-conflicts in Tier 1,
i.e. the mean density score for armed-conflicts occurring within 40 km of a radio installation,
would decline over time at a rate different than the mean density score for armed-conflicts in
Tier 2. Polynomial regression would be applied to offer trend analysis by tier with the
expectation that armed-conflicts would initially peak following the placement of an HF radio
installation before tapering off as more time passed. Further, it was the expectation that Tier 1
would see a larger decline in violence.
Upon coding the Density Score for each armed-conflict, it was discovered that an optimal
and meaningful proximity distance and temporal window could not be found. There was little
variance in the calculated density score values, and in order to increase the variance, the spatial
and temporal windows would needed to have been extended which is counter-intuitive
considering the density score is meant to capture clusters of events occurring nearby and in close
succession.
Compounding this problem was the low number of HF radios and their limited expanse.
For example, three HF radios were installed in 2010. Yet, only a very low ratio of armed-
conflicts occurred within 40 km of these three radios. To assume density score values for armed-
conflicts occurring far from the HF radio installation would omit other external variables such as
local counter-insurgency initiatives or rival forces. Tier values could be modified to only account
46

for armed-conflicts nearer to the HF radios’ location, however, upon mapping every armed-
conflict back to an individual HF radio, the results were found to be lopsided as 24 out of 28 HF
radios had fewer than 50 armed-conflicts map back to them with some HF radios having none.
While the total number of armed-conflicts classified as Tier 1 or Tier 2 is close to being equal,
when examining individual radio locations, the number of armed-conflicts within each radio’s
respective tiers varied widely. This method was abandoned in favor of a more general spatial
distribution model that aggregated armed-conflicts and HF radio locations by country and
territory.
3.4.3 Spatial Distribution of Violence: Summaries of Violence Aggregated by Country, Rates
of Change by Territory, and Comparative Changes using Difference-in-Differences
In light of the challenges associated with these statistical methods, this study takes a
qualitative approach by calculating changes in the distribution of violence over time. First,
violence is aggregated by country to indicate annual trends in the raw counts of violence from
2008 to 2012; this country-level analysis is discussed in Section 3.5. Then, in-depth analysis is
completed by utilizing administrative data for individual territories. By classifying relative rates
of change over time by territory, the resulting level of analysis is compliant with the scales called
for in prior studies (Bahaug and Lujala 2005; Cederman et. al. 2007; Flint et. al. 2009; Tuathail
2010) to detect sub-national trends in insurgency. This territory-level analysis is described in
Section 3.6. Finally, with an objective to analyze the impact of installing HF radios in LRA-
active territories, a Difference-in-Differences descriptive analysis was applied. This approach
was chosen because it has the advantage of measuring the variation in the rates of change
between distinct groups of territories (e.g. a sample group and a control group) while controlling
for an overall, global trend. The inspiration for this method came from Shapiro and Weidman’s
47

study (2012) in which they quantitatively assessed the rate of change in insurgent violence in
Iraq in relation to the installation of cell phone towers and the expansion of mobile infrastructure.
They employed a Difference-in-Differences statistical approach to calculate the variance
between the rates of insurgent attack in the 180 days preceding a cell phone tower installation
and the rates of insurgent violence in the 180 days immediately following the tower’s installation
(Shapiro and Weidman 2012).
While limitations of this study, specifically the low number of territories in which HF
radios were located, prevented a Difference-in-Differences statistical method from being
completed, an analysis of comparative changes represented by the Difference-in-Differences
model is able to describe the changes between the two groups of territories without denoting
statistical significance. The Difference-in-Differences descriptive analysis is discussed in detail
in Section 3.7.

48

3.5 Summaries of Violence Aggregated by Country
Conflict data spatially joined with the “Countries” feature class were aggregated by three
categories of violence: armed-conflicts, civilian murders, and civilian abductions. Values
representing summary totals of violence were coded as attributed values in the “Countries”
feature class and were separated by year. The summary values were exported in table format and
used to create line graphs plotting annual summaries of violence for each country from 2008 to
2012. The background analysis at the national level helps provide context for the in-depth
territory-level analyses that follow in Sections 3.6 and 3.7, which describe sub-national trends in
violence. The detailed flowchart for the country-level process is shown in Figure 15.

Figure 15: Aggregating categories of violence by Country and exporting line graphs

3.6 Annual Rates of Change in the Spatial Distribution of Violence by Territory
With conflict data spatially joined with the “Territories” feature class, Modelbuilder was
used to generate 15 new territory feature classes. Each new feature class represented one
category of violence in a given year (e.g. Armed-conflicts_2008, Civilian-murders_2008, etc.).
Each feature class contained values indicating the normalized rate of violence for each year.
Tables 3, 4, and 5 show the normalized rates of violence by territory for each category.
49

Table 3: Armed-Conflicts by Territory, 2008-2012: Normalized by Population
Country Territory_Name 2008 2009 2010 2011 2012
SSD Raja 0.00 0.04 0.00 0.07 0.00
Wau 0.00 0.00 0.01 0.00 0.00
Nahr Yew 0.01 0.01 0.00 0.00 0.00
Meridi 0.05 0.02 0.05 0.00 0.00
Mundri 0.00 0.02 0.00 0.00 0.00
Tombura 0.05 0.16 0.25 0.16 0.02
Yambio 0.02 0.02 0.10 0.10 0.00
DRC Ango 0.00 4.15 3.69 2.54 1.84
Bambesa 0.00 0.00 0.00 0.06 0.00
Bondo 0.00 0.00 0.05 0.00 0.16
Poko 0.00 0.42 0.21 0.00 0.00
Dungu 1.39 3.08 8.64 5.02 4.34
Faradje 0.20 0.44 0.36 0.96 0.80
Niangara 0.07 1.22 1.86 0.36 0.43
Watsa 0.00 0.06 0.03 0.06 0.00
CAR Djemah 0.00 2.18 0.00 0.00 0.00
Obo 0.08 1.05 0.31 0.39 0.14
Zémio 0.00 0.12 1.19 0.65 0.71
Yalinga 0.00 0.00 0.77 0.00 0.58
Bakouma 0.00 0.00 0.10 0.19 0.29
Bangassou 0.00 0.02 0.03 0.02 0.05
Rafaï 0.00 0.14 3.08 0.43 1.22
Birao 0.00 0.00 0.10 0.00 0.00

Table 4: Civilian Murders, 2008-2012: Normalized by Population
Country Territory 2008 2009 2010 2011 2012
SSD Raja 0.00 0.04 0.00 0.00 0.00
Wau 0.00 0.00 0.01 0.00 0.00
Nahr Yei 0.02 0.04 0.00 0.00 0.00
Meridi 0.07 0.10 0.06 0.00 0.00
Mundri 0.00 0.07 0.00 0.00 0.00
Tombura 0.56 0.11 0.31 0.22 0.00
Yambio 0.21 0.02 0.14 0.09 0.00
DRC Ango 0.00 6.46 13.83 1.38 0.12
Bambesa 0.00 0.00 0.00 0.00 0.00
Bondo 0.00 0.00 0.11 0.00 0.00
Poko 0.00 0.10 0.00 0.00 0.00
Dungu 21.07 12.90 10.96 3.54 0.34
Faradje 7.36 2.00 0.08 0.36 0.12
Niangara 1.43 25.76 7.15 0.29 0.07
Watsa 0.00 3.00 0.00 0.03 0.00
CAR Djemah 0.00 14.17 0.00 0.00 0.00
Obo 0.17 2.39 0.36 0.11 0.08
Zémio 0.00 0.24 2.91 0.18 0.42
Yalinga 0.00 0.00 1.55 0.00 0.00
Bakouma 0.00 0.00 0.38 0.05 0.38
Bangassou 0.00 0.03 0.08 0.00 0.03
Rafaï 0.00 0.14 4.08 0.57 1.15
Birao 0.00 0.00 0.02 0.00 0.00
Ouanda Djallé 0.00 0.00 0.51 0.00 0.00

50

Table 5: Civilian Abductions, 2008-2012: Normalized by Population
Country Territory 2008 2009 2010 2011 2012
SSD Raja 0.00 0.24 0.00 0.20 0.00
Wau 0.00 0.00 0.02 0.00 0.00
Nahr Yei 0.12 0.03 0.00 0.00 0.00
Meridi 0.10 0.07 0.04 0.00 0.00
Mundri 0.00 0.11 0.00 0.00 0.00
Tombura 1.26 0.56 0.65 0.33 0.00
Yambio 0.36 0.03 0.28 0.14 0.00
DRC Ango 0.00 49.34 16.83 13.03 3.34
Bambesa 0.00 0.00 0.00 0.32 0.00
Bondo 0.00 0.00 0.16 0.00 0.97
Poko 0.00 1.77 0.63 0.00 0.00
Dungu 15.26 10.83 11.93 8.35 7.63
Faradje 7.60 0.76 0.84 2.48 2.16
Niangara 0.00 18.60 8.30 1.36 0.79
Watsa 0.00 0.60 0.06 0.06 0.00
CAR Djemah 0.00 15.26 0.00 0.00 0.00
Obo 3.69 3.03 0.75 0.80 0.14
Zémio 0.00 0.48 3.39 4.82 1.96
Yalinga 0.00 0.00 29.18 0.00 5.02
Bakouma 0.00 0.00 1.05 2.57 3.24
Bangassou 0.00 0.02 0.44 0.09 0.78
Rafaï 0.00 2.72 19.48 0.93 2.44
Birao 0.00 0.00 0.76 0.00 0.00
Ouanda Djallé 0.00 0.00 22.63 0.00 0.00

Next, feature classes for the same category of violence were joined from one year to the
next using a table join and a common ObjectID (Territory_Name); Armed-conflicts_2008 was
joined with Armed-conflicts_2009, and Armed-conflicts_2009 was joined with Armed-
conflicts_2010, etc. The table joins allowed for the calculation of rate changes for each year-to-
year transition (i.e. 2008-2009, 2009-2010, 2010-2011, and 2011-2012). The rate change was
calculated for the three categories of violence (i.e. armed-conflicts, civilian murders, and civilian
abductions) for each of the 24 territories. Geographic representations of the rate changes were
generated in a series of maps.
The process of calculating the annual rates of change by territory is illustrated in Figure
16. Included in this figure are the additional steps taken to compare changes between the sample
group of territories and the control group as represented by the Difference-in-Differences
descriptive analysis discussed in Section 3.7.
51

Figure 16: Calculating rates of change by territory and comparative changes from
Difference-in-Differences

3.7 Comparative Changes Represented by a Difference-in-Differences Descriptive Analysis
After the rate changes were coded to each feature class, the territories were divided into
two groups: the first group is the sample (n), which is composed of territories in which HF radios
are located. The second group is the control (m), which is composed of territories without HF
radios. HF radios were aggregated by territory and each territory was coded as either having or
not having one or more HF radios. The mean rate of change for the control group was then
52

subtracted from the mean rate of change for the sample group. In Figure 17 is a
conceptualization of this method, illustrating comparative changes between a sample group and a
control group.

Figure 17: Conceptualizing comparative changes between territory groups

The comparative changes between the two groups are calculated using Equation 1, which
is based on the Difference-in-Differences statistical method (Abadie 2005). In Table 6 are listed
the variables used in this analysis with associated assumptions and limitations. In the
calculations/statistics, territories were only included in each year-to-year iteration if they
sustained one or more conflicts during the designated period (e.g. 2008 to 2009). Resulting
values indicate how the rate of violence in territories with HF radios changed in comparison to
territories without HF radios. The detailed flowchart for this approach is included in Figure 16.
Due to the low n, the significance at local scale of the true Difference-in-Differences statistical
analysis could not be established or validated. Therefore, the numerical results derived from this
53

method are shown in chapter four and described qualitatively focusing on a comparative
description of the results. The discussion in Chapter 5 outlines possible venues in which these
results could be further analyzed given the possibility of being validated in their significance or
impact as discussed in Section 5.3.1.
Equation 1: Comparative Changes ( x1 - x2)

Table 6: Variables referenced in the analysis of comparative changes
Variable Description Assumptions Limitations
i Territory The territory sustained at least
one armed-conflict for each year-
to-year interval.
Territories vary widely in area and
population.
r Rate of change Rate of each assumes each
territory's population is static
through each year-to-year
transition.
The rate of change is calculated for
armed-conflict, civilian murders, and
abductions. It does not consider other
forms of looting or harassment.
y Year Annual interval provides a
summary of violence over the
course of a year.
Interval does not account for spikes or
drops in violence that can occur weekly
or monthly.
n Number of territories in sample
group
All HF radios locations are
known.
Territories are not weighted by how
many radios they contain.
m Number of territories in control
group
Territories in control group are
not affected by HF radios in
adjacent territories.
Some territories in the control group
very seldom experience violence
compared to territories in the sample
group.
za Indicator = 1 if HF radios are
present, 0 if not
If an HF radio is present in a
territory, it is being used.
Does not account for how long a radio
has been present.
zb Indicator = 0 if HF radios are
present, 1 if not
"" ""

3.8 Expected Outcome
Understanding that the number of LRA members has been declining, there is an
expectation that there will be an equal decline in the rate of armed-conflicts, murders, and
 
           



























 
 




m
z r r
n
z r r
x x
n
territory i
bi i y y
n
territory i
ai i y y 1 1
2 1
54

abductions over time. Unilateral military operations in the Congo are thought to have decreased
violence in the areas surrounding the Dungu territory in particular. However, this military push
from the east combined with new uprisings and political instability in the CAR means that the
LRA may find refuge along the southern border of the CAR (Ronan 2012). It is also predicted
that there will be a greater decrease in the number of civilian murders and abductions than
overall armed-conflicts. The rationale for this is in the LRA’s dwindling numbers. Many believe
that the organization is in “survival mode” and the majority of their activities now involve
looting and intimidation rather than systematic murder (DeLaurentis 2012).
It is predicted that territories in which civilian-protection programs are prevalent will
experience a greater decline in incidents than territories without them. While this thesis examines
the use of high-frequency radios, it does not consider the effects of leaflet distribution, FM
broadcasts, or the location and strength of counter-insurgency forces. It is assumed that military
operations are happening in areas with a heavy concentration of LRA attacks; however, those
data are not factored into this analysis. This analysis will simply provide a geographic context for
the rate changes along with a proposed method to compare rate changes between territories.
Readers can use this information and incorporate their own understanding of counter-insurgency
efforts and make an assessment as to the possible impact of programs like the high-frequency
radio network discussed here.
55

CHAPTER 4: RESULTS
This chapter documents the spatial distribution of violence and demonstrates change over time
for each category of violence. Results of the distribution of violence are shown by countries and
by territories.
Aggregate summaries of violence by country are expressed by three plots representing
the raw annual volume of conflicts from 2008 to 2012. These are described in Section 4.1.
Calculating the annual rates of change in violence at the territory-level produced fifteen maps
that are separated into five groups with three maps per group. The five groups include a baseline
(Section 4.2.1) and one group for each year-to-year interval (Section 4.2.2). The three maps
within each group represent the change in the rate of violence for each category. With the
exception of the baseline maps, the figures do not depict the actual annual rates of violence. The
calculated annual rate of violence for each category can be found in section 3.6 in Tables 3-5,
and the raw counts of violent events can be found in Appendix B, Tables 8-10: Raw Counts by
Territory. Calculating the comparative changes using Difference-in-Differences produced four
maps that are described in Section 4.3.
4.1 Summaries of Violence Aggregated by Country
A summary of the raw number of armed-conflicts occurring from 2008 to 2012 is shown
in Figure 18. The year 2008 marks the lowest levels of LRA attacks in the five-year span; only
55 attacks occurred during this period. However, violence increased significantly in 2009 and
again in 2010 with attacks becoming increasingly prevalent in the DRC; The LRA launched 399
attacks in 2010 that resulted in civilian injury, and 276 of those attacks occurred in the DRC.
After 2010, the number of armed-conflicts tapered off with 2012 marking the lowest overall
levels in the study. This decline coincides with MONUSCO’s adoption of Security Council
56

Resolution 1925 in 2010, which authorized the use of force to ensure civilian protection.
Notable, however, is the slight upswing in armed-conflicts in the CAR in 2012. Here, the LRA
launched 10 more attacks than in 2011.

Figure 18: Trends of Armed-conflicts by Country (raw counts)

Analogously, the plot in Figures 19 provides additional context to the changing landscape
of LRA violence. While the overall number of armed-conflicts steadily declined after peaking in
2010, the number of civilian-murders declined even more substantially. Although LRA killed 49
civilians in 2012, this is 96 percent fewer civilian than they killed in 2010 and over 99 percent
fewer than in 2009. In fact, the ratio of civilian murders to armed-conflicts in 2009 was 4.89
while the same ratio in 2012 was 0.25. This decrease implies that LRA attacks are yielding far
fewer casualties than before.
0
50
100
150
200
250
300
2008 2009 2010 2011 2012
Armed-conflicts (raw counts)
CAR DRC SSD
57

Figure 19: Trends of Civilian Murders by Country (raw counts)

The total number of civilian abductions is also declining but with slightly more variance
by country (Figure 20). While the overall number of abductions declined after peaking in 2010,
there was a slight increase in 2012 in the CAR with abductions becoming more prevalent in the
southern territories crowding the Congolese border. This trend is similar to those shown in
Figures 18 and 19. The plot in Figure 20 also shows how the total number of civilian abductions
in CAR (218) is 75 fewer than those that occurred in the DRC (293) despite there being three
times as many armed-conflicts in the DRC during the same temporal period. A comparatively
higher percentage of abductions in the CAR supports the theory that the LRA is fleeing pursuing
forces in the DRC and is capitalizing on regional instability in the CAR to replenish their ranks
via forced conscription strategies.
0
100
200
300
400
500
600
700
800
900
1,000
2008 2009 2010 2011 2012
Civilian Murders (raw counts)
CAR DRC SSD
58

Figure 20: Trends of Civilian Abductions by Country (raw counts)
.
4.2 Annual Rates of Change in the Spatial Distribution of Violence by Territory
The baseline maps (Section 4.2.1) are represented by the rate of violence in each
territory, i.e. the number armed-conflicts, civilian murders, and civilian abductions per 1,000
people. The interval rate change maps (Section 4.2.2) are represented by the increase or decrease
in the rate of violence from one year to the next. Also logged in each interval map are the “global
mean change”, the “greatest increase”, and the “greatest decrease.” The “global mean change”
reflects the average rate change across the study area, while the “greatest decrease” and the
“greatest increase” indicate which territories experienced maximum variance from one year to
the next. An assessment is made of each territory to reveal whether the rate of violence, murders,
or abductions in that territory is worsening or getting better than the overall trend. It should be
0
200
400
600
800
1,000
1,200
2008 2009 2010 2011 2012
Civilian Abductions (raw counts)
CAR DRC SSD
59

noted that territories were only factored into the global mean change if they sustained one or
more armed-conflicts within the two-year span captured by the map. For example, if a territory
sustained zero armed-conflicts in 2011 and 2012, then it was not included in the 2011 to 2012
analysis because it was outside of the LRA’s area of operations during that timeframe.
Reconciling this information provides us with the ability to identify spikes and reductions
in violence and make evaluations regarding the potential causes behind the variance in a
geographic context. Potential causes include changes in political leadership as well as the
implementation of counter-insurgency and civilian-protection initiatives discussed in chapter
two.

60

4.2.1 Baseline Distributions by Category of Violence
2008 Baseline: Armed-conflicts
In 2008, the baseline distribution of armed-conflicts (Figure 21) was concentrated in the
northeastern region of the DRC and the southwestern region of South Sudan. Across all
territories where the LRA was active, the mean rate of armed-conflicts per 1,000 people was
0.24. Beginning Christmas Day 2008, the LRA attacked numerous villages across three
communities within Dungu and Faradje in close succession. This was in conjunction with the
Christmas Day Massacres discussed in section 2.4, explaining why the highest rates of armed-
conflicts are in Dungu and Faradje.

Figure 21: The Rate of Armed-conflicts in 2008

61

2008 Baseline: Civilian murders
In 2008, the average rate of civilian murders in LRA-active territories was 3.86 murders
per 1,000 people (Figure 22). The Christmas Day massacres that took place during this time
caused the rate in that territory to spike to 21.07 making Dungu the deadliest of all territories
(HRW 2009).

Figure 22: The Rate of Civilian Murders in 2008

62

2008 Baseline: Civilian Abductions
In 2008, the baseline rates of civilian abductions (Figure 23) had a pattern similar to the
rates of armed-conflicts and civilian murders. The mean rate of abductions across the area of
operations was 4.06 abductions per 1,000 people. The highest rates were in Dungu and Faradje
measuring 15.26 and 7.60 respectively making the two territories again the most dangerous in
2008. Also evident is the alarming number of murders and abductions that occurred during each
armed-conflict. For example, the rate of armed-conflict in Dungu was 1.39, however, the rate of
murders was 21.07, and the rate of abductions was 15.26. Each armed-conflict, on average,
yielded 15 civilian deaths and 10 civilian abductions.

Figure 23: The Rate of Civilian Abductions in 2008

63

4.2.2 Interval Rate Changes by Category of Violence
2008 to 2009: Armed-conflicts
The map in Figure 24 illustrates the expansion of the LRA’s area of operations from 2008
to 2009. The rate of armed-conflict increased in fourteen territories during this time, and the rate
only decreased in two. The global mean, or the mean rate of change for all territories that saw
violence in 2008 or 2009, was +0.66 incidents per 1,000 people. The rate of armed-conflicts
increased most in Ango (+4.15)and Djemah (+2.18). Rates also increased in Niangara and
Dungu, both located in the DRC. Niangara, as discussed in section 2.4, is the territory in which
the Makomba massacres occurred in 2009 (HRW 2010). The Christmas Day massacres that
began in 2008 carried through into January hence the elevated rates in Dungu and Faradje.

Figure 24: Change in the Rate of Armed-conflicts from 2008 to 2009

64

2008 to 2009: Civilian murders
The rate of civilian murders in 2009 rose dramatically in Niangara partly due to the
Makomba massacres which killed 321 civilians in 10 villages over four days (Figure 25). This
elevated the total rate of murder in Niangara to 25.76 murders per 1,000 people. The rate of
civilian murder actually decreased in the Dungu and Faradje territories, which had traditionally
been the most violent areas, despite the rate of armed-conflicts rising the same year. In these
territories, individual armed-conflicts, on average, yielded fewer civilian murders than in years
past.

Figure 25: Change in the Rate of Civilian Murders from 2008 to 2009

65

2008 to 2009: Civilian Abductions
In 2009, the global mean rate of civilian abductions increased at a rate of +4.47 civilian
abductions per 1,000 people (Figure 26). Driving the upward trend was increased LRA activity
in Ango and a total of 428 abductions. The spike in abductions elevated the rate in Ango by
+49.34. Conversely, Dungu and Faradje saw a decrease in abductions, like murders, despite a
rise in the number of armed-conflicts.

Figure 26: Change in the Rate of Civilian Abductions from 2008 to 2009

66

2009 to 2010: Armed-conflicts
The map in Figure 27 shows how the number of armed-conflicts peaked in 2010 with the
global trend increasing by +0.36 armed-conflicts per 1,000 people. However, we do see many of
the formerly violence areas improving. In fact, a separation begins to reveal itself as violence
increases in South Sudan and CAR in territories which had initially seen little LRA violence.
This indicates a large expansion from the original concentration of LRA activity in the
Northeastern DRC. 2010 is also the first year that MONUSCO began operating with revised
mandate to increase cooperation with the Congolese army and apply the use of force against
LRA members. MONUSCOs forward base of operations is located in Dungu bringing more
attention to the region. 2010 also marks the first year that HF radios are being used to share
intelligence between communities and counter-insurgency forces.

Figure 27: Change in the Rate of Armed-conflicts from 2009 to 2010
67

2009 to 2010: Civilian murders
The year 2010 is very significant as it marks the first year, since 2008, in which the
global mean rate of civilian murders begins to decrease (Figure 28). Even though armed-conflicts
in Dungu rose in 2010, the rate of civilian murders actually decreased. The same pattern
occurred in Niangara. The fact that armed-conflicts are yielding fewer deaths in these territories
may be attributable to the counter-insurgency initiatives based in Dungu. Conversely, the rate of
armed-conflicts in Ango decreased by -0.46 while the rate of civilian murder increased by +7.38,
which is indicative of a westward trending LRA insurgency.

Figure 28: Change in the Rate of Civilian Murders from 2009 to 2010

68

2009 to 2010: Civilian Abductions
In Figure 29, the global mean rate of murders decreased across the study area while the
rate of civilian abductions increased. This is due to a spike of civilian abductions occurring in the
CAR. In Ango, the rate of civilian abductions decreased despite a record number of civilian
murders in the same year. The decrease is misleading because there was still a high number of
abductions in Ango in 2010; there was simply fewer than in 2009.

Figure 29: Change in the Rate of Civilian Abductions from 2009 to 2010

69

2010 to 2011: Armed-conflicts
The transition into the year 2011 is significant as this is the first year, since 2008, in
which the global mean rate of armed-conflicts decreased; the mean rate of change was -0.49
armed-conflicts per 1,000 people (Figure 30). The rate of armed-conflicts decreased in fifteen
territories, and the rate increased in six territories. The rate of violence in Dungu and Rafaii, the
two hardest hit territories in 2010, declined by -3.62 and -2.65 respectively. In 2011, we also
begin to see fragmentation in the organization with LRA violence increasing by very small
amounts in only a few areas. By this time, MONUSCO had engaged increasingly with regional
forces, the HF radios were widely being used in Northeastern DRC, and a series of other counter-
insurgency and LRA initiatives (e.g. leaflet distribution, FM transmissions, and flyovers) were
taking place.

Figure 30: Change in the Rate of Armed-conflicts from 2010 to 2011
70

2010 to 2011: Civilian murders
The steepest decline in civilian murders occurred from 2010 to 2011 (Figure 31). Across
the study area, the rate of civilian murders increased in only two territories: Faradje and Watsa.
Murders declined in the other fifteen territories despite rate of armed-conflicts having increased
in three of them.

Figure 31: Change in the Rate of Civilian Murders from 2010 to 2011

71

2010 to 2011: Civilian Abductions
From 2010 to 2011, the overall rate of abductions dropped significantly, although
abductions rose slightly in different areas throughout the map (Figure 32). Similar to the rates of
armed-conflicts and civilian murder maps for same interval, there is evidence of fragmentation
with disparate groups committing more abductions instead of murders. This may be the result of
a reeling insurgency lacking the strength they had in 2009 and 2010.

Figure 32: Change in the Rate of Civilian Abductions from 2010 to 2011

72

2011 to 2012: Armed-conflicts
In the transition from 2011 to 2012, the rate of armed-conflicts increased in the CAR at a
rate faster than any other country indicating a clear dichotomy in the study area (Figure 33). The
global mean rate of change during this period was -0.03 indicating a continuing decrease in the
overall rate of LRA violence. During this time, violence nearly disappeared from South Sudan
altogether. In 2012, only one LRA armed-conflict resulting in civilian injury occurred in South
Sudan, and there were zero murders or abductions in the entire year. While the rate of armed-
conflicts was still highest in the DRC, there is evidence of a westward trend with the rate of
armed-conflicts increasing in the CAR. This trend may be attributable to MONUSCO who by
this time had gained considerable ground against the LRA. It is also worth noting that, during
this period, new HF radios were being installed nearly every other month.

Figure 33: Change in the Rate of Armed-conflicts from 2011 to 2012
73

2011 to 2012: Civilian murders
In the 2011 to 2012 transition, it is clear that the rate of civilian murders rose only in the
CAR (Figure 34). In 2012, 148 armed-conflicts occurred in the DRC, yet they yielded only 13
murders. At the same time, 46 armed-conflicts occurred in the CAR, however, those conflicts
yielded 36 murders. This marks the first time that conflicts in the CAR resulted in a higher rate
of murders than in the DRC.

Figure 34: Change in the Rate of Civilian Murders from 2011 to 2012

74

2011 to 2012: Civilian Abductions
From 2011 to 2012, the rate of abductions increased in the CAR and decreased in the
DRC, highlighting a greater dichotomy in the study area (Figure 35). The increase in violence in
the CAR is in line with concerns that the LRA is retreating into an increasingly instable region
where they are finding opportunities to replenish their ranks through the abduction and forced-
conscription of children (Stearns 2013). Furthermore, the CAR territories are several hundred
kilometers from MONUSCO’s forward base of operations in Dungu which increases the
logistical challenges associated with pursuing the LRA in a country that has not been an active
participant in the anti-LRA coalition.

Figure 35: Change in the Rate of Civilian Abductions from 2011 to 2012

75

4.3 Comparative Changes Using a Difference-in-Differences Descriptive Analysis
The installation of high-frequency radios reflects one of several civilian protection
methods employed in the region, and a comparative changes from Difference-in-Differences was
used to compare rate changes between the groups to establish whether the presence of radios has
had an impact on the rate of violence against civilians.
Representations of the comparative changes are broken into four maps -- one map for
each year-to-year transition. The sample group, which includes territories with radios, is
highlighted in green and is represented by ( x1) in the inset chart. The control group, which
includes territories without radios, is highlighted in beige and is represented by ( x2) in the inset
chart. Territories were only included in the sample group n and control group m if they sustained
one or more armed-conflicts within each temporal period. For example, if a territory had no
armed armed-conflicts in either 2010 or 2011, then it was not grouped into either n or m for the
2010 to 2011 analysis. Excluding territories with no armed-conflicts allowed the analysis to
focus only on the changes occurring in regions where the LRA was active during each interval.

76

2008 to 2009: Violence Comparative Changes
In 2008 and 2009, there were no HF radios installed and herefore no sample group is
highlighted in Figure 36. However, the LRA increased their activities in all three categories
throughout the study area thereby increasing all rates of violence over time.

Figure 36: Comparative Changes in Violence 2008 to 2009 (no HF radios exist)

77

2009 to 2010: Violence Comparative Changes
In 2010, radios were installed in the Dungu and Poko territories. The mean rate change in
armed-conflicts between these two territories was +2.68 while the control group was +0.13
(Figure 37). This means that violence in the sample group increased at a rate faster than the
control group by +2.54 armed-conflicts per 1,000 people. This is expected because counter-
insurgency initiatives, particularly the installation of the HF radios, targeted those territories with
the highest rate of armed-conflicts. During this year-to-year transition, the rate of murders
decreased in the sample group and the control; however the decrease in the control group was
slightly greater. Conversely, the rate of abductions increased in both groups, but the increase was
greater in the control group.

Figure 37: Comparative Changes in Violence 2009 to 2010

78

2010 to 2011: Violence Comparative Changes
From 2010 to 2011, the rate of armed-conflicts in territories with HF radios decreased at
a greater rate than in territories without them (Figure 38). The same held true for the rate of
murders. Here, the rate of civilian murders in the sample group decreased by -6.62 while the rate
change in the control group was only -0.88. During this time, the rate of abductions also
decreased substantially across the study area; however, the rate of decrease in the control group
was greater than in the sample group.

Figure 38: Comparative Changes in Violence 2010 to 2011

79

2011 to 2012: Violence Comparative Changes
In 2012, the rate of armed-conflicts in territories with HF radios continued to decrease
more than in territories without them (Figure 39). The rate of armed-conflicts in the sample
group decreased by -0.54 while the rate of armed-conflicts in the control group increased slightly
by +0.09. Following this pattern, the rate of murders in the sample group decreased at a rate of
-1.50 murders per 1,000 people while the control group actually increased slightly by +0.03. In
line, the rate of abductions declined in the sample group by -3.69 while the control group, at
+0.31 actually experienced a slight increase. Altogether, this offers evidence that violence in
territories with HF radios has decreased at a rate greater than in territories without them.

Figure 39: Comparative Changes in Violence 2011 to 2012

80

CHAPTER 5: DISCUSSION AND CONCLUSIONS
Chapter Five discusses the impact of the findings, their contribution to existing research on
conflict geographies, and their relevance to ongoing discussions regarding the development of
counter-insurgency efforts in Central Africa. Shortcomings of the data are discussed, and several
recommendations for future geographic studies in insurgency-violence are made.
5.1 Key Observations
Results in this work suggest that the LRA’s area of operations has fluctuated over time in
an apparent reaction to external influences. Since the number of attacks and casualties peaked in
2010, there has been a steady decline in violence committed by the group with the year 2012
marking the lowest levels of violence in the years observed.
In 2012, there were a total of 49 civilian murders. This is a 92.7 percent reduction in
murders since 2010 when MONUSCO first adopted Security Council Resolution 1925. This
represents a 74.8 percent reduction in CAR, a 97.3 percent reduction in the DRC, and a 100
percent reduction in South Sudan. These statistics indicate that the MONUSCO mission, along
with other regional counter-insurgency and civilian-protection initiatives, have been extremely
effective in reducing LRA violence in the region.
There are other spatial changes in the distribution of violence that provide telling insights
to the evolution of the insurgency group. Notable is the upswing in the rate of violence,
particularly abductions, in the CAR in 2012. Since 2010, MONUSCO has prioritized their
mission resources in the Northeastern DRC. This includes the establishment of a 1,200 strong
peacekeeping force in Haut Uele, the Northeastern Congolese district that includes the Dungu,
Faradje, and Niangara territories (UN 2012). MONUSCO missions in the area have included
81

rehabilitating roadways, escorting civilians and humanitarian partners, and participating in leaflet
distribution and FM radio broadcasts designed to convince LRA members to defect.
The increase in violence in the CAR is in line with concerns that the LRA is retreating
into an increasingly instable region where they are finding opportunities to replenish their ranks
through the abduction and forced-recruitment of children (Stearns 2013). Furthermore, the CAR
territories are several hundred kilometers from MONUSCO’s forward base of operations in
Dungu, which increases the logistical challenges associated with pursuing the LRA and
protecting civilians further away.
MONUSCO’s base of operations is not far from the Congolese border with Sudan where,
in 2012, there were no civilian murders or abductions. During this time, there was only one LRA
attack that resulted in civilian injury. This is a significant achievement for a country that
sustained 176 murders and 363 abductions in the four years prior. The reduction may be the
result of a strengthening South Sudan and Dinka leadership following the country’s
independence in July 2011. However, there is the possibility of LRA resurgence following
increasing instability in South Sudan in 2013 and 2014. Ethnic infighting has recently caused
instability in the region, and whether LRA members will resume operations is yet to be seen.
Reviewing the analysis of comparative changes, we can qualitatively describe how the
rate of violence consistently declined within the few territories composing the sample group. As
expected, the fluctuation in violence for territories without radios was inconsistent; however, this
study is unable to ascertain that this is due to the absence of radios. Because of the low sample
size and the inability to test the model at other resolutions, the findings associated with this
procedure are simply descriptive of a small number of cases.

82

This thesis offers an integrated examination of the LRA insurgency in a geographic
framework that acknowledges external influences and derives patterns of insurgency behavior
although with the intrinsic limitations imposed by the data and scale. Moreover, spatial
observations address disjuncture in examining insurgencies as static operations and promote the
efficiency of spatially integrated approaches. Accordingly, the spatial observations described in
Chapter 4 will be provided to the Invisible Children NGO and made available to other parties
interesting in assessing flux in the LRA insurgency and the possible impacts of regional counter-
insurgency and civilian protection initiatives.
5.2 Contrast with Previous Studies
The diminishing of LRA attacks around Dungu suggests that MONUSCO’s forward base
of operations have been influential there as an anti-insurgency organization. This is contrary to
previous analyses by Copeland (2012) and Aboubacar (2012) that cite ongoing violence in
Northeastern Congo as MONUSCO’s inability to work effectively with regional forces and
protect civilians. This work does not suggest that the same counter-insurgency efforts would be
successful against other paramilitary organizations in Africa, but the contrary observations
suggest that the LRA area of operations has reacted to the tailored counter-insurgency strategies
targeting the organization’s structure, priorities, ethnic and ideological foundations.
This study stages variation in violence in social and topographic contexts specific to the
LRA, discussing the challenges associated with interstate cooperation faced by coalition forces
in areas like the CAR, without relying on estimates of violence perpetrated by unrelated
insurgency groups. And, while this study does not provide an examination of the impact of
ethno-linguistic distribution and population densities offered in past GIS conflict analyses
(Buhaug and Lujala 2005; Cederman et. al. 2007), it does adhere to a scale of analysis that
83

allows for sub-national trends in insurgency to be detected. Understanding the non-static nature
of insurgency operations better informs organizations working to improve civilian security. By
adopting intelligent strategies and utilizing geospatial analyses to assess insurgency aggression,
we can avert our research from blanket, generalized statements and begin developing more
comprehensive profiles targeting specific organizations.
5.3 Recommendations for Future Research Directions
Since January, 2014, interethnic violence has increased in South Sudan. After counter-
insurgency forces nearly eliminated the presence of LRA attacks in South Sudan, it is possible
that the political instability there may provide the LRA with the opportunity to regain their
footing. The LRA could potentially coordinate with one of the opposing sides and volunteer their
members for guerilla attacks, like they did in 2005, in exchange for supplies. While this is
unlikely, given that they are historical enemies, LRA activity in the South Sudan should be
closely observed for indications of new ethnic or political alignments and for the assimilation of
former-LRA members into existing South Sudanese paramilitary forces. The adoption of guerilla
fighters in foreign militias is a frequent occurrence in African insurgencies when forces become
disbanded or are defeated.
Ongoing studies should also continue exploring LRA violence in the CAR in the years
2013 and 2014 to see how the rate of violence has evolved. The Invisible Children and the
Resolve, in coordination with UN Agencies, have committed to installing more HF radios in
CAR. Therefore, a continuation of the comparative analysis may yield more significant results
by leveraging the increased sample size.

84

5.3.1 Comparative Analysis for High-frequency Radio Distribution
A drawback of the comparative changes from Difference-in-Differences is the resolution
of the data being used. Many of the benefits of having an HF radio are local and may not be
adequately scaled at the territory-level. To examine the local impacts of utilizing the HF radio
network a municipal-level data would need to be used. An alternative approach could be tested
using a weighted analysis (Hendrix and Glaser 2007) in which the individual territories have an
assigned weight based on the number of radios within them. Hendrix and Glaser (2007)
employed a similar approach to measuring significant covariates of conflict by weighting data
nodes and creating linear time series regressions at multiple scales of aggregation.
An option for applying a comparative analysis to this study would include using linear
regression at annual intervals for each category of violence and classifying an overall trend
within each zone/country. From this, we would be able to infer whether the weight (i.e. the
number of radios within a zone) correlates with rate changes across any of the three categories of
violence (Pers. Com. Dr. Paganelli 2014).
This approach would separate the study area into three zones/countries (e.g. the CAR,
DRC, and South Sudan), aggregate the individual territories for which we already have data,
combine their categorical event totals (e.g. armed-conflicts, civilian murders, and civilian
abductions), and then create a series of plots to form cumulative trends representative of each
category of violence. A linear trend would be created for each country for each year-to-year
transition in which radios were in existence. For example, CAR, DRC, and South Sudan would
each have four distinct linear trends: 2009 to 2010, 2010 to 2011, and 2011 to 2012. Each pair of
years would be plotted on the X-axis while the number of events, i.e. armed-conflicts, murders,
and abductions, would be plotted on the Y-axis. The cumulative trend for each country would be
85

indicative of the overall regional tendency (Pers. Com. Dr. Paganelli 2014). For each temporal
period, we could then consider the individual territories and their relative weight, i.e. the number
of radios within their area, and how the rate of change during the same year-to-year transition
deviated from the cumulative trend. This effect can be displayed geographically indicating how
the rate of change varies from the trend and among territories within the same country.

86

REFERENCES

Abadie, A. 2005. Semiparametric Difference-in-Differences Estimators. In The Review of
Economic Studies, 72 (250): 1-18.
Aboubacar, M. 2012. The MONUSCO Contraction. In Institute for Policy Studies: Foreign
Policy in Focus 9 January.
Andreas, P. 2002. Transnational Crime and Economics Globalization. In Transnational
Organized Crime and International Security. Business as Usual? ed. M. Berdal, M.
Serrano, 37-52, Boulder, CO: Lynne Rienner Publishers, Inc.
Anning, K., A. McIntyre. 2004. From Youth Rebellion to Child Abduction: The Anatomy of
Recruitment in Sierra Leone. In Invisible Stakeholders: The Impact of Children on War,
ed. Institute for Security Studies.
Arquilla, J., D. Ronfeldt, M. Zanini, 1999. Networks, Netwar, and Information-Age
Terrorism. In Strategic Appraisal: The Changing Role of Information in Warfare. ed. Z.
Khalilzad, J. White. Santa Monica, CA: RAND.
Baker, J. 2009. Successful Liberalization Transforms Uganda’s Mobile Market, Pyramid
Research Finds. In Pyramid Research 7 May. http://www.prnewswire.co.uk/news-
releases/successful-liberalization-transforms-ugandas-mobile-market-pyramid-research-
finds-153780105.html (last accessed 2 February 2014).
Baguma, R. 2012. American Politician Puts Bounty on Kony's Head. In New Vision 20 Feb.
BBC. 2014. DRCongo: Key Facts. In BBC News.
http://news.bbc.co.uk/2/shared/spl/hi/guides/456900/456977/html (last accessed 23
February 2014).
Beber, B., c. Blattman. 2011. The Logic of Child Soldiering and Coercion July.
87

http://chrisblattman.com/2011/07/18/the-logic-of-child-soldiering/ (last accessed 23
February 2014).
Berman, E., J. Felter, J. Shapiro, E. Troland. 2013. Modest, Secure and Informed:
Successful Development in Conflict Zones. In National Bureau of Economic Research:
Working Paper Series.
Blakely, R. 2005. Terrorists 'Threaten' Iraq Mobile Operators. In Times Online 22 July.
http://omega.twoday.net/stories/855414/ (last accessed 23 February 2012).
Bontemps, S., P. Defourny, E. Van Bogaert, O. Arino, V. Kalogirou, J. Perez. 2011. Globcover
2009: Land Cover, Africa, and the Arabian. Ed. U. C. Louvain & ESA Team.
Brand, J. 2007. Insurgents Target Water and Electricity, but Spare the Cell Phone." In WITF:
Newshour 29 January.
Buhaug, H., Lujala, P. 2005. Accounting for Scale: Measuring Geography in Quantitative
Studies of Civil War. Political Geography 24 (4): 399-418.
Bunting, I. 2011. Lord's Resistance Army. In The SAGE Encyclopedia of Terrorism, ed. Gus
Martin, 2
nd
ed., 264-367. Thousand Oaks, CA: SAGE Publications, Inc.
Cederman, L., Rod, J., Weidmann, N. 2006. Geo-Referencing of Ethnic Groups: Creating a
New Dataset. In GROW Workship. Peace Research Institute Oslo.
Copeland, T. 2012. Civilian Protection in the Eastern DRC: Evaluation of the MONUSCO
Peacekeeping Operation. Presented to the Naval Postgraduate School.
Cordesman, A. 2005. Iraq's Evolving Insurgency: The Nature of Attacks and Patterns and Cycles
in the Conflict. In Center for Strategic and International Studies 3 February.
Delaurentis, J. 2012. Security Council Briefing on UNOCA/LRA. Presented
for the State Department Documents / FIND 17 July. New York, NY.
88

Economy Watch. 2013. Telephone Statistics: Number of Fixed Lines, Mobile Phone Penetration
Rate. http://www.economywatch.com/economic-statistics/economic-
indicators/Telephone_Statistics/ (last accessed 23 February 2012).
Estefan, F., W. Weber. 2012. Mobile-Enhanced Patricipatory Budgeting in the DRC. In World
Bank: Information and Communications Development 13 February.
Flint, C., P. Diehl, J. Scheffran, J. Vasquez, S. Chi. 2009. Conceptualizing Conflictspace:
Toward a Geography of Relational Power and Embeddedness in the Analysis of Interstate
Conflict. In Annals of the Association of American Geographers, Special Issue:
Geographies of Peace and Armed Conflict 10 November, 99 (5): 827-35.
Fluri, J. 2011. Bodies, Bombs and Barricades: Geographies of Conflict and Civilian (in)Security.
In Transactions of the Institute of British Geographers 36 (3): 280-96.
Goodchild, M. 2007. Citizens as Sensors: Web 2.0 and the Volunteering of Geographic
Information. In GeoFocus 7: 8-10
Hendrix, C., Glaser, S. 2009. Trends and Triggers: Climate, Climate Change and Civil Conflict
in Sub-Saharan Africa. In Political Geography 26 (6): 695-715.
Human Rights Watch. 2009. The Christmas Massacres: LRA Attacks on Civilians in Northern
Congo. 16 February. http://www.hrw.org/node/80773 (last accessed 23 February 2014).
———. 2010. Trail of Death: LRA Atrocities in Northeastern Congo. 28 March.
http://www.hrw.org/reports/2010/03/28/trail-death (last accessed 23 February 2014).
———. 2012. Q&A on Joseph Kony and the Lord's Resistance Army. 21 March.
http://www.hrw.org/news/2012/03/21/qa-joseph-kony-and-lords-resistance-army (last
accessed 23 February 2014).
Ionosphere. 1999. In Focal Dictionary of Telecommunications, Focal Press.
89

Johnston, P. 2008. The Geography of Insurgent Organization and Its Consequences for Civil
Wars: Evidence from Liberia and Sierra Leone. In Security Studies 17 (1): 107-37.
Library of Congress. 1990. Uganda: The Ten-Point Program. In The Library of Congress:
Country Studies. http://lcweb2.loc.gov/cgi-
bin/query/r?frd/cstdy:@field%28DOCID+ug0112%29 (last accessed 23 February 2014).
———. 1990. Uganda after Amin: The Interim Period: 1979-80. In The Library of Congress:
Country Studies. http://lcweb2.loc.gov/cgi-
bin/query/r?frd/cstdy:@field%28DOCID+ug0022%29 (last accessed 23 February 2014).
———. 1990. Uganda: The Second Obote Regime: 1981-85. In The Library of Congress:
Country Studies. http://lcweb2.loc.gov/cgi-
in/query/r?frd/cstdy:@field%28DOCID+ug0023%29 (last accessed 23 February 2014).
Leahy, K. 2005. The Impact of Technology on the Command, Control, and Organizational
Structure of Insurgent Groups. Presented to US Army Command and General Staff
College, 17 June.
Kobayashi, A. 2009. Population Size, Concentration, and Civil War: A Geographically
Disaggregated Analysis. Annals of the Association of American Geographers, Special
Issue: Geographies of Peace and Armed Conflict 99 (5): 819-26.
Menondji, M. 2013. Problematic Peacekeeping in the DRC: From MONUC to MONUSCO. 2
February. http://thinkafricapress.com/drc/problematic-peacekeeping-drc-monuc-monusco
(last accessed 23 February 2014).
Miles, D. 2004. Hotline Succeeding in Foiling Iraqi Insurgents. In American Forces Press
Service 29 December. http://www.defense.gov/News/NewsArticle.aspx?ID=24486 (last
accessed 23 February 2014).
90

Muckian, M. 2006. Structural Vulnerabilities of Networked Insurgencies: Adapting to the New
Adversary. In Parameters Winter 2006: 14-25.
Mukeba, A. 2013. USAID Brings New Communications Options to Remote Communities in
Northeastern Democratic Republic of the Congo. USAID/DRC Development Outreach
and Communications 23 March. http://blog.usaid.gov/2013/05/usaid-brings-new-
communications-options-to-remote-communities-in-northeastern-democratic-republic-of-
the-congo/ (last accessed 23 February 2014).
Pedersen, J., T. Sommerfelt. 2007. Studying Children in Armed Conflict: Data Production,
Social Indicators, and Analysis. In Social Indicators Research 84 (3): 251-69.
Poole, S. 2013. Counter-LRA Program Manager, The Invisible Children. Personal Interview, 24
April.
Posner, D. 2004. Measuring Ethnic Fractionalization in Africa. In American Journal of Political
Science 48 (4) (2004): 849-63.
Riddell, M. 2009. Rebuilding the Lives of Congo's Child Soldiers. In The Telegraph 16 July.
http://www.telegraph.co.uk/news/worldnews/africaandindianocean/congo/5836759/Rebui
lding-the-lives-of-Congos-child-soldiers.html (last accessed 23 February 2014).
Resolve and the Invisible Children. 2012. LRA Crisis Tracker: 2012 Annual Security
Brief, February. http://reports.lracrisistracker.com/pdf/2012-A-LRA-Crisis-Tracker-
Annual-Brief.pdf (last accessed 23 February 2014).
———. 2013. LRA Crisis Tracker (2009-2013). http://www.lracrisistracker.com
———. 2013. Map Methodology & Database Codebook v 1.6 ed. The Invisible Children for the
LRA Crisis Tracker.
Ronan, P. 2013. From the Crisis Tracker: Good News on Defections, but LRA Still Abducted
91

517 People in 2012. The Resolve 4 February. http://www.theresolve.org/2013/02/from-
the-crisis-tracker-annual-security-brief-for-2012-shows-a-surge-in-defections-517-
abducted/ (last accessed 23 February 2014).
Semble, K. 2006. U.S. Backs Hot Line in Iraq to Solicit Tips About Trouble. In The New York
Times 5 November.
http://www.nytimes.com/2006/11/05/world/middleeast/05tips.html?pagewanted=print&_
r=0 (last accessed 23 February 2014).
Shapiro, J., N. Weidman. 2011. Talking About Killing: Cell Phones, Collective Action, and
Insurgent Violence in Iraq. In Newshour.
http://www.princeton.edu/~jns/papers/SW_2011_Cell_Phones_Insurgency_31MAY11.p
df (last accessed 23 February 2014).
———. 2012. Is the Phone Mightier Than the Sword? Cell Phones and
Insurgent Violence in Iraq.
https://www.princeton.edu/~jns/papers/SW_2011_Cell_Phones_Insurgency_07JAN12.pd
f (last accessed 23 February 2014).
Schmidt, E., J. Cohen. 2013. The Future of Conflict, Combat, and Intervention. In The
New Digital Age: Reshaping the Future of People, Nations, and Business 6. New York,
NY: Alfred A. Knopf.
Stearns, S. 2012. CAR Instability Helps Lord's Resistance Army Regroup. In Voice of America,
14 May. http://www.voanews.com/content/car-instability-helps-lords-resistance-army-
regroup/1660577.html (last accessed 23 February 2014).
Tuathail, G. 2010. Localizing Geopolitics: Disaggregating Violence and Return in Conflict
Regions. In Political Geography 29 (5): 256-65.
92

Trofimov, Y. 2010. Cell Carriers Bow to Taliban Threat. In The Wall Street Journal 22 March.
2010.
http://online.wsj.com/news/articles/SB20001424052748704117304575137541465235972
(last accessed 23 February 2014).
U.N. 2010. Factsheet: Child Soldiers. UNICEF.
http://www.unicef.org/emerg/files/childsoldiers.pdf (last accessed 23 February 2014).
———. 2013. Republique Democratique Du Congo: Plan D'action Humanitaire. UN Office for
the Coordination of Humanitarian Affairs.
———. 2012. What MONUSCO Is Doing to Address the LRS Threat. United Nations
Organization Stabilization Mission in the Democratic Republic of the Congo. New York,
NY. http://www.un.org/en/peacekeeping/documents/factsheet_%20MONUSCO.pdf (last
accessed 23 February 2014).
U.N.S.C. 2010. Resolution 1925 (2010). United Nations Security Council. New York, NY.
———. 2012. "Resolution 2053 (2012). United Nations Security Council. New York, NY.
———. 2013. "Resolution 2098 (2013)." United Nations Security Council. New York, NY.
Ungeleider, N. 2011. Cybermapping Africa's Strangest Conflict. In Fast Company 3 October.
Vanvider, J. 2013. Group Tracking LRA Sees Defections Weakening Rebel Force. In McClatchy
– Tribune Business News 4 February.
http://search.proquest.com/docview/1283832123?accountid=14749 (last accessed 7
December 2013)
Whitman, S., T. Zayed, C. Conradi, J. Breau, J. 2012. Child Soldiers: A Handbook for
Security Sector Actors." The Romeo Dallaire Child Soldiers Initiative: Centre for
Foreign Policy Studies.
93

World Bank. 2012. Cell Phones for Citizen Engagement in the DRC. World Bank.
http://wbi.worldbank.org/wbi/stories/cell-phones-citizen-engagement-drc# (last accessed
23 February 2014).

94

APPENDIX A: ARMED-CONFLICT DATA COLLECTION METHODS
Table 7: Data collection codebook methodology
1. Data Collection Reports are gathered from a variety of sources:
• HF radio towers in DRC and CAR
o Civilians report activity to HF radio tower operators
o Over 30 HF radio operators call the Dungu hub twice daily to report armed-group
activity
o Activity is entered into a spreadsheet and then sent to data coders
• UN & NGO reports
• News & media outlets
• Civil society contacts in local communities
• Field research conducted by Resolve and Invisible Children staff
2. Database Entry Database entry:
• Reports are divided between a team of coders from both Invisible Children and
Resolve. Coders determine if the source is reliable or unreliable (see section 4.2B of
the codebook). Before an incident is reported, the coder reads through other incidents
to check for duplicates.
Verification ratings:
• After an incident is categorized, each incident is given a verification rating (see
section 4.2A of the codebook).
• If a coder determines that an incident was potentially committed by the LRA, the
incident is rated on the LRA Actor Verification Scale (see section 4.2C of the codebook).
3. Data Review • A second data coder reviews each incident to catch human errors and duplicate
reports (see section 4.1E).
• IC and Resolve staff with field experience review sensitive incidents immediately and
review all incidents every three months. Should these staff members feel an incident
was misreported, the incident is corrected. External LRA and regional experts are
consulted as necessary.
4. Data Mapping & Sharing • After an incident is entered and approved to be mapped, it appears on the LRA Crisis
Tracker website.
• Data is regularly sent to UN agencies and humanitarian practitioners for comparison
and collaboration.
5. Data Revamp • As the database grows and policies are updated to reflect best practices, data coders
revisit and “revamp” the data when needed.
6. Data Analysis & Reporting • Crisis Tracker staff analyze data for trends and patterns in LRA activity.

• Specific areas and provinces are also analyzed for increases or decreases in the
number and type of attack.

• After analysis has been completed and reviewed, it is reported in various Crisis
Tracker reports that can be found on the LRA Crisis Tracker website.

Source: Resolve and the Invisible Children. "Map Methodology & Database Codebook v 1.6" ed.
The Invisible Children for the LRA Crisis Tracker, 2012.

95

APPENDIX B: RAW COUNTS BY TERRITORY
Table 8: Armed-Conflicts - Raw Counts by Territory
Country Territory 2008 2009 2010 2011 2012
SSD Raja 0 2 0 4 0
Wau 0 0 1 0 0
Nahr Yei 3 2 0 0 0
Meridi 4 2 4 0 0
Mundri 0 2 0 0 0
Tombura 3 9 14 9 1
Yambio 3 3 15 15 0
DRC Ango 0 36 32 22 16
Bambesa 0 0 0 1 0
Bondo 0 0 1 0 3
Poko 0 4 2 0 0
Dungu 33 73 205 119 103
Faradje 5 11 9 24 20
Niangara 1 17 26 5 6
Watsa 0 2 1 2 0
CAR Djemah 0 4 0 0 0
Obo 3 38 11 14 5
Zémio 0 2 20 11 12
Yalinga 0 0 4 0 3
Bakouma 0 0 2 4 6
Bangassou 0 1 2 1 3
Rafaï 0 2 43 6 17
Birao 0 0 5 0 0
Ouanda Djallé 0 0 2 0 0

96

Table 9: Civilian Murders - Raw Counts by Territory
Country Territory 2008 2009 2010 2011 2012
SSD Raja 0 2 0 0 0
Wau 0 0 1 0 0
Nahr Yei 4 8 0 0 0
Meridi 6 8 5 0 0
Mundri 0 6 0 0 0
Tombura 31 6 17 12 0
Yambio 32 3 22 13 0
DRC Ango 0 56 120 12 1
Bambesa 0 0 0 0 0
Bondo 0 0 2 0 0
Poko 0 1 0 0 0
Dungu 500 306 260 84 8
Faradje 184 50 2 9 3
Niangara 20 360 100 4 1
Watsa 0 100 0 1 0
CAR Djemah 0 26 0 0 0
Obo 6 86 13 4 3
Zémio 0 4 49 3 7
Yalinga 0 0 8 0 0
Bakouma 0 0 8 1 8
Bangassou 0 2 5 0 2
Rafaï 0 2 57 8 16
Birao 0 0 1 0 0
Ouanda Djallé 0 0 2 0 0

Table 10: Civilian Abductions - Raw Counts by Territory
Country Territory 2008 2009 2010 2011 2012
SSD Raja 0 13 0 11 0
Wau 0 0 3 0 0
Nahr Yei 25 6 0 0 0
Meridi 8 6 3 0 0
Mundri 0 9 0 0 0
Tombura 70 31 36 18 0
Yambio 55 5 43 21 0
DRC Ango 0 428 146 113 29
Bambesa 0 0 0 5 0
Bondo 0 0 3 0 18
Poko 0 17 6 0 0
Dungu 362 257 283 198 181
Faradje 190 19 21 62 54
Niangara 0 260 116 19 11
Watsa 0 20 2 2 0
CAR Djemah 0 28 0 0 0
Obo 133 109 27 29 5
Zémio 0 8 57 81 33
Yalinga 0 0 151 0 26
Bakouma 0 0 22 54 68
Bangassou 0 1 29 6 52
Rafaï 0 38 272 13 34
Birao 0 0 37 0 0
Ouanda Djallé 0 0 88 0 0

Abstract (if available)

Abstract Understanding the geographic distribution of insurgency violence is critical for assessing where counter‐insurgency and civilian protections operations are effective. It allows researchers and policymakers to detect trends in violence and propose local programs designed to quell insurgency aggression in vulnerable areas. This thesis examines the spatial distribution of armed‐conflicts in Central Africa committed by the Lord’s Resistance Army from 2008 to 2012 and offers a descriptive evaluation regarding the geographic fluctuation of violence throughout the region. Existing counter‐insurgency programs are discussed, and additional analysis is performed on the development of a high‐frequency radio network designed to facilitate information sharing between communities. Resulting geographic representations indicate a steady decline in armed‐conflicts in the Democratic Republic of the Congo and South Sudan with violence becoming more prevalent in the Central African Republic. The revealed fragmentation and variance in the LRA’s operations supplement a growing body of research that seeks to better understand the geographic evolution of conflict, identify why violence may increase or decrease in certain areas, and assess the capacity for civilian protection initiatives in regions afflicted with insurgency (Buhaug and Lujala 2005

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Determining the utility of GIS in border disputes case study: Sudan and South Sudan

PDF

Spatial analysis of urban built environments and vehicle transit behavior

PDF

The geography of voter power in the U.S. electoral college from 1900-2012

PDF

Spatial narratives of struggle and activism in the Del Amo and Montrose Superfund cleanups: a community-engaged Web GIS story map

PDF

Development of a Web GIS for urban sustainability indicators of Oakland, California

PDF

A comparison of urban land cover change: a study of Pasadena and Inglewood, California, 1992‐2011

PDF

Explorations of American churchscape diversity

PDF

Semi-automated visualization of spatial information in unstructured text

PDF

Social media to locate urban displacement: assessing the risk of displacement using volunteered geographic information in the city of Los Angeles

PDF

Spatial delineation of market areas: a proposed approach

PDF

Improving the communication of water allocation decisions using interactive maps

PDF

Using pattern oriented modeling to design and validate spatial models: a case study in agent-based modeling

PDF

Investigating electoral college reform: geography's impact on elections, and how maps influence our perception of election outcomes

PDF

Spatiotemporal visualization and analysis as a policy support tool: a case study of the economic geography of tobacco farming in the Philippines

PDF

Using pedestrian accessibility indicators to locate schools: a site suitability analysis in Greenville County, South Carolina

PDF

Historical temperature trends in Los Angeles County, California

PDF

The role of precision in spatial narratives: using a modified discourse quality index to measure the quality of deliberative spatial data

PDF

An analysis of the North Rainier Elk Herd area, Washington: change detection and habitat modeling with remote sensing and GIS

PDF

Developing and implementing a GIS-based framework to identify optimal locations for clean water wells in sub-Saharan Africa

PDF

Is the likelihood of waterfowl presence greater on conserved lands?

Asset Metadata

Creator Franssen, Robert J., Jr. (author)

Core Title Evaluating spatial changes in the rate of insurgency‐violence in Central Africa: the Lord's Resistance Army 2008-2012

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 05/20/2014

Defense Date 03/12/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Africa,armed‐conflict,GIS,guerrilla,insurgency,LRA,OAI-PMH Harvest,spatial

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Warshawsky, Daniel N. (committee chair), Paganelli, Flora (committee member), Vos, Robert O. (committee member)

Creator Email rfransse@usc.edu,rjfranssen@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-414987

Unique identifier UC11296385

Identifier etd-FranssenRo-2519.pdf (filename),usctheses-c3-414987 (legacy record id)

Legacy Identifier etd-FranssenRo-2519.pdf

Dmrecord 414987

Document Type Thesis

Format application/pdf (imt)

Rights Franssen, Robert J., Jr.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA